Hurricanes, wind speeds, and 911 call volumes: Helping ECCs predict the impact of natural disasters

Hurricane Milton and Helene hit overnight when 911 call volumes were typically lower, so the increases were more significant. Still, it meant gigantic spikes in volume when ECCs are typically at their lowest staffing levels.

For example, right before midnight on October 9, FL_1673 (Titusville) had a roughly 1000% increase in call volumes. Meanwhile, a similar spike occurred at FL_8243 (Clearwater Police) at around 5 am.

The chart below shows call volume activity at the Titusville 911 Center (FL_1673) from Oct. 8 – Oct. 10.

The chart above shows call volume activity at the Clearwater Police Center (FL_8243) from Oct. 8 – Oct. 10.

Similarly, during Hurricane Helene, ECCs, including Taylor County, Pasco County, Pinellas County Regional 9-1-1, and Melbourne Police Department, saw consistently elevated activity between 10 p.m. and midnight on Sept. 26. Overall, call volumes jumped to nearly eight times their typical amount as Helene made landfall and moved across the state.

The chart below represents the 911 call activity across Emergency Communication Centers (ECCs) located in the vicinity of Hurricane Helene’s landfall. It compares call activity over a 24-hour period in the days leading up to, during, and after the weather event. 

One factor that makes emergency response to a catastrophic storm so challenging is the widespread impact. In large natural disasters like hurricanes, neighboring ECCs often experience a spike in calls and likely can’t assist other 911 centers. For example, while the Largo Police Department saw a significant increase in call volumes around 10 pm on Oct. 9, neighboring Pinellas County Regional 9-1-1 had its own spike to address in the same time frame.

The chart below represents the 911 call activity across Emergency Communication Centers (ECCs) located in the vicinity of Hurricane Milton’s landfall. The time window spans from 6:00 AM on October 8, 2024, to 6:00 AM on October 10, 2024. Each row in the heatmap corresponds to a specific ECC, and the calls are grouped into 15-minute intervals, where darker shades indicate a higher density of wireless 911 calls relative to the ECC’s activity during the day.

Meanwhile, a similar phenomenon happened in Hurricane Helene. Some ECCs saw call volumes jump nearly 40-fold in the hours after the storm made landfall, highlighting the intense demands placed on first responders in a very short amount of time.

The chart below represents the 911 call activity across Emergency Communication Centers (ECCs) located in the vicinity of Hurricane Helene’s landfall. The time window spans from 6:00 AM on September 26, 2024, to 6:00 AM on September 27, 2024.Each row in the heatmap corresponds to a specific ECC, and the calls are grouped into 15-minute intervals, where darker shades indicate a higher density of wireless 911 calls relative to the ECC’s activity during the day. 

These thousands of data points represent critical information that could help ECCs maintain situational awareness and prioritize responses. But emergency responders often don’t have access to this type of intelligence—the Safety Gap. To close it, ECCs need the power of real-time, unified data delivered through a platform that ensures that, even in the event of an outage, operators and safety professionals have what they need for safer, faster, and more effective resolutions.

The availability of crucial information even during natural disasters, when power outages are common, can enhance the speed and efficiency of emergency response, potentially saving lives. It’s how ECCs in Florida protected their communities during one of the most active and damaging hurricane seasons in recent memory.

Armed with unified data sets, ECCs can start exploring using cutting-edge methods to prepare for emergencies. For example, imagine if 911 centers could predict the volume of calls they would receive.

RapidSOS worked with Reask, which has extensive experience creating both forward-looking hazard forecasts (LiveCyc) and retrospective analyses (Metryc) to build an AI/ML model that could predict the timing, magnitude, and spatial pattern of emergency calls up to 72 hours before a hurricane makes landfall. could predict up to 72 hours before the timing, magnitude, and spatial pattern of emergency calls during and after a hurricane.

We were also able to identify the likely reason for several deviations that will help us improve the predictions moving forward. For example, our estimates slightly overpredict the call spike across Florida. There could be many reasons for volumes coming in lower than expected, including better preparation on the part of residents and a larger number of people leaving the state ahead of the storm.

The chart below shows a comparison of predicted and observed Call Rate Ratio (CRR) for the 48-hour window surrounding the arrival of the storm. Opacity of each county is proportional to population. Data: RapidSOS (observed), Reask + RapidSOS model (predicted). 

One standout finding from our research is that wind speeds correlate well to 911 calls. This key finding could help ECCs take a more proactive approach to emergency response in a severe weather event.

The call volume is unlikely to change if the ECC is not experiencing hurricane winds or if the winds are relatively light (e.g. below 100 km/hr). However, once speeds reach 100-120 km/hr, call rates increase.

The chart below shows a bin scatter of CRR vs. maximum 3-second gusts. Bins with less than 20 calls in the 3-week baseline period are not shown due to associated high variance in CRR. The size of each point is proportional to daily call rate in the baseline period within each bin. Marginal distributions of CRR and wind speed are shown along the edges of the plot. Data: Reask (wind gust), RapidSOS (CRR).