COMBINING EM AND LIDAR TO MAP COASTAL WETLANDS: AN EXAMPLE FROM MUSTANG ISLAND, TEXAS

Jeffrey G. Paine, William A. White, Rebecca C. Smyth, John R. Andrews, and James C. GibeautBureau of Economic Geology, Jackson School of Geosciences, The University of Texas at Austin

Abstract

We combined airborne lidar and ground-based EM induction measurements with vegetation surveys alongtwo transects across Mustang Island, a barrier island on the Texas coast, to examine whether these methods canbe used to map coastal wetlands and associated geomorphic environments.

Conductivity varied inversely with elevation along both transects. Elevation and conductivity profiles cor-related reasonably well with habitat mapped in the largely imagery-based 1992 National Wetland Inventory (NWI),but they possessed greater detail and identified misclassified habitat. Detail achievable with elevation and conduc-tivity data was similar to that achieved in on-the-ground vegetation surveys. Lowest elevations and highest conduc-tivities were measured in saline environments (marine and estuarine units, forebeach, salt marsh, and wind-tidalflats). Highest elevations and lowest conductivities were measured in nonsaline environments (upland and palustrineunits, dunes, vegetated-barrier flats, and fresh marsh).

Elevation and conductivity data allow better discrimination among coastal wetland and geomorphic envi-ronments than can be achieved from image interpretation alone. Future work should include evaluating the effect ofvegetation density on lidar-beam penetration, quantifying seasonal change in ground conductivity in fresh and salinecoastal environments, examining the geographic variability of elevation and conductivity statistics, and evaluatingthe use of airborne EM sensors to measure ground conductivity at multiple exploration depths.

Introduction

We explore whether two noninvasive geophysical methods — lidar (light detection and ranging) and EM(electromagnetic induction) — can improve the accuracy and resolution of wetland mapping that has historicallybeen based chiefly on analysis of aerial photographs and limited field checks. The importance of monitoring thestatus and trends of coastal wetlands has been increasingly recognized in recent decades as we have become moreaware of the critical role wetlands play in the transitional aquatic-terrestrial environment and have become con-cerned about the rapid change in wetlands resulting from the historic rise in relative sea level. Using Mustang Islandon the central Texas coast as an example (Figure 1), we examine the strong relationships among (1) elevation, soiland water salinity, and coastal habitat and (2) conductivity and salinity. We did this by acquiring lidar-derivedelevations and EM-derived conductivities and comparing these measurements with coastal habitat and geomor-phic environment data across this sandy barrier island.

We selected two representative transects across Mustang Island (Figure 1), where we acquired lidar data,surveyed vegetation type, and measured the apparent electrical conductivity of the ground. Conductivity, which isclosely correlated to soil and water salinity, was measured along the transects using a ground conductivity meter.We evaluated the traditional approach to wetland mapping by comparing habitat types extracted from the mostrecent wetland maps with coastal environments directly observed along the transects. We evaluated the lidar andEM approach by examining the relationships along each transect among mapped wetland units, lidar-derivedelevation, and measured ground conductivity and vegetation type determined during the ground surveys.

We employ terms from two common classification systems to examine the relationship between elevation,conductivity, and coastal vegetation assemblages: the more technical system used by the U.S. Fish and WildlifeService in the National Wetland Inventory (NWI) program and a geomorphic system used in our ground-based

745

Paine, J. G., White, W. A., Smyth, R. C., Andrews, J. R., and Gibeaut, J. C., 2005, Combining EM and lidar to map coastal wetlands: an example from Mustang Island, Texas, in Proceedings, Symposium on the Application of Geophysics to Engineering and Environmental Problems: Environmental and Engineering Geophysical Society, p. 745-756 (CD-ROM).

mapping that includes wetlands and other associated coastal environments. Geomorphic units identified alongthese transects include beach, dune, vegetated-barrier flat (VBF), fresh and salt marsh, and wind-tidal flat (Fig-ure 2), and each have correlative categories within the NWI system (Table 1).

