A cross-scale pattern-process interactions approach for predicting the spread of West Nile at regional and continental scales

Link To Share This Poster: https://cdmcd.co/LdDgzg
Live Discussion Link: https://cdmcd.co/PJ9xXj

Amy R. Hudson, Big Data Initiative and SCINet Program for Scientific Computing, USDA-ARS, Beltsville, MD, Debra P.C. Peters, Jornada Experimental Range, USDA Agricultural Research Service, Las Cruces, NM, John M. Humphreys, USDA-ARS, Sydney, MT, Lee W. Cohnstaedt, USDA-ARS, Manhattan, KS, Justin D. Derner, USDA-ARS, Rangeland Resources and Systems Research Unit, Cheyenne, WY, Kathryn A. Hanley, Biology, New Mexico State University, Las Cruces, NM and Luis Rodriguez, ARS, USDA, Orient Point, NY

Background/Question/Methods
West-Nile Virus (WNV) is a mosquito-borne zoonotic virus of global health and socio-economic concern. Studies of this virus and its disease in humans and horses have primarily focused on processes at individual local, landscape, or regional scales at short time windows, yet interactions across spatial and temporal scales are expected to be important for predicting the spread of infection at the continental scale. Here, we synthesize existing models of WNV for the US with the goal of creating a hierarchical process-based framework to predict WNV spread across several spatial scales, ranging from local to regional to continental. We identified for each model: 1) its spatial (grain, extent) and temporal scales, 2) the multi-scale processes and environmental variables included in the conceptual model, 3) the parameters and associated data sources for these processes and variables, 4) the analyses used to develop the predictive model, and 5) the accuracy and uncertainty of the predictive model. With this synthetic understanding of the processes and environmental variables important to WNV occurrence at different spatial scales, we then leveraged subannual county reports of equine WNV occurrence, and multiple layers of geospatial climate, land cover and management data from 2000 to 2019 for the continental US to produce a multi-scale predictive model.
Results/Conclusions
Our results show that multiple processes interacting across several spatial scales are needed to accurately model WNV occurrence at regional to continental spatial extents. Local processes associated with population dynamics of mosquitoes and birds combined with landscape to regional scale processes associated with climatic drivers were needed to accurately model broad-scale patterns in WNV. This synthesis identified gaps in our knowledge of prediction ecology in a virus context that can be used to generate hypotheses and target data collection campaigns to fill these gaps.