Despite decades of climate change research, we still lack an understanding of how temperature changes will alter community dynamics. This is particularly the case when considering seasonal patterns of variation. Climate-induced changes in phenology have been well-documented, but there is still much to learn about the broader-scale community effects. One challenge is that changing temperatures can influence community dynamics through multiple pathways. Metabolic changes at the level of individuals scale to alter population dynamic rates and community interactions. This can lead to deterministic changes, but also stochastic ones, as relative abundances in small populations tend to exhibit greater ecological drift that generates spatial and temporal variation independent of environmental heterogeneity. Thus, changes in community composition may deviate from those predicted by seasonality. Here, we present a temperature-dependent, stochastic community model. Each simulated species is assigned a “body size” that determines how temperature sets key demographic processes such as reproduction, survival, and competition strengths. The model is parameterized with data from year-long spatial surveys of tropical and temperate stream macroinvertebrates, allowing us to explore how communities might respond to changes in different seasonal thermal regimes. Our analyses focus on temporal β-diversity to quantify compositional changes through time - a key indicator of community assembly processes. We have previously shown that simulated warmer communities tend to have smaller sizes compared to colder ones because warmer temperatures and accelerated metabolic rates lead to lower survival and carrying capacities. This in turn leads to greater compositional variability, reflected in higher temporal β-diversity. Here, we show that β-diversity in warmer temperature communities exhibits lower tracking of seasonal variation than in temperature ones. Preliminary analyses suggest that these differences arise from lower demographic responsiveness to temperature variation in combination with greater stochastic variation. Dispersal recruitment reduces stochastic variation by increasing community sizes across temperatures, also reducing differences in temperature tracking. Our results represent a step towards a framework for predicting changes in community composition in response to changing seasonal temperature regimes across ecosystems.