Determination of Soil Properties Utilizing C-Band Synthetic Aperture Radar (SAR) in Southern Kirkuk Governate, Northern Iraq

This study utilizes Synthetic Aperture Radar (SAR) images to assess and monitor the soil properties. To achieve this, it provides an efficient method for soil resource evaluation in the designated area to increase the information accessibility and minimize the costs and efforts. Sentinel-1 C-SAR images (VV, VH), were collected on 6-Oct-2024 in Taza Khurmatu district, southern Kirkuk, Northern Iraq, which were used as ancillary data. Sentinel-1 C-SAR was joined with field-soil data and machine learning was utilized to classify different kinds of soil properties. The effectiveness of SAR data to predict the soil texture, salt content, and organic properties was confirmed. With this technique, significantly less in-field sampling is required, while it is cost-effective and flexible. Its novelty lies in the discovery that radar data closely correlate with the soil characteristics in a region where this relationship had not previously been explored. Several soil parameters were examined and correlated during the study. These parameters include, texture, organic matter content, salinity, pH, and lime content. The former directly influence the SAR backscatter. The results showed that the study area consists of fine materials where silt appears most frequently, followed by clay, while the sand and gravel contents are minimal. The soil characteristics indicate effective moisture retention, but the sand portion needs evaluation, regarding the drainage performance. However, a medium to low level of plasticity was observed from the plastic and liquid limits. The multiple linear regression analysis revealed strong correlations between the calculated regression coefficients, which ranged from 0.590 to 0.930, and the recorded chemical and physical data. A Unified Soil Classification System (USCS) identified eight soil types throughout the region based on the clay, silt, sand, and gravel content. In addition, C-band radar data were effectively used to predict other soil properties through the K-Dimensional Tree (KD-Tree), K-Nearest Neighbor (KNN), Random Forest (RF), and Minimum Distance (MD) classifiers. The study findings demonstrate that field-based algorithm classification methods can be created from C-band SAR observations. However, this method's suitability for different geographic areas remains untested.