Port Susan Bay Estuary Restoration Project

Final Monitoring Report

June 2018

Roger N. Fuller

Submitted to The Nature Conservancy of Washington

And the Washington State Estuary and Salmon Restoration Program

Photos by R. Fuller

Estuary restoration affects ecological processes like the distribution of freshwater and sediment, inundation patterns, and biophysical interactions. By improving processes, restoration can increase the productivity of tidal wetlands, and their resilience to disturbance and climate change. These photos, both taken in June, show how seacoast bulrush (Bolboschoenus maritimus) wetlands can differ, depending on the status of ecological processes.

1

Table of Contents

Page Section 2 I. Project Synopsis 6 II. Introduction and Overview

a. Project Overview 8 b. Monitoring Design 11 c. Habitat types and summary of biophysical characteristics 15 III. Summary of project objectives, hypotheses, targets, and results 17 IV. Methods and Summary of Data Collection Efforts 23 V. Discussion of Results

a. Objective 1: Restore self-sustaining native tidal wetlands that support estuarine-dependent animals.

23 i. H1.1 Tidal wetlands will develop. 42 ii. H1.2 Accretion will be initially rapid. 45 iii. H1.3 Accretion will keep up with sea level rise 48 iv. H1.4 Complex tidal channels will develop. 49 v. H1.5 Birds will use the site. 52 vi. H1.6 Soils will support benthic invertebrate prey of shorebirds. 55 b. Objective 2: Improve connectivity between the river and northern tidal habitats,

increasing the distribution of freshwater, sediment, energy and other materials. i. H2.1 Accretion rates will increase.

63 ii. H2.2 Salinity will decrease. 66 iii. H2.3 Channel systems will expand. 67 iv. H2.4 LWD will increase. 68 v. H2.5 Marsh area will increase. 76 References A-1 Appendix 1 Lessons Learned from Monitoring the Port Susan Bay Restoration Project A-2 Appendix 2 Description of image archive A-3 Appendix 3 Description of final archival database A-4 Appendix 4 WWU Vegetation and Sediment Monitoring Methods

Suggested Citation: Fuller, Roger N. 2018. Port Susan Bay Estuary Restoration Project: Final Monitoring Report. Report prepared for The Nature Conservancy. 100pp. Author’s contact information: Roger Fuller rfuller.750@gmail.com

2

I. Project Synopsis Overview The Stillaguamish estuary in Port Susan Bay supports extensive brackish tidal marsh habitats that are critical to an array of wildlife species, including dozens of species of salmon, crustaceans, shorebirds, waterfowl, and raptors. Over the past 150 years, humans have altered the estuary, substantially reducing the amount of habitat and number of channels. In addition to the direct loss of habitat, the constructed levee system also constrains the remaining channels between levees until they reach the brackish tidal marsh. This constraining of the channels substantially affects key processes like the pattern of distribution of freshwater and suspended sediment across the estuary ecosystem. After most of the original tidal marshes were diked and converted to agriculture in the 1800’s, rapid accretion and new tidal marsh expansion occurred outside the dike system. Marsh area expanded, peaking in 1990 with about 1,064 acres south of South Pass. However, since 1990 there has been a sharp 26% decline in total marsh area as changes in ecological processes drive marsh retreat. To reverse this habitat decline, restoration of 150 acres at the mouth of the Stillaguamish River was completed in 2012 by The Nature Conservancy. The project removed a dike surrounding the site and carved two channel connections to a small side-channel of the river. With the site located right at the mouth of the river, pre-project modeling indicated that direct river flow across the site would occur at high tide. The habitat objectives of this project were to:

restore tidal marsh and channel habitat on the 150-acre footprint of the restoration site, and to

improve the distribution of freshwater and sediment from the river across the estuary ecosystem in order to increase system resilience and reduce the rate of marsh loss.

This report summarizes the findings of several independent research groups that contracted with The Nature Conservancy to monitor the outcomes of restoration. The research groups include Western Washington University, U.S. Geological Survey-Pacific Coastal and Marine Science Center, U.S. Geological Survey-Western Ecological Research Center, and the Skagit River System Cooperative.

Overall Monitoring Findings Key overall findings are summarized here and described in detail in the body of the report. Site-scale Habitat Marsh Area: Tidal marsh plants underpin the ecosystem services we seek from estuaries by generating the biomass that fuels the food web for fish and birds, and protects human communities from coastal flooding. Tidal marsh on the site initially proliferated rapidly as seacoast bulrush (Bolboschoenus maritimus) and river bulrush (B. fluviatilis) expanded from the smaller non-tidal wetland that existed prior to dike removal. Marsh covered most of the site within two years. As the marsh rapidly expanded upslope, it retreated slowly at first from the lowest elevations, as expected. But in 2015 the marsh almost completely died on the lower 2/3 of the site. Record low river levels had caused very high soil salinity, a stress that we think may have combined with other stresses, leaving the plants vulnerable to insect attack. The stem-boring larvae of a moth wiped out the marsh. 2015 river levels correspond to those projected to be average sometime after 2050 due to climate change. As such, dynamics in 2015 provide us with insights into how climate change affects both the restoration site and the broader estuary. In 2017, with higher than average flows, some marsh areas recovered but others did not. Currently there is less marsh area and more unvegetated mudflat than predicted, and the plant species expected to be most common, Schoenoplectus pungens (3-square bulrush), is only present at trace levels.

3

Accretion: Sediment deposition is critical for the development of productive tidal marsh and to help marshes keep up with sea level rise (SLR). Deposition has led to an annual rate of elevation change on the site of 3.1 cm/year. This is a substantial rate, about 3 times higher than the average rate in reference marshes. Between 2014 and 2015, the restoration site received a sediment load of about 26,000 tonnes, about 1% of the total load from the river. Accretion has been slowing slightly and is expected to slow over time as the subsided site gets closer in elevation to the surrounding natural marshes. If we assume a local sea level rise projection of 82cm by 2100 and that the rate changes non-linearly, current rates of elevation change at the restoration site will keep up with SLR beyond 2100. However, current rates may not reflect long-term background rates of elevation change. Tidal channels: Tidal channels are critical to key ecological processes such as delivery of water and sediment, exchange of the organic matter that fuels the food web, fish access, and drainage of the site. Channel development has exceeded model predictions in terms of total length. After restoration, total channel length increased 10 fold to 23,266 meters in 2015. The prediction for the site was 4,458 m. However, an underlying hardpan layer may be limiting channel depth and site drainage rates. In addition, the linear pathways created by construction truck traffic have become channels, with unknown effects on the developing channel system. Estuary-scale habitat An important objective of the project was to improve connectivity between the river mouth and the rest of the estuary, allowing the river’s freshwater and sediment to regularly reach a larger part of the system. Marshes north of Hatt Slough have been declining in productivity and resilience, and gradually disappearing. Pre-project modeling suggested that restoration at the mouth of Hatt Slough would allow freshwater to be pushed by the high tide across the restoration area and towards the north of the bay. This has not occurred, with river water only flowing from Hatt Slough directly onto the site during high winter flood levels. During spring and summer, when soil salinity limits marsh productivity, there is no extra river flow towards the north during high tides. As a result, marsh decline continues north of the restoration area and there is no indication of improved connectivity at the broader system scale.

Lessons Learned Lessons learned have been summarized in four categories. Some of the main lessons include: 1. Restoration Design Elements

Pre-restoration conditions that affect post-restoration habitat development

Agricultural legacies affect site topography and soils in a way that will strongly influence post-project conditions including tidal drainage and soil physical and chemical conditions. Significant topographic or soil work may be necessary, including channel excavations, re-contouring, sediment addition, or sub-soiling, to achieve desired habitat results. Post-project monitoring should include soil saturation and chemical conditions.

Number and size of breaches The number and size of breaches was insufficient to support tidal flooding and ebbing at velocities and rates similar to reference marshes. New blind tidal channel connections were expected to form across the old dike footprint but have not, due to the highly compacted dike footprint that resists erosion. For the same reason, expansion of the two breach sites and channel downcutting is less than would be expected.

4

2. Design and Construction Process

Documentation of design assumptions and impact pathways

Be explicit about the impact pathways that are expected to deliver the project objectives. Objectives are usually desired habitat functions such as new habitat and improved ecological processes. However, restoration actions alter the physical structure of a site, under the assumption that this will result in the desired functional outcome. It is important to describe in detail the assumptions about the impact pathway between altered structure and new functions. If these assumptions aren’t explicitly documented and included as design constraints, the design process may miss key project elements. For this project, the objective of improved estuary-scale hydrological connectivity has not been achieved because daily flood tides are not sufficient to overtop the Hatt Slough bank along the restoration site as was expected. Key design assumptions made at the beginning of the project did not translate into final construction drawings, resulting in as-built conditions that didn’t match conceptual design model outputs.

3. Monitoring

Tidal wetland hydrology Marsh plain drainage at low tide was assumed to be complete because a hydro logger in the main site drainage channel registered water levels 1m or more lower than the marsh plain. However, much of the marsh surface appears to have retained shallow standing water and saturated soils at low tide for at least three years, likely causing stressful soil conditions for plants. Channel loggers are an insufficient and indirect method of monitoring marsh surface hydrology. Soil saturation may have been a source of stress that contributed to a large marsh dieback event in 2015.

Soil physical and chemical conditions Soil physical and chemical conditions were not part of the monitoring plan, and we thus have no way of testing our hypothesis that overly saturated soils led to soil conditions that were detrimental to plant health and contributed to the marsh dieback event. Specifically we had no direct indicators of saturated soils, oxygen exchange, redox potential, sulfide levels, nutrient processes, or heavy metal species.

Importance of system-scale perspective and multiple reference sites A system-scale perspective and multiple, ecologically varied reference sites are important to understanding restoration responses and possible effects of climate change at site and system scales. Due to the legacy of drainage, farming, or other land-use impacts, most estuary restoration projects will not look similar to reference sites in the near future. However a range of reference sites allow a much improved ability to project likely outcomes at the restoration site. Without multiple reference sites, we would still be befuddled by plant and habitat dynamics at both the restoration site and estuary scale.

Importance of seasonal data on vegetation structure The physical structure of plant communities, in both summer and winter, may be more important to ecological processes and ecosystem functions than metrics like species richness and cover.

Importance of long term reference data sets Long-term reference site data is key, and the Salish Sea needs a network of reference sites that can collectively inform estuary restoration projects throughout the region.

5

Non-native invasives monitoring and response plan

Consider whether invasive non-native plant species such as Typha angustifolia (narrow-leaved cattail) are likely to invade, and even if you don’t expect them, include them in your monitoring plan, and decide beforehand how you would respond if they appear and expand.

4. Species, Habitat, and Ecosystem Responses

Some vegetation responses were a surprise. Tidal marsh retreated farther than expected, the bulrush S. pungens has not colonized the site substantially, and both cattail species (Typha) are playing larger roles than projected.

Bolboschoenus fluviatilis, river bulrush, is a common and important estuarine species. The presumed Bolboschoenus maritimus (seacoast bulrush) meadows in the Stillaguamish estuary turn out to be co-dominated by B. maritimus and B. fluviatilis (river bulrush). The latter species appears to be common in local estuaries, though it has not been documented in Washington or Oregon estuaries previously. This is likely because the two species are nearly identical. It holds potential as an agent of tidal marsh resilience, with high biomass production, very robust above-ground biomass during the dormant but geomorphically important winter storm season, and with potential resistance to summer salinity stress. Its ecological preferences, such as elevation and soil structure, could be considered in developing restoration project design targets in places where B. maritimus or S. pungens are common.

The most important, direct, individual effect of climate change is likely to be declining summer river flows. Declining flows will affect soil salinity, marsh productivity, and resilience during the tidal marsh growing season. Actions that enhance freshwater residence time in the estuary should be priorities.

Interaction of stresses is important, and may be key to understanding climate change impacts. The greatest impacts from climate change will come as a result of the multiple ways that climate change interacts with existing sources of stress and disturbance. Surprising and rapid change can happen from a previously un-recognized source of stress, when the additive effects of multiple sources of stress push a system over a tolerance threshold. These ideas were illustrated by the 2015 dieback of 50 acres of marsh, which involved a previously unknown moth herbivore, extremely high soil salinity levels, and an assumed (but undocumented) legacy of plant stress resulting from over-saturated soils.

6

II. Introduction and Overview Project Overview The Stillaguamish estuary in Port Susan Bay (PSB) supports extensive brackish tidal marsh habitats that are critical to an array of wildlife species, including several species of salmon and dozens of species of shorebirds, waterfowl, and raptors. Over the past 150 years humans have altered the estuary, substantially reducing the amount of habitat and number of channels, and constraining the remaining channels between levees. In addition to the direct loss of habitat through conversion to agriculture, the levee system affects the remaining habitat by altering key processes like the pattern of distribution of freshwater and suspended sediment. These large-scale modifications to the physical character of the estuary have altered its ecological functioning and its capacity to support endemic estuary-dependent species. After most of the original tidal marshes were diked and converted to agriculture in the 1800’s, rapid accretion and new tidal marsh expansion occurred outside the dike system. Marsh area expanded, peaking in 1990 with about 1,064 acres south of South Pass. However, since 1990 there has been a sharp 26% decline in total marsh area as changes in ecological processes drive marsh retreat, particularly in the northern half of the estuary.

Figure 1. The Stillaguamish estuary, showing the position of the 2012 restoration site outlined in white, adjacent to the primary distributary, Hatt Slough.

7

To slow or reverse this habitat decline, restoration of 61 hectares (150 acres) at the mouth of the Stillaguamish River was completed in 2012 (Figure 1). The project removed a dike surrounding the site (Figure 2). Due to subsidence, the restoration site was up to 1m lower in elevation than the surrounding tidal marsh, so simply removing the dike would result in a “bathtub”, limiting drainage, tidal marsh development, and fish movement. As a result, two channel connections (breaches) were carved to connect the site to a small side-channel of the river, to facilitate full tidal exchange at the site. Former farm drainage ditches were filled, with the exception of a remnant of historical tidal channel that had been incorporated into the farm drainage system. A flood and fish bypass structure was built into the southeast corner. This structure is activated when the river overtops the levee system upstream and floods Florence Island farmland east of the restoration site. Floodwaters and associated fish can evacuate the farmland and re-enter the river system through this structure. With the site positioned right at the mouth of the river, pre-project modeling indicated that direct freshwater flow across the site would occur at high tide. This project was expected to have significant ecological effects at both the site and estuary scales.

Figure 2. Elements of the 2012 restoration project.

The primary objectives for the Port Susan Bay Estuary restoration project were to:

1. Restore self-sustaining native tidal wetlands that support estuarine-dependent animals (site scale); 2. Improve connectivity between the river and northern tidal habitats, increasing the distribution of

freshwater, sediment, energy and other materials (estuary scale); 3. Improve juvenile salmon access to restored rearing habitats (site scale); and 4. Improve flood attenuation for neighbors in the lower river valley.

The first two objectives were the subject of an ESRP-funded Learning Project (PRISM record number 11-1650), for which this report summarizes monitoring results. For each of these objectives, the Port Susan Bay Restoration Monitoring Plan (Woo et. al. 2011) identifies a series of hypotheses. The monitoring program was

8

designed to evaluate the hypotheses and measure progress toward the desired objectives. Section III (Table 4) describes the first two objectives, and their associated hypotheses, baseline conditions, restoration targets, and documented progress. The remainder of the report provides details on the progress towards project objectives. Monitoring Context To achieve the two primary site and ecosystem scale habitat objectives requires changing the fundamental physical drivers of habitat formation in a way that facilitates the development and increased functionality of habitats. The project hypothesis was that removing the dike would change those physical drivers in a way that would deliver better habitat at both site and system scales. The main physical drivers of habitat development relate to the processes of sediment and water distribution across the estuary, and to the role of disturbance in modifying those processes. Sediment processes include the distribution of sediment, sorting of particle sizes, and the dynamics of accretion and erosion. Water processes include the distribution of freshwater and its effects on above and below ground salinity, and the hydrodynamics of river and tidal interactions including marsh inundation patterns and soil wetting and drying. Important drivers of disturbance include winter wind storms, river floods, rhizome excavation by snow geese and swans, and to a lesser extent, other forms of herbivory such as by insects. In addition, beyond the outer edge of tidal marsh, distributary and blind tidal channels become shallower and less stable, and frequently migrate across the tide flat, a source of regular disturbance that can affect the process of marsh expansion. However, the outcome of restoration on tidal marsh functions depends not just on the physical processes described, but also on how those processes are modified by the vegetation. For example, different plant species composition in low marsh habitat results in different vertical structure of vegetation during winter. Winter vegetation structure, in turn, affects the amount of suspended sediment that gets captured for accretion, soil vulnerability to winter wave erosion, and accessibility of the marsh to grazing geese. For this reason, monitoring for this project focuses on measuring

how the physical drivers operate at both site and system scales

how vegetation patterns interact with physical drivers to affect long-term habitat development

how restoration affects the outcomes of those biophysical interactions and delivers increased habitat functions.

Monitoring Design Five Study Zones We organized monitoring around five primary zones in the delta (Figure 3) in order to understand the dynamics at the restoration site and the effect of the restoration on different areas of the delta. The five zones encompass the restoration site and four reference sites that span the spectrum of physical drivers of estuarine habitat development. The physical drivers vary substantially over space, with the strongest patterns in relation to distance from the main river mouth at Hatt Slough. The five study zones allow us to capture this range of variability. This structure also provides an excellent baseline that facilitates future monitoring and research related to long-term effects of old and new restoration projects, climate change impacts, and basin-scale land use changes. In addition to the five main study zones, some data was gathered in an additional three study zones in order to better characterize the effects of substrate differences on vegetation. Each study zone is a discrete area capturing, to the extent possible, the entire drainage basin of one or more blind tidal channel systems. The five study zones represent different points on a gradient of hydrogeomorphic connectivity with the river, and allow an analysis of the effects of landscape position on restoration response.

9

As such, various biophysical characteristics are expected to differ among the five zones, including salinity and sediment dynamics, vegetation structure, hydraulic energy, water chemistry, soil particle size and chemistry, invertebrate and vertebrate composition, etc. Zones 1-5 are numbered in order of the relative importance of river flow to their physical conditions, beginning with the southernmost zone. A description of each study zone follows. A summary of general physical conditions and indicators of some key physical processes are provided in Table 1. General soil characteristics are provided in Table 2, and vegetation characteristics are in Table 3. Descriptions of the Study Zones Zone 1 is an island to the south of the Hatt Slough mouth and is the zone under the greatest influence of the river’s freshwater, being closest to the mouth of Hatt Slough (Table 1) and with two of the largest distributaries defining the boundaries of the island. The zone has one major blind tidal channel that drains to the river’s largest distributary, and several smaller channels. Although it is on the south flank of the delta, Zone 1 has one of the lowest wave exposures (Table 1) as a result of a headland that butts out to the south of it. In the original monitoring plan (Woo et al. 2011), this zone was framed as the restoration reference site for the project. However we have found it important to have four reference sites because conditions on the restoration site are different from all four reference sites and having the spectrum of conditions represented by all four sites allows us to better understand and predict what is happening on the restoration site.

Figure 3. 2016 aerial photograph of the Stillaguamish estuary, indicating the locations of the restoration and study zones. Zone 2 is the 150-acre restoration site. Zones 1 and 3-5 are the primary reference sites, spanning a spectrum of connectivity to Hatt Slough, the source of freshwater, sediment, fish and other matter. Additional data was collected in Zones 6-8 to better characterize the relationship between vegetation composition and substrate.

10

Zone 2 is the restoration zone itself. Since it abuts the dike system, it is not a true marsh island, but it is drained by blind tidal channels as are the other sites. Hatt Slough runs along its southern edge, but no flow from Hatt Slough moves directly onto the site except at flood flows. A small distributary channel flows north along much of the western edge and is the channel feeding the two blind tidal channel systems that penetrate the restoration zone. The restoration zone has the gentlest slope gradient (Table 1), as a result of the legacy of farming practices. It experiences relatively low grazing by snow geese, compared to most reference sites (Table 1), which may be related to soil properties. Zone 3 is an island just to the north of the mouth of Hatt Slough and west of the restoration site. This site is similar to zone 1 in that it is surrounded by two distributaries, though they are much smaller than those surrounding zone 1 and so is under less freshwater influence. Being on the same side of Hatt Slough and under less freshwater influence than zone 1, this may be the closest site to the restoration site in terms of reference conditions. Zone 3 is on the north-facing flank of the concave delta, giving it a lower wave exposure than zones 6 or 7 on the south flank (Table 1). It experienced the highest goose disturbance rate, along with the mid-delta zone 4. Two major blind channel systems drain the island, along with a few small channels. Zone 4 occurs in the mid-delta region, about half way between Hatt Slough and South Pass. There are no distributaries in this area so the site includes a marsh plain drained by two blind tidal channel systems. Similar to the restoration site, this site abuts the dike system and so is not a true marsh island. Zone 4 represents the mid-delta marshes that are most disconnected from the river and potentially the most vulnerable to accretion and salinity issues in light of climate impacts. The restoration project may result in improved connectivity for this zone. This marsh plain in this zone has the steepest slope gradient, which may be a characteristic of an eroding marsh. Zone 4 has lost 69% of its area since 1990 (Fuller and McArdle 2014) and continues to erode. It has a medium wave exposure and a high rate of goose disturbance (Table 1), factors that combined appear to lead to rapid erosion. Zone 5 is near South Pass which is under some level of influence from the Old Stilly channel and the south fork of the Skagit via West Pass. However hydrodynamic monitoring by USGS suggests that tidal dynamics dominate in the Old Stilly channel except during high river flows, and riverine influence is quite minor. Hydrodynamically it appears to more closely resemble a blind tidal channel than a river distributary. Growth of barnacles and bladder wrack on hard surfaces in this zone are an indication of the higher salinity and lack of significant freshwater delivery to this zone. Zone 5 has the oldest marsh, recorded in the 1886 U.S. Coastal and Geodetic Survey map, and it has the highest wave exposure and a relatively high rate of goose disturbance (Table 1). It has lost 60% of the marsh area it had in 1964 (Fuller and McArdle 2014). Zones 6-8 were added near the mouth of Hatt Slough when it became evident that vegetation composition and productivity varied substantially among adjacent estuarine islands near Hatt Slough. We wanted to better characterize substrate-plant relationships in order to better understand the fate of vegetation development at the restoration site, and the potential marsh responses to climate change. Zone 6 is the lowest elevation study zone and is on the northwestern flank of the delta developing at the mouth of Hatt Slough. This marsh area is the youngest in the system, having developed since a flood in winter 2006/7 that triggered channel migration away from the area, rapid accretion, and reduced hydraulic disturbance. Being most seaward on the delta, it has the highest wave exposure (Table 1). Zone 7 is on the southern flank of the delta and west of zone 1 from which it is separated by the largest of the Stillaguamish distributaries. It has relatively high wave exposure but low evidence of goose disturbance (Table 1).

11

Zone 8 is on the southern flank of the delta, east of zone 1 and abuts the dike on the lower Stillaguamish floodplain. A smaller distributary flows south along this zone. Study zones on the southern flank of the delta are exposed to greater freshwater influence since this is where the major distributary empties. Zone 8 is adjacent to a protruding headland that protects it from winter southerly winds so it has a low wave exposure index and it showed no evidence of goose disturbance (Table 1). Habitat types and summary of biophysical characteristics Differentiating habitat types is important for understanding how their ecological functions such as food web support and flood protection differ across both space and time. In addition, habitat change in response to restoration or climate change is mediated by the interaction of physical and biological processes, and those interactions differ among habitat types. For example, sediment accretion at a site is a result of physical forces such as the amount of sediment delivered to a site and how the interaction of fluvial and wave regimes affect the balance between accretion and erosion. However these dynamics are strongly modified by biological interactions such as how the seasonal changes in vegetation structure affect the trapping of sediment, and how herbivores alter vegetation structure and disturb soils.

