Proceedings Article | 23 October 2012
KEYWORDS: Soil science, Microwave remote sensing, Backscatter, Radar, Data modeling, Microwave radiation, Dielectric polarization, Correlation function, Reflection, Polarization
This study develops a new algorithm for estimating bare surface soil moisture using combined active / passive
microwave remote sensing on the basis of TRMM (Tropical Rainfall Measuring Mission). Tropical Rainfall
Measurement Mission was jointly launched by NASA and NASDA in 1997, whose main task was to observe the
precipitation of the area in 40 ° N-40 ° S. It was equipped with active microwave radar sensors (PR) and passive sensor
microwave imager (TMI). To accurately estimate bare surface soil moisture, precipitation radar (PR) and microwave
imager (TMI) are simultaneously used for observation. According to the frequency and incident angle setting of PR and
TMI, we first need to establish a database which includes a large range of surface conditions; and then we use Advanced
Integral Equation Model (AIEM) to calculate the backscattering coefficient and emissivity. Meanwhile, under the
accuracy of resolution, we use a simplified theoretical model (GO model) and the semi-empirical physical model
(Qp Model) to redescribe the process of scattering and radiation. There are quite a lot of parameters effecting
backscattering coefficient and emissivity, including soil moisture, surface root mean square height, correlation length,
and the correlation function etc. Radar backscattering is strongly affected by the surface roughness, which includes the
surface root mean square roughness height, surface correlation length and the correlation function we use. And
emissivity is differently affected by the root mean square slope under different polarizations. In general, emissivity
decreases with the root mean square slope increases in V polarization, and increases with the root mean square slope
increases in H polarization. For the GO model, we found that the backscattering coefficient is only related to the root
mean square slope and soil moisture when the incident angle is fixed. And for Qp Model, through the analysis, we
found that there is a quite good relationship between Qpparameter and root mean square slope. So here, root mean square
slope is a parameter that both models shared. Because of its big influence to backscattering and emissivity, we need to
throw it out during the process of the combination of GO model and Qp model. The result we obtain from the combined
model is the Fresnel reflection coefficient in the normal direction gama(0). It has a good relationship with the soil
dielectric constant. In Dobson Model, there is a detailed description about Fresnel reflection coefficient and soil moisture.
With the help of Dobson model and gama(0) that we have obtained, we can get the soil moisture that we want. The
backscattering coefficient and emissivity data used in combined model is from TRMM/PR, TMI; with this data, we can
obtain gama(0); further, we get the soil moisture by the relationship of the two parameters-- gama(0) and soil moisture.
To validate the accuracy of the retrieval soil moisture, there is an experiment conducted in Tibet. The soil moisture data
which is used to validate the retrieval algorithm is from GAME-Tibet IOP98 Soil Moisture and Temperature Measuring
System (SMTMS). There are 9 observing sites in SMTMS to validate soil moisture. Meanwhile, we use the SMTMS soil
moisture data obtained by Time Domain Reflectometer (TDR) to do the validation. And the result shows the comparison
of retrieval and measured results is very good. Through the analysis, we can see that the retrieval and measured results in
D66 is nearly close; and in MS3608, the measured result is a little higher than retrieval result; in MS3637, the retrieval
result is a little higher than measured result. According to the analysis of the simulation results, we found that this
combined active and passive approach to retrieve the soil moisture improves the retrieval accuracy.
