Today, estimating soil moisture using the SMAP satellite with a temporal resolution of less than three days has become feasible; however, the coarse spatial resolution of its pixel size limits its applicability at the field scale. In this study, a downscaling approach was applied by integrating SMAP soil moisture data with Sentinel-1 SAR and MODIS products. For this purpose, soil moisture samples in maize fields were collected from five farms located in the Megsal agricultural complex in the Qazvin Plain during the summer growing season of 2022. The Random Forest (RF) technique was employed to estimate soil moisture due to its strong capability in handling nonlinear relationships and reducing prediction errors. Accordingly, six modeling scenarios were developed based on different combinations of remote sensing indices. The results showed that among the evaluated scenarios in field M1, Scenario RF6 exhibited the highest accuracy with R = 0.97 and RMSE = 0.04 cm³/cm³, followed by Scenario RF5 with R = 0.96. The improved performance of RF5 and RF6 can be attributed to the inclusion of the Leaf Area Index (LAI) and land use map, as LAI demonstrated a strong correlation with measured soil moisture (R = 0.88). Considering the limited availability of point-based soil moisture observations and the need for continuous spatial monitoring in small agricultural fields, applying downscaling approaches can significantly enhance the spatial accuracy of SMAP data and serve as an efficient and cost-effective tool for agricultural research and management.