FIGURE 6' Multidate smoothed composite of soil moisture product on 070602' Red predawn mode 1:30 a'm

1
SYNERGISTIC USE OF AMSR-E AND MODIS DATA FOR
UNDERSTANDING LAND SURFACE PHENOLOGY A CASE
STUDY FROM GRASSLANDS OF GREAT PLAINS Marcela
Doubková, Center for Advanced Land Management
Information Technologies (CALMIT) Graduate
Program in Geography, University of
Nebraska-Lincoln Geoffrey M. Henebry, Ph.D.,
Geographic Information Science Center of
Excellence (GIScCE) South Dakota State University


1. INTRODUCTION
In recent investigations into the response of
native grasslands to global environmental
changes, rainfall variability was offered as a
key factor to explain ecosystem structure and
function. In particular, changes in temporal
patterns of precipitation was shown to alter the
key carbon cycling processes, such as net
photosynthesis and above ground productivity, and
ecological patterns like community
composition. To understand the impact of
rainfall variability in grasslands, an
understanding of soil moisture dynamics is
critical. Here the spatio-temporal trends of two
Advanced Microwave Scanning Radiometer (AMSR-E)
standard data products (vegetation water content
and soil moisture) and high sensitivity to the
precipitation event and land cover composition
are demonstrated. The sensitivities of vegetation
water content and soil moisture retrievals were
found to be dependent on the NDVI value, with an
apparent loss of sensitivity at high NDVI values.
ABSTRACT
2. METHODS
Data Source The AMSR-E L3 land products are
distributed through the National Snow and Ice
Data Center (NSIDC http//www.nsidc.org/data/ae_l
and3.html) We restrict our attention to the
vegetation water content and soil moisture
products. MODIS vegetation index products
(MOD13A2, MYD13A2) in the form of 16-day
maximum-value composites and MODIS Land Cover
Type (MOD12Q1) product were acquired through the
Land Processes Distributed Active Archive Center
(LPDAAC http//edcdaac.usgs.gov/main.asp). Meteo
rological data were gathered through the National
Weather Service Cooperative Observer Network and
provided by the High Plains Regional Climate
Center (http//www.hprcc.unl.edu/coop/home.html).
NEXRAD data were acquired from National
Climatic Data Inventory archive
(http//www.ncdc.noaa.gov/nexradinv/index.jsp).
Recently, retrievals for surficial (lt5 cm) soil
moisture and vegetation water content (VWC) at 25
km spatial resolution became available as
standard data products from the Advanced
Microwave Scanning Radiometer (AMSR-E) on the
Aqua orbital platform (Njoku, 2004). AMSR-E, like
other passive microwave radiometers, offers
synoptic views of cool earthlight, the
terrestrial radiation emitted at longer
wavelengths (0.34 - 4.3 cm). The sensitivity of
the microwaves bands on AMSR-E to the moisture
content of vegetation and soil has been amply
demonstrated (Wang, 1992).
Our study area covers 700 000 km2 and spans from
southeastern Kansas to eastern Montana.
Meteorological data were acquired from 24
representative stations which were distributed
across 24 AMSR-E pixels served as the focal site
for study (Figure 3).
Data Processing Within the study area 24
regularly distributed weather stations were
overlaid with corresponding AMSR-E pixels. Land
cover composition and NDVI values within the
AMSR-E pixel extent were retrieved and plotted
together with the vegetation water content
product. Quadratic models were fit separately to
explain VWC (am and pm retrievals) and NDVI by
Accumulated Growing Degree-Days (AGDD base 0
oC). Timing of the peak occurrences of NDVI and
VWC were calculated from model parameter
coefficients. This method has been successfully
used to model land surface phenology in other
grassland biomes (de Beurs and Henebry, 2005).
Weather data from nine weather stations within
a region influenced by the Nebraska rainfall
event from 7/6/02 were obtained. Moreover,
vegetation water content and soil moisture daily
data were acquired and its sensitivity to the
precipitation was examined.
Vegetation Water Content Land Surface
Phenology
3. RESULTS I
A positive correspondence of canopy water content
(represented by Normalized Differential Water
Index, NDWI) and vegetation biomass (represented
by Normalized Differential Vegetation Index,
NDVI) has been amply demonstrated (Jackson et
al., 2003). Here we used AMSR-E and MODIS NDVI
data to study similar relationships. Data from 24
AMSR-E pixels were investigated. Three of these
located within approximately 160 km from each
other are displayed in Figure 2. These are Tryon
(98 grassland), Oconto (58 cropland and 42
cropland/natural vegetation mosaic), and Holdrege
(98 cropland).
FIGURE 1. Multidate composite of vegetation water
content product (ascending mode) in 2004.
Red 01JUN Green 15JUL Blue 01SEP Overlays
Omerniks level III ecoregions with selected 24
representative stations.
15Jul

4. RESULTS II
Ogallala Flood event Here we concentrate on the
soil moisture and vegetation water content
response to the rapid precipitation event from
Ogallala on 7/6/02. Rainfall data from nine
weather stations on the storm path (Figure 5) and
vegetation water content and soil moisture from
the corresponding AMSR_E pixels were acquired and
plotted in Figure 7.