Table 1. Summary of physical conditions and indicators of key physical processes in the study zones and marsh types. The reference zones are listed in geographical order from south to north. Marsh age is calculated from the base year of 2015 using aerial photos and historic maps. Since historic aerials were irregular in timing, marsh ages are not exact but do allow a relative comparison between study zones. The age of the oldest marsh (130 years) doesn’t represent its actual age but only the first time that the tidal marsh boundary was recorded, in the 1886 US Coastal and Geodetic Survey map. The Wave Exposure Index integrates wind direction, temporal distribution of winds, fetch length, water depth, and surface roughness in the approach. The Grazing Disturbance Index (GDI) integrates three metrics of grazing intensity by snow geese, including grazer footprints, excavations, and exposed rhizomes associated with excavations. The GDI numbers presented here were from observations during the 2014-2015 winter and spring. Hatt Slough is the main mouth of the river, so the Distance to Hatt Slough is a measure of the relative level of freshwater influence. High, middle, and low marsh are defined in the text. Table is from Fuller 2017, which contains additional background information.

12

Many studies differentiate two types of regularly flooded tidal wetlands, high marsh and low marsh. High marsh is usually defined as those wetlands that tidally flood only during the higher high tide or less frequently, and as a result they generally occur above Mean High Water (MHW). High marsh is usually dominated by grasses, sedges, rushes, and a diversity of dicots. Low marsh is usually defined as tidal wetlands that flood daily during all high tides and as a result generally occur between Mean Sea Level (MSL) and MHW. On river deltas in this region, brackish low marshes are usually dominated by species of bulrush or sedge. However, we have found that within the low marsh category, different plant species combinations can result in very different geomorphic effects in terms of how they interact with physical processes. For example, different species of bulrush have very different physical structure during winter which is when critical sediment processes are most active. Winter flood season is when most riverine sediment is delivered. Winter wind storm season is when waves and tides maximize their erosion and sediment re-sorting processes. And winter is when key tidal marsh herbivores, snow geese and swans, are present and actively disturbing the sediment as they dig for bulrush rhizomes. The persistence of senescent above-ground biomass differs substantially among bulrush species and for this reason species composition has an effect on vulnerability to grazing (snow geese and swans avoid tall, dense structure), resistance to wave erosion (structure substantially decreases wave energy), and ability to capture sediment (structure slows water movement, allowing sediment to settle out). For this reason, in order to better characterize tidal marsh differences, we split the low marsh category into two parts, calling them middle marsh and low marsh. The three marsh types, high, middle, and low, are easily distinguished in aerials by their different color and textural signals. High marsh generally occurs above MHW (2.5m NAVD88 in the Stillaguamish delta). This marsh type lacks bulrush species, has high species diversity, and is dominated by Carex lyngbyei, Agrostis stolonifera (formerly Agrostis alba), Juncus balticus, Distychlis spicata, or Potentilla anserina. Vegetation structure is generally shorter but more dense than middle or low marsh. Middle marsh on the Stillaguamish delta generally occurs below MHW, between about 2.0 - 2.5 m NAVD88, and is characterized by the consistent presence of several species of bulrushes and cattails. This marsh is flooded by each high tide, but being closer to the mean high tide line limits the depth and duration of inundation. Middle marsh is tall, often chest high or higher, due to the tall dominant species, including Bolboschoenus maritimus (seacoast bulrush), B. fluviatilis (river bulrush), and Schoenoplectis tabernaemontani (soft-stem bulrush). S pungens (3-square bulrush) is also common, though is less tall than the others. In a few of the fresher parts of the estuary cattails are common in the upper reaches of middle marsh. Typha latifolia (broad-leaved cattail) and Typha angustifolia (narrow-leaved cattail) both occur. Middle marsh species differ in terms of the persistence of their senescent above-ground biomass during winter. S. pungens is the least persistent and is usually completely flattened by December. As a low mat of decomposing biomass, it still plays an important role in trapping and holding sediment at the soil surface, but without erect stems it interacts little with waves or suspended sediment. The remaining bulrush species are listed here in order of increasing above-ground biomass during winter: B. maritimus, S. tabernaemontani, B. fluviatilis. B. fluviatilis can maintain much or all of its senescent biomass throughout winter, having substantial effects on waves and sediment, and resisting grazing by geese. Low marsh on the Stillaguamish delta generally occurs between about 1.3 – 2.0 m NAVD88, and is characterized by low species richness and shorter heights. The dominant species is S. pungens. B. maritimus is occasionally present though at insignificant densities. Low marsh can be very dense and lush, reaching above waist high, and producing a large amount of biomass. S. pungens has smaller stems than other bulrush

13

species, so low marsh stem densities can be much higher than middle marsh densities. During the growing season, low marsh can be very active in sediment capture as a result of the dense stems and longer inundation periods than high marsh. However since S. pungens has relatively weak stems compared to the other bulrush species, it tends to be flattened by tidal and wave action fairly quickly as it begins to senesce in late August. As a result, low marsh is easily invaded by grazing geese and swans during fall and winter, and there is little structure present to slow currents, capture sediment, and reduce erosive wave energy during winter.

Table 2. Soil characteristics in the study zones of the Stillaguamish estuary. Reference zones are listed in geographical order from south to north. Data are from 2015. Methods are described in Appendix 4. Particle size distribution methods are described in Grossman and Curran 2015. Table is adapted from Fuller 2017.

14

Table 3. Vegetation characteristics of the study zones in the Stillaguamish estuary during 2015. Reference zones are listed in geographic order from south to north. Aboveground bulrush biomass volume was estimated mathematically based on plot stem density times the volume per stem, with stem volume calculated from measured heights in each plot, and the average stem diameter calculated from several dozen measurements taken from throughout the estuary. Separate biomass calculations were made for each bulrush species, and each species was adjusted with a leaf biomass factor to account for the different degrees of leafiness among species. All bulrush species in a plot were then summed for each plot to arrive at the total bulrush aboveground biomass volume per plot. Species richness is the total number of species found in a plot. For this study, species diversity was defined simply as the number of dominant and sub-dominant species (definitions provided in the methods in Appendix 4). Table is adapted from Fuller 2017.

15

III. Table 4: Summary of project objectives, hypotheses, targets, and results discussed in this report. Subsequent report sections provide definitions and details on the results reported.

Objective Hypothesis Parameter sampling

Baseline Restoration Target 2013 2014 2015 / 2016

Objective 1: Restore self-sustaining native tidal wetlands that support estuarine-dependent animals.

H1.1 Site elevations, inundation levels, and salinities will support the establishment of high marsh, low marsh and tide flat.

inundation and hydrology

There is no tidal exchange at the site. 151 ac (97%) is in the potential elevation range for low-middle marsh (MSL-MHHW), and 4 ac is in high marsh range.

Re-introduce tidal exchange and tide waters will daily inundate most of site, and evacuate site completely at low tide.

Tidal inundation and evacuation evident in hydro data (Woo et. al. 2014). However, topography and lack of channels inhibits full marsh surface drainage (Fuller 2014a).

Tidal inundation and evacuation apparent in 2014 hydro data (Woo et. al. 2015a). 2014 aerial indicates improved drainage in the northern area of marsh, but not southern.

2015 aerial indicates most of surface now drains, though major marsh dieback. 2016 shows almost complete drainage of tidal floodplain at low tide.

salinity

2004-2005 data show average pore water salinity of 9. Middle marsh salinity in reference zones: Zone 1=3, Zone 3=7, Zone 4=10, Zone 5=13. (Fuller 2016)

Pore water salinity will freshen with tidal flushing and become closer to zone 3 and 1 than to zone 4.

No pore salinity collected. Hydro logger data show surface salinity is closer to zone 1 than 5, except during summer (Woo et. al. 2014).

No pore salinity collected. Hydro logger data show surface salinity is closer to zone 1 than 5, except during summer (Woo et. al. 2014).

2015 spring pore salinity was = zone 4, > zones 1,3. 2015 summer salinity was = zone 4, > zone 1 (Fuller 2016). Avg surface salinity was more similar to zone 1 than 4 (Woo et. al. 2015a).

habitat/ vegetation, remote sensing

2005 non-tidal wetlands: 74 acres (Heatwole 2006a). 2011 wetlands inventory: 0 acres of tidal marsh (Fuller and McArdle 2014).

Restore diked area (150 ac) to tidal marsh and tide flat habitat.

2013 aerials – Continuous marsh 98 ac; Patchy marsh 5 ac; Low density marsh 0 ac (Fuller and McArdle 2014)

Habitat analysis not completed, though continuous marsh area declined, patchy marsh increased, and unvegetated area increased.

Habitats not digitized, though total marsh area has declined substantially.

habitat/ vegetation, vegetation transects

2005 non-tidal wetland plots: 30% total cover, 44 bulrush stems/m2 (Heatwole 2006a)

Restore diked area (150 ac) to tidal marsh and tide flat habitat.

No summer vegetation data collected.

2014 vegetation transect shows 52% plant cover, all middle/low marsh, 160 bulrush stems/m2. No high marsh. (Fuller 2015a, revised)*

2015 vegetation transect shows 34% plant cover, 82 stems/m2. No high marsh. (Fuller 2017)

H1.2 Restored tidal exchange and inundation will result in sediment accretion on the marsh plain, initially rapidly until site elevations approach elevations in adjacent marsh.

hydrology, elevation The restored area was subsiding due to the disconnection from estuarine and riverine processes.

Tidal exchange and sediment elevation will approach that of the reference marsh.

No results to report.

1.1-3.7cm sediment accretion over 22 weeks (Oct 2013 – Apr 2014) from pins (Fuller 2014). SET baseline data collected in 2014.

1.75 – 6.17 cm sediment accreted (at 4 SET sites) in restored area from July 2014 – May 2015 (Rybczyk and Poppe 2015).

H1.3 Restored inorganic and organic accretion will result in accretion rates that will keep up with moderate projections for local sea level rise (SLR).

elevation, sedimentation, soils

The site is subsiding due to disconnection from estuarine and riverine processes.

Inorganic and organic accretion will be similar to that of the reference marsh.

No results to report.

The accretion rates in the reference marsh (Zone 1) exceed the current and predicted rates of eustatic sea level rise. SET data not available for restored area.

All 2016 marsh SETs increased faster than current and some projected SLR rates. Avg annual accretion in restored area (3.8 cm/yr) exceeds reference marshes (0.91 cm/yr).

H1.4 Restored tidal exchange will re-introduce sediment transport and scouring of tidal channels on the project site, resulting in the development of a complex blind tidal channel network.

channel development (channel cross-section, length and head-cutting)

Baseline total channel length was 2,310 meters. The large historical remnant channel cross-section was not monitored in 2011 due to flooded conditions (Hood 2011).

Increased channel network and blind channel habitat types.

Total channel length was 18,447 meters. The historical remnant channel cross-section area was 10.29m2. There were 12 incipient tributary channels observed.

Not monitored.

The 2015 total channel length was 23,266 meters. The historical remnant channel cross-section area was 6.11m2 and continued evidence of channel head-cutting observed.

H1.5. Restored tidal inundation patterns will facilitate site use by a diversity of birds.

bird survey (area by grid)

Site is used by diverse species that vary seasonally. 26 species were recorded March–July 2012.

Restoration will benefit a diversity of estuarine bird species.

Bird surveys were not completed in 2013.

October 2014 – May 2015 there were 35 bird species identified in the restoration area.

No data collected.

H1.6 Restored tidal exchange will result in fine and organic sediment qualities that support primary productivity and benthic invertebrate prey of shorebirds.

invertebrates and invertebrate taxonomic group

42 taxa represented by 14 classes were collected in 2005. Higher elevations had finer substrates (Heatwole 2006a).

Benthic invertebrate prey will not be limiting for estuarine birds in the estuary.

The overall biomass of benthic invertebrates decreased from 2012-2013 however the diversity increased.

Not monitored.

Invertebrate biomass increased 70X post-restoration at the restoration site. Observed increase was a result of growing abundances of bivalves, sabellid polychaetes, and insect larvae (Woo et. al. 2015)

16

Objective Hypothesis Parameter sampling

Baseline Restoration Target 2013 2014 2015 / 2016

Objective 2: Improve connectivity between the river and northern tidal habitats, increasing the distribution of freshwater, sediment, energy and other materials.

H2.1 Accretion rates in existing marsh and tidal flat north of Hatt Slough will increase and accretion rates south of Hatt Slough will not change.

SET & other sediment monitoring

N/A Dike removal will result in sediment accretion in the tide flat north of Hatt Slough.

Analysis not completed.

Marshes to the north of Hatt Slough are accreting sediment however there is also erosion occurring at the high marsh edge.

Marshes north of Hatt Slough are accreting sediment though it is not clear that restoration has influenced rates. The tidal flat SET 1km west of the restoration area eroded 0.32cm/yr from 2012-2015, but accreted ~1.5 cm 2015-2016. (Rybczyk and Poppe 2016)

H2.2 Water column and pore water salinity will decrease north of Hatt Slough and will not change south of Hatt Slough.

water salinity, hydrology

Pore water salinity north of Hatt Slough and south of Hatt Slough TBD (Heatwole 2006a)

Dike removal will result in decreased salinity north of Hatt Slough.

N/A Analysis not completed.

The average pore water salinity in the restored area was higher than the reference sites near the mouth of Hatt Slough. There is insufficient pre-project data to detect an effect of restoration on salinity north of Hatt Slough. (Fuller 2017)

H2.3 Distributary and blind tidal channel systems north of Hatt Slough will expand in size and complexity.

channel allometry

There is a one remnant channel in the restoration footprint that is 620m long and 0.46 ha (Hood 2011).

Increased channel complexity and development in the estuary. Predictive model suggests: Restoration of 60 ha will result in 11 tidal channels, total length = 4,458 m, total area = 0.96ha (Hood 2011)

The total length of tidal channels increased from 2,310m prior to restoration to 18,447 in 2013.

Not monitored.

Total channel length in 2015 was 23,266m and comparable to North and South Fork Skagit deltas but greater than Stillaguamish reference sites and larger than the allometric prediction of 4,015 m for this site. (Hood 2016).

H2.4 The amount of large woody debris will increase in the tidal marsh north of Hatt Slough.

aerial photos used to digitize LWD

TBD: Number of LWD in tidal marsh north of Hatt Slough by 2011 aerial photos.

The amount of LWD in the restoration site and the tidal marsh north of Hatt Slough will increase.

The restoration site had 253 log units and LWD covered 1,500 m2. The tidal marsh north of Hatt Slough Zones 3 – 5 had 176, 1439 and 660 log units respectively and the area of wood ranged from 524 m2 – 10,614 m2.

The restoration site had 237 log units (decrease from 2013) that covered 857 m2. The tidal marsh north of Hatt Slough Zones 3-5 increased in LWD both number of log units and area, except for Zone 5 where a decrease in both were observed.

Not monitored.

H2.5 Tidal marsh area west and north of the restoration footprint will expand.

aerial photo, vegetation

(Zone 4 & 5) Continuous marsh – 90.84 ac Patchy marsh – 0ac Low density marsh – 5.02 ac

Increase acreage of tidal marsh.

(Zone 4 & 5) Continuous marsh – 91.1 ac Patchy marsh – 12.58 ac Low density marsh – 59.83 ac

TBD

Total marsh area in the Stillaguamish estuary continues to decline as northern marshes continue to erode. (Fuller 2017)

17

IV. Methods and Summary of Data Collection Efforts This report summarizes the results of multiple independent contractors who supported The Nature Conservancy’s project monitoring. Methods are generally described in detail in their respective monitoring reports. A short summary of the data and analysis conducted by the contractors is described in this section. A. Tidal marsh biophysical characteristics, processes, and elevation change were monitored by Western Washington University (WWU). Vegetation and environmental data were collected in the field, and soil samples were collected and analyzed in the lab. Vegetation data collected included summer plant species composition and vegetation structure in 2014, 2015, and 2017. Winter vegetation structure was collected during the winter of 2013-2014. Environmental data collected along with the vegetation data included elevation, large woody debris cover, soil salinity, soil penetrability, surface elevation change, sediment accretion, snow goose disturbance, and surface erosion at marsh boundaries. Photos were collected at all sample points, and in addition a formal set of restoration photopoints were established and sampled. Soil surface conditions were qualitatively described (microtopography, relative thickness of diatom cover on sediment, evidence of disturbance or erosion, etc.) Field data collection methods are described in Appendix 4. Soil samples were collected in spring 2015 and analyzed in the lab for soil particle size distribution (PSD), carbon content, and organic matter. PSD and carbon content were analyzed by the USGS Pacific Coastal and Marine Science Center (USGS-PCMSC) Sediment Lab, while organic matter was analyzed at the WWU lab. Field data collection methods are described in Appendix 4. USGS laboratory methods are described in Grossman and Curran 2015. Marsh elevation change and short-term accretion rates were monitored at a series of Sediment Elevation Tables (SETs) and accretion stations (Figure 4). Each zone has four SETs that capture high resolution long term changes in the elevation of the surface. Elevation changes may result from deposition of suspended sediment, sub-surface compaction, vertical land movement, organic matter accretion, decomposition, and erosion. SET and feldspar accretion methods are described in Barber 2014 and Rybczyk and Poppe 2017. Additional short-term accretion stations were installed to span the elevation range of vegetation transects in each study zone. Depending on habitat conditions, accretion stations included feldspar layers, plastic grids, and PVC accretion stakes. A brief description of appropriate field conditions and methods for each type of station is described in Fuller 2014b. An intensive comparative study was initiated in zones 2 and 3 by UW graduate student Brittany Jones, and is partly summarized in her thesis (Jones 2015). Marsh sampling layout Sampling was designed to study the interaction of key physical drivers and vegetation. Five primary study zones were established as described earlier. These zones represent a gradient of

18

hydrogeomorphic connectivity with the river, allowing us to understand how key drivers such as salinity, soil particle size, accretion, and landscape position affect restoration outcomes. Within each study zone, elevation (inundation) is a primary driver affecting plant species distribution, vegetation structure, soil qualities, and sediment processes. For this reason, biophysical monitoring in each zone was organized on transects that sample across the elevation gradient from high marsh to the sea-ward edge of low marsh where the habitat transitions to unvegetated tide flat (Figure 4). Vegetation plots were 20m2 and positioned contiguously along these transects. Each transect was generally positioned in the middle of each study zone, distant from large channels. However some transects crossed blind channels that penetrated the study zone. Short-term accretion plots were positioned alongside the vegetation transect at various elevations, and soil samples were similarly collected along the vegetation transect (Figure 4). For soil sample collection, transects were extended onto the unvegetated tide flat to document soil changes from unvegetated to vegetated habitat. Additional soil samples were collected in other areas to cover the spectrum of variation in both physical and biological conditions, including under-represented areas of the unvegetated delta, and marsh areas dominated by species that were under-represented on vegetation transects, such as Typha.

Figure 4. The left map shows the location of vegetation transects and sediment sampling stations used by WWU. Study zone 2 is the restoration site. The map on the right shows the locations of SETs, short-term accretion, and marsh-front erosion monitoring stations. In most cases, each point represents the location of two or more replicates. Each SET is also associated with three additional short-term accretion stations.

19

Methods and results from vegetation and sediment monitoring were reported in Fuller 2014a,b; Fuller and Thomas 2014; Rybczyk, Barber and Poppe 2014; Fuller 2015a,b,c,d,e,f; Fuller and DeBono 2015; Rybczyk and Poppe 2015a,b; Fuller 2016; Fuller 2017; Rybczyk and Poppe 2017. In addition to field monitoring, WWU also evaluated historic habitat change since 1886 by mapping tidal marsh boundaries from the 1886 US Coastal and Geodetic Survey map and from historic aerial photos (Fuller and McArdle 2014). B. Sediment Transport and Hydrodynamic Processes were monitored by the USGS Pacific Coastal and Marine Science Center (USGS-PCMSC). The sediment load, routing, and physical processes that influence the fate of sediment were monitored, including suspended and bed load transport, and elevation surveys to characterize geomorphic change. Sampling sites are shown in Figure 5. Sampling occurred between 2013 and 2015 and included the use of continuous data loggers as well as discrete sampling methods. This sampling period also captured the time frame during which the massive Oso landslide delivered a large volume of sediment into the system. A hydrodynamic model was also developed to examine the effects of the wave regime on remobilization and redistribution of sediment, and the interaction with vegetation. Methods and results are described in detail in Grossman and Curran 2015.

Figure 5. Map of USGS-PCMSC continuous monitoring and discrete measurement locations for assessing sediment transport and hydrodynamic conditions (figure from Grossman and Curran 2015).

20

C. Hydrology and sediment accretion was monitored by the USGS Western Ecological Research Center (USGS-WERC). Multi-parameter hydrology loggers (water level, temperature, salinity) were deployed pre-restoration in reference sites and post-restoration in the restoration site (Figure 6). USGS-WERC maintained the logger from 2011 to early 2015, at which point WWU took over maintenance of the logger network. Detailed methods and results are provided in Woo et. al. 2014 and Woo et. al. 2015. In addition USGS-WERC installed a network of PVC sediment accretion stakes (Figure 6) similar to those used by WWU. They report their accretion monitoring in Woo et. al. 2014.

Figure 6. Left map shows the hydrology data logger locations monitored by USGS-WERC (map by the author). Right map shows the location of sediment accretion stakes and benthic invertebrate sampling sites sampled by USGS-WERC (map from Woo et. al. 2015).

D. Tidal Channel Development was monitored by the Skagit River System Cooperative. Channels were evaluated using a combination of GIS methods and field elevation surveys using a RTK-GPS unit. GIS methods were used in conjunction with aerial photos from 2003, 2013, 2014, and 2015 to digitize channel and drainage features (Figure 7). Field surveys were conducted to measure changes in channel cross sections and profiles. Changes in channel planform and cross section were described, and compared to reference channels in the Stillaguamish and Skagit deltas. Methods and results are described in detail in Hood 2016.

21

Figure 7. Tidal channel development on the restoration site. Pre-restoration drainage channels (2003) are in yellow, and developing channels are shown for 2013-2015 post-restoration. Map is from Hood 2016.

E. Bird and Invertebrate Populations, and sediment characteristics were monitored by the U.S. Geological Survey-WERC. Bird surveys were completed pre- and post-restoration in and near the restoration site, and included surveys of waterfowl, shorebirds, and passerines. Survey areas included those surveyed historically by Gary Slater (Slater 2003), as well as a grid survey following USGS-WERC standard protocols for the restoration site and nearby areas (Figure 8). Detailed methods and results of the bird surveys are provided in Woo et. al. 2014 and 2015a.

22

Benthic invertebrate prey of waterbirds was monitored by USGS-WERC pre-restoration (2012) and post-restoration (2013 and 2015). In addition, soil conditions were characterized and samples collected for laboratory analysis in order to describe soil temperature, salinity, pH, organic matter, and the percent sand, silt, and clay. Invertebrate samples were collected at the sites shown in Figure 6b, and soil samples were also collected at those sites. Detailed methods and results for invertebrates and soils are reported in Woo et. al. 2015b.

Figure 8. Maps of the bird monitoring areas monitored by USGS-WERC. Left map shows the reference and north bay areas monitored following the protocols of Slater 2003. Right map shows the restoration zone monitoring following USGS-WERC standard protocols. Maps are from Woo et. al. 2015.

23

V. Discussion of Results This section evaluates the restoration hypotheses posed in Table 4, providing details from the various monitoring efforts that describe how well the responses in the restoration site and the rest of the estuary compare to the hypotheses. In the following discussion, a number of common plant species are discussed. Some species have had their scientific names changed over recent years and earlier reports may refer to older names. Some names continue to change, but this report uses the following names and abbreviations.

