AIGR-LSTM: Land Surface Temperature Interpolation Using Adaptive Inductive Graph Representation Learning

Land surface temperature (LST) plays an important role in many fields including urban heat island analysis, climate change studies, and agricultural disaster monitoring. However, achieving high-resolution and continuous LST data remains challenging due to the limitations of remote sensing observations and the sparse distribution of ground sensors. Recently, spatiotemporal kriging has been widely used to address such challenges, but often relies on fixed neighborhood locations and typically ignores temporal dependencies in LST data. To address these challenges, we propose AIGR-long short-term memory (LSTM), an adaptive inductive graph representation learning model based on LSTM for LST interpolation. To better determine the spatial proximity relationship between nodes, we propose a threshold-based adaptive neighbor selection (ANS) module, which effectively captures complex spatial dependencies by adaptively selecting relevant neighbors. Besides, we design a geography-based spatiotemporal modeling (GSM) module that incorporates geographic spatial relationships, combines temporal encoding information with spatial features, and leverages LSTM to achieve accurate spatiotemporal interpolation. The experiment was conducted on LST and three other datasets, and the results showed that AIGR-LSTM outperformed the six state-of-the-art methods in all evaluation metrics. On the LST dataset, the root mean square error (RMSE) was $0.185~^{\circ }$ C and the mean absolute error (MAE) was $0.133~^{\circ }$ C, which improved by 6.09% and 8.28%, respectively, compared with the best baseline model.