The South Dakota Department of Health (DOH) wants to break the model scientists made for it, which might seem an odd goal until one considers the context.
The model uses satellite and weather data with regional mosquito surveys to estimate the number of West Nile infections expected each year—valuable information for a DOH whose state has the highest per capita infection rate in the U.S.
The projections are accurate: The early season estimate for 2017 came in at 83 cases, just 10 shy of the statewide total of 73. This year, the model projects an “outbreak” of approximately 148 cases.
The DOH hopes thrusting the forecast into the public eye will encourage residents to be vigilant in their efforts to prevent disease-spreading mosquito bites and thus a West Nile outbreak.
“We’re trying to bust that model,” South Dakota State Epidemiologist Josh Clayton said.
Clayton’s team will take over the forecasting model next season from researchers at South Dakota State University (SDSU), which created the model in a project funded by NASA’s Applied Sciences Public Health and Air Quality Program.
The remote-sensing piece of the model was created by SDSU’s Geospatial Sciences Center of Excellence (GSCE), a collaboration between SDSU and the U.S. Geological Survey’s Earth Resources Observation and Science (EROS) Center. EROS operates the Landsat satellite program and maintains a massive archive of Earth observation data from multiple satellite sensors.
The South Dakota Mosquito information System (SDMIS) builds on a decade of research by GSCE scientists and graduate students. Previous work, including work on malaria in Ethiopia, combined on-the-ground and remotely-sensed data to study the air and land conditions that give rise to disease-carrying mosquitoes.
Work on the SDMIS began in 2015, with the first weekly forecasts appearing the following year.
Next year, the team will hand DOH a computer script that will allow the agency’s staff to input data on air and vapor pressure, mosquito surveys, and infection reports, then run the numbers and return a weekly, county-level forecast for public health purposes.
Another script uses data from satellites stored in the EROS archive to identify hotspots within those counties, based on factors such as Normalized Difference Vegetation Index (NDVI).
Between the two applications, the DOH will be able to make county forecasts and zero in on particularly West Nile-prone areas.
“What we’re able to do is strengthen the disease collection data by using the Earth observation data,” said Michael Wimberly, one of the GSCE scientists heading up the project. “Part of the goal was always to create something transferrable and sustainable.”
South Dakota builds warning tool to respond to heightened risk
In the past 16 years, 42 people have died from West Nile complications in South Dakota. Public education and prevention efforts run the gamut from public service advertising to funds for mosquito spraying and bug repellent distribution at public events.
The DOH and a handful of cities in the state have long trapped and tested mosquitos to search for infected Culex tarsalis—the primary disease vector—to help triangulate risk. That data is helpful, Wimberly explained, but it’s not complete enough for wide-scale projections.
“You can’t track for mosquitos all across this huge, really sparsely-populated state,” Wimberly said. “That’s where the Earth observations come in. We can look everywhere across the state.”
Mosquitoes can thrive nearly anywhere in the world with little more than warm air, standing water, and a few warm-blooded creatures on which to feed, but species do have preferences. Nuisance mosquitos prefer to feed on mammals, for example, but Culex tarsalis spends the early months of the season feeding on birds, which can carry West Nile virus. They switch to mammals mid-summer, marking the start of the high-risk months.
Culex tarsalis populations don’t spike and disappear in the same pattern as “flood water” mosquitos, either. They tend to lay eggs in shallow pools of standing water, such as those that collect in hoof prints, tire tracks, or pools left by irrigation runoff, and the eggs continue to hatch for longer periods of time.
That means heavy rains are less a factor in West Nile risk than heavy rains followed by hot, humid weather that keeps shallow pools from drying out.
“When those floodwaters go down, they tend to leave pools of water, and that’s where Culex will breed,” Wimberly said.
Projecting risk with the right data
The SDSU and GSCE teams set out to identify the most important climatological risk factors and create a model that sorts through them. Instead of zeroing in on precipitation—a larger factor for malaria risk—Wimberly’s team created a model for South Dakota that pulls air temperature and vapor pressure data into its weekly forecasts.
The model began with land cover and physiography datasets, with a base map for South Dakota and land cover classes from the Landsat-derived National Land Cover Database. Researchers also incorporated information on wetlands, ponding frequency and elevation.
To that, researchers added variables from a daily dataset called gridMET - short for gridded surface meteorological data - which blends gridded climate information from the Parameter-elevation Relationships on Independent Slopes Model (PRISM) with hourly data from the North American Land Data Assimilation System (NLDAS).
PRISM is an interpolation method that combines data from thousands of temperature and precipitation surface stations and factors such as location, elevation and topography to create climate datasets that fit into 800-meter grids.
NLDAS offers near-real time land surface modeling datasets, which simulate the interplay between soil moisture and energy.
“That gives us all the information we need (for the weekly forecasts),” Wimberly said.
That information comes in at a resolution of about 12 kilometers per pixel. To zoom in further, a second operation pulls data from the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard NASA's Terra satellite.
When combined with mosquito surveys and reports of West Nile infection in humans, the higher resolution data will allow the DOH to create snapshots of high-risk areas within a county.
That’s all quality information for the DOH and Dr. Clayton, who took over for longtime state Epidemiologist Lon Kightlinger last fall. Dr. Kightlinger had worked to document cases, pinpoint risk, and hone the public message with data from mosquito surveys. He had championed the SDSU collaboration near the end of his career.
“The information that we’re getting from the project are maps of risk factors associated with things we at the Health Department really can’t track or figure out … I think this will be absolutely an essential tool in controlling West Nile here in South Dakota,” Kightlinger said in a NASA video on the project released in 2016.
The now-optimized modeling tools, Clayton said, will bring an additional layer of context and certainty to those public health efforts. Instead of playing catch-up by reporting cases after they’ve happened and relying on spot-checks from mosquito traps, the state can project risk statewide in advance.
“It helps complete the picture in terms of what we see for the year,” Clayton said.Clayton envisions the weekly forecast becoming a part of the annual statewide dialogue on West Nile.
If more South Dakotans are aware of the potential for an outbreak, the thinking goes, they will take steps to protect themselves by wearing insect repellent, wearing long pants and shirts when practical, and staying inside during the late evening hours when Culex tarsalis prefers to feed.
Higher risk is projected this year, but Clayton said it’s important to be vigilant even in more typical years. There were four deaths in the state last year despite lower infection numbers, according to the Centers for Disease Control’s annual accounting.
“We can’t rest when we’re having a low West Nile year,” Clayton said.
He pointed to the results of the project as evidence of the power of a collective effort toward scientific understanding of a public health problem.
“It highlights the great, integrated, collaborative effort we’ve had in terms of trying to get out as much information as we have,” Clayton said.