Scpu: Schoenoplectus pungens, 3-square bulrush (formerly called S. americanus and Scirpus americanus)

Scta: Schoenoplectus tabernaemontani, soft-stem bulrush (formerly called Scirpus tabernaemontani)

Boma: Bolboschoenus maritimus, seacoast or maritime bulrush (formerly called Schoenoplectus maritimus and Scirpus maritimus)

Bofl: Bolboschoenus fluviatilis, river bulrush (formerly called Scirpus fluviatilis) Tyla: Typha latifolia, broad-leaved cattail Tyan: Typha angustifolia, narrow-leaved cattail (non-native invasive) Caly: Carex lyngbyei, Lyngby’s sedge Juba: Juncus balticus, Baltic rush Agst: Agrostis stolonifera, creeping bentgrass

Objective 1: Restore self-sustaining native tidal wetlands that support estuarine-dependent animals. This objective is focused at the site level and the subsequent hypotheses relate to habitat development on the restoration site. H1.1 Tidal wetlands will develop. Hypothesis: Site elevation, hydrology, and salinity support the development of high marsh, low marsh and tide flats. Pre-restoration Conditions Prior to restoration, the site hosted approximately 50 acres of non-tidal wetland that developed as a result of high groundwater. Flooding extent and depth were greatest in the winter and declined over the summer, through evaporation, lack of precipitation, and the effect of lower river flows on floodplain groundwater levels. Pore water salinities were measured during the summer 2004 and 2005 as part of an estuary-wide habitat assessment (Heatwole 2006a), and summer soil salinities in the restoration area averaged 9 parts per thousand (ppt) across both

24

wetland and upland grassy areas (Fuller 2016). The wetland areas that were not flooded at the time of sampling had pore water salinities averaging slightly higher at 11 ppt. As a result of the brackish soil conditions, the non-tidal wetland was dominated by seacoast bulrush, Bolboschoenus maritimus (Boma), and river bulrush, Bolboschoenus fluviatilis (Bofl), rather than salt-intolerant species such as the Typhas (cattails). Figure 9 shows plot photos taken in 2005 and 2009 within this wetland. As illustrated in the photos, the area of the non-tidal wetland expanded substantially by 2009. The scale of the plants in plot ST093 for 2009 is illustrated by the barely visible head of the six foot tall person standing in the middle of the plot. In this non-tidal wetland, the plants were established at lower elevations than expected by pre-project vegetation models (Yang et. al. 2006, Heatwole 2006b, Hood 2011). We expected Boma and Bofl to rapidly colonize the upper areas of the site when tidal flow was restored and we wondered whether the existing plants would persist at the lower elevations or recede to elevations that matched the natural distribution. With subsequent monitoring across the estuary, it is evident that bulrushes can grow across a broader elevation range than earlier believed (Fuller 2017), and Boma could potentially exist across most of the restoration site.

Figure 9. Prior to restoration, a non-tidal brackish wetland dominated by Bolboschoenus maritimus and B. fluviatilis covered approximately 50 acres. Map is infrared 2003 imagery courtesy of WDFW. Photos of the same sample plots are shown from 2005 and 2009.

25

Water levels in the wetland were controlled to a limited extent with the occasional use of a tide gate pump to maintain shallow water levels for shorebirds. However, major floods starting in 2009, as well as a 2010 dike breach, largely filled the restoration area and maintained higher water levels for much of the last three years prior to restoration in 2012. Post-restoration tidal inundation and hydrology With removal of the old dike and breaching the dike footprint in two places to reconnect with the adjoining distributary channel, the site became fully tidally inundated at high tide. During the first few years post-restoration, complete drainage of the marsh plain did not occur at low tide, though water levels in the major tidal channels dropped well below marsh surface (Figures 10 and 11). However, as of 2016 the marsh surface now drains completely at low tide (Figure 11). All of the vegetation transects in the restoration area fall between MSL and MHW (Table 5), and 95% of the restoration site falls within that zone, with only 3% of the site occurring above MHHW (Fuller 2014a). As a result, the entirety of the restored tidal floodplain is inundated and drained each day.

Figure 10. Through 2014 at least, the almost flat topography, insufficient channel development, and dense vegetation appears to have prevented complete drainage of the restoration marsh plain at low tides. As a result, the “ecological” elevation of the restoration site was likely substantially lower than its actual elevation.

The larger blind tidal channels on the site retain water at low tide, though some of this retained water may be due to perched channels. For instance, the channel profile of the southern breach indicates that there is an elevated lip in the mouth (Figure 12, Hood 2016) that lessens water drainage at lower tides. This lip is likely due to highly compacted soil on the former dike footprint which the channel’s hydraulic energy has not been able to carve through. The surfaces near the northern breach are also hardpan and may similarly reduce the rate of downcutting

26

onto the site which would allow the site to drain more fully and quickly. As a result, the larger blind channels on the site retain water at low tide.

Figure 11. In 2013, the tidal floodplain within the black outline retained shallow water throughout the tidal cycle, as can be seen in areas not covered by vegetation in the aerial. Shallow surface ponding throughout the area was noted by the author in 2013-2014. By 2016, low tide resulted in essentially complete drainage of the tidal floodplain. Also note the substantial marsh retreat that occurred within the area where the tidal floodplain did not fully drain during the first two years. In 2013 only the linear truck traffic routes were bare, along with a larger bare area at the lowest elevations in the NW corner of the outlined area. In 2016, close to 70% of the outlined area is bare.

Low tide drainage of the new marsh surface was incomplete during the early years post-restoration, likely due to a combination of low slope (Figure 13), lack of channels (Figure 7), and high density vegetation (Figure 14). The flooded marsh plain at low tide is evident in the 2013 aerial in Figure 11. A large area of the restoration marsh plain retained 5-15cm of water on the surface even at the lowest tides of the year (author’s personal observation). Many areas that did fully drain retained so much pore water in the soil that a 20cm soil pit would fill to the top within minutes, indicating that groundwater was at the surface. In contrast, soil pits in the reference zones accumulated at most a few mm’s of water in the bottom. This water level in

27

the restoration zone persisted despite the fact that water levels in the downstream blind tidal channels were two to four feet lower than the marsh surface.

Table 5. Physical characteristics of the five study zones and transects, as measured in 2014. Zone 2 is the restoration zone. Table is from Fuller 2014a.

Figure 12. Longitudinal channel profile of the mouth of the southern breach. The right side of the graph is upstream in the channel draining the restoration site, and the left side is at the confluence of the channel where it joins the adjacent distributary channel from the river. Note the perched lip that limits channel evacuation at low tide. Figure from Hood 2016.

Lack of drainage on the marsh plain was likely caused by at least three factors. The historical legacy of farming, including plowing, leveling, and isolation from river-borne sediment has resulted in a site with a much gentler slope than any of the reference marshes (Figure 13, Table 5), resulting in a less rapid evacuation of water during ebbing tides. Secondly, the lack of channel development in early years (Figure 7) resulted in proportionally more time spent in slow, shallow sheet flow across the marsh plain during the ebb, and less time spent in channel flow. As channels developed within the restoration marsh (Figures 7 and 11), marsh drainage during the ebb tides improved. Lastly, the vegetation in the lower two thirds of the site was very dense during the early years (Figure 14, plots 21041 and 21050), and on an essentially flat marsh plain without channels, this density of vegetation would create a very high friction to surface flow, helping to retain standing water during low tide. At low tide, the author observed

28

motionless standing water in vegetated marsh, while in nearby channels the ebbing water level was 60 – 120 cm (two to four feet) lower than the marsh plain.

Figure 13. The slope and shape of the vegetation transects in each study zone. The colored circles mark the early 2015 seaward boundary of the middle marsh which is dominated by a mixture of Boma (Bolboschoenus maritimus) and Scpu (Schoenoplectus pungens). Note that the restoration site (zone 2) has a much gentler slope than any of the natural reference marshes. Also note the steep marsh slope and inflection point at the seaward marsh edge in zones 4 and 5 which are experiencing rapid marsh erosion. Data represent the conditions on the primary vegetation transect in each study zone which are shown in Figure 2. Figure is from Fuller 2017.

It should be noted that assumptions regarding marsh plain inundation and drainage patterns are often made from either hydrology loggers located in nearby tidal channels or on a GIS analysis of LiDAR elevations. However neither of these approaches directly measure marsh plain hydrology and would therefore result in the false conclusion that the restoration marsh plain was fully draining during the early years post-restoration. This is an important lesson learned and has implications for understanding the potential condition and trajectory of marsh development on a restoration site because marsh drainage affects soil conditions that impact plant growth and survival. A lack of understanding of actual conditions on the marsh plain can lead to surprise outcomes. Continuously saturated soils have ecological implications because the saturation results in reduced gas exchange and altered physical and chemical conditions in the soil. Lack of oxygen leads to redox conditions and sulfite levels that may create substantial stress levels for plants. Soil chemistry was not monitored so we are not able to confirm soil chemistry conditions. However a 50-acre marsh dieback in June 2015 resulting from an insect outbreak suggests the plants may have been highly stressed and susceptible (Fuller et. al. 2015, Fuller 2016). Almost every plant within the dieback area was dead by the end of June and most plants examined had an insect bore hole near the base (Figure 15). We were unable to find historical records of any other insect-mediated marsh die-back of that scale on the west coast, despite contacting experts throughout California, Oregon and Washington. Since we were not monitoring soil

29

chemistry, we cannot directly link the insect outbreak to high environmental stress levels experienced by the plants, so the cause of the dieback remains uncertain. Additional discussion is provided later in this report to support the idea that cumulative stress may make marshes vulnerable to diebacks of this nature.

Figure 14. The same plots from zone 2, in 2014, 2015, and 2017. They are ordered with the highest elevation plot shown in the upper trio of photos and successively lower elevation plots below. The black arrows in each photo indicate the location of the same feature on the horizon, to aid in orientation.

30

Finally, the high level of soil saturation on the marsh plain of the restoration site also affects soil physics, making the soil very soft (and walking more difficult), which has unknown implications for plants. An indicator of the soft soils is the high soil penetrability found in the restoration area (Table 2), considerably higher than any of the reference sites. Continuously saturated soil with low sand content is less compacted and more soupy. In the two years since 2015, anecdotal evidence and aerial photos (Figure 11) show there to be less flooding of the surface at low tide and walking also appears easier. Blind tidal channels have developed and complete drainage of the marsh surface now appears to occur, though soil saturation remains high.

Figure 15. A 50-acre marsh dieback in the restoration zone occurred as a result of a moth that lays its eggs in the stems of bulrushes. Infrared imagery shows the decline in vegetation cover in the white outlined area where all plants died by July 2015. Figure from Fuller et. al. 2015.

Salinity Pore water salinity on the marsh plain was measured by WWU during 2015 (methods described in Appendix 4) and surface salinity was monitored in tidal channels by USGS-WERC with continuous loggers (Woo et. al. 2015). Surface salinity is relatively easy to monitor with loggers and is a critical parameter for monitoring habitat response. Pore water salinity is labor intensive to monitor but is more closely related to plant response since it reflects actual soil conditions where plant roots inhabit. Surface salinity varies substantially at short time scales in response

31

to tides and rain events. Pore water salinity is more stable and changes slowly over the seasons in response to changes in the relative balance between fluvial and tidal contributions. Surface Salinity Estuarine surface water salinity increases during the summer as river flow decreases, as is evident at the reference site hydrology loggers in Figure 16, where surface salinity tends to peak in August-September. Salinity data in zone 4 is only available 2011-2013 but interestingly there are occasional smaller salinity peaks during winter, a pattern seen once in zone 5, but not seen in other reference zones. The restoration zone follows a different pattern than other sites, appearing to have a delayed annual peak in salinity and temperature, both reaching their maximums in October-November in 2014 and 2015. It is not clear why the restoration site would have a delayed peak in salinity and temperature.

Figure 16. Mean monthly temperature (blue bars) and surface salinity (orange lines) patterns across the Stillaguamish estuary. The vertical dashed lines in some graphs indicate the date of restoration. Note that the time periods of record differ among the loggers. Figure from Woo et. al. 2015.

32

As expected, the highest salinities are regularly seen in zone 5 where annual peaks range from 18-25 psu (Figure 16). In 2013-2014, the zone 5 salinities peaked at around 18 psu. In the restoration zone during those years, 2013 reached a similar level of about 18 psu, but in 2014 the restoration site was slightly lower, peaking at about 15 psu. The zone 1 marsh had the lowest surface average salinities, as expected, but followed a different pattern, peaking in 2013 at about 6 psu, but increasing in 2014 to about 16 psu. Pore Water Salinity Despite being connected to a distributary channel and near Hatt Slough, pore water salinity on the restoration site was more similar to zone 4, the mid-delta marsh, than to either zones 3 or 1 (Fuller 2016, salinity methods are described in Appendix 4.). This was true for both spring and summer in 2015. Pore water salinity levels appear to have been extremely high during the 2015 growing season as a result of record low river flows throughout most of spring and summer (Fuller 2016, and described more fully under Objective 2 below). Spring 2015 salinity on the restoration site was slightly lower than during the summers of 2004 and 2005 when the site was a non-tidal brackish wetland. The earlier salinity levels were assumed to be high as a result of evaporation and lack of either tidal or riverine flushing. The 2004-2005 values were similar to the values seen in the more isolated mid-delta zone 4. But summer 2015 pore water salinity was two times higher than recorded in 2004-2005 (Fuller 2016). Spring 2015 pore water salinity was 7ppt, and summer salinity was 20ppt (Table 2) which was well above the published 15ppt salinity tolerance level for bulrush species (Fuller 2016). 2015 river levels were at record lows for several months and the mixed-bulrush marsh in all zones except zone 1 exceeded 20 ppt. Summer values in 2005 were 11 ppt in the pre-restoration zone, and the highest average salinities that year were in zone 5 at 13 ppt (Fuller 2016). During the summer of 2015, average vegetation height in the restoration zone decreased by 56% compared to 2014, and stem density decreased by 48% (Table 6). Height in the middle marsh (mixed-bulrush) in all zones decreased substantially in 2015, though nowhere as much as in the restoration zone. The greater decrease in the restoration zone is likely due to greater levels of environmental stress. Initially we thought that Boma and Bofl on the restoration site were at lower elevations than they naturally occur in the other zones and that this may have resulted in plants already under stress from over-inundation. However, in reference zones 3-5 Boma occurs at elevations similar to the restoration site, so we think it more likely that stresses related to soil conditions described earlier (over-saturation) may have combined with the new 2015 salinity stress to reduce aboveground biomass. In addition, there was a large outbreak of insect herbivores in 2015 (Figure 15), which was likely made possible by plants that were already under stress from the soil and salinity issues described. The insects were found in all zones, however only in the restoration zone did they attack 75-100% of bulrush stems. Only the highest part of the restoration zone, near the river, was unaffected by the insects and had tall, lush Boma/Bofl. In zones 3-5, insects attacked 5-50% of the stems, occasionally higher in zone 4. In zone 1, closest to the river and least affected by the high salinity levels, affected stems were generally fewer than 5%. These numbers are

33

estimates based on the author’s field notes of counts of plant stems with insect bore-holes. However, counts were haphazardly recorded during the process of normal vegetation sampling, and were not part of a formal scientific study.

Table 6. Summary of vegetation structural changes 2014-2017, across the Stillaguamish estuary. For each parameter in each zone, the year of lowest value is shaded. Bulrush biomass volume includes only aboveground biomass.

One question coming out of the 2015 growing season was whether or not the restoration marsh would recover and re-establish where it had been killed back. No vegetation data was collected in 2016, although general observations and aerials indicated that the areas affected by the dieback in 2015 remained mostly bare. River flows in 2017 were higher than normal and in the restoration zone we saw a substantial increase in average vegetation height, density, and aboveground biomass (Table 6), though it did not recover to 2014 levels. The patterns differed across the elevation spectrum and are described in more detail below. Habitat and Vegetation Development Patterns Predicted responses: Two sources of predictions exist for post-restoration marsh vegetation. Battelle (Pacific Northwest National Laboratory) generated a 3-D hydrodynamic model to guide the development of restoration alternatives, and included projections developed in collaboration with TNC (Yang et. al. 2006, Heatwole 2006b). Habitat predictions were based on elevations and calculated inundation patterns. They estimated that about 55% of the site would be marsh, with 39 acres of graminoid-dominated high marsh and 48 acres of bulrush-dominated low marsh. About 40% of the site (63 acres) was predicted to be ponded at low tide, with an additional 8 acres of bare mudflat. These predictions were developed before the project design incorporated breaches through the former dike footprint, hence the large pond predicted, and the rationale for adding breaches. In addition, the Skagit River System Cooperative (Hood 2011) developed a vegetation model for the Stillaguamish and made vegetation predictions based on the final project design. Hood’s model assumed elevation as the primary control on species distributions, though he did note that salinity differences likely drove different species compositions between the area near Hatt

34

Slough and more distant areas to the north. At the highest elevations, roughly overlapping the area predicted by Yang et. al (2006) as high marsh, Hood (2011) predicted 20-50% 3-square bulrush (Schoenoplectus pungens, Scpu, which was formerly known as Schoenoplectus americanus), 10-40% soft-stem bulrush (S. tabernaemontani, Scta), and 20-30% Lyngby’s sedge (Carex lyngbyei, Caly). Below this elevation, in a 40 acre area roughly similar to the area predicted by Yang et. al. (2006) to be low marsh, Hood (2011) predicted 60-70% Scpu, and 20-30% Boma. Finally, Hood (2011) predicted about 80 acres of flooded area that would have 20 acres unvegetated, and 60 acres with 40-90% cover Scpu and 10-60% bare ground. He predicted that Baltic rush (Juncus balticus, Juba) and narrow-leaved cattail (Typha angustifolia, Tyan) would not colonize the site. Potential species ranges: The elevation range of the important dominant plant species in the Stillaguamish estuary is shown in Figure 17, from vegetation data in zones 1-7, collected in 2015 (methods described in Appendix 4). Elevations for each sampling plot were extracted from the 2013 LiDAR. The elevation range of the restoration site is also shown in the figure. Based on this vegetation model, which assumes that elevation is the only driver of species distribution, the restoration

Figure 17. Elevation range of dominant plant species in the Stillaguamish estuary, based on vegetation sampled across seven study zones in 2015. The boxes indicate elevations where the species was found to be common (covers at least 20% of a plot), and the whiskers on the boxes show the elevations at which the species was present, but never exceeded 20% cover. The red lines indicate the elevation range in the Restoration area (zone 2). The listed species are: Scpu = Schoenoplectus pungens, Boma = Bolboschoenus maritimus, Scta = Schoenoplectus tabernaemontani, Bofl = B. fluviatilis, Caly = Carex lyngbyei, Juba = Juncus balticus, Tyla = Typha latifolia, Tyan = Typha angustifolia, Disp = Distychlis spicata, Agst = Agrostis stolonifera.

35

site should be nearly completely vegetated, with Scpu at high levels throughout the site, sharing dominance at the lower end with Boma, and at the upper end with Boma, Scta, Bofl, and Caly. The vegetation transects used to develop this model have little presence of Typha species on them, though both Tyla and Tyan are common in areas off the transects, so this model does not adequately reflect the distribution of Typha. The actual Bofl (river bulrush) elevation range overlaps much more with Boma than Figure 17 implies. Bofl is almost identical to Boma in the field, except under ideal conditions when it is much larger. When it is slightly stressed, for example by being at lower-than-ideal elevation, it is difficult to separate from Boma. Figure 17 shows only its ideal elevation where its size makes it easy to differentiate from Boma. We speculate that its’ true elevation range includes at least the upper half of Boma’s range. Bofl was not identified as being distinct from Boma at the site until after the restoration project. To our knowledge it has never been reported from other estuaries in Puget Sound or Oregon. We expect that it is actually common in estuaries, but has been mistaken for Boma. It is not referenced in any northwest wetland plant field guide, and is not even mentioned as a similar species in the Boma entries. Pre-project photos suggest that it was present and likely dominated the non-tidal wetland on the restoration footprint, based on plant heights (Figure 9, see the head of a standing human in the 2009 photo of plot ST093). Old, dead stem bases from pre-restoration still remain on the site and their diameter far exceed diameters reported for Boma. We believe the current marsh is co-dominated by both Boma and Bofl, though Bofl has only been easily identifiable in 2017. High river levels in 2017, and presumably lower pore water salinity, have allowed some clones in the restoration marsh to stand out as obviously taller, with wider leaves, and with very thick stem base diameters. In previous years, presumably due to higher stress levels, bulrush clones have not appeared sufficiently distinct for rapid differentiation. Descriptions of differences between Boma and Bofl can be found in many online floras such as the Flora of North America (www.efloras.org). The two species overlap broadly in many physical parameters, though in ideal conditions Bofl tends to be larger in most dimensions including stem diameter and height. A summary of their differences is provided in Table 7, based largely on the Flora of North America. Response of Schoenoplectus pungens: After restoration, some things happened as expected and some things were a surprise. The biggest surprise has been the lack of colonization by Scpu, which all predictions expected to be the most common species. It is the most common dominant species in the estuary and its elevation range is the broadest, encompassing the upper range of all other bulrush species and extending well below the range of other bulrushes down to slightly below MSL. Our first hypothesis was that it was related to soil particle size because Scpu reaches its maximum aboveground productivity (height and stem density) with 35-75% sand content (Fuller 2017). Other bulrushes have a distinct preference for finer soils, with Boma for example

36

declining as the % sand content increased, and dropping out completely by 45% sand. At 5% sand, the restoration site has the lowest % sand of all the study zones (11-67%, Table 2).

Table 7. Summary of differences between Bofl and Boma.

However, reference zones 1, 3 and 4 all had areas with 0-10% sand where Scpu was still common and productive (Fuller 2017), so particle size can’t be the only deterrent to Scpu colonization of the restoration site. At this point, our best guess is that its’ absence from the restoration site is related to the physical and chemical properties of the soil that are likely related to site drainage. Sandy areas tend to drain well at low tide and facilitate gas exchange, which may explain why Scpu’s preference for sandier soils may help it to grow down to and even below MSL. However the lack of drainage in the restoration soils up through 2015, combined with the finer soil particle distribution, may have limited colonization by Scpu by reducing gas exchange, lowering soil redox potential, or increasing soil sulfides above levels tolerated by Scpu. In addition, the finer particle size distribution and high saturation levels in the restoration site resulted in soils that were slightly more “soupy” than other zones, as measured by penetrability (Table 2). These physical and chemical soil differences may have prevented Scpu colonization. However, soil chemistry, except for salinity, was not monitored, so we cannot evaluate the Scpu hypothesis.

37

Response of Bolboschoenus maritimus and B. fluviatilis: As mentioned earlier, we haven’t been able to differentiate Boma and Bofl in most years on the restoration site, so this discussion lumps Boma and Bofl together and refers only to Bosp (Bolboschoenus species). As expected, Bosp rapidly colonized the higher elevations of the restoration site. One year after restoration, Bosp had expanded east to the new dike and advanced south. The highest elevations to the south were occupied by creeping bentgrass (Agrostis stolonifera, Agst) prior to restoration and Bosp began moving into this zone immediately. By 2014 much of the Agst had senesced but the biomass still covered 10-12 acres of ground. By 2015 the grass meadow was completely replaced by dense, tall Bosp. In 2013, the overall cover of Bosp expanded as it moved into the higher elevations and only receded from a small area near the northern breach (Figure 18). We lack quantitative data from summer 2013, but observations suggest its height and density declined at lower elevations. In 2014, it continued to recede slowly from the lower elevations, expanding the area of bare mud in the northern part of the restoration site (Figure 18).

