We use neural networks and large climate model ensembles to explore predictability of internal variability in sea surface temperature anomalies on interannual (1-3 year) and decadal (1-5 and 3-7 year) timescales. We find that neural networks can skillfully predict SST anomalies at these lead times, especially in the North Atlantic, North Pacific, Tropical Pacific, Tropical Atlantic and Southern Ocean. The spatial patterns of SST predictability vary across the nine climate models studied. The neural networks identify “windows of opportunity” where future SST anomalies can be predicted with more certainty. Neural networks trained on climate models also make skillful SST predictions in reconstructed observations, although the skill varies depending on which climate model the network was trained. Our results highlight that neural networks can identify predictable internal variability within existing climate datasets and show important differences in how well patterns of SST predictability in climate models translate to the real world.

• Neural networks can learn predictable signals of internal sea surface temperature variability at 1-3, 1-5, and 3-7 year lead times • Neural networks trained on climate model output can skillfully predict sea surface temperature variability in reconstructed observations • Neural network skill in predicting observed sea surface temperature variability depends on the climate model used for training

Soil moisture influences near-surface air temperature by partitioning downwelling radiation into latent and sensible heat fluxes, through which dry soils generally lead to higher temperatures. The strength of this coupled soil moisture-temperature (SM-T) relationship is not spatially uniform, and numerous methods have been developed to assess SM-T coupling strength across the globe. These methods tend to involve either idealized climate-model experiments or linear statistical methods which cannot fully capture nonlinear SM-T coupling. In this study, we propose a nonlinear machine learning-based approach for analyzing SM-T coupling and apply this method to various mid-latitude regions using historical reanalysis datasets. We first train convolutional neural networks (CNNs) to predict daily maximum near-surface air temperature (TMAX) given daily SM and geopotential height fields. We then use partial dependence analysis to isolate the average sensitivity of each CNN’s TMAX prediction to the SM input under daily atmospheric conditions. The resulting SM-T relationships broadly agree with previous assessments of SM-T coupling strength. Over many regions, we find nonlinear relationships between the CNN’s TMAX prediction and the SM input map. These nonlinearities suggest that the coupled interactions governing SM-T relationships vary under different SM conditions, but these variations are regionally dependent. We also apply this method to test the influence of SM memory on SM-T coupling and find that our results are consistent with previous studies. Although our study focuses specifically on local SM-T coupling, our machine learning-based method can be extended to investigate other coupled interactions within the climate system using observed or model-derived datasets.

Heat related illnesses are one of the leading causes of weather-related mortality in the United States, and heat extremes continue to increase in frequency and duration. Public health interventions include population mobility, including travel to central cooling centers or wellness checks on vulnerable populations. Using anonymized cellphone data from Safegraph’s neighborhood patterns dataset and gridded temperature data from gridMET, we explored the mobility-temperature relationship in the San Francisco Bay Area at fine spatial and temporal scale. We leveraged spatial variability in median income and temporal variability in COVID-19 related policies across two summers (2020-2021) to analyze their influence on the mobility-temperature relationship. We completed quantile regressions for a dataset stratified by income and year. We found that mobility increased at a higher rate with higher temperatures in 2020 than 2021. However, in 2021, the relationship reversed for several wealthier income groups, where mobility decreased with higher temperatures. We then augmented the analysis and calculated a panel regression with fixed effects to characterize the mobility-temperature relationship while controlling for temporal and spatial variability. This analysis suggested that all areas exhibited lower mobility with higher summer temperatures. However, similar to the results of the quantile regression, the rate of decrease in mobility in response to high temperature was significantly greater among the wealthiest census block groups compared with the least wealthy. Given the fundamental difference in the mobility response to temperature across income groups, our results are relevant for heat mitigation efforts in highly populated regions in current and future climate conditions.