Methods

We acquired surface elevations using airborne lidar and made ground-based conductivity measurementsand vegetation observations along the Mustang Island State Park and Port Aransas transects (Figure 1) across theisland. We compared elevation and ground conductivity data with vegetation assemblages as depicted on NWImaps and geomorphic units as determined from on-the-ground observations.

���������

���� �

���������

�����

����� ����

��

����

��� �����������

��������������

�

�

������

������

������������������

� � � ��

� � ��

������������������������������

Figure 1. Mustang Island study area, Texas Gulf Coast.

Figure 2. Generalized profile of a barrier island showing common geomorphic features that may be presentbetween the bay (left) and gulf (right) shorelines.

746

Lidar SurveyThe University of Texas at Austin Bureau of Economic Geology lidar team acquired and processed air-

borne, scanning laser terrain (lidar) data in September and October 2003. Lidar x, y, and z points representing theground surface were generated by combining laser range and aircraft attitude data (Wehr and Lohr, 1999) col-lected using an Optech Airborne Laser Terrain Mapper 1225 with once-per-second aircraft positions collectedusing geodetic quality airborne and ground-based GPS receivers. The lidar point data have a vertical accuracy ofabout 15 cm and are spaced at about one per 0.5 m2.

We used lidar point data to produce digital elevation models (DEM’s) along swaths from the gulf to thebay. We also constructed gulf-to-bay elevation profiles by averaging lidar data points located within 1.5 m of atransect station where we also measured ground conductivity and vegetation characteristics.

EM SurveyWe used the frequency-domain EM (FDEM) method to measure apparent electrical conductivity. FDEM

employs a changing primary magnetic field created around a transmitter coil to induce current to flow in the ground,which in turn creates a secondary magnetic field that is sensed by the receiver coil (Parasnis, 1986; Frischknechtand others, 1991; West and Macnae, 1991). The strength of the secondary field is a complex function of EMfrequency and ground conductivity (McNeill, 1980), but it generally increases with ground conductivity at constantfrequency.

We used a Geonics EM38 ground conductivity meter to measure the apparent conductivity of the ground.This instrument operates at a primary frequency of 15 kHz, measuring apparent conductivity to a depth of about0.8 m (horizontal dipole [hd] orientation) and 1.5 m (vertical dipole [vd] orientation). The instrument has a range ofabout 1 millisiemen/m (mS/m) to more than 1,000 mS/m. We acquired measurements at 234 sites on Mustang

NWI code Classification description Common (geomorphic) description

U Upland Not a wetland

PEM1A Palustrine emergent persistent wetland,

temporarily flooded

Fresh or interior marsh, persistent

vegetation, topographically high

PEM1C Palustrine emergent persistent wetland,

seasonally flooded

Fresh or interior marsh, persistent

vegetation, topographically low

E2EM1P Estuarine intertidal persistent emergent

wetland, irregularly flooded

Salt- to brackish-water marsh, persistent

vegetation, topographically high

E2EM1N Estuarine intertidal persistent emergent

wetland, regularly flooded

Salt- to brackish-water marsh, persistent

vegetation, topographically low

E2AB1P Estuarine intertidal aquatic bed, algal,

irregularly flooded

Tidal and wind-tidal flats, with algal

mats, topographically high

E2USP Estuarine intertidal unconsolidated

shore, irregularly flooded

Tidal and wind-tidal flats,

topographically high

E2USN Estuarine intertidal unconsolidated

shore, regularly flooded

Tidal and wind-tidal flats,

topographically low

E1UBL Estuarine subtidal unconsolidated

bottom, subtidal Estuarine open water

M2USP Marine intertidal unconsolidated shore,

irregularly flooded Backbeach along Gulf shore

M2USN Marine intertidal unconsolidated shore,

regularly flooded Forebeach along Gulf shore

Table 1. Classification system (Cowardin and others, 1979) used by the U.S. Fish and Wildlife Service in the1992 National Wetland Inventory (NWI). This partial list of units includes only those mapped along the Mus-tang Island transects (Figure 1).