Figure 18. Infrared imagery of the restoration zone post-restoration. After first expanding rapidly to cover the site, the Boma/Bofl marsh then receded substantially, leaving the restoration site more than 60% bare mud (pinkish grey color) in 2016. The areas bounded by black are largely bare and in 2016 even the vegetated areas in the northern 2/3 were very sparsely vegetated as suggested by the less intense red.

38

A 50-acre marsh die-off occurred in 2015, the year of record low river flows, and Bosp thinned and receded substantially (Figures 15, 18). Average aboveground biomass declined by a remarkable 70% in the restoration area between 2014 and 2015 if we compare just the 12 plots that were sampled in both years (Fuller 2017). In 2014 we only sampled 12 plots, essentially every fifth plot, with each plot more or less evenly spaced every 100m along the transect. In 2015 we sampled all 69 plots along the transect. If we compare the average of the 12 plots sampled in 2014 with the average of all 69 plots sampled in 2015, there is an even larger biomass decline of 82% (Table 6). This difference is because there was greater biomass loss in lower part of the transect and so only looking at the 12 plot sample underrepresents the impact of the biomass loss across the lower two thirds of the transect. This loss of the plant biomass that fuels the rest of the food web in the estuary has substantial implications for the estuary food web. This is particularly concerning in the context of climate change, because the 2015 record extreme low river flows are expected to become the average summer condition within 40-50 years (Fuller 2016). The pattern of marsh recession continued in 2016, with a majority of the restoration site being barren (Figure 18). However, 2017 river levels were higher than normal, and parts of the restoration zone partially recovered. Bosp responded differently at different elevations (Figure 19). The massive biomass decline in 2015 throughout most of the restoration area is clear in Figure 19, except for a small area near the highest elevations, close to Hatt Slough which actually saw an increase in biomass. In 2017 the lowest parts of the transect, below approximately 1.8m, did not improve in terms of bulrush biomass. These areas did however see a significant expansion in the biomass and cover of small annual species like Cotula coronopifolia (brass buttons) and Subularia aquatica (awlwort). The 2017 photos of plots 21041 and 21050, shown in Figure 14, show the shift in dominance from bulrush to short annual species. There was some minor

Figure 19. Spatial and temporal patterns of change in Bosp biomass in the restoration area.

recovery of bulrush biomass in the middle of the transect, between about 1.9-2.01m elevation, though it was still a fraction of 2014 biomass (Figure 19). Between 2.02-2.09m elevation there

39

was substantial recovery of biomass. However, interestingly at the highest level of the transect, above 2.28m, there was a large drop in bulrush biomass in 2017, particularly at 2.28m which was the only elevation that saw an increase in biomass in 2015. The patterns of retreat of Bosp and partial recovery is illustrated in Figure 14, which shows the same series of plots in 2014, 2015, and 2017. The plots are ordered by elevation, with the highest site at the top and the lowest site at the bottom. The large decline in bulrush biomass at the lowest elevations, and the lack of recovery is clear in the photos, as is the large increase in the cover of small annual species in 2017. The full recovery at plot number 21016 in 2017 is obvious, and in plot 21006 the photo shows an increase in cover of the non-native invasive Typha angustifolia (Tyan) and a decline in bulrush in 2017, which was consistent in the upper 6 plots of the transect. These upper plots were first colonized by bulrush in 2014. In 2015 that area had the second highest bulrush biomass volume in the entire estuary, with dense and tall Bosp. But in 2017 the bulrush biomass volume dropped, despite the greater freshwater influence. The decline appeared to be due to a stem-boring insect attack, presumably the same moth that attacked the lower marsh in 2015. However, the attack was restricted to the upper 6 plots where Tyan expanded very substantially in 2017. More details are provided below. Response of Typha species: Both the native broad-leaved cattail (Tyla) and the non-native invasive narrow-leaved cattail (Tyan) have colonized the restoration site and are expanding rapidly, another result that was not predicted by vegetation models. Typha species are taller and have larger leaves than the bulrushes, and in infrared imagery they show up as brighter red. Figure 20 illustrates the locations of some of the Typha patches in 2016. 2017 data indicate that Typha species have expanded very substantially in the past two years. The large crescent of Typha evident in the southern part of the restoration zone in Figure 20 appears to have now replaced much of the Bosp all the way south to the old dike footprint next to the river (the area marked “A” in Figure 20). Interestingly this is the former Agrostis meadow that Bosp only colonized in 2014. Sampling in June 2017 revealed that not only has Typha expanded across this area, competing with Bosp, but the Bosp in this area is much reduced in size and more than half the stems were browning down already because they had been attacked by a stem-boring insect. All the senescing stems that were inspected had the same small bore holes evident near their base as had been observed in the 2015 event at lower elevations. The result is much lower bulrush biomass in this upper area compared to 2015 (Figure 19). The spatial pattern of insect attack in 2017 was very different from 2015. The earlier infestation attacked the lowest elevation Bosp, leaving the upper elevations relatively lush and flowering. The 2017 infestation attacked the upper elevation Bosp, and the lower elevation Bosp appears relatively lush and is flowering. I expect this is related to relative stress levels. 2015 was the year of record high pore water salinity, with greater salinity at lower elevations. In the restoration zone, the higher elevations were closer to Hatt Slough and had the lowest salinity, and hence the highest productivity. In contrast, 2017 river levels were high and the lower

40

elevations are lush. It is not clear whether there is a source of stress at the higher elevations for Bosp, other than the moth and the rapidly expanding, competitive, and shade-casting Tyan.

Figure 20. The bright red patches in the infrared imagery, indicated by the arrows, are clones of Typha. Tyla is common at the higher elevations in the southern part of the site, and Tyan occupies both high and lower elevations. Tyan is a non-native invasive. In 2017 Typha had expanded south from the large, bright red crescent towards the river, covering much of the area marked “A” which had been dominated by Bosp.

Interestingly, however, there is greenhouse experimental evidence that Tyan, though not Tyla, has allelopathic capabilities that were specifically tested on Bofl (Jarchow and Cook 2009), and presumably would also affect Boma. Tyan exuded different phenolic compounds into the soil than Tyla, and had substantial negative impacts on Bofl growth. Those impacts were ameliorated when activated carbon was incorporated into the soil to sequester the phenolics. It is possible that allelopathic and competitive effects of the rapidly expanding Tyan in 2017 may have generated enough stress to weaken the Bosp and make it more susceptible to insect attack. Tyan is an invasive, non-native species and chemical control should be considered at the restoration site. It has a considerably broader ecological amplitude, at least in terms of inundation levels and salinity tolerance, than the native Tyla (Fuller 2015b), and its distribution at the restoration site already indicates a much more rapid expansion, and colonization at lower elevations than Tyla. If it expanded throughout the elevation range within which it is currently found at the site, it would occupy close to half of the site. Typha species, because of their tall, dense, robust structure, prevent over-marsh access by most fish and bird species.

41

Response of Schoenoplectus tabernaemontani: Soft-stem bulrush (Scta) does not dominate in any part of the restoration area, but its presence has been gradually expanding throughout the same area being colonized by Tyan, the upper half of the restoration area. I expect it to continue expanding and may eventually reach co-dominant levels with Boma as it does in zone 3. Other habitat responses: With the additional tidal prism added by the restoration site, the adjacent distributary channel has widened and deepened downstream of the northern breach (Grossman 2015, Hood 2015). Much of the excavated substrate has been pushed into the restoration site, forming a large sand splay inside (Figure 21). This sand splay is higher in elevation than the surrounding marsh plain and has yet to be colonized by plants. Given its particle size distribution, Scpu is most likely to colonize, unless hydraulic energy is sufficient to limit plant colonization in that area. One of the vegetation predictions projected a substantial area of high marsh in the restoration zone. No high marsh has developed yet, except on parts of the old dike footprint. Agst, the most common high marsh species, has been mostly displaced from the restoration site by Bosp and Tyan. Caly has established a few robust clones scattered in the highest 20 acres near Hatt Slough, but is competing with Typha, Bosp, and Scta.

Figure 21. Widening and deepening of the distributary channel west of the restoration site has led to flood tide deposition of a large sand splay in the restoration area.

42

H1.2 Accretion. Restored hydrology will result in marsh plain accretion. Hypothesis: Restored tidal exchange and inundation will result in sediment accretion on the marsh plain, initially rapidly until site elevations approach elevations in adjacent marsh. Annual rates of elevation change in the restoration zone have been substantial, ranging from 1.8 to 6.2 cm/year (Rybczyk and Poppe 2017). The highest average rates of elevation change over the past three years were recorded at the lowest elevation SETs near the breach sites (Figure 22). These sites averaged 3.93 and 3.54 cm/year of elevation increase. The SETs at the southern end of the site near the river, at higher elevations, have seen average increase in elevation of 1.43 and 3.51 cm/year (Figure 23). Prior to restoration, as a result of diking and plowing inside the dike, and accretion outside the dike, the site was about 1m lower in elevation than the adjacent tidal wetlands (Figure 24, Grossman and Curran 2015). This resulted in a bathtub like morphology for the restoration site and contributes to the high rate of accretion at the site. At three of the four SETs in the restoration area, the rate of elevation change has slowed slightly over the past three years. This slowing of elevation change may or may not be related to the 2014 Oso landslide. It is likely that the landslide continues to affect sediment delivery to the estuary and current rates of accretion may not reflect “normal” background rates.

Figure 22. Mean elevation changes in the lowest elevation SETs in the restoration area, near the breach sites. Figures from Rybczyk and Poppe 2017.

43

Figure 23. Mean elevation changes in the highest elevation SETs in the restoration area. Figures from Rybczyk and Poppe 2017.

Figure 24. Map and graph showing elevation change along three transects, with about 1m subsidence within the PSB estuary restoration area compared to outside of the area. Figure from Grossman and Curran 2015.

44

The elevation change rates in the restoration area are remarkably high and have already necessitated the installation of SET extensions on three of the four SETs to avoid burial. Rates of change are expected to decline over time as the site elevation increases towards that of the adjacent marshes, and less time is spent inundated. Accretion rates would likely increase substantially if there was greater direct flow from Hatt Slough into the site. Tidal exchange and sediment delivery: Tidal exchange into and out of the restoration site shows a complex asymmetry (Grossman and Curran 2015). Current velocities during onshore flood tides are 25-50% stronger than ebb flows, but the weaker offshore ebb flows generally lasted a few to several hours longer each week. The majority of the restoration site is inundated 30-70% of the time (Figure 25). Flood tides are the dominant process by which sediment is moved onto the site and given the inundation periods, tidally-driven sediment transport onto the site occurs 15-35% of the time (Grossman and Curran 2015).

Figure 25. Map showing the inundation frequency across the restoration site. Note that processes leading to sediment flux into the area which principally occur during flood tide transport, would be represented by half (50%) of these inundation values. Figure from Grossman and Curran 2015.

45

During their study, approximately 3/2014 – 6/2015, Grossman and Curran estimate a total sediment load of 26,200 tonnes (range 10,200 – 34,300 tonnes), with a volume of 9,500 – 13,300 m3, was delivered to the restoration site. This represents about 1% of the total Stillaguamish sediment load during the sampling period. If we assumed that this sediment was spread evenly across the 150-acre site, it would result in a mean vertical accretion rate of 1.5 cm (Grossman and Curran 2015). Given that accretion is not happening in the developing complex channel system, the actual accretion rates would be higher on the marsh surfaces. These values are in general agreement with observed accretion rates at the SETs (Rybczyk and Poppe 2017). H1.3 Sea Level Rise vulnerability Accretion will keep up with sea level rise. Hypothesis: Restored inorganic and organic accretion will result in accretion rates that will keep up with moderate projections for local sea level rise. Accretion on the restoration site is substantial and currently exceeds sea level rise. We can evaluate the site ability to keep up with sea level rise (SLR) by assuming a SLR projection of 82cm by 2100, which is the value projected by the National Research Council (NRC 2012) for Seattle, Washington, adjusted for local rates of land subsidence of -1 mm/year.

Box 1. Derivation of Sea Level Rise Projection The 2012 NRC report provides regional SLR projections for four west coast cities. Its projection for Seattle in 2100 is 62cm (range 10-143cm) above the 2000 sea level, however the NRC used a general rate of positive vertical land motion of 1mm/year for all of the west coast region north of Cape Mendocino in California. This is because much of the outer coast is rising, which has the effect of reducing the local relative rate of sea level rise. At the sub-regional level VLM can be highly variable, depending on position relative to tectonic plate convergence zones, glacial isostatic adjustment rates, the presence of deep alluvial and lahar deposits which may undergo deep compaction and subsidence, and on local rates of groundwater withdrawal. The NRC report covers some of this sub-regional variation in their report, but for their SLR projections, they settled on a single NW regional rate of uplift. Several sources of information suggest that the Port Susan Bay area is subsiding. In northern Puget Sound, tectonic movement is projected by the CAS3D-2 spherical earth model to result in future subsidence at a rate of about -0.87mm/yr (NRC 2012, table 4.4) assuming the interseismic phase continues (no large earthquake). Models of glacial isostatic adjustment suggest there may be a small uplift in northern Whatcom County, but limited effects or slight subsidence is projected for areas south of there (NRC 2012, table 4.3). Continuous GPS measurements of VLM in the Puget Sound area are generally negative and sometimes substantially so (NRC 2012, Figure 4.14). One source of continuous GPS estimates of VLM is the Pacific Northwest Geodetic Array (PANGA), which has documented subsidence rates of 1.4-2.0 mm/year in the Camano Island, Mount Vernon, Fidalgo Island area (data downloaded from http://www.panga.org/data). Other potential sources of subsidence that are possible but have not been measured are the long-term compaction of alluvial and lahar soils which dominate the soils on Puget Sound river deltas. Most, if not all, continuous GPS stations are positioned on

46

bedrock and therefore may not accurately reflect VLM rates on the alluvial soils of river floodplains and estuaries. As a result of the above indicators of subsidence near Port Susan Bay, for this report we use the NRC Seattle projection of 62cm, and adjust it by replacing their regional VLM estimate of +1mm/year uplift with a local estimate of -1mm/year subsidence. This results in the addition of 20cm of SLR to their projection with a final projection of 82cm by 2100. It should be noted that we are referencing the NRC 2012 central projection, but that their full range in projections is 10-143cm (or 30-163cm if you adjust VLM to -1mm/yr). As a result, the impacts of SLR on Stillaguamish tidal marshes could be either less than we project below, or higher. It should also be remembered that any stress to plants from SLR will interact with other climate-related stresses such as lower summer river flows, as well as other existing sources of natural and anthropogenic stress.

However, SLR rates are not expected to increase linearly, but to continue to accelerate over time (see Figure 5.5 in NRC 2012). Using the NRC 2012 projections for 2030, 2050, and 2100 adjusted for local subsidence, we can generate a 2nd order polynomial curve to evaluate changes in the rate of sea level rise between the present and the total projected sea level increase of 82cm by 2100. Using this approach, we assume that the rate of SLR will increase from about 0.3cm/year early this century to over 1.8cm/year at some point after 2075. At those rates of SLR, all Stillaguamish study zones have elevation change rates that would exceed SLR through the middle of the century, with the exception of zone 5 which would begin to lose ground by then. Sometime after 2050, the other reference zones would begin to lose ground and by 2075 only the restoration site would be keeping up with SLR, and would continue to do so beyond 2100. However several factors may affect those projections. First, accretion rates in the restoration zone are expected to slow over time as the site elevation approaches that of the adjacent marsh. Second, we don’t have many years of SET data yet, and so current rates may not reflect longer term trends. Indeed, the 2014 Oso landslide may have elevated the rates of elevation change in recent years, given its effect on sediment load (Grossman and Curran 2015). Site-specific factors may also affect accretion rates in the short term. For example the highest reference rates, found in zone 4, may be associated with the high rates of seaward marsh front erosion in that area, with waves picking up the sediment at the marsh edge and pushing it higher into the marsh where the SETs occur. The lower pair of SETs do show a loss of elevation. As this site illustrates, high accretion rates do not necessarily mean a resilient marsh, or that the sediment source is the river. Both zones 4 and 5 are eroding at the seaward marsh edge, but their SETs record average positive elevation gain. If erosion continues to remove those marshes at present rates, elevation gains at the SETs would reverse direction. Grossman and Curran (2015) developed a wave model and calculated shear stresses across the shallow intertidal zone of the estuary based on the distribution of soil particle sizes. They

47

identified areas where the critical shear threshold of the substrate is exceeded by waves, which would lead to sediment mobilization (Figure 26). It should be noted that the calculation of critical shear in the model does not include cohesive properties of the sediment including the effects of biogenic polysaccharides produced by biofilm. Nevertheless, it is intriguing that the areas of greatest excess shear stress coincide with the observed areas of greatest erosion in zones 4 and 5, along the seaward marsh front north of the restoration area, and extending all the way through the marsh to the LWD wrack line in zone 5.

Figure 26. Map of calculated excess shear stress associated with the transformation of 2m deep water waves approaching PSB from the south at 4 sec period under high tide (3 m) relative to the mean particle size results observed across PSB. Higher values indicate greater potential for mobility. Figure from Grossman and Curran (2015).

Another site-specific factor affecting accretion is species-related. The zone 4 upper SETs showing the greatest elevation change rates are associated with river bulrush (Bofl), which has the most robust and persistent aboveground biomass during the winter sediment/wave season. Their structure not only captures more sediment, but also wrack, which may further increase sediment capture capacity. Bofl’s relatively narrow distribution in the estuary may mean that similar high rates of accretion may occur in only limited places. The restoration zone also appears to support extensive Bofl, and this may affect accretion rates. However, as the

48

environment changes, summer river flow declines, increasing pore water salinity, the distribution of Bofl may change. Elevation change rates won’t stay the same over decades either. At moderate rates of SLR, the increased inundation period would increase accretion, creating a feedback loop that would help marshes keep up. However, small increases in sea level also result in increased nearshore winter wave height and energy, and could accelerate marsh-front erosion in western and northern parts of the estuary, regardless of accretion rates in the marsh interior. Sediment recruitment from the watershed is expected to increase substantially as a result of higher winter flood flows (cf. Lee et al. 2016 for the Skagit), which can be expected to increase sediment delivery and accretion rates in the estuary. However, lower summer flows may cause a decline in aboveground biomass which could lessen the rate of sediment capture in tidal marshes. Finally, future restoration or adaptive management projects could alter hydrodynamics in the estuary in ways that increase delivery or retention of sediment. There are therefore a number of ways in which sediment delivery and retention could increase accretion, as well as dynamics that could increase erosion. How these processes will interact to affect the net change is unknown. In conclusion, at current rates of elevation change, the restoration site will keep up with SLR but the reference marshes will lose ground sometime past the middle of the century. However, erosion of the marsh in zones 4 and 5 is expected to eliminate much of the northern marsh before the middle of the century, regardless of SLR. Whether marshes near Hatt Slough keep up with SLR will depend on how the dynamics of increased sediment delivery, increased water levels, and increased pore water salinity interact with the engineering biota that either build elevation (plants) or erode it (snow geese). These interactions are discussed in more detail in Fuller 2017. H1.4 Channel network A complex blind tidal channel network develops. Hypothesis: Restored tidal exchange will re-introduce sediment transport and scouring of tidal channels on the project site, resulting in the development of a complex blind tidal channel network. A large and complex channel network is developing on the restoration site, particularly in the lower elevations of the area (Figure 5). Initially, the site developed shallow, broad drainage areas with ill-defined channels, but each year these became narrower, deeper, more sinuous, and finally recognizable channels with well-defined edges (Hood 2016). Total channel length increased from 2,310 m prior to restoration to 18,447 m in 2013, and 23,266 m in 2015 (Hood 2016). The current lengths are comparable to reference data from the North and South Fork Skagit deltas, but they are greater than Stillaguamish reference conditions. The allometric prediction for the site was 4,458 m of tidal channels, with an upper 95% confidence limit of 64,000 m (Hood 2011). The high values seen post-restoration are likely due to the subsided topography of the site that reduces flow across the marsh surface and forces more of the water into the channel network. A larger channel network forms when over-marsh flow is reduced and a greater portion of the tidal prism is constrained to the channel network (Hood 2016).

49

No blind tidal channel or distributary connections have formed across the old dike footprint to increase connectivity between the site and nearby channels or marshes. Most of the hydraulic energy therefore continues to be focused on the two under-sized breach connections with the adjacent distributary. These breaches have expanded in cross sectional area by 14-32% (Hood 2016), though their expansion is likely reduced by the hard, compacted nature of the old dike footprint. The effect of the hard-packed channel bottom is evident in the channel profiles, which have a steep slope on the western end where the site drains into the adjacent distributary, but then perch on the inside of the old dike footprint. For example, see the channel profile for the southern breach shown in Figure 12, clearly showing the perched nature of the channel mouth which would inhibit ebb flows at low tide. Initially we assumed that additional channels would form across the old dike footprint, but the footprint is proving to be very compacted and resistant to erosion. This results in slower drainage on the ebb tide and longer effective inundation periods on the marsh than would occur in natural marshes at the same elevation. Lack of additional channels across the dike footprint also reduces the connectivity for fish access into the site. There are half a dozen places at least where a blind tidal channel in the adjacent tidal marsh comes very close to the restoration site, but channels have not connected yet. This may be an opportunity for adaptive management (Fuller 2017). H1.5. Bird use Birds will use the site. Hypothesis: Restored tidal inundation patterns will facilitate site use by a diversity of birds. Birds were sampled in zones 1 and 2, and throughout the northern marsh and tide flats from the restoration area north to zone 5 (Figure 8, and Woo et al. 2015). Observers walked to pre-set survey locations and recorded weather, all species of birds, habitat type, and behavior. The restoration area, zone 2, had the highest density of birds in the fall, compared to other sites (Figure 27). In October, the site had particularly high numbers of dabbling ducks and in November, it had high numbers of snow geese. High tide was when geese were most prevalent on the restoration site, and they moved to agricultural lands when the tide went out. The restoration site is not hunted, unlike the reference areas, which could explain the high numbers of game birds on the restoration site during the fall. The fact that overgrazing by geese is a cause of marsh loss causes one to consider whether the protected nature of the restoration site might end up increasing the rate of grazing disturbance on the restoration marsh. In the winter, the restoration site had higher bird densities than the zone 1 reference marsh, but lower densities than the northern marsh and tideflat areas (Figure 27). In winter, snow geese continued to dominate the bird densities in the restoration zone, though shorebirds and dabbling ducks also contributed. In the northern marshes, snow geese dominated in the tidal marsh environments, where goose-assisted erosion is eliminating marsh. Meanwhile on the tide flats adjacent to the northern marshes, shorebirds dominated with by far the highest densities of any group of birds in the sampling areas of the estuary. In April, the reference zone 1 marsh supported the highest densities of birds, dominated by shorebirds. Overall, the zone 1

50

marsh supported the greatest diversity of bird species, although most of the birds were recorded as flying over the site, rather than roosting or feeding (Figure 28). In contrast, five times as many birds were observed roosting or foraging on the restoration site compared to the reference zone 1 (Figure 28, and Woo et al. 2015), no doubt dominated by the large numbers of geese and dabbling ducks using the site during hunting season.

Figure 27. Post-restoration seasonal bird community structure at four separate areas of the Stillaguamish estuary. Bird numbers are presented as densities to account for differences in survey unit size. The top two graphs represent habitats near study zones 4 and 5. See Figure 6 for their locations. Note that Y-axis scales differ. This figure is from Woo et. al. 2015.

Bird use of the site varies not just seasonally, but also tidally. Figure 29 indicates how the abundance and distribution of dabbling ducks, diving ducks, geese/swans, and shorebirds varies across seasons and tides. The composition of the bird community changed post-restoration. Before restoration it was dominated by dabbling ducks and passerines, while after restoration it was dominated by snow geese, shorebirds, and a lower number of dabblers. In the north marshes and tide flats, shorebird numbers were much higher in 2015 than in 2012.

