Precision agriculture delivers a promising solution to address water, food, and environmental security concerns. Accurate soil moisture prediction is crucial for calculating optimal irrigation timing and water quantity, to prevent issues related to excessive or insufficient watering. It also leads to an augmentation in the quantity of cultivated regions, hence enhancing air purification and agricultural productivity. In situ, soil moisture measuring techniques are characterized by being point measurements that are time-consuming, laborious, labor-intensive, and costly. Recent progress in artificial intelligence and satellite data possesses the capacity to enhance the efficiency of soil moisture prediction. Nevertheless, analyzing soil moisture using synthetic-aperture radar (SAR) imagery and ground data remains challenging due to higher computational time and lower accuracy. A novel deep learning model known as the efficient convolutional neural network for SAR (ECNN-SAR) data processing is proposed. The ECNN-SAR model aimed to provide the automatic and efficient prediction of soil moisture while lowering the computational burden. The ECNN-SAR architecture takes Sentinel-1 images of different polarisations. The investigation of soil moisture is conducted using Sentinel-1 SAR data in the C-band and with multiple polarization. The proposed CNN model consists of four convolutional layers using different activation functions and regularization techniques. Activation functions and regularization handle vanishing gradient and overfitting issues. After four convolutional layers, the fully connected layer is linked to the multi-class classifier in the classification layer. The ECNN-SAR model is evaluated using real-time Sentinel-1 SAR data from different stations in Maharashtra state, India. The results reveal the efficiency of the proposed model over the existing methods.