Soil temperature (ST) plays a key role in the processes and functions of almost all ecosystems, and is also an essential parameter for various applications such as agricultural production, geothermal development, and their utilization. Although numerous machine learning models have been used in the prediction of ST, and good results have been obtained, most of the current studies have focused on daily or monthly ST predictions, while hourly ST predictions are scarce. This paper presents a novel scheme for forecasting the hourly ST using weather forecast data. The method considers the hourly ST prediction to be the superposition of two parts, namely, the daily average ST prediction and the ST amplitude (the difference between the hourly ST and the daily average ST) prediction. According to the results of correlation analysis, we selected nine meteorological parameters and combined two temporal parameters as the input vectors for predicting the daily average ST. For the task of predicting the ST amplitude, seven meteorological parameters and one temporal parameter were selected as the inputs. Two submodels were constructed using a deep bidirectional long short-term memory network (BiLSTM). For the task of hourly ST prediction at five different soil depths at 30 sites, which are located in 5 common climates in the United States, the results showed the method proposed in this paper performs best at all depths for 30 stations (100% of all) for the root mean square error (RMSE), 27 stations (90% of all) for the mean absolute error (MAE), and 30 stations (100% of all) for the coefficient of determination (R²), respectively. Moreover, the method adopted in this study displays a stronger ST prediction ability than the traditional methods under all climate types involved in the experiment, the hourly ST produced by it can be used as a driving parameter for high-resolution biogeochemical models, land surface models and hydrological models and can provide ideas for an analysis of other time series data.

Accurate estimation of solar radiation is of great significance for solar energy development and climate research. However, in China, the scarcity and uneven distribution of observation stations often cause deep learning models to overfit and suffer from accuracy degradation under small-sample conditions. To address this issue, this paper proposes a deep learning framework that integrates transfer learning and multi-scale time series modeling for predicting hourly global solar radiation at target meteorological sites. The method employs representation learning and clustering to select source domain sites with similar climatic characteristics. It integrates wavelet transform convolution, depthwise separable convolution, and a Transformer encoder–decoder to achieve multi-scale feature extraction and long-term dependency modeling. Experimental results demonstrate that the model achieved a coefficient of determination (R²) of 0.9710 in tests conducted in the Ningxia region. It maintained good predictive performance even in a cold-start scenario with only one month of training data and exhibited stable accuracy across all four seasons, effectively mitigating seasonal bias. This provides a reliable solution for solar radiation estimation in data-scarce regions, and its modeling approach can also be extended to other climate-related time series prediction tasks.