Using Microwave Radar for Soil Moisture Inversion Under Soybean Canopies

The successful radar mapping of soil moisture under vegetation canopies would be a boon to agriculture, global climate change studies, water resource management and other areas. Toward this end, a series of polarimetric radar measurements of soybeans over the entire growing season of 1996 were made at Land C-band at the Long Term Ecological Research site at the Kellogg Biological Station, Hickory Corners, MI. The radar data obtained has been successfully modeled with a simplified radiative transfer expression, including the dependence of the VV, HH, and cross-polarized backscattering coefficients on the angle of incidence, vegetation water mass, vegetation height, and volumetric soil moisture. In addition, a simple inversion model based on the forward model has been obtained which yields vegetation water mass and volumetric soil moisture for select polarization and frequency combinations. DESCRIPTION OF MEASUREMENTS During the summer of 1996, an extensive set of radar measurements were conducted at the Long-Term Ecological Research site of the Kellogg Biological Station in Hickory Corners, MI, for the purpose of developing algorithms which would relate the soil moisture content of agricultural fields to the microwave radar backscatter of those fields. Three primary types of crops were chosen for measurement: corn, which represents agricultural fields in which a stem or stalk is a dominant feature at microwave frequencies, and soybeans and alfalfa, which represent agricultural fields which lack a dominant stem. In addition, measurements were also made of a field which has lain fallow for many years and is now populated with many native species. This report is restricted to the analysis of the soybean data. Target characteristics Soybean measurements commenced on 20 Jun 1996 and were completed on 26 Oct 1996, after the harvest for the season. During this period, the vegetation fresh biomass ranged from negligible, during the early summer and post-harvest, to a maximum of 1.337 kg/m on 30 Aug. The volumetric soil moistures ranged from 0.036 cc/cc on 5 Sep to a maximum of 0.249 cc/cc on 28 Sep. Despite a relatively wet spring and fall, western Michigan was plagued with a drought in the midst of the summer, resulting in lower biomass values than normal for the region. Radar characteristics and techniques Polarimetric radar measurements were made at 1.25 GHz (Lband) and 5.4 GHz (C-band). The choice of frequencies was based on previous observations that the L-band response was sensitive to both the soil moisture and biomass of vegetated surfaces, and the C-band response was largely dependent only on the biomass, and the hope that together, unequivocal inversion of measurements for soil moisture would be possible. The angle of incidence was restricted to 45 from nadir in order to reduce the volume of the required dataset to an obtainable size. This is an angle at which radar backscatter is sensitive to the primary ground parameters of soil moisture and biomass, and also is amenable to SAR deployment for large scale observations. Special care was taken to reduce the effect of the agricultural row structure on the radar backscatter. Anomalous scattering is observed parallel and perpendicular to the row directions, with relatively smooth transition halfway between these two extremes [1, 2]. Therefore, the platform on which the antennas were mounted was translated adjacent to the fields, and the antennas were always azimuthally oriented at 45 with respect to the row direction and at a range of 12 m to the ground. To reduce fading, the number of independent samples was kept to at least 205 at L-band and 157 at C-band for each measurement. The Single Target Calibration Technique [3] was used to correct raw radar data to differential radar cross sections. Calibrations were performed prior to any measurements (but only after the system had stabilized), and again after the measurements for the day were completed. Calibration precision is estimated at dB for co-polarized backscatter, dB for cross-polarized backscatter. A season-long systematic bias appears to be present in the data, but its presence does not effect the conclusions about the accuracy with which a radar sensor can invert soil moisture. This bias remains under investigation. A CLOUD MODEL FOR BACKSCATTERING The scattering mechanisms of the agricultural canopies are modeled as a cloud of water vapor over a rough ground [4]. This model is similar to the Michigan Microwave Canopy Scattering (MIMICS) model [5], developed for predicting backscatter from forests. The model for the rough ground backscattering component is an empirical model for bare rough ground as derived by Oh et al.[6]. Various interaction terms between the ground and the canopy “cloud” are included. The model predicts the scattering to depend on the values of 0.01 0.1 1 -13. -12. -11. -10. -9. -8. -7. -6. -5. -4. 0.02 0.05 0.2 0.5 2 C -b an d xp ol ra tio σ 0 vh /σ 0 v v (d B ) Vegetation water mass mw (kg/m ) mv=0.01 cc/cc mv=0.05 cc/cc mv=0.10 cc/cc mv=0.15 cc/cc mv=0.20 cc/cc mv=0.25 cc/cc mv=0.30 cc/cc Figure 1: C-band cross-polarization ratio for soybeans. The curves represent the model developed in the previous section for different values of soil moisture. The points are the data used to develop the model coefficients. the area density of the vegetation water mass, the volumetric soil moisture, the soil rms surface height. This model also contains a number of free parameters, describing the strength of the various scattering and extinction mechanisms. These free parameters were chosen to provide the best fit of the model to the measured backscatter. With the model coefficients found, the polarizations and polarization ratios are found which have the largest sensitivity to biomass while insensitive to soil moisture, as well as the opposite. The polarization ratio with the largest sensitivity to biomass is the C-band HV to C-band VV ratio, and the model predictions as well as the data used to derive the model coefficients are shown in Fig. 1. The polarization ratio with the largest sensitivity to soil moisture while relatively insensitive to biomass is the L-band HV to C-band HV ratio. It is shown in Fig. 2. It is important to note that the data itself has more sensitivity to the soil moisture, especially at low moistures, than is predicted by the model derived from the data.