One of the recent challenges in population ecology is to identify the spatial scales at which bottom-up and top-down processes operate. However, few studies have examined scale-dependent effects of these processes by using hierarchical (multi-level) models which integrate multiple spatial scales. We focused on the spatial variation in temporal dynamics of the larval frogs (mainly Japanese Tree Frog Hyla japonica) in rice fields during the flooding period (from late May to late June). We hypothesized that the initial density and subsequent population changes in the larval frogs are affected by both bottom-up (flooding management and density dependence) and top-down forces (abundance of Dojo Loach Misgurnus anguillicaudatus and Intermediate Egret Egretta intermedia). Forty-eight fields were sampled with spatially-nested hierarchical design. The effect of each process was estimated by using hierarchical state-space models. Results showed that bottom-up and top-down forces were prominent at the field (50-100 m) and the larger (200-300 m) spatial scales, respectively. This study showed the usefulness of multi-level models to integrate scale-dependent effects of various processes on population dynamics in the field.