7/27/11 Peggy O’Neill, NASA GSFC IGARSS’11, Vancouver, BC, Canada National Aeronautics and Space Administration Utilization of

7/27/11 Peggy O’Neill, NASA GSFC IGARSS’11, Vancouver, BC, Canada

National Aeronautics and Space Administration http://smap.jpl.nasa.gov

Utilization of Ancillary Data Sets for SMAP

Algorithm Development and

Product Generation

IGARSS’11, Vancouver, BCJuly 27, 2011

Peggy E. O’Neill, NASA GSFCErika Podest, JPLEni G. Njoku, JPL


BACKGROUND

• SMAP is a planned NASA Earth Science Decadal Survey Mission

• Launch currently scheduled for October 2014 into a 6 am / 6 pm sun-synchronous orbit

• Will use an L-band radar & radiometer to measure global soil moisture & freeze/thaw every 2-3 days

• Baseline SMAP data products include:

-- radar-derived F/T at 3 km resolution

-- radiometer-only SM at 40 km resolution

-- combined radar/radiometer SM at 9 km resolution

-- value-added products (root zone SM, carbon NEE) at 9 km

• All SMAP products output on nested 1, 3, 9, 36 km EASE grids


Product Short Description Resolution/Grid

Latency

L1A_S0 Radar raw data in time order – 12 hours

Instrument Data

L1A_TB Radiometer raw data in time order – 12 hours

L1B_S0_LoRes Low resolution radar σo in time order 5x30 km 12 hours

L1B_TB Radiometer TB in time order 36x47 km 12 hours

L1C_S0_HiRes High resolution radar σo 1-3 km 12 hours

L1C_TB Radiometer TB 36 km 12 hours

L2_SM_A Soil moisture (radar) [research product] 3 km 24 hoursScience Data (Half-Orbit)

L2_SM_P Soil moisture (radiometer) 36 km 24 hours

L2_SM_A/P Soil moisture (radar/radiometer) 9 km 24 hours

L3_SM_A Soil moisture (radar) [research product] 3 km 50 hours

Science Data (Daily Composite)

L3_F/T_A Freeze/thaw state (radar) 3 km 50 hours

L3_SM_P Soil moisture (radiometer) 36 km 50 hours

L3_SM_A/P Soil moisture (radar/radiometer) 9 km 50 hours

L4_SM Soil moisture (surface & root zone) 9 km 7 days ScienceValue-AddedL4_C Carbon net ecosystem exchange (NEE) 9 km 14 days

SMAP Data Products


Algorithm Needs

• All baseline SMAP products have associated algorithm(s) which require a variety of ancillary data to meet retrieval accuracies:

-- 0.04 cm3/cm3 for soil moisture within SMAP land mask

-- 80% classification accuracy for binary F/T in boreal latitudes

• Areas of snow/ice, frozen ground, mountainous topography, open water, urban areas, and dense vegetation (> 5 kg/m2) are excluded from SM accuracy statistics • Static ancillary data do not change during mission

-- permanent masks (land/water/forest/urban/mountain), DEM, soils

• Dynamic ancillary data require periodic updates ranging from daily

to seasonally

-- soil T, precipitation, vegetation, surface roughness, land cover


Ancillary Parameters

1 Soil Temperature2 Surface Air Temperature3 Vegetation Water Content

(VWC)

4 Sand & Clay Fraction5 Urban Area6 % Permanent Open Water7 Crop Type8 Land Cover Class9 Precipitation

10 Snow11 Mountainous Area [DEM]12 Permanent Ice13 b, ω, & τ Vegetation Parameters14 h Roughness Parameter

Table 1. Ancillary Parameters • 14 ancillary data parameters identified as needed by one or more SMAP algorithms

• choice of source of each parameter driven by:

-- availability

-- ease of use

-- inherent error

-- latency

-- temporal & spatial resolution

-- global coverage

-- positive impact on SMAP retrieval accuracies

-- compatibility with SMOS choices

• choices documented in a SMAP Ancillary Data Report for each parameter

• data from each primary source will be used now in pre-launch simulations

• choices will be revisited as new information becomes available


Soil Temperature

Accuracy of synchronized NWP forecast surface soil temperature compared against in situ temperatures for the Oklahoma Mesonet for 2004 and 2009.

SMAP 6 am descending orbit • SMAP soil moisture products will be retrieved using data from the 6 am descending orbits

• the 6 am 0-5 cm TS is the most dynamic

ancillary parameter needed -- it is updated every orbit for each location

• SMAP error budgets currently carry 2 K as the error in ancillary TS

• data from the Oklahoma Mesonet indicates

that at the 6 am overpass time, all NWP TS products have errors just below 2 K

• initial global estimates of NWP TS error

against in situ point measurements are less optimistic, more in the range of 2.5 – 3.0 K; analysis on global TS error is continuing


Vegetation Water Content

Annual climatology of NDVI for Walnut Creek, IA

snow

• a new 10-year (2000-2010) MODIS NDVI climatology has been created at 1 km resolution

globally

• VWC calculated using NDVI-based water contributions from both foliage and stem components, adjusted for IGBP land cover classes

VWC (kg/m2) over the continental U.S. for the month of July on a 1-km EASE grid as constructed from a 10-year MODIS NDVI climatology and land cover products.


