Seasonality exerts strong pressures on biodiversity patterns. Yet, temporal beta-diversity is poorly studied in tropical systems, and the drivers of variability in amphibian activity and seasonality remain largely unknown. We quantified intra- and interannual variation in temporal beta-diversity relying on a nine-year, year-round survey (51 species, n > 23,000) performed in a protected area (Betampona, Madagascar). We assessed the dependence on climate of beta-diversity and abundance using a distance-based redundancy analysis and generalised linear mixed models, respectively. Despite the majority of species being preferentially active during one specific period, beta-diversity and abundance were more variable between years than within years. Temporal variation in beta-diversity was best explained by temperature (but climate accounted for only 2% of variation). Species abundance was best explained by temperature (for 32% of the tested species), monthly humidity (30%) and monthly rainfall (24%). We found no climatic dependence for 24% of the species. Our results suggest that studies focusing on species phenology can be misleading when based on single-year surveys even in seasonal systems. The high interannual variability in diversity may be due to an adaptive responses to an important regime of stochastic events. Given the direction of the relationships between weather and abundances, we predict that a large proportion of amphibians would suffer from climate change in Madagascar. We emphasise the need to account for multiple temporal scales in studies of tropical species composition and abundance to better understand species phenology and their response to climate change, and make targeted conservation actions more effective.