The estimation of reservoir temperature is critical in geothermal exploration for the classification and utilisation of geothermal systems. A marginal increase in reservoir temperature predictive accuracy has the potential to significantly improve the efficiency of geothermal energy exploration and production. The goal of this study was to create five machine learning models to predict reservoir temperature using a hydrogeochemical dataset from Western Anatolia geothermal systems, in Turkey. The models developed include natural gradient boosting (NGB), extreme learning machine (ELM), group method of data handling (GMDH), generalised regression neural network (GRNN) and back propagation neural network (BPNN). The NGB model proved superior over the remaining models by obtaining the highest R2 of 0.9959 and the lowest RMSE and MAE values of 4.5938 and 3.9678, respectively. Further, this study utilised a model explainable technique known as Shapley additive explanation (SHAP) to evaluate why certain predictive decisions are reached by the NGB model in order to convey trust and to foster its adoption in real-time applications. By assessing the relationship between hydrogeochemical variables and reservoir temperature, the SHAP interpretation revealed that SiO2 was the most important variable in predicting reservoir temperature. It was also found an increase in SiO2 concentration leads to increased temperature and vice versa. The high predictive performance and the consistency between the SHAP interpretation and empirical knowledge indicate the reliability of the NGB model for estimating reservoir temperature.