436
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
Especially, the MVI-provided woody part of vegetation information can be the complementary information to NDVI in global vegetation monitoring from space.
18.5 CONCLUSIONS
In this chapter we demonstrated a new set of MVIs using AMSR-E observations.
The basis of the approach is that the bare surface emission signals of a surface for
two adjacent AMSR-E frequencies are highly correlated and can be well described
by a linear function with coefficients that only depend on the frequency pair used
and are polarization independent. This important finding then leads to the assumption that the two adjacent AMSR-E frequencies in vegetated surfaces can also be
described as a linear function by canceling out the background emission signal components. The intercept, A, and the slope, B, of this linear function are the MVIs.
This minimizes dealing with background soil emission signal problems encountered
with other microwave indices. We demonstrated that A is positively correlated to
NDVI and that it is affected not only by the vegetation properties but also by the
surface physical temperature. On the other hand, B is negatively correlated to NDVI
and is only affected by the vegetation properties. Both can be directly derived from
AMSR-E measurements under the assumption that there is no significant polarization dependence of the vegetation emission and attenuation properties.
To evaluate the microwave vegetation indices, we compared MVIs with the NDVI
measurements derived from MODIS for the year 2003. Comparisons of the vegetation indices derived from the optical and microwave sensors showed:
• The general global distribution and the seasonal change patterns of the
MVIs derived by the microwave sensor are consistent with those of NDVI
derived by the optical sensor. However, their range of values in a region
(considering the overall range for each index) and responses to seasonal
change can be significantly different. These variations are associated with
land cover type and are due to the differences in sensitivity of the optical
and microwave observations to different parts of vegetation canopy.
• For monitoring vegetation phenology, the MVIs, in general, show less seasonal variability than NDVI, since the woody part of vegetation exhibits
less seasonal variation, while the larger seasonal variations of the NDVI
are mainly due to the seasonal changes of the leafy part of the vegetation.
Our overall impression from the limited analyses conducted here is that the
high frequency pair–derived MVIs are more sensitive to seasonal vegetation change than those derived by the low-frequency pair because the highfrequency microwave signals are more sensitive to the crown and leafy part
of vegetation properties while the low-frequency microwave measurements
are more sensitive to the stems and woody part of vegetation properties.
Therefore, the high frequency pair–derived MVIs are likely to be more
useful in monitoring vegetation phenology.
• It is clear that optically based vegetation indices such as NDVI are responsive
to a thin layer of the canopy (leaves, the living green material). Microwave
sensors can provide significant new vegetation information due to greater
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
Especially, the MVI-provided woody part of vegetation information can be the complementary information to NDVI in global vegetation monitoring from space.
18.5 CONCLUSIONS
In this chapter we demonstrated a new set of MVIs using AMSR-E observations.
The basis of the approach is that the bare surface emission signals of a surface for
two adjacent AMSR-E frequencies are highly correlated and can be well described
by a linear function with coefficients that only depend on the frequency pair used
and are polarization independent. This important finding then leads to the assumption that the two adjacent AMSR-E frequencies in vegetated surfaces can also be
described as a linear function by canceling out the background emission signal components. The intercept, A, and the slope, B, of this linear function are the MVIs.
This minimizes dealing with background soil emission signal problems encountered
with other microwave indices. We demonstrated that A is positively correlated to
NDVI and that it is affected not only by the vegetation properties but also by the
surface physical temperature. On the other hand, B is negatively correlated to NDVI
and is only affected by the vegetation properties. Both can be directly derived from
AMSR-E measurements under the assumption that there is no significant polarization dependence of the vegetation emission and attenuation properties.
To evaluate the microwave vegetation indices, we compared MVIs with the NDVI
measurements derived from MODIS for the year 2003. Comparisons of the vegetation indices derived from the optical and microwave sensors showed:
• The general global distribution and the seasonal change patterns of the
MVIs derived by the microwave sensor are consistent with those of NDVI
derived by the optical sensor. However, their range of values in a region
(considering the overall range for each index) and responses to seasonal
change can be significantly different. These variations are associated with
land cover type and are due to the differences in sensitivity of the optical
and microwave observations to different parts of vegetation canopy.
• For monitoring vegetation phenology, the MVIs, in general, show less seasonal variability than NDVI, since the woody part of vegetation exhibits
less seasonal variation, while the larger seasonal variations of the NDVI
are mainly due to the seasonal changes of the leafy part of the vegetation.
Our overall impression from the limited analyses conducted here is that the
high frequency pair–derived MVIs are more sensitive to seasonal vegetation change than those derived by the low-frequency pair because the highfrequency microwave signals are more sensitive to the crown and leafy part
of vegetation properties while the low-frequency microwave measurements
are more sensitive to the stems and woody part of vegetation properties.
Therefore, the high frequency pair–derived MVIs are likely to be more
useful in monitoring vegetation phenology.
• It is clear that optically based vegetation indices such as NDVI are responsive
to a thin layer of the canopy (leaves, the living green material). Microwave
sensors can provide significant new vegetation information due to greater