Integrating Landsat with MODIS Products for Vegetation Monitoring - Satellite-based Applications on Climate Change - page 248

Fig. 16.5 MODIS surface reflectance ( a ) and Landsat ETM + surface reflectance derived from
the reference-based approach ( b ) and LEDAPS physical approach ( c ) for Washington DC area
(Landsat WRS-2 path 15 and row 33). Black areas represent clouds, cloud shadows, and missing
or poor quality data
The basis of this approach is that homogeneous pixels of the same land cover
type have the same surface reflectance regardless of patch size and that the seasonal
and bidirectional reflectance changes of those pixels should also be the same for
different patch sizes given each land cover type does not split into two or more
types (1-to- n ) during a short period. The relationship of each land cover type
between acquisition date and target date remain approximately same for different
resolution images. Therefore, the relationships built on the MODIS data and the
aggregated medium resolution data can be applied to medium resolution data to
produce surface reflectance at a target date (e.g., MODIS acquisition date).
Differing from traditional empirical approaches that need to measure surface
reflectance on the ground, this approach takes ground “truth” from MODIS data.
The high-quality MODIS data represent one of the most accurate data records we
achieved today and are thus an appropriate data source to use as a reference data set.
The approach uses the cluster-based pure homogeneous pixels as samples; this
allows the approach to be generalized and deal with various seasonal changes from
different surface types and can be extended to other biophysical parameters.
Figure 16.5 shows the Landsat surface reflectances derived from the physical-
based LEDAPS approach (Fig. 16.5c ) and the reference-based normalized approach
(Fig. 16.5b ). They are very similar in surface reflectance composition from near-
infrared, red and green bands.
The normalization approach can be used to normalize satellite images from
different acquisition dates to a close and clear MODIS acquisition date. Figure 16.6
illustrates the processing result over Chesapeake Bay. In the test, 14 Landsat Global
Land Survey (GLS)-2000 data were used. Although the GLS-2000 Landsat scenes
have been selected from growing seasons, the differences are inevitable due to the
different acquisition dates. Figure 16.7a is the mosaic of Landsat surface reflectance
produced by the LEDAPS system. The differences of seasonality are obvious on
this map. However, those differences have been reduced in the mosaic map of the
normalized Landsat images by using MODIS surface reflectance as a correction
reference. The remaining differences in water body are due to the lack of samples
from MODIS NBAR product.