51

Figure 28. Prevalent bird behaviors in zone 2, the restoration site, and the zone 1 reference marsh. Five times as many birds were observed foraging or roosting in the Restoration site as compared to reference zone 1. Figure from Woo et. al. 2015.

Figure 29. Monthly abundances (2014-2015) of dabbling ducks, diving ducks, geese, and shorebirds at PSB Preserve during high and low tide.

52

H1.6 Benthic invertebrates Shorebird prey. Hypothesis: Restored tidal exchange will result in fine and organic sediment qualities that support primary productivity and benthic invertebrate prey of shorebirds. USGS-WERC collected benthic macroinvertebrates using benthic cores at stations throughout the estuary (Figure 6) and their results are discussed here (Woo et. al. 2015b). Samples were collected in April-May of 2012 (pre-restoration), 2013, and 2015. Samples were collected in the restoration site (zone 2) and in zone 1 (their “Reference” site). In addition they sampled in areas roughly similar to study zones 5 and 4 (their sites A and B, the “North Bay” sites), and in areas roughly similar to zones 6 and 7 (their sites C and D, the “Hatt Slough” sites). The North Bay sites (zones 4 and 5), with silty soils, had on average the highest densities of invertebrates, dominated by worms (polychaetes, nematodes, and oligochaetes), corophiid amphipods, and bivalves (Figure 30), all important prey for shorebirds and waterfowl. In contrast, the Hatt Slough sites (zones 6 and 7), with sandy soils, had low invertebrate densities (Figures 30, 31). The restoration site and zone 1 both had relatively high densities of invertebrates. Zone 1 had many polychaete, nematode, and oligochaete worms, as well as amphipods, isopods, and invertebrate larvae (Figure 30). Zone 1 had the highest single event invertebrate density, in the 2012 sample (Figure 31). The restoration site changed substantially with restoration. Pre-restoration it had invertebrate larvae and nematodes, and after restoration the densities of amphipods, polychaetes, oligochaetes, and bivalves increased. Total invertebrate density across the estuary declined over time, though individual sites differed. The benthic community structure at each site each year is shown in Figure 31. With the exception of the restoration site, general benthic community structure remained relatively stable across years. Across all years, the North Bay (zones 4 and 5) had the highest total invertebrate biomass, though the restoration site increased substantially in biomass after restoration, and saw a 500-fold increase in biomass in 2015 (Figure 32). Total biomass patterns were driven by larger taxa such as the bivalve Macoma balthica. In summary, invertebrate prey resources were greatest at the North Bay sites (zones 4 and 5), in terms of density, diversity, and total biomass. Densities were also high in zone 1, but the total prey biomass for water birds was low.

53

Figure 30. Prey species densities through time (2012-2015) at sampling sites throughout the Stillaguamish estuary. Sites A and B are “North Bay” sites, roughly equivalent to zones 5 and 4 respectively. Sites C and D are the “Hatt Slough” sites, roughly equivalent to zones 6 and 7 respectively, and the Reference site is zone 1. Figure from Woo et. al. 2015b.

54

Figure 31. Port Susan Bay-wide invertebrate communities through time (2012-2015). Figure from Woo et. al. 2015b.

At the restoration site, pre-restoration prey biomass was low and similar to the Reference Marsh (Figure 32), but after restoration, biomass increased to levels similar to the North Bay sites. After restoration, bivalves quickly colonized and provide a substantial portion of the invertebrate biomass. Between 2013 and 2015 there was a large increase in invertebrate density in the restoration site (Figure 31), driven by an expansion in the numbers of polychaetes and oligochaetes. However, in terms of biomass the community shifted from one dominated by bivalves to one dominated by insect larvae, with bivalves and polychaetes as secondary dominants (Figure 32). At the zone 1 reference site, insect larvae also increased in both density and biomass in 2015. Across the three years, the surface soil at the restoration site shifted, with a decline in the relative % sand and more than doubled in % clay (Woo et. al. 2015b). It is perhaps an interesting coincidence that 2015 was also the year of the large marsh dieback in the restoration site. Whether similar factors were driving the changes in both marsh and invertebrate community structure is unknown.

55

Figure 32. Total invertebrate biomass for invertebrate prey species by site and year. Figure from Woo et. al. 2015b.

Objective 2: Improve connectivity between the river and northern tidal habitats, increasing the distribution of freshwater, sediment, energy and other materials. While Objective 1 was focused on habitat development at the restoration site, Objective 2 is focused on impacts of the restoration on estuary-wide processes and functions. H2.1 Accretion Rates Increase N of Hatt Slough, don’t change to S. Hypothesis: Accretion rates in existing marsh and tidal flat north of Hatt Slough will increase and accretion rates south of Hatt Slough will not change. This hypothesis was based on the assumption that the restoration would result in increased flow of freshwater, and the materials it carries, to the north over the restoration site. A second assumption is that there is sufficient sediment load in the river that currently bypasses the tidal marsh system, that an increase in flow towards the north will not decrease the amount of sediment deposited on marshes to the south. In other words, restoration will allow a greater proportion of the total sediment load to be captured by the tidal wetlands. There isn’t enough pre-project accretion data to determine whether rates changed as a result of the restoration project. Sediment export from the river varies annually depending on flow

56

patterns related to annual and decadal climate patterns. For this reason, accretion rates are also expected to vary annually. Accretion of new sediment onto marsh surfaces is a critical component of elevation change, but not the only component. To monitor the actual change in elevation, Surface Elevation Tables (SETs) are the best method, by integrating both aboveground processes such as mineral soil deposition, and belowground processes such as organic matter accumulation, decomposition, and compaction. The oldest SETs were installed in zones 1, 3, and 5 in 2011 (Rybczyk and Poppe 2017), the year before restoration. At this time, a SET was also installed in the lower unvegetated mudflat west of zone 3. In zones 1, 3, and 5, a pair of SETs were installed in high marsh, which is infrequently inundated and dominated by Juncus balticus (Juba), and a pair was installed in low marsh, dominated by Scpu and Boma. SETs were installed in the restoration zone 2 and in zone 4 in 2014, which was also the year of the Oso landslide. In the restoration zone, there is no high marsh so 4 SETs were distributed across the elevation gradient of the site. In zone 4, the higher SETs were not installed in typical high marsh, but were installed in the upper reaches of middle marsh, in an area dominated by Bofl. SET locations are shown in Figure 4. At most of the sites shown in the figure, each point represents the location of a pair of SET replicates. Overall, the restoration zone is accreting at a rate about three times the reference zones (Figure 33, 3.1 vs. 0.97 cm/yr). On average, all of the reference zones are currently keeping up with sea level rise, though individual SETs vary. The only SETs not keeping up are the low marsh sites in zone 4, and the mudflat SET.

Figure 33. Mean annual elevation change, by zone, for the entire period of record (2011 – 2017) with standard error bars. Zone 2 is the restoration area. Figure from Rybczyk and Poppe 2017.

In reference zones 1 and 3, near the mouth of Hatt Slough, low marsh habitats which are inundated for longer time periods than high marsh, accrete at higher rates than high marsh (Figure 34). This is consistent with most published literature on tidal marsh accretion rates. High

57

marsh is generally inundated less frequently than daily, or during floods. On average low marsh in reference zones near Hatt Slough accrete 1.2 cm/year, while high marsh accretes 0.8 cm/yr (Rybczyk and Poppe 2017).

Figure 34. Typical examples of elevation change in study zones near the mouth of Hatt Slough. The top graph is from a high marsh SET and the bottom a low marsh SET, both in reference zone 3. Figures from Rybczyk and Poppe 2017.

In contrast, in the northern zones 4 and 5, high marsh is accreting faster (0.8 cm/yr), than low marsh (0.14 cm/yr) (Rybczyk and Poppe 2017). These numbers don’t include the highest two SETs in zone 4 which are placed in a Bofl marsh rather than true high marsh. In fact, the zone 4 low marsh SETs are actually eroding and are the only marsh SET locations in the estuary that are not keeping up with current sea level rise (Figure 35). The fact that the lower elevation SETs in zones 4 and 5 are not accreting more than the high marsh SETs is likely indicative of an area that is not resilient, particularly in the context of climate change. Both zones have been eroding from the seaward marsh front for many years and the lower SETs in both zones are in areas that are very dynamic and undergoing erosion. In these areas, elevation tends to be more variable than in other areas because erosion doesn’t happen uniformly, but in rills and ponds. Some of the sediment removed from these rills and ponds is deposited nearby or pushed up into the high marsh. For this reason, SETs in the low marsh may fluctuate more widely than other SETs, sometimes eroding and sometimes accreting rapidly (Figure 36). The high marsh SETs in these areas may similarly vary, depending on whether winter storms in any particular year happen to deposit a load of sediment at the SET location. The greater variability in zones 4 and 5 is evident in the wider standard error in the average values for these sites (Figure 33) compared to reference zones 1 and 3.

58

Figure 35. Typical elevation change in zone 4. The top graph is a SET located in a dense, tall Bofl marsh at the higher elevation end of middle marsh. The bottom graph is from the low end of the middle marsh where Boma dominates. Figures from Rybczyk and Poppe 2017.

The highest elevation SETs in zone 4 are placed in a Bofl meadow and so are not comparable with either high marsh or the typical low marsh habitats in any zone. The Bofl SETs in zone 4 had the highest accretion rates of all reference marsh sites, averaging 2.7 cm/yr (Rybczyk and Poppe 2017). This unusually high rate may be due to two factors. First the seaward edge of the middle marsh is eroding, with rills like fingers extending from the marsh front up into the marsh, cut by winter waves. The excavated sediment is picked up and pushed higher into the marsh. During the late winter and early spring, patches of soft new sediment depositions are found in the higher part of the marsh, while the seaward eroded gullies cut 10-30cm or more deep into the soil, exposing bulrush rhizomes. As a result, the local sediment delivery rate to the upper middle marsh in this area is likely higher than would be predicted by just including river inputs. Secondly, accretion at the upper SETs is influenced by the Bofl meadow, the only area in the estuary where Bofl forms a dense, monotypic meadow. Bofl is the tallest middle marsh species, and it has the most robust stems and leaves that, despite senescing in the fall, often remain vertical through the winter storms. In contrast, Scpu stems break down during the fall and play a relatively minor role in winter sediment trapping, while Boma and Scta stems break down gradually during fall and through winter, reducing their role in winter sediment trapping (Fuller 2014, Stellern 2016). For this reason, Bofl is likely much more actively involved in attenuating waves and trapping sediment throughout the winter sediment season (Stellern 2016). Bofl was

59

only recently identified as a separate species from Boma in this estuary (Fuller 2015a). It has different environmental preferences than Boma, apparently plays a bigger role in sediment trapping, and in 2015 appeared to be more tolerant of high pore water salinity (Fuller 2017). For these reasons, Bofl needs to be studied for its potential roles in increasing marsh resilience to climate change. Its’ ecological preferences could be used to guide the design of restoration projects in ways that maximize the biomass productivity and climate resilience of the restored marsh.

Figure 36. Typical SETs from zone 5. Top graph is a high marsh SET and bottom graph is a low marsh SET. Figures from Rybczyk and Poppe 2017.

The effect of restoration on the accretion rates of reference zones can’t be determined with the limited pre-project accretion data that we have. However, the project is not likely to have significantly increased sediment distribution to the northern marshes. First, based on the 3D hydrodynamic model developed to inform the selection of a preferred alternative during preliminary design, we expected that dike removal would result in flow from Hatt Slough directly onto the restoration site during most high tides (Yang et al. 2006). However this has not occurred post-restoration, another unexpected outcome. Direct flows from the river onto the restoration site only occur during winter flood flows, resulting in less delivery of sediment to the north than expected. In addition, site subsidence and the resulting bathtub configuration of the restoration site will likely also result in higher trapping of sediment on the restoration footprint, lessening the amount of sediment that would make it past the site to the north. During major floods, which carry very large amounts of sediment and enough water to push sediment in sheet flow over and past the restoration site, there may indeed be an increase in the flow of sediment to northern marshes. But these conditions will be limited in frequency.

60

Sediment delivery, distribution, and erosion: USGS monitored turbidity, suspended sediment, and flows in the lower river, Old Stilly Channel, and the restoration site between late 2013 and 2015 (Grossman and Curran 2015). During this period, they calculated a total suspended sediment load of 2.6 million tonnes (Figure 37), with about 0.8 - 0.9 Mt resulting from the Oso landslide, or about 35% of the total load. Of this amount, about 1%, or 21,200 tonnes, entered the restoration area. During one 7-day period in November 2014 there was net sediment flux out of the restoration area, correlating with a coastal storm event with high tidal surge and winds that apparently re-suspended sediment from the marsh surface, which then advected out of the site (Grossman and Curran 2015).

Figure 37. River flow and cumulative sediment load for the Stillaguamish River during the USGS-PCMSC study. Flow (discharge) is from USGS Site 12170300. Note the timing and influence of the SR530 (Oso) landslide. Figure from Grossman and Curran 2015.

As discussed earlier, the northern marshes are actively eroding away, and have been eroding for decades (Figure 38). The maximum extent of Stillaguamish tidal marsh, after the initial diking that eliminated the original marsh, was in 1990, based on aerial photos (Fuller and McArdle 2014). Zone 5 marsh reached its maximum extent in about 1964 (99 acres) and lost almost 40% area by 1990, at a rate of about 1.3 ac/yr. Zone 4 marsh reached its’ maximum extent in about 1990 (179 acres), but lost a remarkable 70% of its area by 2011 (54 acres), a loss rate of about 6 ac/yr (Fuller and McArdle 2014). While the northern study zones have been losing marsh, reference zones 1 and 3 were growing in size up until 1990 and since then have not grown appreciably (Figure 38). The change in maximum width of the marsh in each study zone is shown in Table 8. As can be seen, the marsh in zones 4 and 5 has been retreating at 10-

61

40m per year since 1990. After advancing at 4-10m per year during the latter half of the 20th century, zones 1 and 3 now appear to advancing at less than 1m per year.

Figure 38. Trends in the area of continuous tidal marsh in the 5 study zones. 1933 and 1947 data are lumped because 1933 aerials only cover zones 4 and 5, and 1947 aerials only cover zones 1-3. Note that the X axis shows photo years and is not to scale. Figure from Fuller and McArdle 2014.

This process of erosion is continuing and signs of erosion are now evident in zone 5 throughout the middle marsh, up to the LWD line. Most of the high marsh in zone 5 is gone and the LWD line armors the shrub zone. At the shrub line, buried beneath the LWD is an approximately 1m vertical bank that drops from the shrub elevation directly to the middle marsh elevation. All that remains of high marsh are a collection of eroding pedestals in the middle marsh, and near South Pass a larger wedge of high marsh. In zone 4, erosion rills are evident throughout the middle marsh zone of Boma, extending up to the monotypic Bofl meadow which occupies a horizontal band of about 30m width immediately below MHHW where the high marsh begins.

Table 8. Historic changes in the maximum width, in meters, of tidal marsh in the 5 study zones. Zone 2 is the restoration zone which was diked in the late 1950’s and restored in 2012. Table is from Fuller and McArdle 2014.

62

The wave model developed by Grossman and Curran (2015) indicated that the area of maximum excess shear stress during winter storm waves coincides with the areas of active erosion (Figure 26). The rapid loss of marsh since 1990 is likely driven by the interaction of several factors, including declining river flows through the Old Stilly channel to South Pass, channel migration at the mouth of Hatt Slough, increasing pore water salinity, and particularly the interaction of grazing disturbance by snow geese and winter waves (Fuller 2017). These factors are discussed below under hypothesis 2.5. Although Hatt Slough has been the primary river distributary since the early 1900’s, the amount of river flow in the Old Stilly channel continues to decrease over time, resulting in less delivery of freshwater and sediment via South Pass. In addition, as a result of two major floods in 1990, the distribution of water among the Hatt Slough distributary network shifted, with the main flow diverting southwards via the channel between Goose and Clown Islands (Figure 39). The result of this shift diverted the primary flow of freshwater and sediment south on the shortest path to the bay’s deeper water, and thereby bypassing the north tidal marshes. In 2007, another flood delivered a large amount of LWD to the western end of Hatt Slough, trapping a large volume of sediment, blocking a northerly distributary and shifting even more flow to the south around the west side of Goose Island (discussed further below in Hypothesis 2.5, in Figure 43, and in Fuller and DeBono 2015).

Figure 39. The Hatt Slough distributary network shifted after the major floods of 1990, diverting the primary flow of freshwater and suspended sediment to the south.

63

Despite these changes in sediment routing, most of the SETs in zones 4 and 5 are recording sufficient sediment accretion to keep up with current sea level rise. However what we don’t know is how much of that sediment is direct delivery from the river and how much is sourced locally from the eroding marsh front. One of the reasons for the restoration objective to restore processes to the northern marshes was the hypothesis that the old levee on the TNC property at the mouth of Hatt Slough was partially responsible for reducing the flow of sediment to the northern marshes and was thereby facilitating marsh loss due to insufficient sediment supply. The old levee was angled southwest (see levee location in Figure 39) and may have played a role in shifting the thalweg of Hatt Slough to the south. We know that other factors such as snow goose grazing contribute to marsh erosion in the north (Fuller 2017), but it remains unknown whether, in the absence of geese, sufficient river sediment is reaching the northern marshes for them to keep up with SLR. H2.2 Salinity Decrease N of Hatt Slough, don’t change S. Hypothesis: Water column and pore water salinity will decrease north of Hatt Slough and will not change south of Hatt Slough. As with sediment, this hypothesis is based on the assumption that the restoration project would result in regular flow at high tide from Hatt Slough onto the restoration site, increasing the volume of fresh water that goes north from the mouth of the river. The pre-project 3D hydrodynamic model of the restoration indicated that restoration would push water north, resulting in a greater residence time for fresh water in the estuary and a larger area where brackish tidal marsh would be supported (Figure 40, and Yang et al. 2006). Model animations using river flow levels that were near the annual average flow rate showed regular flow from Hatt Slough across the restoration site during high tides. This change in flow distribution resulted in a substantially larger estuarine area with surface salinity lower than 22 ppt which is near the upper salinity tolerance level for most tidal marsh plant species (e.g. Hutchinson 1988). The project hypothesis was that this change in fresh water distribution should substantially improve marsh resilience in northern marshes by reducing the stress of higher pore water salinity. The northern marshes are close to the salinity thresholds that affect productivity and ultimately survival. Pore water salinity from summer 2004-5 indicated that the northern marshes experienced 10-13 ppt salinity (Fuller 2016). Those years had summer flows that were close to the long term average summer flow for the Stillaguamish. As described in more detail in Fuller 2016, the science literature suggests that Scpu (called S. americanus in earlier reports) biomass production declines by 60% at 12 ppt salinity, and growth is reduced by 62% at 15 ppt. For Boma, senescence began by 15 ppt, with complete senescence at 20 ppt. Observed differences in aboveground biomass between the different study zones correlates with these literature values, with biomass/m2 in the northern Boma and Scpu marshes being substantially less than biomass quantities in the Hatt Slough reference marshes (Table 3, and Fuller 2017).

64

Figure 40. Salinity maps at high tide, generated by a pre-project hydrodynamic model showing the predicted effect of dike removal on estuary salinity. Warm colors are high salinity. The dark grey line marking the extent of cooler colors indicates the approximate location of the 22 ppt salinity line which is the maximum salinity tolerance threshold for brackish tidal marsh. Most bulrush species, including Scpu and Boma, have substantially reduced productivity at 15 ppt. The model projection for pre-project conditions (A) shows a brackish zone that matches remarkably well with the actual extent of tidal marsh. The prediction for post-project (B) shows a substantially larger area where lower salinity would support marsh resilience and expansion. Maps were taken as screen shots from model animations created by Battelle (Yang et al. 2006).

In summer 2015 river levels reached record lows beginning in late March and extending at least through August. The conditions were considered analogous to the projected average conditions beginning sometime after 2050 (Fuller 2016), as a result of declining snow accumulations during winter. During 2015, pore water salinity was two to three times higher than in 2004-2005 when flows were near average (Figure 41). August pore water salinity exceeded 20 ppt in all parts of the estuary except the middle marsh of zone 1 (Table 9). These high salinity levels led to an overall 34% reduction of plant height in the mixed bulrush of the middle marsh, and a 43% reduction in the low marsh with monotypic Scpu (Fuller 2016). In the restoration zone, aboveground biomass declined by 70% compared to 2014. In addition to biomass reductions, most of the bulrush marshes failed to flower or set seed due to stress levels, and Scpu and Boma were fully senesced about a month earlier than normal throughout the estuary except at the highest elevations of the Hatt Slough marshes (author’s personal observations). These observations support our earlier finding that climate change impacts on summer river flow will likely have much larger and earlier impacts on the Stillaguamish estuary, and likely most Puget Sound estuaries, than will sea level rise, at least in the next 5-10 decades (cf. Clough and Larson 2010). The large decline in biomass production in summer 2015 represents a very substantial

65

impact on the estuarine food web which depends on plant biomass to fuel the detrital food web that produces the invertebrates that feed the salmon, shorebirds, and other wildlife.

Figure 41. Average seasonal pore water salinity across the Stillaguamish estuary. River flows were close to average in 2004-2005, but hit record low levels for several months in 2015.Figure is from Fuller 2016.

Regarding the project hypothesis, there is insufficient pre-project data to determine whether the restoration project has changed salinity levels. Salinity varies substantially from year to year depending on river flow patterns and air temperature effects on the snow pack, so long term data are needed to be able to tease out agents of change. However, the fact that high tides do not overtop the Hatt Slough bank and flood the restoration zone indicates that there is likely very little effect of the restoration site on salinity in the northern marshes. Increased overtopping during winter floods is irrelevant to summer salinity.

Table 9. Pore water salinity values in the five study zones of the Stillaguamish estuary. Table is from Fuller 2016.

66

H2.3 Channel systems Expand N of Hatt Slough. Hypothesis: Distributary and blind tidal channel systems north of Hatt Slough will expand in size and complexity. This hypothesis is based on the assumption that increasing the tidal prism north of Hatt Slough will increase the hydraulic energy available to scour channels. The expansion and development of the channel system on the restoration footprint was described above under hypothesis 1.4. The principal distributary west of the restoration site expanded as a result of the additional tidal prism it now carries, both widening and deepening downstream of the northern breach at the restoration site (Figure 42, and also Fuller 2014 and Grossman and Curran 2015). Upstream of the northern breach as well as the southern breach, the distributary channel does not appear to have changed substantially in size. Flow velocity and discharge volume at the northern breach was substantially higher than the southern breach (Grossman and Curran 2015) due to a greater cross-sectional area and deeper breach, resulting in the larger downstream change in the distributary channel.

Figure 42. Restoration increased the tidal prism and enlarged the distributary channel near the restoration site. Figure adapted from Grossman and Curran 2015.

