Surface Soil Moisture Retrieval over Heterogeneous Agricultural Plots Using Sar Observations

Abstract

Soil moisture is a basic component of meteorological cycle and in the determination of agricultural crop yield. Spatial information about soil moisture over agricultural crops is required for efficient irrigation, which in turn helps in saving water and increases the crop yield. However, capturing spatiotemporal field measurement of soil moisture is time consuming and not a practical approach. Synthetic Aperture Radar (SAR) remote sensing is a valuable tool for retrieving surface soil moisture over agricultural fields owing to its great sensitivity to surface soil moisture. The objective of the research is retrieval surface soil moisture over typical heterogeneous agricultural plots of a semi-arid region of India using C and L band polarized SAR data. A methodology is developed to retrieve surface soil moisture over different agricultural fields at different crop stages. To implement the methodology, a typical agriculture-dominated landscape has been selected. For the study, different agricultural plots of Malavalli village in Karnataka, were selected. Agricultural crops include; crops like Paddy, Tomato, Maize, Sugarcane and a reference bare field. Agricultural plots of size 1 acre approximately, were selected and sampling grids were made according to SAR ground resolutions. Field measured data like surface soil moisture, surface roughness, soil texture, vegetation height and vegetation water content were collected from every grid of the agricultural plots in synchronization with satellite pass. Sentinel-1a, C-band data and ALOS PALSAR-2, L-band SAR data products are used to retrieve surface soil moisture. The developed models were compared with existing models and validated using field measure values. Surface soil moisture was retrieved using L-band SAR across agricultural plots at two distinct crop stages. Initially, processed SAR images are decomposed using Freeman Durden, Yamaguchi and Van-Zyl decomposition techniques to know the major scattering components (like surface, dihedral, and volume scattering). In vegetative crop stage, surface scattering (>34%) is dominating scattering component, which shows less interaction of vegetation with radar backscattering energy. iSurface scattering component of Yamaguchi decomposition has dependence on field measured surface soil moisture with R2> 0.5 good correlation. Multilinear regression (MLR) is carried out in which soil moisture (Mv) is a dependent variable and 𝑠𝑢𝑟𝑓𝑎𝑐𝑒 , 𝑉𝐻−𝑉𝑉 and 𝐷𝑖ℎ𝑒𝑑𝑟𝑎𝑙 are considered as independent variables and validated. To assess the resilience of the developed models, it is compared with existing models like Oh 1992, Oh 2004, X-Bragg and WCM. RMSE of developed model varies from 0.82 to 2.51 cm3/ cm3 for two distinct crop stages. Whereas, in case of sugarcane at grand growing stage none of models performed well (RMSE= 3.644.7 % gm/ cm3). X-Bragg model is underestimating surface soil moisture in two distinct crop stages of paddy, maize, tomato and sugarcane field plots (RMSE= 1.214.23 % gm/ cm3). In the same way, surface soil moisture is retrieved using C-band SAR across above mentioned agricultural plots for whole crop cycle of each crop at an interval of 12 days. Each crop cycle is divided into vegetative, maturity, yield formation stage and surface soil moisture of each crop stage is estimated. The relationship between backscattered energy and soil moisture, roughness and vegetation parameter (RVI) is analyzed and MLR analysis is carried out to develop semi empirical model (SEM) and validated against grid sampled field data (RMSE= 1.38.1 % gm/cm3). The developed model found to be better when compared with Oh model, 1994. In grand growing stage of sugarcane and yield formation stage of maize and sugarcane, the RMSE values were found to 4.18.1 % gm/cm3. Which shows the vegetation attenuation increased as the crop matures and affecting soil moisture retrieval beneath it. Performance of C-band dual polarized data with L-band quad polarized data at two different crop stages were compared for surface soil moisture retrieval. Quad polarized data is found to performing better than dual polarized data. At various crop stages, the proposed semi-empirical model for retrieving surface soil moisture functions effectively. In future, the developed model can be simplified by introducing constant parameters based on crop stage and type of crop. This study helps to understand the spatial variation of soil moisture within the small plots thus helping marginal farmers and local irrigation departments for better allocation of water resources.