• Even though the precipitation patterns at three
  studied weather stations provided similar
  patterns, the VWC response to precipitation
  varied rapidly
• Strong correspondence was found only in between
  VWC and NDVI values in Tryon (grassland).
• High NDVI values were associated with lower VWC
  values.
• The difference in peak occurrences of VWC
  (predawn) and NDVI increase rapidly once maximum
  NDVI values exceeded value of 0.5.
• At the beginning of the growing season when soil
  moisture does not limit evapotranspiration, the
  afternoon acquisitions exhibit higher VWC than
  predawn. Once soil moisture becomes limiting to a
  mature canopy later in the season, predawn VWC is
  higher than afternoon.

On Saturday, July 6, 2002, a large storm front,
moving east through the state of Colorado and
into Nebraska dumped a large amount of rain in
the vicinity of Ogallala, Nebraska. The storm
front resulted in as much as 28 cm of rain in 10
hours.
National Climatic Data Center, extreme weather
and climate events Highway 61 was closed due to
water over the road. Water runoff occurred
quickly as heavy rain from thunderstorms fell on
already saturated ground from flash flooding
early that morning.
FIGURE 5. Ogallala flood as recorded by NEXRAD.
These data were obtained from the Doppler radar
located in North Platte, Nebraska. The range of
the radar reaches only to the border of Nebraska
and South Dakota, hence precipitation beyond this
limit is not displayed above.
FIGURE 6. Multidate smoothed composite of soil
moisture product on 07/06/02. Red pre-dawn mode
(130 a.m.), Green afternoon mode (130 p.m.),
Blue pre-dawn mode.
FIGURE 3. MODIS Land Cover Type product overlaid
with Omerniks ecoregions (Chapman et.al. 2001)
and 24 weather stations with AMSR-E pixels .
FIGURE 2. AMSR-E VWC retrievals from ascending
(afternoon) and descending (pre-dawn) orbits
averaged over 16 days, MODIS NDVI time series,
and total precipitation in 2003 and 2004.
Strong response to the extreme precipitation
event was encountered in both soil moisture
(Figure 6) and VWC products in all 9 stations.
Results from Ogallala, Hyannis, and Gettysburg
are displayed in Figure 7. We were able to fit
negative exponential curves to data from all nine
stations. The behavior of these curves was
influenced by land cover composition. Both soil
moisture and VWC rapidly decreased after the
rainfall in those pixels with mostly grassland. A
slower decrease was found in regions covered with
mixture of croplands and grasslands.
The increase of total NDVI values was followed by
the shift of the peak occurrence of VWC and NDVI
towards higher AGDD. Simultaneously, high NDVI
values cause the peak in predawn water content to
shift to significantly higher AGDD than NDVI peak.
All 24 AMSR-E pixel located within Sand Hills and
five adjacent ecoregions were studied as
demonstrated in Figure 2. The peak occurrences in
Accumulated Growing Degree-Days (AGDD) are shown
in Figure 4.
FIGURE 7. Soil moisture and vegetation water
content products (afternoon acquisitions)
responding to an extreme precipitation event.

• Significant correspondence of temporal evolution
  of daily vegetation water content retrievals and
  the land surface phenology as captured by 16 day
  composites of MODIS NDVI was found for areas
  dominated by herbaceous vegetation.
• The sensitivity of the vegetation water content
  is limited for MODIS NDVI gt 0.5. Predawn and
  afternoon VWC retrievals provided important
  information about diel changes in the canopy
  water content.
• A high correspondence between vegetation water
  content and soil moisture was evident following
  an extreme precipitation event.
• The rapidity of the exponential drydown was
  modulated by land cover type.

5. CONCLUSIONS
FIGURE 4. NDVI and peak occurrences of VWC and
NDVI in 2003. (Stations arranged from North to
South.)

6. ACKNOWLEDGMENTS
This research was supported in part by the NSF
Sand Hills Biocomplexity Project and the USDA RMA
Grasslands Ecological Monitoring System project.
AMSR-E data were acquired from the National Snow
and Ice Data Center.
Chapman, S. S., Omernik, J.M., Freeouf, J.A.,
Huggins, D.A., McCauley, J.R., Freeman, C. C.,
Steinauer, G., Angelo R.T. and Schlepp, R.L.
(2001). Ecoregions of Nebraska and Kansas (2
sided color poster with map, descriptive text,
summary tables, and photographs). 11,800,000.
U.S. Geological Survey, Reston, VA.. de Beurs,
KM, and Henebry, GM. (2005). A statistical
framework for the analysis of long image time
series. International Journal of Remote Sensing,
26(8) 1551-1573. Jackson, T., Chen, D., Cosh,
M., Li, F., Anderson, M., Walthall, C.,
Doraiswamy, P., Hunt, E.R. (2004). Vegetation
Water Content Mapping Using Landsat Data
Normalized Difference Water Index (NDWI) For Corn
And Soybean. Remote Sensing Of Environment.
92(4)475-482. Njoku, E. (2004, updated daily),
AMSR-E/Aqua Daily L3 Surface Soil Moisture,
Interpretive Parms, QC EASE-Grids V001, March
to June 2004. Boulder, CO, USA National Snow and
Ice Data Center. Digital media. Wang, J. R.
(1992). An overview of the measurements of soil
moisture and modeling of modeling of moisture
flux in FIFE. Journal of Geophysical Research,
97(D17)18,955-18,959.
8. REFERENCES

7. CONTACT INFO
Marcela Doubková , M.A. student,
(mdoubkova_at_calmit.unl.edu) Geoffrey M. Henebry,
Ph.D., (Geoffrey.Henebry_at_sdstate.edu)