Multi-hazard improves app retention: Comparison of alerting and attrition for the multi-hazards SD Emergency and the single-hazard QuakeAlert

One of the targets of the Sendai Framework for Disaster Risk Reduction is to ‘substantially increase the availability of and access to multi-hazard early warning systems and disaster risk information and assessments to people by 2030’. A key problem in implementation of warning systems is retention: ensuring users retain the technological pathways for receiving warnings. We investigated trends in the installation and uninstallation of the SD Emergency multi-hazard alerting app and whether they were related to alerting patterns in the County of San Diego, California, USA. Evaluation of alerts from 2017-2023 revealed a wide range of alert types with fire being the most common, accounting for more than half of alerts. We found a temporal correlation between alerting with SD Emergency installation rates, with clear peaks in installation rates that matched the timing of alerts. Individuals were ∼4 times more likely than the median (33 installs/day) to install the app during an alerting period than outside of one. Uninstallation rates had only a modest increase with the number of alerts and were far outpaced by the increase in installation rates as the number of alerts increased. This finding indicates that overalerting may not be an issue as previously thought. Most importantly, the SD Emergency app maintained a retention rate of ∼55%, which was ∼15% higher than the single-hazard alerting app QuakeAlert . Higher retention rates for multi-hazard alerting apps improve the likelihood that people will retain other hazard alerting capabilities, especially for low frequency, high impact events like earthquakes or tsunami. • Fire is responsible for over 50% of alerts for County of San Diego, California, USA • App installation rates closely mirror the timing of hazard events • Multi-hazard apps retain users at a 15% higher rate compared to single-hazard apps • Higher retention rates improve the likelihood that people will receive an alert for low-frequency, high-impact events, like earthquakes