Other channels west and north of the restoration site do not appear to have changed as they are not as yet associated with the expanded tidal prism of the restoration site. We expected that existing blind tidal channels near the restoration site would connect across the old dike footprint and would then increase in overall size and complexity as a result of the greater drainage area. However the old dike footprint is very compacted and resists erosion. At this time, no channels have formed across the old dike footprint though signs of shallow erosion are evident. Hind sight suggests that more of the old dike footprint should have been excavated

67

during restoration, the soil on the old dike footprint should have been treated in some way to reduce compaction, or additional breaches of various sizes should have been cut across the dike footprint to facilitate increased connectivity with nearby blind tidal channel networks. In addition, regular high tide flow directly from Hatt Slough onto the restoration site has not occurred, though was predicted by pre-restoration modeling. This would have delivered more water and hydraulic energy towards the north and would likely have expanded the volume of the channel network. H2.4 LWD Increase N of Hatt Slough. Hypothesis: The amount of large woody debris (LWD) will increase in the tidal marsh north of Hatt Slough This hypothesis was based on the assumption that over-site flow from Hatt Slough would occur and would carry with it additional LWD from the river, particularly during floods when most LWD is recruited from the watershed. Aerial imagery was used to estimate wood volumes in 2011 (Woo et. al. 2014) and in 2013 and 2014 (Fuller and DeBono 2015). However, the two efforts evaluated LWD at different geographic scales and we have not yet synthesized the two approaches. Analysis of the 2011 imagery found 2,316 pieces of LWD or log jams (Woo et. al. 2014), however the spatial scale of analysis is not clear. They found that 65% of LWD was associated with vegetated habitats, 23% was in channels, and 6% on mudflats. In 2013, there were 6,788 log units in the entire estuary south of South Pass, increasing to 8,696 in 2014 (Table 10, and Fuller and DeBono 2015). Log units may be individual logs where they are easy to separate visually in aerials, or they may be log piles where many logs are intertwined and overlap. However between 2013 and 2014, the average size of log units appears to have decreased because the total area covered by the wood declined from 68,835 m2 to 57,203 m2. In the Hatt Slough marshes of zones 1 and 3 as well as in the restoration zone 2, both the number of wood pieces and the area covered by wood declined from 2013 to 2014. However in the northern marshes of zones 4 and 5, both parameters increased, potentially the result of winter storms. Zone 4 accumulates that greatest amount of wood, occupying the pocket of the Port Susan Bay “catcher mitt”. LWD numbers and dynamics likely vary substantially from year to year depending on winter flood and storm patterns. With only one year of pre-project LWD and two years post-project digitized, there isn’t enough data to attribute these changes to restoration or to any other factor at this time.

68

Table 10. Amount of large woody debris across the estuary and in each study zone in 2013 and 2014, digitized from aerial photos. Table is from Fuller and De Bono 2015.

H2.5 Marsh area Expand N and W of restoration site. Hypothesis: Tidal marsh area west and north of the restoration footprint will expand. Marsh area continues to shrink north of the restoration area. As was described earlier, marsh loss has been in progress in zones 4 and 5 for many years (Figure 38, and Fuller and McArdle 2014). To the west of the restoration area, including zone 3, there is evidence of marsh thinning and some erosion at lower elevations (personal observation). After the dikes were first built, wiping out most of the estuary, marsh expanded seaward of the dikes. South of South Pass, the tidal marsh area peaked in the 1990 aerial at 1,064 acres and then declined by 26% to 792 acres in 2011, largely due to the erosion of marsh in zones 4 and 5. The 2012 restoration of zone 2, as well as marsh expansion in zone 6 brought the total up to 972 acres in 2013 (Fuller and McArdle 2014). The role of “ephemeral” LWD in marsh expansion The marsh expansion in zone 6, west of zone 3, began after a flood in the winter of 2006-7 triggered channel migration and stabilization at the western terminus of Hatt Slough (Figure 43, and Fuller and DeBono 2015). Zone 6 was bare mudflat in 2006, with highly dynamic, mobile channels shifting across the area. The 2007 flood event deposited about 15 acres of logs at the far west end of Hatt Slough, triggering rapid accretion and sufficient channel stabilization to allow low marsh to begin to colonize zone 6 (Fuller 2017). This marsh expansion is not related to the restoration project. The 15 acre debris field of logs in 2007 was ephemeral, with most logs floating off within one or two years. I call a debris field rather than a log jam because the logs were loosely and thinly piled over a large area, and most logs disappeared within a year. This same situation occurred following a large flood in 2003, though that event did not appear to trigger rapid accretion or channel change. LWD dynamics along Hatt Slough were monitored by TNC in detail following the 2003 event (Selleck 2006), and most of the log field was disbursed within two winters. No similarly extensive log fields appear to have been deposited since the 2007 event.

69

Considering that, aside from restoration, the only major source of wetland expansion since 1990 has been related to a large, ephemeral, flood-related log debris field, this suggests an important and unrecognized role of LWD in estuaries. The geomorphic and fish habitat roles of stable log jams are much better understood, thanks to extensive research on rivers, but these broad, temporary log debris fields appear to be unrecognized as an important geomorphic control in estuaries. Though ephemeral, they may trigger channel migration, rapid accretion, and marsh development. Their ephemeral nature make their role difficult to discern.

Figure 43. Role of ephemeral log debris fields or “dumps” in estuary geomorphology. As a result of major floods, large 15-20 acre log debris fields were left in the estuary at the west end of Hatt Slough during 2003 and 2007. In the 2007 event the logs trapped enough sediment to fill a major channel, shift it from NW to SW, and initiate a large area of marsh expansion. This is the only area of significant marsh expansion in the Stillaguamish estuary since 1990.

The role of geese and winter waves in marsh erosion Our current best understanding of the erosion dynamics west and north of the restoration site is that marsh loss results from the interaction of several factors: over-grazing of rhizomes, relative exposure to winter waves, different effects of plant species on sediment/wave dynamics, soil particle size, and pore water salinity.

70

The dominant forces seem to be the interaction of snow goose grazing with winter waves. Excavation of rhizomes loosens the soil during the stormy winter when there is the least amount of aboveground biomass to reduce wave scour. Excavation also disturbs the diatom layer on the surface which plays a key role in estuary soil stabilization (cf. Wood and Widdows 2002 and Lucas et. al. 2003). Tide and waves may then re-suspend loosened soil, particularly the finer particles, and move them either higher into the marsh or offshore. Swans also excavate rhizomes, though their population is much lower than the goose population. On the Fraser River delta, tidal marsh plots where snow geese were excluded accreted sediment faster than the rate of SLR, but plots where geese were not excluded were eroding at a rate of about 1cm/year (Kirwan et. al. 2008). To evaluate relative differences in both winter wave exposure and goose grazing disturbance, we developed indexes of each, described in the caption to Table 1 and in Fuller 2017. Figure 44 illustrates the physical evidence used to evaluate plots for grazing disturbance. The level of disturbance was ranked according to the presence, number, and size of excavation pits, density of footprints, and presence of exposed rhizomes associated with the pits. To compare study zones in terms of the grazing-wave interaction, we can simply add the wave and grazing indexes together for what we might call a biophysical erosion risk index.

Figure 44. Examples of physical evidence used in the grazing disturbance index. Footprints of various ages can be seen, with the most recent ones being deeper and free of the darker brownish diatom growth. Pits of various ages are also evident, where rhizomes have been excavated. The most recent pits are deeper, with steeper sides and adjacent sediment piles, indicating that tides have yet to smooth off the edges. Exposed rhizomes and roots are evident in the area of most recent excavation in the lower center of photo.

71

Zones 4 and 5 have eroded the most quickly over recent decades (Table 8). Zone 5 has the highest wave exposure index, and the third highest grazing index (Table 1, and Fuller 2017). When we add the indexes, zone 5 has the highest erosion risk value (3.92). Zone 4 has also eroded quickly, but only has the fourth highest wave index. However, it has the second highest grazing index which is enough to give it the second highest erosion risk value (3.54). Zone 3 shows signs of heavy grazing, with the highest grazing index value (Table 1). Stem density may be declining in this zone (though we don’t have enough data yet to confirm one way or the other). However, unlike zones 4 and 5, there aren’t signs of active wave erosion. Zone 3 has a lower wave exposure index than zones 4, 5, 6, and 7 (Table 1) resulting from its presence on the north slope of the concave prograding delta at the mouth of Hatt Slough. Being on the “lee” side of the delta, storm waves from southerly quarters first have to travel up the gradient of the delta, cross the high marsh and banks of driftwood in zones 1, 7, and 3, before arriving at the middle marsh of zone 3. In contrast, southerly storm waves have no biophysical roughness to impede their access to zone 5 (Fuller 2017). Zone 3 has the third highest combined erosion risk index value at 3.24. In terms of the combined erosion risk index, the remaining study zones fall in the following order of decreasing risk: 6, 7, 1, 2, 8. However, the following caveats should be understood. The grazing disturbance index was developed using data collected in the late winter/early spring of 2015 only. The data included physical evidence of grazing disturbance but did not include actual bird observations. Some physical evidence like grubbing pits remain visible for years. However other evidence like goose footprints only remains evident for weeks, and a major storm could potentially erase it more quickly. As a result, seasonal differences in goose distribution may not be picked up in the current grazing index values. This may be important, particularly in regards to the restoration zone as described below. USGS WERC monitored bird use of some study zones once per month during several months of 2012 and 2014-2015 (Woo et. al. 2015). They found high snow goose and swan numbers in the area of the zone 4 and 5 tidal marsh and tidal flats (Figure 27) in January and February, after the hunting season. It should be noted that their “tidal flats” sample area largely overlaps an area we have called low marsh due to the occasional sparse Scpu patches that are present, and that in recent decades was thickly vegetated. The USGS observations of heavy geese presence north of the restoration area supports our grazing index values that are based on physical evidence. In zone 1, snow geese and swans were recorded in November in relatively low numbers, and were only present at trace levels in other months (Figure 27). This also aligns with the relatively low grazing disturbance index we found for zone 1. The lower grazing intensity may be due to a higher sand fraction in the soil (see below for discussion). Zone 1 also has a relatively low wave exposure index, so wave action is less likely to re-suspend disturbed soil (Table 1, and Fuller 2017).

72

However, Woo et. al. (2015) found the highest concentrations of snow geese and swans in the estuary to have occurred in the restoration area during the fall (hunting season). They also found large numbers there in January and February. But we found that the restoration zone had the second lowest grazing disturbance index value (Table 1). This could be for several reasons. First, our index was based on data gathered at the restoration site in March, so there would have been time for some of the physical evidence in the restoration zone to be filled by new sediment or eroded by storms. Secondly, as described earlier, the bath tub shaped topography of the restoration site results in far higher accretion rates, so evidence could be buried more quickly than in other zones. And finally, soils in the restoration zone are physically very soft due to being perpetually saturated (compare soil penetrability values in Table 2). As a result, evidence of physical grazing would be more likely to disappear quickly than in the firm soils of other zones. These caveats should frame the interpretation of our grazing disturbance index, particularly our understanding of the potential role of snow geese on the trajectory of marsh development on the restoration site. Additional data on seasonal spatial distributions of geese could likely be gleaned from the WDFW aerial counts, which could further inform our understanding of geomorphic dynamics. The grazing-wave interaction may be the primary factors of erosion, but are not the only contributing factors to erosion and marsh loss. Different plant species have different vulnerabilities to grazing. Scpu appears to be the most vulnerable since its aboveground biomass senesces to ground level in early Fall and it therefore offers little resistance to goose access and excavation. The degree to which it is vulnerable, however, may relate to the sand content of the soil. Zones 4 and 5 low marsh have the lowest sand content (15 and 19%) and highest penetrability (8.0 and 8.3) of the all the reference low marsh areas (Table 2, Fuller 2017). The Hatt Slough reference zones (1, 3, 6, 7, and 8) are closer to the mouth of the river and all have higher sand content (33.8-66.2%) and lower penetrability (mostly less than 5.8). These differences suggest that geese would find it less metabolically costly, and therefore more attractive to grub rhizomes from the northern marshes. Sean Boyd with the Canadian Wildlife Service, has experimentally tested goose grazing efficiency in different soil types and found that soils with greater sand content were more difficult for model goose bills to probe, and supported a substantially lower rate of rhizome recovery per unit effort (Boyd, pers comm). This pattern may partly explain why the low marsh in zones 4 and 5 is mostly bare, with very sparse Scpu (Fuller 2017), and has higher grazing disturbance index values compared to similar areas of the Hatt Slough reference marshes. This finding also suggests that the restoration site could be more vulnerable to goose grazing than the Hatt Slough reference marshes because it has very low sand content and high soil penetrability (Table 2). In contrast to Scpu, Boma retains more of its aboveground biomass in the fall, and slowly breaks down over the course of winter (Fuller 2014, Stellern 2016). Early in the season, the biomass is likely tall enough to inhibit goose access since they avoid tall vegetation that limits views of predators. However as it breaks down over the winter season, geese move in from the seaward edge. While Scpu can grow well on any soil including almost pure sand, Boma only

73

occurs on soils that have less than 45% sand, and is most productive on fine soils (silt/clay) without sand (Fuller 2017). It seems likely that one of the first impacts of overgrazing may be the disappearance of Boma because as tidal action and waves re-suspend disturbed soil, the fines are the most likely component to be winnowed away. This coarsening of the soil may make the site unfavorable to Boma despite occurring at an appropriate elevation, leaving the more grazing-vulnerable Scpu as the only marsh plant species. This may explain why the low marsh in zones 4 and 5 is far lower in stem density and height than in other reference areas (Table 3). Even at the species level, Scpu is much shorter and lower density in the northern marshes than in the southern, despite the higher sand fraction in the southern marshes (Fuller 2017). Bofl is the largest bulrush, with the most rigid stems and persistent aboveground biomass through winter. Oftentimes old, dead Bofl stems, leaves and even old florets will be fully intact and still present the summer following their growth (personal observation). This large aboveground biomass is an effective barrier to snow geese who likely avoid them due to the predator detection issue. Bofl is also available throughout the winter to interact with waves, decreasing their energy and trapping sediment. We are just beginning to study the distribution of Bofl, but it is tallest and most productive at the highest elevations in the middle marsh, near the MHHW line, and it appears to prefer the finest soil grain size (Fuller 2017). However, it is unclear if it grows only in places with fine soil particle size distribution or if it successfully traps more suspended fine particles, resulting in finer soils. At lower than ideal elevations, Bofl is very difficult to tell apart from Boma and it may therefore be more common at lower elevations than we know. In 2017 when river flows were high and pore water salinity lower, it became clear that Bofl was extensively present in the restoration site, intermixed with Boma at elevations well below MHHW. Bofl clones were larger, flowered later, and appeared more robust. But in previous, higher salinity years, there was no obvious morphological difference. The seasonal progression of the physical structure of the major bulrush species, and their interaction with sediment and waves have been investigated in more detail by Stellern (2016). Finally, the last factor that appears to affect marsh loss and erosion is salinity. The volume of aboveground biomass of each bulrush species is related to pore water salinity, with declining biomass as salinity increases (Fuller 2017). The amount of aboveground biomass affects the winter physical structure that is available to interact with waves and trap sediment. Zones 4 and 5 have the highest salinities in both spring and summer, and their aboveground biomass volume in the Scpu/Boma marsh is considerably less than half that of zones 1 and 3 (Table 3, and Fuller 2017). It is likely that zones 4 and 5 experience summer salinities in excess of the growth threshold for bulrushes with some regularity, at least during years of lower than average stream flow. This would help explain the considerably lower biomass values for the northern zones. Lower winter biomass results in lower resistance to grazing and to wave energy, and greater vulnerability to erosion. It is likely that the interaction of salinity stress with

74

grazing and waves affects marsh dynamics and the pace of change. Low flow years may result in faster marsh loss in northern marshes than other years. One surprise from summer 2015 was that Bofl appeared to be the least affected of all bulrush species by high salinity. We had assumed that river bulrush would be sensitive to salinity since it has not been reported from estuaries elsewhere in Washington and Oregon. However at the end of August, at maximum soil salinities over 20ppt, Bofl was the only bulrush that had not already senesced. Though its leaves were often yellowing at the tips, it was generally green and almost full sized. This contrasts with Scpu and Boma which had completely senesced throughout the estuary and were about 35% smaller in height than in summer 2014 (Fuller 2016). Even Bofl in zone 5 was still green and robust, at least at the upper edge of the middle marsh (Fuller 2017). Erosional dynamics are also evident in the morphology and slope of the transects in each zone (Figure 13). Erosional dynamics tend to introduce slope breaks where habitat shifts from unvegetated tidal to vegetated marsh. In Figure 13, the seaward edge of the middle marsh, dominated by both Boma and Scpu, is marked with a colored circle on each transect. Middle marsh maintains some aboveground marsh biomass during winter to interact with waves and sediment. Seaward of this point is the low marsh dominated by pure Scpu which is essentially absent during the winter storm season. In zones 4 and 5, the sharply steepened slope at the colored circle is very evident. Below this point, the winter-bare low marsh and unvegetated intertidal slope is much gentler and smoother. In contrast, in zone 1 the transect is much less steep, and more uniform across its entire length. Zone 3 has a slope that is part way between zone 1 and zone 4 in terms of steepness. The early part of the transect in zone 3 has a relatively steep slope and the fact that zone 3 had the highest index of goose grazing disturbance for the winter of 2014-15 (Table 1) suggests there may be some cause for concern regarding the trajectory for marsh erosion in zone 3 (Fuller 2017). In summary, regarding the project hypothesis, it was hoped that the restoration site would increase connectivity between Hatt Slough and the northern sites, allowing more freshwater, sediment, organic matter, juvenile fish, wood and other material to get to the northern part of the estuary. The desired result was northern wetlands that were more productive and more resilient to climate change. As described earlier, the level of connectivity between Hatt Slough and the restoration site is less than expected, with no direct connection except during winter floods. During spring and summer, when pore water salinity appears to be a limiting factor for marsh productivity in the northern marshes, there is no direct flow from the river across the restoration site, and no apparent change in connectivity between Hatt Slough and the northern marshes. For this reason, it is unlikely that restoration has improved salinity or sediment conditions in the northern marshes, and there is no evidence that the restoration has reduced the rate of marsh loss in the northern and western marshes, or affected their overall productivity.

75

VI. Final notes

Interaction of stresses The interaction of stresses is critical to marsh resilience (Fuller 2017). Climate change impacts from river flow, sea level rise, higher storm surge as a result of sea level rise, and increasing goose populations won’t affect the estuary independently, but will interact with each other and with existing stresses imposed by levee configuration, land use changes in the basin, and other little-known stresses like insect herbivory. Climate change will amplify the effects of existing stresses and increase the frequency with which multiple stresses co-occur. This would accelerate the rate of marsh loss we already experience. Sediment delivery from the watershed is likely to increase, with the projected increase in winter floods, which could help marshes keep up with sea level rise. However the marshes already get enough sediment to build elevation and they’re still disappearing, so additional sediment may not help the marshes, at least under current conditions. The exception is the restoration area where the bath tub shape would allow it to capture some of the additional sediment, accelerating accretion and allowing the site to return to greater productivity more quickly. In order for marsh area to be sustained or to increase in the face of climate change, the “plumbing” of the delta (the levees and channels) needs to be realigned in a way that increases the area of tidal marsh, particularly at higher marsh elevations, maximizes the residence time of freshwater in the estuary, and captures a larger proportion of the finer suspended soil particles in the marshes.

76

References

Barber, A. 2014. Sediment and vegetation monitoring during a levee removal project on the Stillaguamish River Delta at Port Susan Bay, WA. Master’s thesis. Western Washington University. Bellingham, WA 98225. 102 pp. WWU Masters Thesis Collection. Paper 379. Clough, J. S. and E. C. Larson. 2010. Application of SLAMM 6 to Port Susan Bay including remedial alternatives. Report to The Nature Conservancy. Warren Pinnacle Consulting, Inc. Warren, VT. 05674. 128 pp. Fuller, R. 2014a. Stillaguamish estuary monitoring report summary September 2013 – July 2014: Field monitoring of biophysical conditions. Report for The Nature Conservancy. Western Washington University. Bellingham, WA 98225. 38 pp. Fuller, R. 2014b. Draft Sediment accretion monitoring guidance for restoration projects. Prepared for The Nature Conservancy. Western Washington University. Bellingham, WA 98225. 5 pp. Fuller, R. 2015a. Stillaguamish estuary monitoring report summary July 2014 – May 2015: Field monitoring of biophysical conditions (Revised). Report for The Nature Conservancy. Western Washington University. Bellingham, WA 98225. 23 pp. Data summaries were revised from the original submitted report, due to a change in the way that low and middle marsh habitat categories were defined. Fuller, R. 2015b. 2015 photopoint monitoring at the Stillaguamish estuary. Report to The Nature Conservancy. Western Washington University. Bellingham, WA 98225. 19 pp. Fuller, R. 2015c. Erosion pin monitoring at Port Susan Bay: Summary report 2015. Report to The Nature Conservancy. Western Washington University. Bellingham, WA 98225. 7 pp. Fuller, R. 2015d. Sediment and Vegetation Associations. Report to The Nature Conservancy. Western Washington University. Bellingham, WA 98225. 5 pp. Fuller, R. 2015e. Juvenile Salmon Access to the Port Susan Bay Restoration Site. Memo to The Nature Conservancy. Western Washington University. Bellingham, WA 98225. 3 pp. Fuller, R. 2015f. Summer 2015 Intensive Vegetation Sampling. Memo to The Nature Conservancy. Western Washington University. Bellingham, WA 98225. 3 pp. Fuller, R. 2016. 2015 Extreme Summer Salinity in the Stillaguamish Estuary. Monitoring memo prepared for The Nature Conservancy of Washington. Western Washington University. Bellingham, WA. 98225. 19pp. Fuller, R. 2017. Stillaguamish estuary monitoring report: Summary of current conditions and trends. Report to The Nature Conservancy. Western Washington University. Bellingham, WA. 98225. 34 pp. Fuller, R. and S. Thomas. 2014. Systematic Qualitative Monitoring at the Stillaguamish Estuary. Report to The Nature Conservancy of Washington. Western Washington University. Bellingham, WA. 98225. 10 pp.

77

Fuller, R. and D. DeBono. 2015. Large woody debris distribution in the Stillaguamish estuary: aerial imagery analysis. Report to The Nature Conservancy. Western Washington University. Bellingham, WA. 98225. 5 pp. Fuller, R. and J. McArdle. 2014. Stillaguamish estuary habitat change: 1886-2013. Report to The Nature Conservancy. Western Washington University. Bellingham, WA. 98225. 19 pp. Fuller, R.N., K. Poppe, J. Rybczyk. 2015. Marsh dieback in Puget Sound: hungry insects, the 2015 “drought”, and implications for the future. Poster presented at Coastal and Estuarine Research Federation 2015 Conference. Portland, Oregon. Grossman, E.E., George, D.A., Lam, A. 2011. Shallow stratigraphy of the Skagit River Delta, Washington, USA derived from sediment cores. USGS Open File Report 2011-1194, 123 pp. Grossman, E. E. and C. A. Curran. 2015. Draft-Sediment transport processes in response to estuary and salmon recovery efforts and the SR530 (Oso) landslide, Port Susan Bay, Washington 2013-2015. U. S. Geological Survey, Pacific Coastal and Marine Science Center. Scientific Investigations Report Draft. 127 pp. Heatwole, D. 2006a. Habitat mapping and characterization in Port Susan Bay: Summary of 2004 and 2005 Monitoring. The Nature Conservancy of Washington. Seattle, WA. 98121. 22 pp. Heatwole, D. 2006b. Methodology for predicting post-restoration habitats at Port Susan Bay. The Nature Conservancy of Washington. Seattle, WA. 98121. 3 pp. Hood, W. G. 2011. Development of empirical models to predict channel geometry and vegetation distribution in the Stillaguamish delta tidal marshes. Report prepared for The Nature Conservancy. Skagit River System Cooperative. La Conner, WA. 20 pp. Hood, W. G. 2016. Tidal channel monitoring in 2015 for the Port Susan Bay restoration project. Report prepared for The Nature Conservancy. Skagit River System Cooperative. La Conner, Wa. 16 pp. Hutchinson, I. 1988. Salinity tolerance of plants of estuarine wetlands and associated uplands. Report to the Washington State Shorelands and Coastal Zone Management Program. Simon Fraser University. Burnaby, B.C., Canada. V5A 1S6. 81 pp. Jarchow, M.E., and B.J. Cook. 2009. Allelopathy as a mechanism for the invasion of Typha angustifolia. Plant Ecology. 204:1:113-124. Kirwan, M.L., A.B. Murray, and W.S. Boyd. 2008. Temporary vegetation disturbance as an explanation for permanent loss of tidal wetlands. Geophysical Research Letters. 35:L05403. doi:10.1029/2007GL032681. Lucas, C.H., J. Widdows, and L. Wall. 2003. Relating spatial and temporal variability in sediment chlorophyll a and carbohydrate distribution with erodibility of a tidal flat. Estuaries. 26:4A:885-893. National Research Council. 2012. Sea-level rise for the coasts of California, Oregon, and Washington: past, present, and future. National Academies Press, Washington, D.C. 217 pp.

