In eastern Siberia, topography controls the abundance of the larch forest via both drought and flooding stresses. For the reconstruction of these topographical effects, we modified a dynamic vegetation model to represent soil water relocation owing to within-grid heterogeneity of elevation, over-wet-kill of trees, and air temperature differences within-grid. After calibration, the model reasonably reconstructed the geographical distributions of observation-based-estimates of fundamental properties of plant productivity and thermo-hydrology. Thus, the model appropriately responded to environmental gradients in eastern Siberia. The modified model also partially reconstructed the topography control on tree abundance and thermo-hydrology status in eastern Siberia, although its geographical distribution was not always good. In the modified model, soil water redistribution increased the risk of over-wet-kill in lower elevation classes, whereas it reduced the risk of over-wet-kill for larch trees in higher elevation classes. We demonstrated that without considering the latter effect, forest collapse due to over-wet stress would happen throughout eastern Siberia under a forecasted climatic condition during the 21st century, which will deliver a much moister environment throughout eastern Siberia. Therefore, modeling the over-wet-kill of trees without considering topographical heterogeneity would result in the overestimation of forest collapse caused by the over-wet-kill of trees under an expected climate trend in eastern Siberia.