Application of Seasonal Climate Forecasts to Predictions of Regional Crime Anomalies in the United States

Recent studies have found consistent evidence of a relationship between seasonal temperature anomalies and crime rates at a variety of spatial scales. While some papers have produced future projections of changes in crime rates attributable to climate change, none have attempted to develop shorter-term (seasonal) predictions that leverage existing seasonal forecasting capacity and may be of more immediate application in law enforcement.

Seasonal predictions of near-surface air temperature produced as reforecasts of the National Centers for Environmental Prediction (NCEP) Climate Forecast System model (CFS) were first compared to output from the NCEP North American Regional Reanalysis (NARR) to examine the skill of the CFS reforecasts as relevant to this application. Next, data from the FBI Uniform Crime Reporting Database (UCR) and the NARR were combined to empirically determine the sensitivity of crime rates to temperatures at a regional level using the recently developed methodology of Harp and Karnauskas (GeoHealth, in revision). The regional-level empirical models were applied to CFS monthly temperature predictions at forecast lead times of one, three, and six months to generate experimental hindcasts of temperature-driven anomalies in crime, which are then compared to actual crimes as recorded in the UCR database for verification. Further development and testing of such an applied prediction system has the potential to be beneficial to law enforcement operations and decision-making.