Soil Texture

• soil sand & clay fraction needed by dielectric models used in SM retrieval

• best available source used for any given region

• resulting global map a combination of different data sets

• potential for discontinuities at data set boundaries (e.g., US / Canada)

Global sand fraction at 0.01 degree resolution based on a composite of FAO, HWSD, STATSGO, NSDC, and ASRIS datasets using best available source for a given region.


Urban Areas

• GRUMP urban data (Columbia U.) gridded to SMAP 9 km EASE grid

• better delineation between urban & rural areas

• urban fraction > 0.5 shown

• however, urban flag likely to be set much lower since TB cannot be

corrected for presence of urban areas

Global Rural-Urban Mapping Project


Open Water Fraction

• use SMAP HiRes radar to determine open water fraction

• a 3 dB threshold is applied to HH to VV ratio to distinguish water from land

• this SMAP parameter can be supplemented by static permanent water body data sets like MODIS MOD44W and JERS-1/PALSAR (for boreal latitudes)

• the water fraction is then used to correct TB for a mix of land & water in the

grid cell

Partial UAVSAR ratio image of Mono Lake. ~7% detection error

Open water (both permanent & transient)

in a SMAP footprint is a potential large

error source for SMAP retrieval algorithms if

its presence is not detected & corrected for


Topography / DEM

JPL Global DEM

-- compiled from different sources

-- 1 arc-secondresolution

-- GMTED2010 will eventually replace GTOPO30

-- above will be useful in assessing any discontinuities between existing data sets

-- elevation and slope variance (TBC) could be used to set topography flag

Input Data Set: US SRTM SRTM GTOPO Alaska DEM Canada DEM

Coverage: United States 56 °S to 60 °N Global Alaska Canada

Source: NASA-JPL NASA-JPL USGS USGS GeoBase

Resolution: 1 arc-second 3 arc-seconds 30 arc-seconds 2 arc-seconds 3x6 arc-seconds

Horz. Datum: WGS84 WGS84 WGS84 NAD27 NAD83

Vertical Datum: EGM96 EGM96 EGM96 NAVD29 CVGD28

Projection: Geographic Geographic Geographic Geographic Geographic

Acquisition Date: February 2000 February 2000 Late 1996 1925 - 1999 --


Error Analysis

L2_SM_P Error Analysis

• Errors in ancillary data are factored into the SMAP soil moisture retrieval algorithm error budget

Simulated error performance of candidate retrieval algorithms for the radiometer-derived

soil moisture product using one year of simulated SMAP H- and V-pol TB with indicated

errors in model and ancillary parameters.


Ancillary Parameter Choices

1 Soil Temperature GMAO or ECMWF forecast temperatures (TBD)

2 Surface Air Temperature GMAO or ECMWF forecast temperatures (TBD)

3 Vegetation Water Content (VWC) MODIS NDVI [R. Hunt approach]

4 Sand & Clay fractionHWSD (global), regional data sets (STATSGO-US, ASRIS-Australia, NSD-Canada), FAO

5 Urban Area GRUMP data set – Columbia University

6 Open Water Fractiona priori static water fraction from MODIS MOD44W to be used in conjunction with open water fraction from SMAP HiRes radar

7 Crop Type combination of USDA Cropland Data Layer, AAFC-Canada, Ecoclimap-Europe

8 Land Cover Class MODIS IGBP; crop class will be further subdivided into four general crop types

9 Precipitation ECMWF total precipitation forecasts (or GPM once launched)

10 Snow Snow & Ice Mapping System (IMS) - NOAA

11 Mountainous Area [DEM] combination of SRTM, Alaska USGS DEM, Canada Geobase DEM, and GTOPO30

12 Permanent Ice MODIS IGBP

13 b, ω, and τ Vegetation Parameters land cover-driven table lookup

14 h Roughness Parameter land cover-driven table lookup

Anticipated Primary Sources of Ancillary Parameters


Summary

• All ancillary data will be resampled to the SMAP EASE grids at 1, 3, 9, 36 km

• Preliminary choices have been made for primary source of each ancillary parameter

-- these choices will be used pre-launch for SMAP simulations and algorithm development

-- choices will be re-examined as new information becomes available

-- will leverage SMOS data and experience

-- SMOS / SMAP consistency desirable

• Choices documented in SMAP Ancillary Data Reports

• Wise choices in ancillary data will help SMAP to provide accurate global measurements of SM & F/T

Documents

7/27/11 Peggy O’Neill, NASA GSFC IGARSS’11, Vancouver, BC, Canada National Aeronautics and Space Administration Utilization of