American Journal for Young Scientists ยท ISSN 3068-9015
When the Night Stops Cooling: Modeling Extreme Warm Nights Across California Weather Stations
Authors: Rohit Krishnan
Published: October 29, 2026
Category: Environmental Science
Abstract
This paper investigates which factors influence extreme warm nights across California using NOAA GHCN-Daily station observations. We defined extreme warm nights using station-specific 95th percentile summer TMIN thresholds based on a historical baseline period. For each station-year, the percentage of valid summer nights classified as extreme warm nights was calculated. A random forest regression model was then used to explore the effect of year, latitude, longitude, elevation, and distance to coast on extreme warm-night frequency. Partial dependence plots and permutation feature importance results suggest that year has a disproportionately large influence on nighttime heat across California, and that it positively influences the percent of extreme warm nights a station experiences. The next most important variable was distance to the coast, which also had a positive influence on the percent of extreme warm nights. These results showcase how both temporal changes and geographic characteristics may play an important role in explaining differences in extreme nighttime heat across California. In terms of predictive power, the random forest regressor resulted in an R2 value of 0.410, an MAE of 2.513, and an RMSE of 3.765 when used to predict on stations within a withheld test set and considerably outperformed a naive baseline. This performance showcases how the selected geographic and temporal variables provide meaningful predictive information. Overall, these findings suggest that station geography and climate context may help identify areas where nighttime heat is more common or changing, while also providing a foundation for further investigation into the environmental factors contributing to extreme warm-night events across California.