78

Rybczyk, J., A. Barber, and K. Poppe. 2014. Task 1.a. Sediment Elevation Tables. Report to The Nature Conservancy. Western Washington University. Bellingham, WA. 98225. 17 pp. Rybczyk, J. and K. Poppe. 2015a. Task 6. Port Susan Bay: Sediment Elevation Tables. Report to The Nature Conservancy. Western Washington University. Bellingham, WA. 98225. 22 pp. Rybczyk, J. and K. Poppe. 2015b. Task 14. Port Susan Bay: Estimating Long Term Rates of Accretion. Report to The Nature Conservancy. Western Washington University. Bellingham, WA. 98225. 10 pp. Rybczyk, J. and K. Poppe. 2017. Task 1.b: Port Susan Bay: Sediment Elevation Tables. Report to The Nature Conservancy. Western Washington University. Bellingham, WA. 98225. 22 pp. Selleck, J.R. III. Characteristics of Large Woody Debris in the Hat Slough Distributary Channel of the Port Susan Bay Preserve: Results from the 2004-2006 Monitoring Data. TNC Internal Report. Seattle, WA. 36pp. Slater, G.L. 2003. Waterbird abundance and habitat use in estuarine and agricultural habitats of the Skagit and Stillaguamish River Deltas. Ecostudies Institute Final Report to The Nature Conservancy. Stellern, C. 2016. Emergent wetland plant biophysical characteristics associated with wave attenuation and sediment retention. Master’s thesis. Western Washington University. Bellingham, WA 98225. 154 pp. Wood, R. and J. Widdows. 2002. A model of sediment transport over an intertidal transect, comparing the influences of biological and physical factors. Limnology and Oceanography. 47:3:848-855. Woo, I., R. Fuller, M. N. Iglecia, K. L. Turner, J. Y. Takekawa. 2011. The Nature Conservancy: Port Susan Bay Restoration Monitoring Plan. Report to The Nature Conservancy. U. S. Geological Survey, Western Ecological Research Center. Vallejo, CA. 94592. 115 pp. Woo, I., S. De La Cruz, K. Lovett, M. Davis, and A. Smith. 2014. Port Susan Bay Monitoring Brief: June 2014. Report to The Nature Conservancy. U. S. Geological Survey, Western Ecological Research Center. Vallejo, CA. 94592. 45 pp. Woo, I., S. De La Cruz, M. Davis. 2015a. Data summary report: Restoration Monitoring at Port Susan Bay and the Stillaguamish River delta. Report to The Nature Conservancy. U. S. Geological Survey, Western Ecological Research Center. Vallejo, CA. 94592. 19 pp. Woo, I., S. De La Cruz, and M. Davis. 2015b. Data summary report: Restoration monitoring at Port Susan Bay and the Stillaguamish River delta. Addendum: Benthic invertebrate community structure and biomass. Report to The Nature Conservancy. U. S. Geological Survey, Western Ecological Research Center. Vallejo, CA. 94592. 16 pp. Yang, Z., K.L. Sobocinski, A. Borde, T. Khangaonkar, and R. Thom. 2006. Hydrodynamic and ecological assessment for Port Susan Bay Restoration Project. Prepared for The Nature Conservancy. PNWD-3674. Battelle, Pacific Northwest Division. Richland, WA. 99352. 86 pp.

A1-1

Appendix 1 Lessons Learned from Monitoring the Port Susan Bay Restoration Project June 2018

Written by Roger N. Fuller Submitted to The Nature Conservancy of Washington

1. Restoration Design Elements

a. Pre-Restoration conditions that affect post-restoration habitat development Many pre-restoration conditions were taken into account during the project design phase, but there were a few conditions that we did not consider that, in retrospect, have likely had a substantial impact on habitat development. There are treatment options we could have considered that may have moderated their effect on habitat outcomes. The biggest sources of surprise were site characteristics that were likely the outcome of the site’s recent land-use history as an agricultural field. First, the slope of the site was much gentler than the slope of reference marshes, and this likely impacted marsh plain drainage post-restoration. Secondly, beneath the top layer of soft soil, 30-100cm deep, is a hardpan layer that appears to underlay most of the site. This layer affects channel development and marsh plain drainage. Thirdly, the dike footprint is highly compacted soil that is not allowing blind tidal channels to erode across it, and is inhibiting the downcutting of the mouths of the two major drainage channels. The effect of this compacted rim around the site is no development of increased channel connectivity with the exterior marshes, and prolonged site drainage times. Finally, the long rows that were formed the last time the site was plowed for row crops still persist and influence vegetation patterns. More details on each of these four characteristics are below. Site Slope The slope of the restoration site is much more gentle than any reference marsh, as described in the monitoring report. Tidal wetland slopes are typically very gentle to begin with, and without measuring the slope of the restoration site, we did not realize the difference, or its importance. Slope affects drainage patterns and rates, and a very gentle slope appears to have contributed to a lack of complete drainage from the marsh plain during low tide for at least the first few years. This resulted in site inundation patterns that were probably more characteristic of lower elevations in the estuary, with impacts on the vegetation response. As described in the report, standing water was observed in the restoration marsh at the lowest tides, even though the water level in the associated tidal channel was up to a meter or more lower than the marsh surface. Drainage appears to have been inhibited by a combination of gentle slope, lack of channels, and very dense vegetation that provided enough friction to maintain 10-15cm of water on site. Lack of marsh surface drainage appears to have affected soil qualities such as soil saturation and penetrability. Continuous soil saturation also affects things like oxygen exchange, redox potential, iron chemistry, and sulfide levels, and these kinds of changes can be stressful or even lethal to plants. These soil characteristics were not part of the monitoring program, so unfortunately we don’t know how drainage and soil characteristics affected the vegetation. We do know that substantially larger areas of marsh on the site have disappeared than was expected, and we think this may be partly related to drainage issues. Another surprise is that Schoenoplectus pungens, 3-square bulrush, has not colonized

A1-2

the site, despite seemingly excellent conditions for elevation, soil particle size, and pore water salinity. It was projected to become the dominant species on the site. We hypothesize that soil conditions such as oxygen exchange have prevented its establishment. At the lower elevations of some reference tidal marshes, S. pungens does particularly well on sandy soils, which drain well during low tide. Problems associated with site drainage could have been addressed during construction by some level of re-contouring of the site, construction of a small network of blind tidal channels, sediment nourishment to raise elevations or coarsen the soil slightly with the addition of sand, or other potential treatments. Treatments could have been done experimentally to test the effectiveness of treatments on post-restoration drainage, soils, and plants, to inform future projects. In retrospect, some level of tidal channel construction would have been good, not just for habitat values, but to insure adequate site drainage at low tide. At the time of design, the rationale was that “nature knows best”, and natural hydraulics would do a better job of developing a complex channel network suited to the site than humans could design and build based on our limited understanding at the time. We did not wish to err in the channel design and mistakenly lock the site into an unnatural and less than ideal channel configuration. However, because the soils and topography were far from natural to begin with, we should not have expected a channel system comparable to reference marshes to develop. The site failed to drain adequately for at least the first three years, affecting soil properties and likely the colonization and productivity of plants. Developing channels are probably too shallow and continue to fall short in terms of drainage as a result of the underlying hardpan that was not broken up prior to restoration. Some level of channel or marsh plain treatment would likely have eased post-restoration drainage problems and accelerated functional site development. Underlying hardpan

The subsoil of the site appears to have a layer of hardpan about 0.5+ m below the surface. This may be a natural feature, but is more likely a legacy of compaction resulting from decades of plowing. While the surface soil is regularly turned and broken up, the soil below the depth of plowing compacts over time from the regular traffic of heavy farm machinery. Now, as the new tidal channel network develops on the site, it appears to cut down to, but not beyond this layer. Near channel junctions with the primary drainage channels small waterfalls form as a result of the retarded pace of downcutting through this hardpan. What this means in terms of channel development (particularly channel depth and volume capacity), low tide wetted area, and marsh drainage is unknown. However it is likely that this affected the marsh drainage issue mentioned above. Furthermore, the channel bottom in the two breaches appears to be resisting downcutting, and the south breach in particular appears to have a lip that slows flow at the lower end of the tidal cycle. These channel characteristics may slow the rate of drainage during the ebb and maintain higher channel water levels than would otherwise occur. During construction we could have considered incorporating subsoiling to break up compaction, or a similar treatment, and we could have tested this at different scales experimentally to inform future projects. For example, we could have tested it as a way to facilitate channel formation without full channel excavation, or we could have tested a broader treatment at the scale of a marsh island. Compacted dike footprint We chose full removal of the levees down to the grade of marsh external to the site because it was clear from our restoration assessment that full hydrological connectivity across the marsh plain was key to accomplishing our ecosystem-scale objectives of improved distribution of freshwater and sediment.

A1-3

With full dike removal, we expected blind tidal channels to erode across the old dike footprint and increase connectivity over time between the restoration site and adjacent marshes and channels. This has not happened because the dike footprint is densely compacted. The dike was removed down to the adjacent marsh elevation, however, due to the accretion that occurred on the outside of the dike after the dike was built, dike removal down to marsh grade did not fully remove all of the imported dike material. Subsequent erosion has exposed a great deal of rock and buried lumber used in the original construction. As a result of this hard packed remnant dike material, and subsoil compaction from dike construction and decades of driving on the dike, new channels have not eroded across the dike footprint. The site continues to be hydrologically connected only through the two original breach sites, which are too few and undersized for the size of the site. There is over-marsh flow only during higher tide levels, insufficient hydraulic energy to carve channels. With better knowledge, we could have excavated more of the dike material, cut crossings across the dike footprint, or treated it in some other fashion to facilitate natural erosion of channels. Row crops Another agricultural legacy is that the new marsh plants are still, after five years of tidal action and at least 16 years since the site was last plowed, arranged in the long rows reminiscent of agricultural mono-cropping (Figure A1-1). The long field rows were not very evident before restoration, as the field had not been plowed since TNC purchased the site in 2001, and supported dense, tall pasture grasses and herbs. Had the rows been noticed, they would have been ignored under the assumption that a couple of tidal cycles would reshape them. However, perhaps due to high silt and clay content and many years of compaction from rain, the rows still persist. They became quite obvious in the aerials after restoration because wetland plants established on the raised ridges while the furrows may have remained too saturated to encourage healthy root systems. The furrows may also have affected the development of channel systems, possibly reducing their extension or rate of formation into the upper areas of the restoration site. However, another perspective may be that the ridges allowed more vegetation to develop than otherwise might have, by providing a place for plants to escape some impacts of the poor drainage. An un-furrowed surface might have been uniformly poorly drained and therefore uniformly bare. The actual role of the furrows is unknown since it was not monitored. This question could be addressed experimentally on the site even now, to determine whether the furrows facilitate increased plant productivity. Perhaps furrows or ridges of some other low design is a feasible treatment to encourage greater wetland productivity when restoring flat sites where drainage is expected to be a potential problem.

b. Number and size of breaches Another design element that also relates to marsh drainage issues, the compacted dike footprint, and the compacted soil sublayer is the number and size of breaches in the design. The very compacted old dike footprint has inhibited natural enlargement of the existing breaches and has prevented the formation of new channel crossings that we expected to connect with blind tidal channels adjacent to the site. As a result, no adjacent blind tidal channels have connected across the old dike footprint, neither of the two breaches have expanded much in width, and the limited downcutting of the breach channel mouths has limited the expected expansion of the primary channels on the site. The project design incorporated too few breaches, and the two breaches are too small in size for the site. This is likely the cause of the observation that the ebb tidal flows take far longer than the flood flows. Site drainage is slowed due to the insufficient connectivity with external channel systems.

A1-4

Figure A1-1. The legacy of farming and site construction are still visible. Black arrows point to the long, straight furrows left from the last time the site was plowed more than 16 years ago, with bulrush plants growing on the long ridges between furrows. White dotted lines parallel the pathways of heavy construction traffic during dike removal. Compacted truck routes, despite tilling to break them up after construction, have turned into long, straight channels that reduce the complexity of the developing tidal channel network. IR imagery from 2016.

c. Bolboschoenus fluviatilis (river bulrush) holds potential as an agent of tidal marsh resilience.

B. fluviatilis is a robust, highly productive species apparently common, though unrecognized because of its similarity to B. maritimus (sea-coast bulrush), in Puget Sound estuaries. Based on our post-restoration observations at Port Susan Bay, it appears to resist grazing by geese, reduce wave energy and erosion, increase sediment capture, and may be more tolerant of higher pore water salinity than B. maritimus or Schoenoplectus pungens (3-square bulrush). It appears to have a narrower ecological niche in estuaries in terms of inundation tolerance (elevation) and soil particle size, but deserves further study to evaluate its ecological distribution and potential as a design target in restoration projects. In the context of snow geese and climate change, it may help with wetland resilience.

2. Design and Construction Process

a. Documentation of design assumptions and impact pathways Connectivity across the restoration site, between Hatt Slough and the northwestern marshes, is much less than predicted by the preliminary design hydrodynamic model. Regular flow at high tide from Hatt Slough across the restoration site does not occur. Several factors may have contributed to this, including insufficient resolution in the elevation data for the narrow channel bank between the old dike and the Hatt Slough channel, lack of specificity in the assumptions underlying the project’s objectives, and an almost complete change in project managers, science support, and contractors between the initial and final design phases. Over-marsh flooding from Hatt Slough at high tide was an assumed outcome based on the model, but it was not an explicit design criteria for the final design. As a result, design elevation surveys along Hatt Slough were not cross-checked with the model’s elevation map or with tidal

A1-5

inundation assumptions. Project objectives are usually functional in nature, as they were in this case: new habitat for valued species and improved ecological processes at site and system scale. However restoration actions alter the physical structure of a site, with the assumption that physical alterations will result in the desired functional outcomes. In retrospect, the project objectives were good except they lacked a sufficiently specific description of the impact pathway between the physical restoration action and the desired functional outcome. It was assumed, based on the modeling, that dike removal would allow high tide flow from the river across the site. As a result, the objective of improved connectivity was based on the assumption that dike removal was a sufficient physical change to achieve connectivity, and therefore the final design criteria was full removal of the dike. However, one link in the impact pathway was missing, which is that connectivity required the berm elevation between Hatt Slough and the restoration site to be low enough to allow over-marsh flow at higher tide levels. Dike removal was assumed to be sufficient, but the underlying physical impact pathway was not described in enough detail to elicit the right criteria for the final design process. Adaptive management can still address this deficiency in outcomes.

b. Effects of construction traffic pathways on future channel development During the removal of the old dike and construction of the setback dike, heavy truck traffic crossed the marsh plain of the restoration site. Initial construction design called for truck traffic to remain on the footprint of the old dike, but sharply increased costs due to a peak in oil prices at the time of construction led to a compromise for cross-marsh traffic pathways. The heavy traffic compacted the soil and, despite post-construction tilling, those pathways became straight, wide channels, reducing the complexity of subsequent channel development (Figure A1-1). In retrospect, traffic pathways could likely have been designed with some gentle curves in them to at least moderately improve the sinuosity of subsequent channel development. They could also have been positioned with more forethought for where future channels would be most desirable.

3. Monitoring

a. Tidal wetland hydrology As has been described, drainage on the restoration marsh plain surface may have had a substantial effect on subsequent marsh development. There has been substantially more marsh retreat than expected, and the species projected to be dominant (Schoenoplectus pungens) is only present in trace amounts, despite apparently favorable soil particle size, salinity, and elevation. Marsh plain drainage was not something directly monitored in this project because it was assumed that a hydro logger in the main site drainage channel would be sufficient to reflect marsh plain drainage. Based on the logger which registered channel water levels up to a meter or more lower than marsh plain elevations, and LiDAR site elevations which mapped the difference in elevation between marsh surface and channels, that the marsh surface was draining completely during each tidal cycle. However, it was observed during sampling that many marsh plots retained a shallow 2-15cm layer of ponded water on the surface throughout the tidal cycle. Even plots whose surface did drain completely had soils that were far more

A1-6

saturated than reference marshes, as indicated by groundwater levels in shallow pits dug for soil salinity measurement. The lesson learned is that indirect measures of marsh plain hydrology, such as hydro loggers in channels and LiDAR elevations, are not sufficient to characterize hydrology or soil saturation on the marsh plain. We did not directly measure soil hydrology and so we have no direct evidence that soil drainage, and resulting impacts on soil physical and chemical conditions, were a cause of vegetation responses. The assumption that a channel water level logger accurately quantifies marsh plain hydrology may be a common error in the restoration community, and has implications for understanding the potential condition and trajectory of wetland development on a restoration site. Soil drainage has significant effects on soil conditions that impact plant growth and survival.

b. Soil physical and chemical conditions Soil saturation has physical and chemical effects on soil conditions, including compaction, oxygen exchange, and redox potential, all of which can have substantial effects on plants. Redox has implications for biogeochemical cycles of N, S, Fe, and other elements, which can in turn affect nutrient availability and whether elements occur in toxic or bioavailable forms. We measured soil penetrability which is a measure of soil firmness or compaction, though we intended it as an indicator associated with relative ease of goose grazing. However the project monitoring plan did not include other soil physical or chemistry characteristics related to soil saturation, so we lack data to document whether poor drainage yielded poor soil qualities that contributed to the large reduction in plant cover. Similarly, we cannot link soil conditions to the failure of Schoenoplectus pungens to colonize the site, though that is our hypothesis. In retrospect, we should have monitored at least some indicators of soil saturation at low tide, redox potential, oxygen/gas exchange, sulfide or nutrient levels, or other potentially informative correlates with site drainage.

c. Importance of system-scale perspective and multiple reference sites This project began with a system-scale perspective at the core of its objectives and this perspective was key to the amount we’ve learned about ecosystem restoration in general and estuary resilience to anthropogenic impacts specifically. In estuaries, where key drivers of habitat development are controlled by system-scale processes, it is critical to select sites and approach project design from a system perspective. Implicit in a system perspective is the use of a range of reference areas that encompass the full spectrum of the environmental gradients that drive habitat development on restoration sites. The original reference site for this project (zone 1), turns out to be very different from the restoration site in several key physical and biological ways and if we relied only on zone 1 monitoring, we would still be befuddled about why habitat is developing as it is on the restoration site. Having multiple, very different reference sites has provided a far clearer understanding of both how the site will develop and in what ways the site will be vulnerable to climate change impacts and other stresses in the future.

d. Importance of seasonal data on vegetation structure An important aspect of marsh vegetation structure, and one of the least commonly collected data, is seasonal vegetation structure such as stem density and height, including during the geomorphically active winter season. Sediment is both delivered by winter river floods and re-distributed by winter storm waves. Plant species composition and winter physical structure matters to accretion rates, wave attenuation, herbivory resistance, and other factors. Although

A1-7

our geomorphically important marsh species are perennials and are senescent in the winter, they never the less differ substantially in terms of winter geomorphic roles as a result of the variability in the resistance of their biomass to degradation during the winter. Although only four species of bulrush dominate the brackish marshes and broadly overlap ecologically, different marsh areas can vary tremendously in their winter structure and their resulting geomorphic and functional roles, as a result of the relative composition of bulrush species. These differences will affect ecological functions, timelines of response to restoration, and climate change response scenarios.

e. Importance of long term reference data sets There are several key questions about how the site is performing and how it affects system-scale processes that we can’t answer due to the lack of sufficient pre-project data. Puget Sound needs a network of estuary reference sites that can develop long term data sets to inform restoration projects throughout the region. In particular, hydrology and sediment dynamics are highly variable from year to year, and as a result marsh productivity, species composition, and overall functions can vary a lot as well. We need data spanning inter-annual climate extremes so that we can better understand how wetlands may respond to climate change, and what is needed to improve their adaptability. Climate change impacts don’t act alone but interact with other existing stresses of which we aren’t always aware, such as stem-boring moth larvae. If we don’t understand these interactions we face some unhappy surprises in the future, and more rapid wetland changes than we expect. For example, an estuary climate change impact analysis that only examines sea level rise may substantially underestimate change rates and impact pathways.

f. Non-native invasives monitoring and response plan Typha angustifolia, narrow-leaved cattail, was not a widely recognized threat to restoration sites at the time of project design. But even so, the pre-project vegetation model did not predict cattail colonization of the site. While there was a program to monitor and eliminate Spartina, as required by state law, the state does not legally require control of T. angustifolia despite its potential for large scale invasion and ecological impact. T. angustifolia has the potential to occupy much of the restoration area if allowed. It has a considerably broader ecological amplitude in terms of inundation levels and salinity tolerance than the native broad-leaved cattail (T. latifolia), and it appears to have allelopathic capabilities that have been shown to significantly reduce bulrushes specifically. In retrospect, even though the predictive vegetation model suggested that it wouldn’t be a problem, there should have been a monitoring plan and decision-framework for response in place.

