IEEE Geoscience and Remote Sensing Magazine - September 2015 - 56

from fused MODIS-Landsat reflectance time series in
the red and NIR bands during the 2011 growing season
in central Iowa (subset of Landsat path 26 and row 31).
As shown in the 2011 cropland data layer (lower right
image), this area is a rain-fed agricultural region and
mostly planted corn and soybeans. Other natural vegetation types include deciduous forest and grasslands. The
corn and soybean fields are normally rotated in consecutive years. Different temporal trends of NDVI can be observed for different surface types during the growing season. Deciduous forest (in northeast of the image) shows
an earlier green-up, followed by corn and soybeans. Most
of the crop fields show maximum NDVI around August
1. Crop NDVI values drop quickly in early October as the
crops senesce.
These daily NDVI time-series were analyzed using
the TIMESAT phenology program [34]. Crop phenological metrics such as the dates of the start of the season
and end of the season were extracted and compared to
crop progress reports collected by the USDA National
Agricultural Statistics Service (NASS). USDA NASS collects crop progress and condition data reported at the
county by trained observers. These individual reports
are aggregated over a larger area (state or sub-state) before public release. High correlations were found between
the NASS-reported soybean emergence date and the start
of the season from TIMESAT. The end of the season date
agrees with the reported harvest time [20]. By using the
data fusion approach and connecting the remote sensing

detected phenology metrics to the crop growth stages, we
will be able to map crop growth at field scales and provide useful information for monitoring crop conditions
and forecasting yields.

Cumulative ET (mm)

C. Daily FielD-sCale evapotranspiration
Evapotranspiration (ET) describes the net exchange of water vapor between the Earth's surface and the atmosphere.
Over land, ET is comprised primarily of water transpired
by vegetation and water evaporated directly from the soil
or leaf surfaces. Remotely sensed maps of ET can provide
spatial and temporal information about how water is consumptively used over landscapes, and can be a valuable
tool for managing freshwater resources at field to basin
scales. In particular, land surface temperature retrieved
from thermal infrared (TIR) satellite imagery has proven
to be a useful diagnostic of surface moisture conditions
and evaporative fluxes [35].
For agricultural water management, ET information is
most beneficial at sub-field scales (~100 m resolution or
finer) and daily timesteps [36]. Unfortunately, TIR imagery
is not routinely available globally with this spatiotemporal
frequency. A single fully functional Landsat provides images at 16-day intervals, or 8 days in the overlap regions.
However, in regions of persistent cloudiness, we may be
fortunate to acquire 2 or 3 clear scenes during a growing
season. To improve temporal sampling in seasonal water
use curves at Landsat scale, STARFM has been employed
to fuse 1-km ET retrievals from MODIS TIR imagery at
approximately daily timesteps with periodic
Landsat-based retrievals at 30-m resolution,
using TIR imagery sharpened to the short10
wave band resolution [7]. While STARFM
8
was originally designed for fusing reflectance
data, experiments have demonstrated utility
6
for fusing ET data as well.
4
The ET fusion approach has been evaluated over several agricultural landscapes in
2
comparison with ground-based measureRAIN
0
ments collected at micrometeorological flux
150
160
170
180
190
200
210
220
230
240
tower sites in both irrigated and rainfed crop300
ping systems [22], [23]. The greatest benefit of
Reference ET
STARFM over a Landsat-only interpolation to
250
Observed ET
daily timesteps has been observed when sigLandsat Retrievals
200
Landsat-Only
nificant precipitation occurs between succesLandsat-MODIS Fusion
150
sive clear-sky Landsat overpasses, resulting in
100
a discontinuous change in soil moisture and
evaporative flux. Assuming the rainfall event
50
is reasonably resolved at the 1-km scale, MO0
150
160
170
180
200
210
220
230
240
190
DIS TIR is able to convey timely information
Day of Year
regarding the change in moisture status (expressed by a depressed surface temperature
signal) during the gap between Landsat overFigure 7. Daily (top) and cumulative (bottom) ET retrieved using Landsat-MODIS
passes (Fig. 7). Landsat, in turn, provides high
data fusion and with a Landsat-only interpolation compared to observed fluxes.
The fused timeseries better captures ET response to the rainfall event around day of spatial frequency structure to the resulting
fused ET timeseries.
year 190.
56

ieee Geoscience and remote sensing magazine

september 2015

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - September 2015