747

Island in December 2003, recording apparent conductivity in the hd and vd orientations at stations spaced 20-mapart from the gulf beach to the bay shore.

Where the apparent conductivity of the ground was within the instrument’s range, we recorded measure-ments with the instrument on the ground. In areas where apparent conductivity approached or exceeded the upperlimit of the instrument’s range, we made one set of measurements with the instrument on the ground and another setwith the instrument 0.6 m above the ground. We then corrected the out-of-range values by extrapolating the lowerapparent conductivities recorded with the instrument at a fixed height according to the relationship observedbetween the ground-level and fixed-height measurements made over ground having lower apparent conductivities.

In the vd orientation, we determined the relationship between ground-level and raised-instrument mea-surements using 24 data pairs that had apparent conductivities at ground level of less than 1,300 mS/m (Figure 3).The relationship,

σg = 1.89 × σr + 34.7,

where σg is the apparent conductivity at the ground surface and σr is the apparent conductivity with the instrument0.6 m above the ground surface, has an r2 value of 0.95. We then extrapolated a ground-level apparent conductiv-ity from the raised-instrument conductivity for stations where the measured conductivity at ground level exceeded1,300 mS/m, the instrument's maximum linear limit in the vd orientation.

In the hd orientation, the 22 data pairs having apparent conductivities less than 1,400 mS/m produced asimilar relationship,

σg = 4.03 × σr − 85.5,

that yields an r2 value of 0.97.

���

���

���

�

����

� ��� ��� !�� ��� ���

��"��##�������������$�����������%��&�'

��"��##�������������$��������()���%��&�'

���*� (+,������-�./(!���*��(,���*��/

������

0������

.�����&�

Figure 3. Relationship between apparent conductivity measured in the vd orientation at an instrument height of0.6 m above ground (σr ) and apparent conductivity measured at ground level (σg ) determined using onlyground-level measurements below 1,300 mS/m.

748

Vegetation SurveyAt each transect station, we recorded plant species, percent cover, vegetation height, and soil wetness.

We combined aerial photograph signatures and field observations to classify the locations into the following geo-morphic environments: beach, dune, VBF, fresh marsh, salt or brackish marsh, and wind-tidal flat.

Mustang Island State Park Transect

The Mustang Island State Park (MISP) transect is located on the southwest part of Mustang Island(Figures 1 and 4), extending 2.2 km from the gulf beach to the Corpus Christi Bay shore. We surveyed vegetationand measured apparent conductivity at 112 stations along this transect and obtained elevations at these stationsfrom the lidar point data.

Elevation and VegetationElevations range from 0.01 to 5.5 m above the 1988 North American Vertical Datum (NAVD88) (Fig-

ures 5 and 6). We found the highest elevations (2 m or more) across the fore-island dunes within about 300 m ofthe gulf shoreline and midisland dunes between about 800 and 1500 m from the gulf shoreline. We found lowestelevations (0.3 m or less) on the beach and bayward of the midisland dunes to the bay shoreline.

At one-third (38) of the locations, vegetation was sufficiently dense for us to question whether the lidar-derived elevation represented the ground surface or the top of the vegetation mass. At these locations, height ofmassed vegetation averaged 0.5 m, ranging from 0.1 to 1.4 m. These heights can be subtracted from the lidar-derived elevation profile to produce a corrected ground-surface elevation profile, assuming lidar was unable topenetrate the vegetation. In densely vegetated areas, vegetation mass might cause significant overestimation ofland-surface elevation and potential misclassification of environments on the basis of lidar data alone.

Figure 4. Aerial photomosaic of the MISP transect showing habitats (Table 1) identified on the 1992 NWI map.

