Soil moisture plays a key role in agriculture, drought monitoring, and developing resilience to extremes. Accurately measuring and predicting how much water is in the soil helps farmers use water more efficiently and prepares communities for dry conditions. In this study, we use a computer model called pyWBM to estimate soil moisture in Illinois. To improve the model’s accuracy, we apply machine learning, which is a way for computers to learn patterns from data. We trained the model using a combination of satellite data, weather records, and real-world measurements from soil sensors. By adjusting how the model represents water movement through soil and the effects of meteorology, we were able to reduce prediction errors by 15 percent. We also tested different ways of training the model to better capture local soil and meteorological differences. While the model does well overall, it still struggles to detect very dry conditions during droughts. Our results show that combining satellite and ground data with machine learning can improve soil moisture tracking. This approach could help support better water management and drought response.