437
Microwave Vegetation Indices from Satellite Passive Microwave Sensors
penetration of the canopy and sensitivity to both the leafy and woody parts
of the vegetation. It was shown that the MVIs can have a large dynamic
range for a given NDVI value, which results from differences in vegetation
fractional cover, structure, size, and water content or wet biomass, especially due to the differences in the woody part of the vegetation that optical
sensors have virtually no sensitivity to. Based on these results, we concluded
that the microwave vegetation indices derived here can provide new and
complementary information (to NDVI) on vegetation that can improve our
ability to monitor global vegetation and ecosystem properties from space.
Due to the sensor limitations (frequencies of AMSR-E) in penetrating different vegetation covers and the complexity of the earth’s surface, we realize that the
microwave vegetation indices need further study. We have only evaluated MVIs in a
qualitative way by comparing them with NDVI measurements. There are likely to be
further caveats that have not been discovered yet. The following section summarizes
potential issues or weakness of the MVIs and their computation:
• In deriving MVIs, we clearly stated the assumption that there was insignificant polarization dependence on vegetation at the frequencies and scales of
AMSR-E. At coarse resolution, this assumption is very reasonable for most
land surfaces. However, this needs to be further examined for vegetation
types with a preferred orientation structure such as in forests.
• The technique is based on the specification that both frequencies in the
pair can penetrate the vegetation cover. For other cases, the values-derived
MVIs are expected to have noncomparable values and unreliable estimations when one or both of frequencies cannot penetrate the vegetation cover
because of the sensor limitation. As a result, the relation between MVIs
and vegetation is not monotonic, as demonstrated in Section 18.3.2. This
limitation will be less significant when lower frequency sensors become
available, that is, SMOS and SMAP.
• As part of the data processing in deriving MVIs, we used a very simple
technique (a median filter in the time domain) to reduce the uncertainties
caused by the atmospheric effects. The significance and implications of
atmospheric effects needs further study.
• The evaluation of applying the MVIs was demonstrated qualitatively, in
a manner similar to the interpretation of NDVI. Further studies utilizing
modeling and field verifications are needed for the quantitative descriptions
on how to separate the vegetation signals between leafy and woody parts
and how to derive the important useful vegetation parameters such as biomass or other properties.
REFERENCES
Asrar, G., Fuchs, M., Kanemasu, E. T., and Hatfield, J. H. (1984). Estimating absorbed photosynthetic radiation and leaf area index from spectral reflectance in wheat. Agronomy
Journal, 76, 300–306.
Microwave Vegetation Indices from Satellite Passive Microwave Sensors
penetration of the canopy and sensitivity to both the leafy and woody parts
of the vegetation. It was shown that the MVIs can have a large dynamic
range for a given NDVI value, which results from differences in vegetation
fractional cover, structure, size, and water content or wet biomass, especially due to the differences in the woody part of the vegetation that optical
sensors have virtually no sensitivity to. Based on these results, we concluded
that the microwave vegetation indices derived here can provide new and
complementary information (to NDVI) on vegetation that can improve our
ability to monitor global vegetation and ecosystem properties from space.
Due to the sensor limitations (frequencies of AMSR-E) in penetrating different vegetation covers and the complexity of the earth’s surface, we realize that the
microwave vegetation indices need further study. We have only evaluated MVIs in a
qualitative way by comparing them with NDVI measurements. There are likely to be
further caveats that have not been discovered yet. The following section summarizes
potential issues or weakness of the MVIs and their computation:
• In deriving MVIs, we clearly stated the assumption that there was insignificant polarization dependence on vegetation at the frequencies and scales of
AMSR-E. At coarse resolution, this assumption is very reasonable for most
land surfaces. However, this needs to be further examined for vegetation
types with a preferred orientation structure such as in forests.
• The technique is based on the specification that both frequencies in the
pair can penetrate the vegetation cover. For other cases, the values-derived
MVIs are expected to have noncomparable values and unreliable estimations when one or both of frequencies cannot penetrate the vegetation cover
because of the sensor limitation. As a result, the relation between MVIs
and vegetation is not monotonic, as demonstrated in Section 18.3.2. This
limitation will be less significant when lower frequency sensors become
available, that is, SMOS and SMAP.
• As part of the data processing in deriving MVIs, we used a very simple
technique (a median filter in the time domain) to reduce the uncertainties
caused by the atmospheric effects. The significance and implications of
atmospheric effects needs further study.
• The evaluation of applying the MVIs was demonstrated qualitatively, in
a manner similar to the interpretation of NDVI. Further studies utilizing
modeling and field verifications are needed for the quantitative descriptions
on how to separate the vegetation signals between leafy and woody parts
and how to derive the important useful vegetation parameters such as biomass or other properties.
REFERENCES
Asrar, G., Fuchs, M., Kanemasu, E. T., and Hatfield, J. H. (1984). Estimating absorbed photosynthetic radiation and leaf area index from spectral reflectance in wheat. Agronomy
Journal, 76, 300–306.