�

� �(� ��

� �(. ��

��������� �

��������� �

��

��

��

��

��

��

��

����

��

��

��

��

��

����������

������������

������������

����������

������������

����������

����������

����������

����������

����������

����������

����������

����������

����������

����������

�� ��

�����

�����

������������

��������������������

����������

���

��

���

�� �����

�����

������

������

������

����

������

����

749

Transect locations with the highest elevations generally correlated with upland or high palustrine units andlocations with the lowest elevations generally coincided with estuarine units (Figure 6a). Average elevation washighest for upland (U) locations (Table 2), but elevation for this unit overlapped with elevation ranges for otherNWI units. The highest of the palustrine units (PEM1A) had the next highest average elevation. Unit PEM1C,topographically lower than PEM1A, had a slightly lower average elevation. Estuarine units E2EM1P, E2EM1N,and E2USP have similar average elevations that are considerably lower than those for the upland and palustrineunits. Elevation limits for the mapped upland and palustrine units overlap, as do ranges for the estuarine units(Table 2). Nevertheless, there is a distinct difference in average elevation (and little overlap in elevation range)between the palustrine and estuarine units.

During the ground-based survey, we classified each transect station into a coastal geomorphic unit (Fig-ure 2) on the basis of field characteristics such as vegetation (Figure 6b and Table 3). Most common were dune,VBF, and wind-tidal flat. These ground-based surveys produced a classification with higher spatial resolution thanthat depicted on the NWI maps and one that better represents the variability evident from the topographic profile(Figure 6b). The dune environment has the highest average elevation and the largest elevation range, overlapping atthe low end with the VBF, fresh marsh, and beach environments (Table 3). Relatively high elevation averages areassociated with VBF, high fresh marsh, low fresh marsh, and beach environments, which all have some degree ofoverlap in elevation ranges. Distinctly lower elevation averages are associated with high and low salt marsh andhigh and low wind-tidal flat environments. Elevation ranges for these estuarine environments overlap with eachother, but not with fresh marsh, VBF, or dune environments.

Conductivity and VegetationApparent ground conductivity varies over three orders of magnitude, ranging from resistive ground at a

few mS/m to relatively conductive ground at more than 2,000 mS/m (Figure 6 and Table 2). We found highapparent conductivities (greater than 100 mS/m) within a few tens of meters of the gulf shoreline, along two

1�$������%�'

�

2��

�

� �(� ��

� �(. ��

��������� �

��������� �

��

��

��

��

��

��

��

����

��

��

��

��

��

����������

������������

������������

����������

������������

����������

����������

����������

����������

����������

����������

����������

����������

����������

����������

�� ��

�����

�����

������������

��������������������

����������

���

��

���

�� �����

�����

������

������

����

��

����

����

��

����

Figure 5. Lidar-derived DEM swath along the MISP transect showing habitats depicted on the 1992 NWI map.

750

Figure 6. Elevation and conductivity (vd orientation) profiles superimposed on (a) 1992 NWI units and(b) surveyed coastal environments along the MISP transect. Distances are measured from the gulf shoreline.B = beach or berm, D = dune, other units as described in Tables 1 and 3.

3�������%�'� ��� ��� !�� ��� ��� ��� !�� ���� ����

�

�

4

+

/

)

1�$������%�

'

��������������������� �##�������������$����%��&�'

�

�5���

��

������ �����!��� �����!��� �����!

�1� �

1�1

� � 1�6��

1�1

� �

1�1

� �

1�1

� �

1�6

��

��6��

6 6 6�1� � �1� �

�1� 0

�1� 0

3�������%�'� ��� ��� !�� ��� ��� ��� !�� ���� ����

�

�

4

+

/

)

1�$������%�

'

��������������������� �##�������������$����%��&�'

�

�5���

��

���

7

3

33879

��:

;9<

879 ;9:

��<

;9: ��:

;9: ;9<

��<

;9:

;9<

��<

7 3333333 879

879

879

879

879

879

879

879

879 8879

�9:

�9<

�9<;�9<

�9:

�9:

7

3

33879

��:

;9<

879 ;9:

��<

;9: ��:

;9: ;9<

��<

;9:

;9<

��<

7 3333333 879

879

879

879

879

879

879

879

879 879

�9:

�9<

�9<;�9<

�9:

�9:

��� �����!��� �����!��� �����!

