Time series modeling is a way to predict future values by examining temporal data. The present study analyzes the monthly mean soil moisture data at various depths: surface, profile, and root soil moisture, spanning from 1981 to 2022. The analysis employs two distinct approaches: the statistical seasonal autoregressive integrated moving average (SARIMA) and a deep learning long short-term memory (LSTM). The models are trained on a data set, covering the period from 1981 to 2021, acquired from the agricultural site at Andhra Loyola College in Vijayawada, Andhra Pradesh, India. Subsequently, the data from 2021 to 2022 is reserved for testing purposes. The study provides comprehensive insights into the design of both SARIMA and LSTM models, along with an evaluation of their performance using established error metrics such as the model mean absolute percentage error (MAPE), mean absolute error (MAE), and root mean squared error (RMSE). In the context of surface soil moisture prediction, the LSTM model demonstrates superior performance compared to SARIMA. Specifically, LSTM achieves a notably lower MAPE of 0.0615 in contrast to SARIMA’s 0.1541, a reduced MAE of 0.0316 compared to 0.0871, and a diminished RMSE of 0.0412 as opposed to 0.1021. This pattern of enhanced accuracy persists across profile and root soil moisture predictions, further establishing LSTM’s supremacy in predictive capability across diverse soil moisture levels.