In light of climate change and biodiversity loss, modeling and mapping soil moisture at high spatiotemporal resolution is increasingly crucial for a wide range of applications in Earth and environmental sciences, particularly in boreal forests, which play a key role in global carbon cycling, are highly sensitive to hydrological changes, and are experiencing rapid warming and more frequent disturbances. However, modeling and mapping soil moisture dynamics is challenging due to the nonlinear interactions among numerous physical and biological factors and the wide range of spatial and temporal scales at play. This study aims to identify key spatial and temporal controls on soil moisture using an empirically based modeling approach. We focused on a boreal forest landscape in northern Sweden, where we monitored surface soil moisture with dataloggers at 78 locations during the summer of 2022. We investigated the relationships between observed soil moisture variations and numerous environmental and meteorological predictors from multiple sources at varying spatial resolutions and temporal scales, and we assessed how these relationships changed over time. Spatial variation in soil moisture was influenced not only by topography and by the spatial resolution used to represent it, but also by soil properties, vegetation, and land use/land cover (LULC). In addition, the relative importance of these factors changed over time, with topography generally explaining more spatial variation during wet periods, while soil and vegetation were more relevant during dry periods. This suggests that current soil moisture maps relying primarily on topographic indices could benefit from integrating soil, vegetation, and LULC information to better capture spatial variability under different wetness conditions, as well as from selecting the optimal spatial resolution for the specific area of interest. Temporal variation in soil moisture was better explained by hydrological and meteorological variables averaged over 5 to 7 d preceding soil moisture measurements, highlighting the importance of accounting for both lagged and cumulative effects of weather conditions. Field predictors generally outperformed remote sensing and modeled predictors, indicating that soil moisture mapping based solely on spatially continuous predictors requires improving spatial detail of maps describing soil texture, structure, and organic matter content. Our findings contribute to improving the accuracy and interpretability of data-driven methods, such as machine learning, for mapping soil moisture across space and time for forest management and nature conservation.