Spatiotemporal analysis and deep-learning based forecasting of land surface temperature in the UNESCO world heritage sinharaja rainforest

A deep-learning framework using Convolutional Long Short-Term Memory (ConvLSTM) has successfully forecasted temperature dynamics in the Sinharaja Rainforest, a critical UNESCO World Heritage Site and biodiversity hotspot. The model achieved high predictive accuracy with a root mean square error (RMSE) of 1.55°C, capturing complex, nonlinear spatiotemporal patterns from two decades of satellite-derived Land Surface Temperature (LST) data (2000–2021). While the nonparametric Mann-Kendall test did not indicate a statistically significant trend, time series decomposition revealed a subtle but persistent long-term warming signal. This study provides essential insights for climate adaptation and conservation policy decisions in tropical ecosystems, where accurate temperature prediction is challenging due to sparse ground observations and strong spatial heterogeneity.

Silva, K. I., Fonseka, P. U., Gamage, S., Kantamaneni, K., & Rathnayake, U. (2026). Spatiotemporal analysis and deep-learning based forecasting of land surface temperature in the UNESCO world heritage sinharaja rainforest. Trees, Forests and People, 25, 101308. https://doi.org/10.1016/j.tfp.2026.101308