(a)

(b)

midisland segments, and near the bay shoreline (Figure 6). We measured lowest apparent conductivities (about10 mS/m or less) just inland from the gulf shoreline and along two midisland segments.

Conductivity correlates reasonably well with NWI units (Figure 6 and Table 2). Upland (U) and highpalustrine (PEM1A) units tend to occur where apparent conductivity is low (less than about 100 mS/m), whereaslower palustrine (PEM1C), estuarine (E2EM1P, E2EM1N, and E2USP), and marine (M2USN) units are mappedwhere apparent conductivity is relatively high (greater than 100 mS/m). Among the more conductive NWI units,average apparent conductivity is highest for the topographically lowest estuarine unit (E2USP), decreases slightlyfor the next lowest estuarine unit (E2EM1N), and decreases further for the highest of the mapped estuarine units(E2EM1P). Marine-influenced (M2USN) and lowest palustrine (PEM1C) units have lower conductivities that donot overlap with those measured for the more conductive estuarine units. Among the relatively nonconductive NWIunits, the lowest average conductivity is associated with upland (U) locations. Slightly higher average conductivityis associated with the highest palustrine unit (PEM1A). The conductivity range measured for locations within

751

Environment n

Elev. avg.

(m)

Elev. range

(m)

App. con.

avg., vd

(mS/m)

App. con.

range, vd

(mS/m)

App. con.

avg., hd

(mS/m)

App. con.

range, hd

(mS/m)

Dune 21 2.64 0.8-5.49 59 2-767 38 1-505

VBF 37 1.31 0.5-2.92 76 4-561 57 2-514

MFH 4 0.86 0.7-1.06 145 42-329 99 34-221

MFL 9 0.77 0.52-1.06 242 43-408 202 30-270

MF (all) 13 0.8 0.52-1.06 212 42-408 170 30-270

MSH 3 0.29 0.22-0.38 1175 852-1392 1150 842-1445

MSL 4 0.17 0.01-0.25 1223 1106-1345 1263 1119-1429

MS (all) 7 0.22 0.01-0.38 1202 852-1392 1214 842-1445

WTFH 20 0.23 0.1-0.46 1489 1157-1783 1565 1046-2021

WTFL 7 0.2 0.17-0.23 1397 1279-1592 1477 1224-1715

WTF (all) 27 0.22 0.1-0.46 1465 1157-1783 1542 1046-2021

Beach, berm 4 0.79 0.26-1.55 578 91-1192 604 72-1219

Water 2 0.78 0.74-0.82 33 25-40 29 27-31

Table 3. Elevation and apparent conductivity ranges measured at 112 locations for geomorphic units (Figure 2)along the MISP transect. VBF = vegetated-barrier flat, MFH = high fresh marsh, MFL = low fresh marsh,MSH = high salt marsh, MSL = low salt marsh, WTFH = high wind-tidal flat, WTFL = low wind-tidal flat.

NWI

unit n

Elev. avg.

(m)

Elev. range

(m)

App. con.

avg., vd

(mS/m)

App. con.

range, vd

(mS/m)

App. con.

avg., hd

(mS/m)

App. con.

range, hd

(mS/m)

U 25 2.62 0.52-5.49 26 2-288 21 1-260

PEM1A 40 1.12 0.38-1.97 94 10-852 75 5-842

PEM1C 10 0.9 0.54-1.23 266 160-408 207 106-270

E2EM1P 3 0.2 0.18-0.22 1254 1157-1326 1293 1163-1405

E2EM1N 11 0.19 0.01-0.26 1318 1106-1592 1386 1119-1715

E2USP 21 0.26 0.1-0.8 1467 767-1783 1516 505-2021

M2USN 2 0.68 0.34-1.02 515 322-707 530 298-828

Table 2. Elevation and apparent conductivity ranges measured at 112 locations for 1992 NWI units mappedalong the MISP transect (Figure 4). Elevations are relative to NAVD88. Apparent conductivities were measuredusing a Geonics EM38 in the vd and hd orientations.

upland (U) units overlapped with ranges measured for locations within palustrine units, but not with marine orestuarine units.

