Microwave remote sensing provides a direct measurement of the surface soil moisture for a range of vegetation cover conditions. Two basic approaches are used, passive and active. In passive methods, the natural thermal emission of the land surface (or brightness temperature) is measured at microwave wavelengths using very sensitive detectors. Only passive microwave methods are treated in this chapter. Chapter 8 provides details on active microwave systems.
The microwave region of the electromagnetic spectrum consists of frequencies between 0.3 and 30 GHz. This region is subdivided into bands, which are often referred to by a lettering system. Some of the relevant bands that are used in earth remote sensing are K (18-27 GHz), X (8-12 GHz), C (4-8 GHz), and L (1-2 GHz). Frequency and wavelength are used interchangeably. Wavelength (in centimeters) approximately equals 30 times frequency (in GHz). Within these bands only small ranges exist that are protected for scientific applications, such as radio astronomy and passive sensing of the earth's surface. A general advantage of microwave sensors, in contrast to visible and infrared, is that observations can be made through cloud cover because the atmosphere is nearly transparent, particularly at frequencies <10 GHz. In addition, these measurements are not dependent on solar illumination and can be made at any time of the day or night.
Microwave sensors operating at very low frequencies (<6 GHz) provide the best soil moisture information. At low frequencies, attenuation and scattering problems associated with the atmosphere and vegetation are less significant, the instruments respond to a deeper soil layer, and a higher sensitivity to soil water content is present.
Low-frequency passive sensors provide information on the surface reflectivity. An examination of relationship between reflectivity and soil moisture is essential to estimate soil moisture from microwave remote sensing data.
Soil Moisture and Reflectivity Assuming that the earth is a plane surface with surface geometric variations and volume discontinuities much less than the wavelength, only refraction and absorption of the media need to be considered. This situation permits the use of the Fresnel reflection equations as a model of the system (Ulaby et al., 1986). These equations predict the surface reflectivity as a function of dielectric constant (k) and the viewing angle (0) based on the polarization of the sensor (H = horizontal or V = vertical).
Polarization refers to the orientation of the electromagnetic waves with respect to the surface. The dielectric constant of soil is a composite of the values of its components (air, soil, and water). The basic reason that microwave remote sensing can provide soil moisture information is that there is a large difference between the dielectric constants of water (~80) and the other components (<5).
Based on an estimate of the mixture dielectric constant derived from the Fresnel equations and soil texture information, volumetric soil moisture can be estimated using an inversion of the dielectric mixing model (i.e., Hallikainen et al., 1985). The depth of soil contributing to the measurement is about one-quarter the wavelength (based on a wavelength range of 2-21 cm). As noted above, it is desirable to use low frequencies because the measurement at these frequencies provides more information on the soil column.
Soil Moisture and Brightness Temperature Passive microwave remote sensing uses radiometers that measure the natural thermal microwave emission within a narrow band centered on a particular frequency. The measurement provided is the brightness temperature in degrees Kelvin, TB, which includes contributions from the atmosphere, reflected sky radiation, and the land surface. Atmospheric contributions are negligible at frequencies <10 GHz, and the cosmic radiation contribution to sky radiation has known values that vary only slightly in the frequency range used for observations of soil water content.
The brightness temperature of a surface is equal to its emissivity (e) multiplied by its physical temperature (T).
The emissivity is equal to 1 minus the reflectivity, which provides the link to the Fresnel equations and soil moisture for passive microwave remote sensing. Figure 7.1 illustrates the relationships between emissivity and soil
moisture that can be expected at a high and a low microwave frequency. If the physical temperature is determined independently, the emissivity can be determined from T B. The physical temperature can be estimated using surrogates based on satellite surface temperature, air temperature observations, or model predictions (i.e., Owe and van de Griend, 2001).
Microwave Measurement and Vegetation For natural conditions, a variation in vegetation type and density is likely to be encountered. The presence of vegetation has a major impact on the microwave measurement. Vegetation reduces the sensitivity of the relationship to changes in soil water content by attenuating the soil signal and by adding its own microwave emission to the measurement. This attenuation increases with increasing microwave frequency, which is another important reason for using lower frequencies. Attenuation is characterized by the optical depth of the vegetation canopy. Jackson and Schmugge (1991) presented a method for estimating optical depth that used information on the vegetation type (typically derived from land cover) and vegetation water content, which is estimated using visible or near infrared remote sensing.
