Resource selection is fundamental to understanding the ecology of a species because it is directly related to its survival and reproduction. The scale-dependent process for the habitat use of Siberian roe deer (Capreolus pygargus) was investigated in temperate forest to determine their resource priorities. Since imperfect detectability of fecal pellet count method can lead to false zeroes and overdispersion issues, we accounted for these issues using a zero-inflated model. The abundance of roe deer was constant between the deciduous forest and larch plantation. This might be due to the high productivity of temperate forest in the study area. However, the microhabitat use by roe deer differed between the forest types because of resource availability. Roe deer exhibit consistent pattern of microhabitat use in the larch plantations, while they exploited higher food quality in deciduous forests. Due to high resource availability, including food and covers, in larch plantation, habitat use of deer may not be limited by spatial variation on the resources. Moreover, we confirmed that weather, forest type, and ground conditions affected spatiotemporal variation in fecal pellet detection on a habitat and microhabitat scale. The zero-inflated model could make the fecal pellet count method more efficient and reliable without additional field surveys by correcting the detectability of fecal pellets. This simple statistical model can facilitate future research on ungulate and lagomorph populations.