Geomorphic environments also correlate well with measured apparent conductivity (Figure 6b and Table 3).Highest apparent conductivities measured in the vd orientation occur in beach, low and high salt marsh, and lowand high wind-tidal flat environments. Lowest apparent conductivities occur in dune, VBF, and low and high freshmarsh environments. Dune locations have the lowest average conductivity, but their measured range extends abovethe average values observed for low and high fresh marshes. Low average conductivities are also found in VBFenvironments. Gulf beach and bay berm environments have higher average apparent conductivities than are foundin dune and fresh marsh environments. Salt marsh and wind-tidal flats have the highest apparent conductivities.

752

Average apparent conductivity increases from high to low salt marsh and from low to high wind-tidal flat. Rangesof measured conductivities overlap for the salt marsh and wind-tidal flats and for the dunes, VBFs, and freshmarshes, but there is little or no overlap in observed conductivity range between these two groups of relativelysaline and nonsaline environments.

Elevation, Conductivity, and VegetationIn general, elevation and apparent conductivity vary inversely (Figure 6), reflecting the strong negative

correlation between elevation and salinity in coastal environments. As elevation decreases, the frequency of flood-ing by saline water increases. At higher elevations, infrequent saline flooding, infiltrating fresh precipitation, andrelatively dry soil combine to produce less electrically conductive soil. Conductivity values vary over a greaterrange than do elevations, but both vary significantly and measurably across the island.

We can attempt to better discriminate NWI and geomorphic units that may have overlapping elevation orconductivity ranges by combining elevation and apparent conductivity. For example, locations within the NWIupland (U) unit generally have both low apparent conductivities and high elevations, whereas the highest palustrineunit (PEM1A) generally has lower elevations and higher conductivities (Figure 7a). High and low palustrine unitsPEM1A and PEM1C have minor differences in elevation but more distinct differences in apparent conductivity.Estuarine and marine units have both very low elevations and very high apparent conductivities.

Among the geomorphic units, dunes have high and highly variable elevations but have low conductivitiesthat vary over a relatively small range (Figure 7b). VBF environments generally have lower elevations than duneenvironments and higher and more variable conductivity values. High fresh marshes have elevations that are indis-tinguishable from VBF environments but have apparent conductivities that tend to be higher than those measuredin the VBFs. Salt marshes and wind-tidal flats have very low elevations and very high apparent conductivities.

Advantages and Limitations

Airborne lidar offers detailed and accurate elevation measurements that can be used to help classify wet-lands and associated habitats more accurately than classifications based on aerial photographs alone. Compari-sons of mapped NWI units with elevation profiles across Mustang Island show that elevation detail achieved withlidar allows more precise discrimination of wetland and upland units than appears on NWI maps (Figure 6a).Furthermore, some NWI units on both island transects are misclassified; some units mapped as wetland containupland habitat, and some units mapped as upland contain wetland habitat. Comparisons of lidar-derived elevationswith geomorphic units delineated during the field survey show similar levels of detail (Figure 6b). This suggests thatlidar can be used to map coastal environments at the same level achievable with labor-intensive ground-basedsurveys, which are impractical over large areas. Lidar-derived elevations can complement aerial photographicanalysis by helping to distinguish coastal environments, as well as upland, palustrine, estuarine, and marine habitatsthat may have ambiguous photographic signatures.

Most NWI habitats and coastal geomorphic environments have statistically distinct average elevations butrather wide elevation ranges that overlap to varying degrees with other habitats and environments. Furthermore,lidar pulses may not reach the ground surface in densely vegetated areas, producing anomalously high elevations atthose points and leading to potential misclassification of habitat or environment on the basis of elevation alone.

