Skip to main content
ARS Home » Research » Research Project #440495

Research Project: Predicting Insect Contact and Transmission Using Historical Epidemiological Data

Location: FOREIGN ARTHROPOD BORNE ANIMAL DISEASE RESEARCH

Project Number: 3022-32000-023-001-S
Project Type: Non-Assistance Cooperative Agreement

Start Date: Sep 1, 2021
End Date: Sep 30, 2025

Objective:
A decision support tool is needed to optimize provisioning and alignment of resources based on estimated risk for arthropod-transmitted pathogens. Estimating pathogen transmission risk will reduce waste of limited shelf-life products and the movement of resources between locations. Planners will use PICTURED to evaluate mosquito-borne virus risk of a geographic location at specific time or duration of time. Users can also monitor conditions at specific locations to decide when and which products to reorder. Predictive models will estimate arthropod-borne pathogen transmission risk in locations using abiotic, biotic, and dynamic real-time data. Abiotic data examples are environment (urban, silvatic, etc.), elevation, weather and seasonal patterns, and climatic region. Biotic data consists of vegetation, disease vector species distributions, historic pathogen transmission in the area, and host/reservoir presence. PICTURED will be modular to accommodate additional data as it becomes available such as routine surveillance or user data sets of mosquitoes or human cases. The decision support platform will use mathematical models to classify landscapes into three levels of transmission current risk–high, medium, and low in any location in the world. The platform will also have a forecasting tool to assess the trend of active outbreaks in a specific place into increasing, constant, and decreasing cases, provided users input the incidence data for the selected location or surrounding areas.

Approach:
To forecast or predict mosquito-borne disease outbreaks, risk will be evaluated based on three factors: (1) Vector arthropod species distribution (Genus level vector distribution records and niche modeling to infer vector distribution in poorly sampled areas). Species distributions will be mapped based on published records and linked to habitat characteristics. The habitat characteristics can then be used for niche habitat analysis to infer the distributions in un-surveyed areas. Therefore, an area can be classified as reported presence, predicted presence, and unsuitable. These shape files will be updated as new records become available. (2) Past cases based on health records. Historical records of past pathogen circulation in the human and animal reservoirs can be used to determine which pathogens may be present in an area. However, vector species distributions can change, along with the pathogens they carry. (3) Estimated vector abundance based on weather (rainfall and temperature) and seasonal data. Abundance and distribution of disease vectors are driven by weather conditions. Monitoring current weather and comparing it to past weather patterns during outbreaks will help signal elevated risk. This reduces the need to survey for disease vectors year-round, which may not be possible in some locations globally. The model will be validated and updated when arthropod survey data is available. The innovation of PICTURED is the risk estimation comes from the combined use of multiple data streams to estimate current pathogen transmission. Similarly, the model smoothing methods will fill gaps in incomplete data sets by interpolating geographic and temporal gaps in disparate data sources. Ultimately, the output of the cleaned data will result in three levels of risk: high, medium, and low for a given pathogen in a geographic area. When used for short-term prediction of a current outbreak, PICTURED is a synthesis of several interdependent and extendable layers of data that model the pathogen, environment, and hosts (human and animal) and their interactions with disease vectors using a combination of three forecasting approaches: Ensemble Kalman filters (in space and time), deep neural networks, and stochastic networks approaches. This work will focus on Culex mosquitoes and specifically, Culex tarsalis which is a mosquito vector of the West Nile, Rift valley fever, and Japanese encephalitis viruses.