413
Microwave Vegetation Indices from Satellite Passive Microwave Sensors
It is well known that passive microwave sensors are sensitive to variations in
vegetation properties in a relatively thick layer of the canopy. However, current satellite sensors such as Advanced Microwave Scanning Radiometer (AMSR-E), the
Tropical Rainfall Measuring Mission Microwave Imager, and WindSat have coarse
spatial resolutions (a few tens of kilometers). In comparison with optical sensors,
passive microwave sensors can be used both day and night, can penetrate clouds (all
weather), and are less affected by atmospheric conditions. Collectively, these satellites have a long period of record and might be useful as either a primary or complementary tool for assessing the impacts of global climate change on carbon cycling
and ecosystem variability, or vice versa.
Deriving vegetation information from passive microwave instruments has been
explored in previous studies. Some of the earliest investigations (Choudhury and Tucker
1987; Choudhury et al. 1987) showed that microwave polarization difference tempera; Choudhury et al. 1987) showed that microwave polarization difference temperaet al. 1987) showed that microwave polarization difference temperashowed that microwave polarization difference temperatures (MPDT) at 37 GHz were highly correlated to NDVI in arid and s emiarid regions
and related to variations in leaf water content (Pampaloni and Paloscia 1986; Kerr and
Njoku 1990; Jackson and Schmugge 1991; Le Vine and Karam 1996; Njoku and Li
1999). Based on the microwave radiative transfer theory and field measurements, we
know that MPDT is affected not only by the vegetation properties but also by the suraffected not only by the vegetation properties but also by the surby the vegetation properties but also by the surbut
surface effective reflectivity (soil moisture and roughness) and the physical temperature.
To minimize the physical temperature effects, Becker and Choudhury (1988)
p roposed the normalized microwave polarization difference index C ∙ (T Bv – T Bh )/
(T Bv + T Bh ) for a given frequency, where C is a scale factor. Here T Bv and T Bh are the
brightness temperature for horizontal h and vertical (v) polarization. This is also
referred to as the normalized polarization index (PI; Paloscia and Pampaloni 1992).
They derived a microwave vegetation index based on the difference in normalized
brightness temperature (normalized by thermal infrared measurements) at two frequencies, ∆Tn = Tn(f 2 ) – Tn(f 1 ). This approach was used for detecting the biomass
and water conditions of agricultural crops using data at 10 and 36 GHz. The impact
of physical temperature on vegetation properties derived using this technique is
minimal (Justice et al. 1989; Paloscia and Pampaloni 1992; Njoku and Chan 2006).
The microwave vegetation indices described above can be useful if all other perturbing factors are uniform. However, depending on the sensor frequencies and the
level of vegetation present, they can be significantly affected by soil emission variations resulting from soil moisture and surface roughness conditions. This problem can
limit the value of such a product in global vegetation monitoring. In a recent study,
Njoku and Chan (2006) developed a combined vegetation and surface roughness
parameter using multitemporal AMSR-E data analyses. Most of the variation of this
estimated parameter could be attributed to changes in vegetation water content.
Another index, defined as the microwave emissivity difference vegetation index
(EDVI), 2(T Bp (f 1 ) – T Bp (f 2 ))/(T Bp (f 1 ) + T Bp (f 2 )), was proposed by Min and Lin (2006).
It was intended for application to dense forest conditions using 19- and 37-GHz
observations, where both measurements do not “see” the ground surface. It was demonstrated that the EDVI was more sensitive to and correlated with evapotranspiration
than the NDVI and that it could be used to estimate turbulent flux.
The advantages and disadvantages of the optical-based NDVI and microwavederived vegetation indices in monitoring vegetation properties (MPDT and PI) were
Microwave Vegetation Indices from Satellite Passive Microwave Sensors
It is well known that passive microwave sensors are sensitive to variations in
vegetation properties in a relatively thick layer of the canopy. However, current satellite sensors such as Advanced Microwave Scanning Radiometer (AMSR-E), the
Tropical Rainfall Measuring Mission Microwave Imager, and WindSat have coarse
spatial resolutions (a few tens of kilometers). In comparison with optical sensors,
passive microwave sensors can be used both day and night, can penetrate clouds (all
weather), and are less affected by atmospheric conditions. Collectively, these satellites have a long period of record and might be useful as either a primary or complementary tool for assessing the impacts of global climate change on carbon cycling
and ecosystem variability, or vice versa.
Deriving vegetation information from passive microwave instruments has been
explored in previous studies. Some of the earliest investigations (Choudhury and Tucker
1987; Choudhury et al. 1987) showed that microwave polarization difference tempera; Choudhury et al. 1987) showed that microwave polarization difference temperaet al. 1987) showed that microwave polarization difference temperashowed that microwave polarization difference temperatures (MPDT) at 37 GHz were highly correlated to NDVI in arid and s emiarid regions
and related to variations in leaf water content (Pampaloni and Paloscia 1986; Kerr and
Njoku 1990; Jackson and Schmugge 1991; Le Vine and Karam 1996; Njoku and Li
1999). Based on the microwave radiative transfer theory and field measurements, we
know that MPDT is affected not only by the vegetation properties but also by the suraffected not only by the vegetation properties but also by the surby the vegetation properties but also by the surbut
surface effective reflectivity (soil moisture and roughness) and the physical temperature.
To minimize the physical temperature effects, Becker and Choudhury (1988)
p roposed the normalized microwave polarization difference index C ∙ (T Bv – T Bh )/
(T Bv + T Bh ) for a given frequency, where C is a scale factor. Here T Bv and T Bh are the
brightness temperature for horizontal h and vertical (v) polarization. This is also
referred to as the normalized polarization index (PI; Paloscia and Pampaloni 1992).
They derived a microwave vegetation index based on the difference in normalized
brightness temperature (normalized by thermal infrared measurements) at two frequencies, ∆Tn = Tn(f 2 ) – Tn(f 1 ). This approach was used for detecting the biomass
and water conditions of agricultural crops using data at 10 and 36 GHz. The impact
of physical temperature on vegetation properties derived using this technique is
minimal (Justice et al. 1989; Paloscia and Pampaloni 1992; Njoku and Chan 2006).
The microwave vegetation indices described above can be useful if all other perturbing factors are uniform. However, depending on the sensor frequencies and the
level of vegetation present, they can be significantly affected by soil emission variations resulting from soil moisture and surface roughness conditions. This problem can
limit the value of such a product in global vegetation monitoring. In a recent study,
Njoku and Chan (2006) developed a combined vegetation and surface roughness
parameter using multitemporal AMSR-E data analyses. Most of the variation of this
estimated parameter could be attributed to changes in vegetation water content.
Another index, defined as the microwave emissivity difference vegetation index
(EDVI), 2(T Bp (f 1 ) – T Bp (f 2 ))/(T Bp (f 1 ) + T Bp (f 2 )), was proposed by Min and Lin (2006).
It was intended for application to dense forest conditions using 19- and 37-GHz
observations, where both measurements do not “see” the ground surface. It was demonstrated that the EDVI was more sensitive to and correlated with evapotranspiration
than the NDVI and that it could be used to estimate turbulent flux.
The advantages and disadvantages of the optical-based NDVI and microwavederived vegetation indices in monitoring vegetation properties (MPDT and PI) were