4. Species, Habitat, and Ecosystem Responses

a. Some vegetation responses were a surprise. Plants respond to many aspects of their environmental conditions. Our predictive models relied on the best available published science and local experience to develop predictions of habitat response. The best knowledge at the time was largely related to elevation, which of course is only one, albeit an important factor. Schoenoplectus pungens has not colonized significantly, but was predicted to be one of the most common species. Our hypothesis is that soil saturation alters oxygen exchange, soil physics, or chemistry in a way that inhibits S. pungens. We did not

A1-8

expect extensive Typha colonization, but the two species, particularly the non-native invasive T. angustifolia, are beginning to play significant roles. Better information about the ecological requirements of the dominant estuarine plant species would have allowed us to better project habitat development, and perhaps to incorporate restoration treatments that would have improved habitat functional outcomes.

b. Bolboschoenus fluviatilis, river bulrush, is a common and important estuarine species. We discovered that the meadows of B. maritimus were in fact a mixture of B. maritimus and B. fluviatilis, and in some areas B. fluviatilis forms an almost monotypic band near the MHHW line. The two species are extremely similar except under ideal conditions for B. fluviatilis when it is substantially larger. B. fluviatilis is entirely missing from common wetland plant field guides which is likely the reason it appears to have been missed by coastal biologists. We haven’t found a published or herbarium record of B. fluviatilis from estuaries, and have not found any estuarine worker in Washington or Oregon who is familiar with B. fluviatilis. However, after finding it in the Stillaguamish estuary we have since found it in other estuaries as well. B. fluviatilis is worthy of further study because it may offer important services such as higher rates of accretion and greater biomass production than other bulrushes. Its biomass appears to be more resistant to breakdown during the winter and for this reason it also appears to resist goose grazing, as well as offer greater friction to winter waves, thus protecting soils, shorelines, and nearby infrastructure from erosion and overtopping. Finally, in 2015 it appeared to be less impacted by the extreme soil salinities than other bulrush species. In the context of climate change, all these characteristics may suggest it can play a role in marsh resilience. With sufficient knowledge of its environmental requirements, restoration treatments could be implemented that would favor its development.

c. Ephemeral log debris fields may play a key role in marsh expansion. Aside from the restoration project, the only substantial increase in marsh area since 1990 in the estuary occurred in zone 6 where a 2007 flood deposited a large, 15 acre debris field of loosely aggregated logs. Most of the logs floated off within one or two years, but their presence was enough to trigger channel migration, stabilization, rapid accretion, and marsh initiation. In 2003 a similar, ephemeral log field was left by a flood, though it did not trigger marsh expansion. Log jams and smaller log debris fields are more frequent but don’t appear to cause large scale marsh formation, though they may provide fish habitat when inundated. Large, ephemeral log debris fields appear to be infrequent, associated with large floods, and may suggest an unappreciated role for large wood in the estuary. Since they are ephemeral, their role in initiating marsh may not be obvious.

d. Interaction of stresses is important, and may be key to understanding climate change impacts. The site experienced a 50 acre marsh dieback in 2015, with a stem-boring moth larvae being the direct cause. Although we have little data other than soil salinity, the marsh may have been made more susceptible to attack by the interaction of multiple stressors that weakened the plants’ natural defense abilities. As described in more detail in the report, high soil saturation levels may have been stressing the plants since the restoration, and the added stress of salinity may have helped make the plants vulnerable. The moths were found in some of the reference marshes as well, but never at the extremely high densities found in the restoration zone. We contacted estuarine experts throughout Washington, Oregon and California and could find no

A1-9

other record of a large marsh dieback associated with insects. Joy Zedler in California did provide a reference with documentation of a moth herbivore in tidal marsh, though it did not result in a large dieback event.

e. Climate change interacts with existing stresses, and can alter food web relationships. The greatest impacts from climate change will likely come as a result of the multiple ways that climate change interacts with existing sources of stress and disturbance, and the differential ways in which it alters relationships among species. Many of these interactions and potential triggers of change remain unknown and unrecognized by us. The insect-mediated dieback event is instructive because it was caused by a native species whose presence and role as a natural source of stress to plants was unknown. The site, as most sites do, lacks an invertebrate inventory, let alone an inventory of all species that feed on or have a symbiotic relationship with the dominant plant species. The moth had never been noticed at the site, and never observed to cause diebacks at any site. The surprising and rapid habitat change it triggered was likely because the unusually low river levels and high soil salinity combined with other stresses to push the overall marsh stress level over a tolerance threshold. Being aware of these kinds of interactions can help us to imagine and potentially prepare for climate-related surprises in the future. Climate change affects estuaries in several different ways, amplifies existing stresses such as pore water salinity, increases rates of disturbance such as winter wave erosion and snow goose population sizes, and increases the frequency and co-occurrence of various stressors such as drought, extreme temperatures, extreme precipitation events, and insect populations. Surprising, rapid, and large-scale habitat changes can occur when the cumulative stress level passes a threshold. The proximal cause of a change may appear to be unrelated to climate change, though it may be the additive effect of climate that pushes the system past its tolerance level. Climate change vulnerability assessments that only look at the effect of a single impact pathway, such as the gradually rising sea level, are likely to underestimate the vulnerability from the cumulative effects of multiple, interacting stressors.

f. The most important, direct, individual effect of climate change is likely to be declining summer river flows. Reduced summer flows will affect soil salinity, marsh productivity, and the ability of marsh to resist and be resilient to other sources of stress. Stillaguamish summer river flows have been declining for years and are projected to continue to drop in coming decades as our winter precipitation shifts from snow to rain dominance. Lower summer river flows will likely impact the tidal marshes much earlier and to a greater extent than sea level rise, though eventually sea level rise will become a more important stressor. One impact of sea level rise will be to amplify the salinity stress initiated by the declining summer river level. Restoration and adaptive management actions that enhance freshwater residence time and distribution in the estuary should be management priorities. In addition, actions upriver that moderate declines in summer flows should be a priority, including restoring hydrologically mature forests, and floodplain connectivity that increases water storage capacity in hyporheic zones. Areas of the tidal marsh that are farther from primary sources of freshwater like Hatt Slough are likely to be more vulnerable to changes in flow. Monitoring should pay particular attention to those areas, including the various sources of stress that may interact with rising soil salinity to affect plants.

A2-1

Appendix 2 Port Susan Bay Restoration Project - Image Archive Eighteen photopoints were initially set up by The Nature Conservancy around the restoration site to photo-document changes before and after restoration. Pre-construction photos were obtained by TNC. In 2014, 2015, and 2017, TNC contracted WWU to resample the photopoints. WWU added 10 additional photopoints and shifted the position of some slightly, in order to better track the appearance and development of key structural elements on the restoration site. Photopoint locations are shown in Figure A2-1. WWU developed a protocol for information-rich photopoint monitoring called Systematic Qualitative Monitoring (Fuller 2014), combining photos with a rich narrative describing features that might not show up well in a two dimensional photo. Written observations at each site were recorded for four categories of ecologically important features: land, water, vegetation, and woody debris. Key characteristics of those features such as size and distance from the photopoint, and evidence of associated ecological processes, were recorded initially, and in subsequent years any changes were noted. In general, photos were taken in a complete circle around the compass points. Photo nomenclature: Photos are named following this format: “photopoint number_compass direction_date of photo.jpg”, with the date in year-month-day order to facilitate sorting of files. For example: “R1_E_20150405.jpg” is from photopoint R1 (Restoration 1) and was taken looking towards the east on April 5, 2015. Reference: Fuller, R. 2014. Information-rich photopoint monitoring: Systematic qualitative monitoring methods. Western Washington University. Bellingham, WA. 98225. 6 pp.

A2-2

Figure A2-1. Locations of photopoint monitoring stations around the restoration site.

A3-1

Appendix 3 Port Susan Bay Restoration Project - Description of final archival database A final data set for archival has been submitted in a single Excel file called “PSB Restoration Data Archive”. Multiple contractors collected various types of data over the course of the project. Each type of data is provided in separate worksheets, and each has its own data dictionary. For each data type, the first worksheet is a data dictionary that defines the column headings and describes the content for each data column when the data is not self-explanatory. Subsequent worksheets contain the data. Hydrologic data collected from continuous data loggers by USGS WERC and WWU has been submitted for archival to the Washington Department of Ecology’s Environmental Information Management (EIM) database. Due to the file size for data from continuous loggers, it is not included here, but can be retrieved from the WDOE’s archival database (http://www.ecy.wa.gov/eim/). We do not have copies of all raw data collected for this project, including from the Skagit River System Cooperative and from the United States Geological Survey (Pacific Coastal and Marine Science Center). Data summaries from those two contractors are available in their reports submitted to The Nature Conservancy. Worksheet nomenclature is as follows: Data dictionaries: “Data type_DD_Contractor name” Data: “Data type_Contractor name”

List of worksheets in order: Worksheet titles are in italics, data dictionaries in bold. Instructions: Contains the overall archive instructions as provided here. Vegetation_DD_WWU: data dictionary for vegetation data collected by Western Washington University. Vegetation_2017 Summer_WWU Vegetation_2015 Summer_WWU Vegetation_2015 Summer Sub_WWU Vegetation_2014 Summer_WWU Vegetation_2013 Winter_WWU Sediment_DD_WWU: data dictionary for sediment data collected by Western Washington University Sediment_WWU Elevation change_DD_WWU: data dictionary for elevation change (Surface Elevation Tables) collected

by Western Washington University. Elevation Change_WWU Sediment_DD_USGS WERC: data dictionary for sediment data collected by United States Geological

Survey, Western Ecological Research Center. Sediment_USGS WERC Sediment Accretion_USGS WERC Birds_DD_USGS WERC: data dictionary for bird data collected by USGS WERC Birds_USGS WERC Invertebrates_DD_USGS WERC: data dictionary for benthic invertebrate data Invertebrates_USGS WERC Fish_DD_ST: data dictionary for fish data collected by the Stillaguamish Tribe Fish_ST

A3-2

Monitoring contractors and contact information: Western Washington University: vegetation, sediment, elevation change, hydrology, photopoints Roger Fuller, John Rybczyk, Katrina Poppe

roger.fuller@wwu.edu 360-333-7490 Western Washington University Huxley College of the Environment, MS 9181 516 High Street Bellingham, WA 98225

USGS Pacific Coastal and Marine Science Center: sediment transport processes (raw data not included in the data archive) Eric Grossman

egrossman@usgs.gov 206–526-6282 x334 USGS Western Fisheries Research Center 6505 NE 65th Street Seattle, WA 98115

USGS Western Ecological Research Center: sediment, accretion, hydrology, birds, benthic invertebrates Isa Woo, Susan De La Cruz, Melanie Davis

iwoo@usgs.gov 707-562-2001 USGS Western Ecological Research Center San Francisco Bay Estuary Field Station 505 Azuar Drive Vallejo, CA 94592

Stillaguamish Tribe of Indians: fish Jennifer Sevigny

jsevigny@stillaguamish.com 360-631-2372 Stillaguamish Tribe of Indians P.O. Box 277 Arlington, WA 98223

Skagit River System Cooperative: tidal channel development (raw data not included in the data archive) W. Gregory Hood

ghood@skagitcoop.org 360–466-7282 Skagit River System Cooperative P.O. Box 368 La Conner, WA 98257–0368

A4-1

Appendix 4 WWU Field Methods Vegetation Methods

Equipment List

Field data sheets GPS Pencils, Sharpie pens Camera 30m or similar tape measure Refractometer (for field salinity measurement) Penetrometer (a metal stake pointed on one end, weighing about 75g, 33cm long x 6.5mm diameter) PVC pole, 1.5 m long (penetrometer is dropped through this to keep it vertical) 1m2 and 0.25m2 PVC quadrat frames Timing of Sampling Estuarine vegetation in Puget Sound should be sampled when most plants are near their maximum size. This generally occurs around mid to late June, depending on the year. However, we’ve recently discovered that Bolboschoenus fluviatilis is common in some Puget Sound river deltas, and it matures later that B. maritimus, the species most commonly recognized. Recent experience suggests it may still be growing in late July. Schoenoplectus pungens has weak stems and can become completely covered by adhered sediment by the end of August, which appears to hasten senescence. It has weak stems and begins to break and flatten in response to tidal and wave action early. As a result we recommend sampling before mid-September if possible to be able to accurately capture the growing season’s vegetation structure. Pore water salinity measurements should be taken whenever field work is done, but are most informative when taken during the period of minimum river flow, usually between late July and early September. Pore water salinity measurements should be taken in years of differing hydrological conditions. For understanding potential climate change interactions, intensive vegetation and pore water salinity sampling should occur during hydrologically extreme years for high and low flows. This will allow a measurement of the effect of low flow years on biomass production which underpins the estuarine food web. Winter vegetation sampling is strongly encouraged, at least as a baseline, to capture the structure of vegetation that is present during the geomorphically active period when aboveground biomass will aid in sediment capture and retention. In the field:

Check GPS Settings: WAAS enabled, units: hddd.dddddo, UTM Zone 10 NAD83, metric units On the field sheet, record date, names of crew members, GPS id#, Camera id#, name of data recorder.

A4-2

Navigate to the center of the first plot and take a GPS waypoint (try for an accuracy of +/- 2m or less). Also from the center, take four photos at each plot, pointing approximately in cardinal directions in the order South, West, North, East. On the field sheet, record the plot number, GPS waypoint, and photo numbers. Plots are 20mx20m, centered on GPS point. Upon arrival at the plot, dig a 10-15 cm deep pore water salinity pit in an area representative of the dominant vegetation pattern in the plot. It often takes many minutes for pore water to accumulate in the pit. General plot data: Species: note which species occur under the following categories:

dom: dominant, the species with the most cover, generally covering >40% of the plot co-dom: co-dominant, sometimes there may be 2-4 species that all have +/- the same cover (there can’t be both a dom and a co-dom species) sub-dom: species that cover at least 20% of the plot, but not as much as the dominant species present: species that are present, but at sufficiently low cover that they are unlikely to affect the overall physical structure of the habitat compared to the other species. It isn’t necessary to capture every single, low-cover species that is present since the objective of this sampling is vegetation structure and dominant species, so don’t spend too much time searching for rare species.

Height: record the bin number (see below) for the average full height of plants. Don’t record the height of the tallest plant, just the height that appears to be the average max for the dominant species. Record the actual structural height of the vegetation biomass regardless of time of year. In other words, don’t record the total stem or leaf length, but the distance from the soil surface to the top of the vegetation, regardless of how bent over it is. During the fall/winter the plant biomass may be bent over or prostrate. If the vegetation is substantially bent over from wave or wind action, make a note to that effect.

Height bin numbers. Note that the body part notations are estimates and should not be used unless calibrated against the individual sampler’s body. An alternative is to use a meter stick or a PVC pole whose length has been marked in 10cm increments. 0 = 0 1 = <10cm (prostrate) 2 = 30cm (mid-calf) 3 = 50 cm (knee) 4 = 75 cm (thigh) 5 = 100 cm (waist) 6 = 125 cm (diaphragm) 7 = >135cm (chest – make a note if it is head height or higher)

Stem Density: For larger species in the low or middle marsh habitats, including bulrushes and cattails (Scam, Boma, Bofl, Scta, Tyla, Tyan), estimate the average density of stems and assign to one of the following bins. At the beginning of the field season, this should be done only after first measuring actual

A4-3

stem densities in the sub-plots as described below, in order to train the eye to be able to accurately estimate stem density at the full plot scale.

Density bins: 0 = 0 VL = <1 L = 1-25 M = 26-75 H = 75+

% Vegetation Cover: Estimate total % plant cover by recording the appropriate bin number below (e.g. if it was 75%, then you would be able to see 25% bare sediment when looking down from head level).

% Cover Bin Numbers 0 = <5% cover 1 = 25ish (i.e. 10 to ~35) 2 = 50ish (i.e. ~35-60) 3 = 75ish (~60-90+) 4 = 100ish

% LWD Cover – Estimate and record the % of the plot that is covered by large woody debris. % Plot Bare – Record the % of the plot that is unvegetated. NOTE that this parameter is NOT the inverse of % Vegetation Cover, but is the area of plot that is barren of any vegetation. It is intended to capture patchiness of the marsh and lower elevation areas where bulrush clones have not coalesced into a solid meadow. This parameter is measured where marsh vegetation does not uniformly cover the surface of the plot. For example, a plot that is uniformly vegetated may have total vegetation cover of 75% but have no “% Plot Bare” recorded. However if there are bare patches of at least 1m2 in size scattered in the plot, the total area of barren patches should be estimated. Soil Penetrability: Collect 5 penetrometer readings in 5 random places in the plot. If there is a mixture of substrate types (e.g. firm marsh surface and soft-bottom eroded rills or pannes), take 5 readings in each substrate type. Hold the 1.5m tall PVC pole perfectly vertical and drop the penetrometer through it to keep it vertical. Record how deeply the penetrometer penetrates the soil. Be sure the bottom of the PVC pole is resting on sediment, not on vegetation. Clean off penetrometer after each use to keep the sediment from making it stick to the inside of the PVC pole. Salinity: Following the field method document for use of the field refractometer, measure the salinity of the pore water from the pore water pit. Also measure the salinity of nearby surface water and make a note if the surface water is in an isolated puddle or in a drainage feature connected to a channel. Notes: make a note of anything unique to the plot, such as the presence of large wood or channels in the plot, or large parts of the plot that are different than other parts (e.g. if 25% of plot is bare, or if there are signs of recent disturbance, etc.) Sub-plot data Measure vegetation structure in 5 sub-plots (a-e), chosen randomly by tossing a flag or something blindly over the shoulder. However, if the plot is not uniform in vegetation coverage, stratify the

A4-4

samples to ensure that the samples are distributed in proportion to the relative distribution of vegetation density. Place at least one sub-plot within each of the four quarters of the plot. Height (1m2 subplot): record the bin number (see below) for the average full height of plants. Don’t record the height of the tallest plant, just the height that appears to be the average max for the dominant species. Record the actual structural height of the vegetation biomass regardless of time of year.

Height bin numbers. Note that the body part notations are estimates and should not be used unless calibrated against the individual sampler’s body. An alternative is to use a meter stick or a PVC pole whose length has been marked in 10cm increments.

0 = 0 1 = <10cm (prostrate) 2 = 30cm (mid-calf) 3 = 50 cm (knee) 4 = 75 cm (thigh) 5 = 100 cm (waist) 6 = 125 cm (diaphragm) 7 = >135cm (chest – make a note if it is head height or higher)

Stem Density (50cm2 subplot = 1/4m2): Used for larger species including bulrushes and cattails (Scam, Boma, Bofl, Scta, Tyla, Tyan) that typically dominate the middle and low marsh habitats. Record the density of stems for the species identified as either dominant and sub-dominant at the plot scale. Count density in ¼ of the 1m2 subplot. Depending on funding (time), either count actual stem density or estimate densities and assign to one of the following bins.

Density bins: 0 = 0 1 = 1-25 2 = 26-50 3 = 51-100 4 = 101-150 5 = 151+ (if more than 200, note approx. actual density)

For high marsh habitats and where shorter or lax species (grasses, Carex, Juncus, etc.) dominate, record the following data:

Density (10cm2 subplot): Record the number of stems of Carex, Juncus, Triglochin, etc. in a small 10cm2 subplot approximately in the center of the 50cm2 plot. Record the species next to the stem count. Do not count grass stems (e.g. Agal, Disp, etc.) Mat: (50cm2 subplot): For lax species such as grasses that provide a dense, tangled matrix of thin stems, mostly in height bin 1 or 2. Record the height of the mat, note if it is loose (easy to see the ground underneath) or dense (can’t see ground surface), and note if it covers less than ½ of the 50cm2 subplot.

A4-5

Pore Water Salinity: Field refractometer method

Equipment Needed:

Refractometer Plastic syringe (without needles) Plastic syringe filter holder (25mm) 25mm glass microfiber circular filters (e.g. Whatman GF/C, cat no. 1822-025) Optional: coffee filters or large glass microfiber filters cut to size, potato ricer or garlic press (or filter holder, or other mechanism for compressing soil) Sampling pore water salinity:

Dig a shallow pit and allow pore water to seep into the pit from the soil. Don’t allow water on the surface to drain into the pit. The pit should be in the root zone of plants, preferably 10-30cm deep.

The pore water will likely contain suspended sediment which should be filtered out to minimize interference with the refractometer. To do this:

While the pore water is collecting in the pit, separate the two halves of the syringe filter holder, being careful not to lose the rubber O-ring. Place a 25mm circular glass microfiber filter into the filter holder and screw the filter holder together. (not recommended, but a cheaper alternative to the syringe filter holder and Whatman circular filters would be to use Whatman #2 filter paper cut to size, or coffee filters cut to appropriate size, with the filters placed within the syringe or in a potato ricer or garlic press, used to compress the soil and expel the pore water. The key is to pick a method and use the same method for all samples so that results among sites are comparable).

To make sure the refractometer and equipment is clean and functioning, place a small amount of DI water into the syringe and expel it through the filter onto the refractometer, and take a salinity reading (should be 0 - record the “blank” reading at least once per day to document that the unit is correctly calibrated. Note that most refractometers can be calibrated in the lab with a calibration screwdriver, with the calibration reading usually taken at or close to 68 degrees F).

Collect the pore water from the pit using the syringe (without filter).

Attach the filter holder to the syringe.

Press the water through the filter and onto the refractometer’s prism. Press enough pore water to flush the DI water completely off the prism and daylight plate, so that the DI water is not diluting the sample.

Hold the refractometer horizontal while looking through it towards the light. Read the salinity.

Place more water on the refractometer and repeat until 3 consecutive measurements show the same reading. Record the value.

Remove the filter from the syringe holder, then rinse the refractometer, syringe, and filter holder with DI water.

If there is surface water near the pore water sampling location, sample it as well, whether in a channel, puddle, panne, etc. Note the type of surface water sampled, particularly whether it is

A4-6

flowing, regularly flushed (tidally or riverine), or potentially subject to significant evaporation between wettings (e.g. a panne in high marsh that is only flushed by spring tides or less often).

At the end of the sampling day, be sure to clean the refractometer and rinse off any dirt and salty residue.

If no pore water accumulates in the pit after a few minutes, try the following:

With sandier soils, place some moist sediment directly into the syringe, attach the filter, and

press the plunger. This may extract enough pore water for sampling.

For any soil, use a garlic press and coffee filter to compress the sediment and force it through a

coarse filter. If necessary, filter the resulting liquid through the normal syringe apparatus.

If no pore water can be extracted, collect a sample in a plastic bag for testing in a laboratory

equipped with a centrifuge or other pore water extraction methods.

Refractometer use. Salinity alters the way that light passes through the prism, commonly resulting in part of the view being blue and part white, with the dividing line indicating the level of salinity.

For instructions on use of a refractometer, see:

http://coastalroots.lsu.edu/Documents/Nursery%20Information/Seed%20Preparation%20&%20Nursery

%20Production/Reading%20a%20Salinity%20Refractometer.pdf

or: http://www.lumcon.edu/bayousideclassroom/CR_workshop_handout_2010.pdf )

A4-7

Sediment sample collection Equipment Needed Camera Refractometer (for field salinity measurement) Penetrometer (a metal stake pointed on one end, weighing about 75g, 33cm long x 6.5mm diameter) PVC pole, 1.5 m long (penetrometer is dropped through this to keep it vertical) Small ruler (15+ cm) Sample bags (ziplocks), quart, labeled. PVC corers

6cm diameter (outer) x 25cm length for main samples, with a 10cm depth line marked on the outside, and one end sharpened with a file. With a #11 rubber stopper for top to ease extraction.

6-inch wide PVC cap for collecting top 2cm sample for USGS Flat, stiff plate for covering the bottom of core while extracting (or just use hands) Trowel Water squirt bottle to rinse off sediment from inside core between samples Labels, interior and exterior. Interior printed on write-in-rain paper. Include ID number, lat/long, date. Data Collection 1. Complete field sheet including personnel, date, study zone. WWU samples are collected near the

WWU vegetation transect. At each plot: 2. Record dominant vegetation (species) 3. Take a perspective photo with the collection point in foreground, and the horizon visible, taken facing

down-transect (down the elevation gradient). Record photo number on fieldsheet. 4. Take close up photo of soil surface (~25cm height) with ruler or coin in photo. Record photo number

on fieldsheet. 5. Record surface conditions (e.g. benthic algae present, or other sign of polysaccharide biofilm ‘glue’

stabilizing surface), evidence of disturbance or erosion, or other general observations. 6. Cut 15cm+ profile with trowel and note if there is evidence of visually-distinct layers, measure layer

depths (e.g. a different top layer, generally finer than lower layers), note if evident grain size or other difference between layers. Take photo if there are layers. Record photo number.

7. Measure and record pore water salinity of water collecting in the soil profile pit, using refractometer, and following the directions in the Pore Water Salinity method document. Also measure salinity of nearby surface water and note whether surface water is in an isolated puddle or in a drainage feature connected to a channel. Due to evaporation, isolated puddles may have higher salinity.

8. Measure and record penetrability (5 replicates), following methods described in the general vegetation sampling protocol document.

9. Collect soil sample (3 replicates, about 0.5m apart, mixed into same bag). Volume should be about 1/3-1/2 of a quart ziplock bag. • Use a 6cm diameter PVC core (2.375 inches) • Remove rubber stopper when inserting PVC into sediment. Push it into soil until 10cm depth, then replace the rubber stopper. Use something stiff and flat to cover the end of the core while extracting it – use trowel to dig down a little until you can slip the flat under the core before extracting core, or just use a gloved hand to cover bottom.

A4-8

Label samples with ID #, lat/long, date, sampler initials

For the USGS shallow samples (2cm), use a PVC cap with a 2cm line drawn on the outside to indicate depth. Depress the cap into the sediment to the right depth, use a stiff flat object or a hand under it as you extract it.

10. Upon return to the lab, freeze samples until processing.