Note: A version of this article was originally published on Sept. 1, 2017, by Data-Smart City Solutions, an initiative of the Harvard Kennedy School Ash Center for Democratic Governance and Innovation.
Cities in southeastern Texas finally saw dry weather on Aug. 30, after days of unprecedented rainfall from Hurricane Harvey. But while floodwaters started receding from many Houston neighborhoods, explosions at a Crosby chemical plant and Beaumont’s lost water supply showed how vulnerable the area remains to health and safety threats.
Among these post-Harvey issues, storm-related flooding could increase risk for diseases spread by mosquitoes, which breed in standing water. Key mosquito-borne diseases include West Nile Virus, which is already a significant health problem in the Gulf Coast region, as well as Zika and the closely related dengue virus, which has circulated in the Houston area as recently as 2005.
Yet, as tropical disease expert Peter Hotez told The Atlantic, “the impact of floods on mosquito-borne viruses is still understudied.” As news outlets like Newsweek and The Guardian reported on the threat, I consulted open data sources to investigate the risk. I found that mosquitoes will proliferate within a week or two, likely after a steep drop in their abundance, and that the several weeks remaining in mosquito season represent a dangerous period for disease transmission.
What citizens report: 311 data sources
In many cities, residents can report mosquito problems to authorities using non-emergency 311 systems. In New Orleans, one of the low-lying cities that felt effects from Harvey this week, the city’s open data portal publishes several years of 311 call records labeled by type, including mosquito control requests.
Plotting these requests against storm-related flooding (as tracked by the National Oceanic and Atmospheric Administration’s Storm Events database) shows that within a week or two of floods, citizens’ mosquito-related complaints jump higher than the levels observed before the storm.
Weather matters, as well. When floods occur close to, or during, cooler months, mosquitoes do not proliferate. In Louisville, Kentucky, the pattern of floods and mosquito-related 311 calls appears inconsistent, until plotted against weekly temperatures. In this chart, the area below zero represents values (for mosquito call volume or temperature) that are below the overall average. During mosquito season, floods precede significant increases in mosquito-related calls; near or during the offseason, the effect is minimal.
What cities catch: Mosquito trap data
Do these calls accurately represent mosquito populations? As far as I can tell, 311 data has never been used in published research assessing mosquito burden, and thus never rigorously validated as a measure of mosquito burden. Perhaps since floods affect people, too, 311 calls are skewed towards over- or underestimating mosquitoes.
Many U.S. jurisdictions use mosquito traps to monitor mosquito abundance. While almost all these records are locked away, Chicago publishes 10 years’ worth of trap data, and these show the same pattern as Louisville: Floods plus warm weather appears to increase the mosquitoes trapped citywide.
But while mosquitoes are a nuisance — and their biting alone hampered recovery efforts after Hurricane Andrew in 1992 — their presence does not guarantee disease incidence. When researchers failed to find increases in West Nile Virus (WNV) after Hurricane Katrina, they speculated that high winds might have destroyed mosquito habitats and dispersed birds (which are also involved in the WNV transmission cycle). Indeed, in Chicago, traps testing positive for WNV were highest in years without flooding, and mostly trivial in years with flooding.
This pattern does not necessarily indicate safety for Harvey-affected areas, however. A team that followed up on Katrina found, first, that WNV case reports had actually increased in the three weeks after the hurricane’s landfall, then sharply dropped. They speculated that infections might have happened in the days during and after the storm, with residents stuck in damaged housing or outdoor shelters; residents’ relocation and aerial insecticide spraying might explain the drop thereafter.
Like Katrina, Harvey arrived with several weeks remaining in mosquito season, which leaves enough time for a rebound in risk — perhaps most comparable to Chicago’s 2008, 2013 and 2015 seasons above, when WNV-positive traps appeared several weeks after early-season and midseason storms. While those years still saw relatively few WNV-infected mosquitoes, those insects would be more dangerous with more residents outdoors rebuilding or living in makeshift conditions. (Note, moreover, that the mosquito season lasts deeper into the fall in Gulf Coast cities.)
Perhaps more worrisome, in the year after Katrina, storm-affected areas experienced double their normal WNV disease rates. Next year could be pivotal, therefore, for mosquito control authorities around Houston and elsewhere in the region.
What cities can learn
The Houston area’s mosquito control chief promised aerial spraying if his teams discovered enough mosquitoes after Labor Day, and the health department conducted those tests in the Houston area Labor Day week. The following week, authorities began large-scale aerial spraying with help from the Air Force.
After Hurricane Andrew, this strategy was highly effective, with each spraying reducing mosquito counts by as much as 96 to 97 percent. There is controversy and limited evidence, however, as to how well aerial spraying kills the Aedes mosquitoes that transmit Zika and dengue. Those mosquitoes, which are more likely than their WNV-carrying cousins to live in and around buildings, might survive. Homeowners, and especially shelter managers, should be careful to remove any standing water sites within 500 meters, the mosquitoes’ typical flight range.
The second takeaway from this analysis is the relevance of open data sources for disaster situations. During the storm, Harris County’s open data map apprised residents of when and where bayous overflowed; city agencies have used open data sources to provide information about the status of city services, transit and more. For this analysis I used data generated in reverse—citizens informing government about the locations of mosquitoes.
These datasets could be better. Houston, for example, publishes 311 calls to its open data portal, but mosquito control requests appear to be combined with other “Nuisance” health code reports. This aggregation made it impossible to measure precisely how citizens’ mosquito reports have corresponded with the city’s many recent floods. To solve problems like these, a group of chief data officers is working towards uniform data standards that would make analyzing 311 data easier and allow better comparisons across jurisdictions.
Mosquito trap datasets, for their part, should be openly available. Most cities can access these records, yet Chicago was the only U.S. city I could find with both mosquito counts and WNV test results, and one of just a few that publish either. Far more cities publish data on tough topics such as their police-involved shootings. Arguments against sharing mosquito data, such as those raised by county officials near Syracuse, New York, that negative trap results might create a false sense of security, fall woefully short. In Harvey-affected areas and beyond, residents need governments to help protect them from mosquitoes — not to protect them from mosquito data.