Soil respiration is expected to increase with rising global temperatures but the degree of response may depend on soil moisture and other local factors. Experimental climate change studies from single sites cannot discern whether an observed response is site-dependent or generalizable. To deconvolve site-specific vs. regional climatic controls, we examined soil respiration for 18 months along a 520 km climate gradient in three Pacific Northwest, USA prairies that represents increasingly severe Mediterranean conditions from north to south. At each site we implemented a fully factorial combination of 2.5–3 °C warming and 20% added precipitation intensity. The response of soil respiration to warming was driven primarily by the latitudinal climate gradient and not site-specific factors. Warming increased respiration at all sites during months when soil moisture was not limiting. However, these gains were offset by reductions in respiration during seasonal transitions and summer drought due to lengthened periods of soil moisture limitation. The degree of this offset varied along the north–south climate gradient such that in 2011 warming increased cumulative annual soil respiration 28.6% in the northern site, 13.5% in the central site, and not at all in the southern site. Precipitation also stimulated soil respiration more frequently in the south, consistent with an increased duration of moisture limitation. The best predictors of soil respiration in nonlinear models were the Normalized Difference Vegetation Index (NDVI), antecedent soil moisture, and temperature but these models provided biased results at high and low soil respiration. NDVI was an effective integrator of climate and site differences in plant productivity in terms of their combined effects on soil respiration. Our results suggest that soil moisture limitation can offset the effect of warming on soil respiration, and that greater growing-season moisture limitation would constrain cumulative annual responses to warming.