Conductivity is highly inversely correlated to lidar-derived elevation on the Mustang Island transects. EM-derived conductivities correlate well with both mapped NWI wetland and upland habitat and coastal geomorphicunits identified in the field. EM and lidar achieve similar levels of detail exceeding that depicted on the NWI maps.Conductivities closely track changes in coastal environment, suggesting that EM data can be used to distinguishhabitats and geomorphic units to the same level achievable with ground-based vegetation surveys. Comparisons ofmapped NWI units with conductivity data reveal apparent misclassifications in the NWI maps, both where mapped

753

6#����

�1� 0

��6��

1�6��

1�1� �

1�1� �

�1� �

+

�

4

� �� �5��� ���

�

/

)

1�$������%�

'

�##�������������$����%$�=���&�'

3��

879

�9:

;9:

��<

�9<

��:

7���&7��

;���

;9<

+

�

4

� �� �5��� ���

�

/

)

1�$������%�

'

�##�������������$����%$�=���&�'

Figure 7. Elevation and apparent conductivity of (a) 1992 NWI units and (b) coastal geomorphic units alongthe MISP transect.

(a)

(b)

wetland units enclose areas having conductivities that indicate an upland habitat and where mapped upland habitatsenclose areas having conductivities that indicate wetlands.

Average conductivities for each NWI and coastal geomorphic unit are distinct (Tables 2 and 3), but theranges of conductivities measured within these units overlap to varying degrees. Upland and fresh-water environ-ments are most easily distinguished from estuarine and marine environments because conductivity strongly re-sponds to changes in salinity. Overlap in ranges could lead to misclassification of units if the classification is basedon conductivity alone.

754

Classifying Wetland and Coastal Environments

Correlations among wetland habitat, geomorphic unit, elevation, and conductivity suggest that lidar andEM data can be used to improve the accuracy of coastal habitat classification and perhaps partly automate theprocess. One approach would be to combine photographic, elevation, and conductivity data in a common spatialenvironment, using elevation and conductivity as a supplement to aid classification of ambiguous habitat signatureson aerial photographs.

A more quantitative and automatable approach would be to establish statistical elevation and conductivitycharacteristics for all habitat and geomorphic types. One could then use measured elevations and conductivities toclassify locations according to proximity of each measurement to average elevation and conductivity for eachhabitat or environment. Because the statistical characteristics (such as average, range, and standard deviation)could be calculated for each habitat or environment type, probabilities of accurate classification could be assignedfor each point. Because elevations and conductivities are easily distinguished between upland and fresh-waterhabitats and estuarine and marine environments, likelihood of misclassification at this level would be low. Likeli-hood of misclassification among habitats with more elevation and conductivity overlap, such as between someestuarine and marine units and saline environments, would be higher.

Future Work

Preliminary results are encouraging, but many uncertainties remain to be investigated before lidar and EMcan be used routinely and accurately in coastal habitat classification. For example, further work is needed todetermine where vegetation density is great enough to prevent lidar from detecting the top of vegetation rather thanthe ground surface. In coastal areas, where errors of less than one meter can lead to significant habitat misclassification,methods of correcting for vegetation height become critical.

We measured ground conductivity late in the fall and examined the relationship with habitat and geomor-phic units on the basis of those measurements. It is likely that conductivities within the uppermost meter of thesubsurface will change seasonally with precipitation and ambient temperature. We hope to reoccupy the same sitesin different seasons, especially following precipitation or flooding events, to examine the magnitude of these changesand to identify the environments that are most susceptible to seasonal change.

We made our conductivity measurements using a ground-based instrument that explores 0.8 to 1.5 m inthe subsurface. Ground surveys are adequate for limited field investigations but are too labor-intensive to maplarger areas. Similar instruments can be towed beneath low-flying helicopters to rapidly and remotely acquireconductivity data. Measurements from airborne EM instruments can be made simultaneously at multiple explora-tions depths, enabling shallow data to be used for vegetation mapping and deeper data to be used for complemen-tary purposes such as monitoring saltwater intrusion into coastal aquifers and characterizing fresh-water lensesunderlying many coastal barriers.

