Determining a relationship between measured ground soil moistures and SAR data in the Tigris basin and estimating the soil moistures on the basis of this relationship

The remote sensing technology is used in many areas such as determining the land parts using on the earth, monitoring the rapid changes and obtaining instant records in lands, detecting and usage of natural resources. Due to the increasing of population and agricultural areas, the capacity of water resources will not be sufficient for irrigation. Therefore; the remote sensing technology will be needed either now or in the future to ensure proper information about using of water resources in the agricultural lands. Since the water content in the soil significantly affects the backscattering coefficient of the soil, the relationship between ground soil moisture measurements and the remote sensing data enables the soil moisture estimation in a short period of time. Moreover, the use of SAR sensors in soil moisture estimation is more appropriate because these sensors operating in the microwave range of electromagnetic spectrum which is sensitive to changes in the soil moistures. Therefore, SAR based Radarsat-2 was used in this study for soil moisture estimation. The main purpose of the thesis study is to determine a relationship between the ground soil moisture measurements and Radarsat-2 data; estimating the soil moisture over bare and/or vegetated agricultural areas on the basis of the determined relationship. The study consists of four stages. In the first stage; the Radarsat-2 data was obtained at different dates and the ground measurements were carried out simultaneously with the Radarsat-2 data acquisition. In the second phase; the Radarsat-2 data has been pre-processed and the GPS coordinates of the points where each soil sample was taken were transferred to this data. After pre-processing step; the standard sigma backscattering coefficients with the Generalized Freeman Durden and H/A/α polarimetric decomposition models were utilized to extract feature vectors and a feature vector with 10 backscattering coefficients was formed for each pattern. In the last phase, a nonlinear machine learning model: Generalized Regression Neural Network (GRNN) was used to estimate the regional soil moisture content from the obtained feature vectors. As a result of the study, the proposed system performed good results for single C-band SAR data over the bare and vegetated agricultural fields. Moreover, the results showed that the radar is a powerful remote sensing tool for the soil moisture estimation, with mean absolute errors around 2.31%, 2.11 and 2.10 vol.% on datasets 1-3, respectively; and 2.46 %, 2.70 %, 7.09%, and 5.70 vol.% on datasets 1&2, 2&3, 1&3, and 1&2&3, respectively.