Forests interact with the local climate through a variety of biophysical mechanisms. Observational and modelling studies have investigated the effects of forested vs. non-forested areas, but the influence of forest management on surface temperature has received far less attention owing to the inherent challenges to adapt climate models to cope with forest dynamics. Further, climate models are complex and highly parameterized, and the time and resource intensity of their use limit applications. The availability of simple yet reliable statistical models based on high resolution maps of forest attributes representative of different development stages can link individual forest management practices to local temperature changes, and ultimately support the design of improved strategies. In this study, we investigate how forest management influences local surface temperature (LSTs) in Fennoscandia through a set of machine learning algorithms. We find that more developed forests are typically associated with higher LST than young or undeveloped forests. The mean multi-model estimates from our statistical system can accurately reproduce the observed LST. Relative to the present state of Fennoscandian forests, fully develop forests are found to induce an annual mean warming of 0.26 °C (0.03/0.69 °C as 5th/95th percentile), and an average cooling effect in the summer daytime from -0.85 to -0.23 °C (depending on the model). On the contrary, a scenario with undeveloped forests induces an annual average cooling of -0.29 °C (-0.61/-0.01 °C), but daytime warming in the summer that can be higher than 1 °C. A weak annual mean cooling of -0.01 °C is attributed to forest harvest from 2015 to 2018, with an increased daytime temperature in summer of about 0.04 °C. Overall, this approach is a flexible option to study effects of forest management on LST that can be applied at various scales and for alternative management scenarios, thereby helping to improve local management strategies with consideration of effects on local climate.