Recent efforts to develop research and operational methods for estimating soil moisture retrieval algorithms for the advanced microwave scanning radiometer (AMSR) instruments onboard the NASA Aqua and NASDA ADEOS-II satellites (Njoku et al., 2000) have resulted in the formaliza-tion of several alternative approaches. For the most part, all these methods are based on the same basic relationships but are implemented differently. Series of equations used to estimate soil moisture involve many variables related to frequency, polarization, and viewing angle of the sensor, describing physical temperature and atmospheric profile (Njoku and Li, 1999). These equations are solved using forward calculations of TB or inversions for soil moisture.
Most research and applications involving passive microwave remote sensing of soil moisture have emphasized low frequencies (L band). In this range, it is possible to develop soil moisture retrievals based on a single H polarization observation (Jackson, 1993). It is well known that with H polarization TB is more sensitive to soil moisture than V polarization. This approach relies on ancillary data on temperature, vegetation, land cover, and soils. Atmospheric corrections are assumed to be negligible at these frequencies. The single channel/ancillary data approach has been tested and calibrated using aircraft L-band observations (Jackson et al., 1999) and higher frequency satellite measurements (Jackson, 1997; Jackson and Hsu, 2001). Standard error of estimated values in L-band aircraft experiments were on the order of 3% volumetric soil moisture. The higher frequency satellite studies had larger errors (>5%). As noted previously, the optical depth computation approach used in Jackson and Schmugge (1991) has
a high degree of uncertainty at higher frequencies. In addition, at high frequencies other investigators have found that the single scattering albedo must be considered (Owe et al., 1992; Njoku and Li, 1999).
The weakness of the vegetation correction has led to a reconsideration of the application of the single channel/ancillary data algorithm when using high-frequency microwave data for soil moisture estimation. As a result, several investigators have examined a variation of the approach proposed in Njoku and Li (1999). In this approach, a series of equations for predicting the brightness temperature for several channels is solved iteratively. Several channels (polarizations and/or frequencies) can be used subject to the constraint that the number of unknowns is less than or equal to the number of equations. Based on practical considerations and existing data sources, a soil moisture retrieval algorithm uses the lowest microwave frequency available for both polarizations. Therefore, there will be two independent observations. By controlling and specifying parameters, the unknowns in the radiative transfer equations can be reduced to the dielectric constant and another free variable, such as the vegetation optical depth. A more exhaustive validation of this approach is needed.
Another multiple-channel approach to soil moisture retrieval was proposed by Wigneron et al. (2000). They suggested using measurements made at several viewing angles. A challenge in this approach is obtaining simul taneous, consistent multiple angle measurements. Finally, there have been attempts to use the polarization difference (PD; the difference between the vertical and the horizontal brightness temperature i.e., TV — T^i) for soil moisture retrieval. As described by Njoku et al. (2000), an advantage of this approach is that it normalizes out the physical temperature. However, because it does not use physically based equations, the retrieval ultimately relies on calibration. The polarization difference approach has been explored in a number of studies. These investigations did not involve actual soil moisture but rather related indices.
Microwave Polarization Difference Index and Vegetation Parameters
The microwave polarization difference index (MPDI), an alternative to the polarization difference, is expressed as:
where c is a scaling factor. Tucker (1989) and Tucker and Choudhury (1987) have attempted to exploit high-frequency passive microwave remote sensing in monitoring drought. They used 37-GHz polarization difference data collected by the scanning multifrequency microwave radiometer (SMMR). At this frequency the brightness temperature is dominated by vegetation effects and, in particular, vegetation water content. It was hypothesized that these observations could enhance products based on the normalized difference vegetation index (NDVI; chapters 5 and 6). Comparisons were made between the PD and NDVI for several study areas during periods of drought and nondrought. It was observed that the PD was very sensitive to changes in the NDVI at lower NDVI levels. However, as NDVI increased, the PD saturated at a low value. The polarization difference approaches zero as vegetation level increases.
Teng et al. (1995) summarized much of the work that has been done to relate microwave polarization information to the NDVI. They found the PD to be more sensitive to vegetation cover in regions of sparse vegetation and to NDVI in densely vegetated regions. These authors further used satellite 37-GHz PD data over the U.S. Midwest and found that drought years could be differentiated from normal years during the preplanting and early stages of crop growth. At a point in the growth cycle through harvest there was little information contained in the PD between years or months. The type of crop also had an influence on the potential information available using the PD. They concluded that NDVI provides more information on vegetation conditions; however, the PD provides unique information during periods when the NDVI is not useful.
Teng et al. (1995) concluded that PD and NDVI data are complementary for drought analyses. They noted that in some situations (the 1988 drought in the U.S. Midwest in particular) the NDVI can detect drought early in the season for severely affected regions. However, other regions that are not as severely affected are often not detected until later in the growth cycle. The PD information would be of particular use in these situations.
Was this article helpful?