UC Irvine

Abstract Soil moisture‐precipitation coupling critically shapes Earth's water and energy cycles, yet it remains difficult to quantify because land‐atmosphere interactions are nonlinear, multivariate, and state‐dependent. This study applies the High‐Dimensional Model Representation (HDMR) framework to observation‐based data sets over the contiguous United States to diagnose temporal SM‐precipitation amount coupling, defined here as the influence of morning land‐atmosphere states on same‐day afternoon rainfall. HDMR decomposes structural, correlative, and cooperative controls of morning precursors and shows that traditional linear approaches, such as correlation analysis, underestimate coupling strength, explaining only 6%–8% of precipitation variance in regions where HDMR identifies soil moisture (SM) contributions of up to about 20%. The first‐order decomposition reveals a direct wet‐soil advantage in the eastern Great Plains, while convective available potential energy (CAPE) over Texas and land‐surface temperature over Louisiana dominate in separate instability‐ and surface‐temperature‐controlled regimes, respectively. Beyond direct effects, the strongest second‐order signal is a localized SM‐CAPE interaction over southern Oklahoma, where dry soils combined with high CAPE enhance afternoon precipitation through a thermodynamic‐triggering pathway. Thus, positive and negative SM effects can coexist over the Great Plains within temporal precipitation‐amount coupling through distinct physical pathways. Finally, using the observation‐based HDMR diagnostics as a benchmark, we evaluate the convection‐permitting CONUS404 simulation. CONUS404 qualitatively reproduces the main functional structures of SMPC, including the wet‐soil first‐order response and dry‐soil‐high‐CAPE synergy, but systematically underestimates their coupling strength. These findings underscore the need for nonlinear diagnostics to improve the representation of sub‐daily land‐atmosphere coupling in weather and climate models. Plain Language Summary Soil moisture plays a critical role in local weather by controlling the release of heat and water vapor into the atmosphere. However, because these land‐atmosphere interactions are complex, standard statistical tools often struggle to measure them accurately. We apply a nonlinear functional decomposition method to observations across the United States to quantify how morning soil conditions influence afternoon rainfall. We then use these results to evaluate whether convection‐permitting models accurately simulate these relationships. We find that soil moisture explains up to 20% of rainfall variability compared to only 6%–8% estimated by traditional methods. While we identify distinct regions where either wet or dry soils favor rainfall, the convection‐permitting model correctly simulates the patterns but systematically underestimates their strength. These results suggest that current models undervalue the land surface's influence on precipitation. Capturing these complex interactions is essential for improving precipitation forecasts and climate projections. Key Points High Dimensional Model Representation disentangles structural, correlative, and cooperative effects in soil moisture‐precipitation coupling Observations reveal wet‐soil coupling in the eastern Great Plains and dry‐soil, thermally driven coupling over southern Oklahoma Convection‐permitting simulations reproduce the spatial sign of coupling but systematically underestimate the coupling strength