Characterizing the spatio-temporal behavior of extreme events remains statistically challenging since they are inherently rare. This study quantifies the response of temperature and precipitation extremes across Spain to global warming using a hierarchical Bayesian framework that jointly estimates spatially varying model parameters and their uncertainty across a dense network of weather stations over 1981–2022. Summer daily temperature extremes (TXx) increased markedly with global warming, especially in central and western Spain, whereas autumn daily precipitation extremes (Rx1day) showed no detectable response. Therefore, TXx required a non-stationary model, whereas Rx1day could be fairly described assuming stationarity. Concerning spatial dependence, two strategies were compared: a two-step approach combining local model fitting with universal kriging, and a fully spatial Bayesian model, with the latter providing better-calibrated predictions and a more coherent quantification of uncertainty, especially for the non-stationary TXx model. Temperature and precipitation extremes also showed contrasting geographic patterns: TXx magnitude decreased with elevation and coastal proximity, though its interannual variability peaked along coasts. Conversely, both the mean intensity and variability of Rx1day increased with elevation and closeness to the sea. These contrasting behaviors have direct implications for future risk. Under a scenario of 2.5 °C of global warming, 100-year return levels for summer TXx are projected to approach 50 °C across southwestern Spain, whereas the selected stationary model estimates autumn Rx1day return levels exceeding 200 mm along the Mediterranean coast. Climate adaptation planning must therefore address both a rapidly intensifying heat signal and a precipitation hazard that remains severe.