A transdisciplinary framework for predictive disease ecology based on cross-scale interactions: Insights from long-term data

Authors: Peters, Debra; BURRUSS, N. DYLAN - New Mexico State University; Rodriguez, Luis; McVey, David; Elias, Emile; PELZEL-MCCLUSKEY, ANGELA - Animal And Plant Health Inspection Service (APHIS); Derner, Justin; Pauszek, Steven; Savoy, Heather; Peck, Dannele

Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 7/7/2019
Publication Date: N/A
Citation: N/A

Interpretive Summary:

Technical Abstract: The availability of long-term environmental data for many variables at multiple scales across large spatial extents provides opportunities for novel questions to be addressed as well as new insights into unresolved questions. These questions can pertain to regional- to continental-scale dynamics that affect animal and human health, and are driven by interactions among processes occurring at multiple spatial and temporal scales. We are developing a strategic framework based on pattern-process integration and interactions across scales with human guided-machine learning to identify, harmonize, analyze, and interpret big data involving a variety of variables from online and local sources to meet these challenges. We illustrate our framework with questions related to drivers of spatial and temporal patterns in the invasion by vesicular stomatitis virus (VSV), a vector-borne, zoonotic RNA virus that affected > 1500 livestock premises from 2004-2016 across 10 states in the western US. In addition to incidence and phylogenetic data, we obtained online data for 9 environmental drivers and host density data. For each driver, we selected variables for analysis based on hypothesized relationships with disease processes. The geo-referenced maps of the >50 variables were harmonized in time and space. Multivariate analyses of the resulting data cube showed that the initial incursion of VSV from Mexico into the southwestern US in 2 separate years (2004, 2014) occurred under similar conditions (low surface water in summer and fall, above-average summer vegetation, below-average winter precipitation) that were different from conditions when VSV expanded throughout the 10-state region (2005, 2015: below-average summer temperatures on locations containing soils with high water holding capacity). Watershed-based analyses showed that VS incidents were distributed near streams with 72% located within 1 km of stream habitat. All first incidents (n = 35) occurred following peak annual streamflow, with 89% of these occurring after streams returned to baseflow. Our big data-model integration framework is being applied to other disease systems that are temporally variable and spatially heterogeneous across large spatial extents, e.g. West Nile Virus in the conterminous US.