Conclusions

Elevations measured using airborne lidar correlate well with NWI upland, palustrine, estuarine, and marineunits. Lidar-derived elevation profiles provide greater detail than is present in NWI maps produced from aerialphotographs and can be used to map wetland habitat more accurately and in greater detail than is feasible fromaerial photographs and limited field checks. Mapping detail achievable with lidar approaches that of ground-basedinvestigations. Where vegetation is dense, lidar-derived elevations may represent the top of massed vegetationrather than the ground surface, leading to potential habitat misclassification.

Measurements of shallow electrical conductivity using a ground-based EM instrument range over threeorders of magnitude and correlate well with both NWI habitats and coastal geomorphic units. Highest conductivi-

755

ties are measured within marine and estuarine NWI units and in salt marsh, wind-tidal flat, and forebeach environ-ments. Lowest conductivities are measured within upland and palustrine NWI habitats and in dune, VBF, and freshmarsh environments. Conductivity changes are consistent with, but more detailed than, changes depicted on NWImaps.

Lidar-derived elevation and EM-derived conductivities are inversely correlated, and each method hasadvantages and disadvantages. Both methods readily discern saline- and fresh-water environments and comple-ment traditional, photograph-based wetland classification by helping classify distinct coastal environments thathave similar signatures on aerial photographs. Overlap in elevation and conductivity among some habitats andenvironments suggests that a statistical approach to wetland classification based on integrated data from lidar, EM,and aerial photographs could achieve greater detail and accuracy than methods based on limited field checks ofboundaries mapped on aerial photographs or remotely sensed images.

Further evaluation of lidar and EM in coastal environment classification should include (1) characterizingand minimizing land-surface elevation error where vegetation is dense; (2) determining the variation in measuredconductivity with seasonal changes in ambient temperature and precipitation; (3) evaluating whether elevation andconductivity statistics measured and calculated in one area can be applied to classifying similar environments inother, geographically distinct areas; and (4) migrating conductivity measurements to an airborne platform wherelarge areas can be surveyed rapidly and multiple depths can be explored simultaneously.

Acknowledgments

This project was partly funded under contract number 03-005 from the Texas General Land Office to TheUniversity of Texas at Austin, Jeffrey G. Paine, principal investigator. The Texas Coastal Coordination Councilfunded the project under the Texas Coastal Management Program (CMP). The Army Research Office partlysupported lidar data acquisition. This article is a publication of the Coastal Coordination Council pursuant toNational Oceanic and Atmospheric Administration (NOAA) award number NA17OZ2353. The views expressedherein are those of the authors and do not necessarily reflect the views of NOAA or any of its subagencies. Lidardata were acquired and processed by John R. Andrews, James C. Gibeaut, Roberto Gutierrez, Tiffany L. Hepner,and Rebecca C. Smyth. Rachel L. Waldinger assisted with vegetation analysis. Publication approved by the Direc-tor, Bureau of Economic Geology.

References

Cowardin, L. M., Carter, V., Golet, F. C., and LaRoe, E. T., 1979, Classification of wetlands and deepwaterhabitats of the United States: Washington, D.C., USA, U.S. Department of Interior, Fish and WildlifeService, 131 p.

Frischknecht, F. C., Labson, V. F., Spies, B. R., and Anderson, W. L., 1991, Profiling using small sources, inNabighian, M. N., ed., Electromagnetic methods in applied geophysics—applications, part A and partB: Tulsa, Society of Exploration Geophysicists, p. 105–270.

McNeill, J. D., 1980, Electromagnetic terrain conductivity measurement at low induction numbers: Mississauga,Ont.,Geonics Ltd., Technical Note TN-6, 15 p.

Parasnis, D. S., 1986, Principles of applied geophysics: London, Chapman and Hall, 402 p.Wehr, A. and U. Lohr, 1999, Airborne laser scanning—an introduction and overview: ISPRS Journal of Photo-

grammetry and Remote Sensing, vol. 54, no. 2-3, p. 68–82.West, G. F., and Macnae, J. C., 1991, Physics of the electromagnetic induction exploration method, in

Nabighian, M. N., ed., Electromagnetic methods in applied geophysics—applications, part A and partB: Tulsa, Society of Exploration Geophysicists, p. 5–45.

756