Using weather to explain and predict West Nile Virus risk in Nebraska
Abstract
Associations between dry and wet conditions, temperature, and West Nile Virus infection (WNV) rates vary by place due to factors such as climate and prevalence of different mosquito species. We tried a method for using dense data to predict an annual outcome variable, to produce information useful for early warning of years with higher WNV risk in Nebraska. We used monthly values of drought indices and mean-centered standardized temperature by county to retroactively explain and then predict the effects of weather on human cases of WNV. We used generalized additive models with a negative binomial distribution and smoothing curves to identify the most influential combinations of extremes and timing, experimenting with all combinations of temperature and drought data, lagged by 12, 18, 24, 30 and 36 months. We fit models to 7 sets of data starting with 2002-2011 and ending with 2002-2017, using AIC to pick the best-fit model, and used the subsequent year for out-of-sample prediction. We found that warmer winters and a dry year preceded by a wet year were the strongest predictors of cases of WNV. Our models did better than random chance and better than the naïve model (assuming each county would have the same number of cases as the previous year) at retroactively explaining the contributions of weather to the number of human cases, but did not do as well at predicting future numbers of cases. Our models did however provide new information about which counties would have any cases, valuable in sparsely populated areas where most county-years have no cases. Using the model for retroactive attribution, we ran a simulation that eliminated drought and warm winter temperatures by setting standardized values of dry conditions and warm temperatures to zero in our data, and comparing the predicted number under the hypothetic scenarios with the actual number of cases. The model predicted that without drought, there would have been 38% fewer cases of WNV in Nebraska through 2018 (using preliminary Arbonet data for 2018); without warm temperatures, 41% fewer; and with neither drought nor warmth, 54% fewer. This method for assessing the influence of different combinations of extremes at different time intervals is likely applicable to diseases other than West Nile, and to other annual outcome variables such as crop yield.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFMGH33B1188S
- Keywords:
-
- 0230 Impacts of climate change: human health;
- GEOHEALTH;
- 0240 Public health;
- GEOHEALTH;
- 0245 Vector born diseases;
- GEOHEALTH;
- 1616 Climate variability;
- GLOBAL CHANGE