Soil Moisture Effects on Afternoon Precipitation Occurrence in Current Climate Models

Soil moisture-precipitation feedbacks in a large ensemble of global climate model simulations are evaluated. A set of three metrics are used to assess the sensitivity of afternoon rainfall occurrence to morning soil moisture in terms of their spatial, temporal, and heterogeneity characteristics. Positive (negative) spatial feedback indicates that the afternoon rainfall occurs more frequently over wetter (drier) land surface than its surroundings. Positive (negative) temporal feedback indicates preference over temporally wetter (drier) conditions, and positive (negative) heterogeneity feedback indicates preference over more spatially heterogeneous (homogeneous) soil moisture conditions. We confirm previous results highlighting a dominantly positive spatial feedback in the models as opposed to observations. On average, models tend to agree better with observations for temporal and heterogeneity feedback characteristics, although intermodel variability is largest for these metrics. The collective influence of the three feedbacks suggests that they may lead to more localized precipitation persistence in models than in observations. Plain Language Summary Not only does rainfall influence soil moisture, but soil moisture can also actively influence rainfall. Current climate models do not represent such two-way relationships correctly, mainly due to uncertainty in the latter. Our understanding of models' weaknesses in simulating these processes is relatively low, and this is the focus of this study. Here we investigate how afternoon rainfall occurrence is affected by morning soil moisture conditions from three perspectives: relative soil moisture of the region where it rains compared to (1) surrounding regions (spatial feedback), (2) its long-term mean (temporal feedback), and (3) the spatial heterogeneity of soil moisture (heterogeneity feedback). In models, the afternoon rainfall preferably occurs over regions that are wetter than their surroundings, as opposed to observations. Models show better agreement with observations in the temporal and heterogeneity feedback, but large differences across the models remain. We suggest that the combined effect of these three relationships in models may contribute to their biases in the persistence of precipitation.