Abstract
<title>Abstract</title> <p>Southern Florida (SoFL) wet-season rainfall is shaped by coastal geometry, land–sea interactions, and large-scale moisture convergence, making it a challenging target for climate models and downscaled products. This study evaluates SoFL wet-season precipitation in HighResMIP coupled and uncoupled simulations, the MESACLIP CESM-HR ensemble, and three statistically downscaled CMIP6 products—BCCAQ, LOCA2, and NEX-GDDP-CMIP6—against PRISM and ERA5 during historical (1981–2005) and recent SSP5-8.5 (2015–2024) periods. Statistically downscaled products best reproduce the observed annual cycle and wet-season rainfall magnitude, although some methods show reduced daily variability and frequency–intensity tradeoffs. HighResMIP coupled simulations capture wet-season onset but underestimate rainfall amplitude and intensity, while uncoupled simulations produce broader rainfall variability and a more realistic wet-season demise but retain coastal dry biases. MESACLIP shows the largest errors, including weak seasonal amplitude, delayed transitions, and excessive low-intensity rainfall. Across GCM-based categories, rainfall occurs too frequently but with insufficient intensity, especially in high-rainfall coastal regions. These results indicate that nominal 25-km resolution alone is insufficient for representing SoFL wet-season precipitation, highlighting the need for improved convection, air–sea coupling, and mesoscale land–sea interaction processes.</p>