README.md

RVF_Mauritania

Repository for modeling of Rift Valley Fever (RVF) in Mauritania.

The analysis includes

• calculation of potential risk areas (suitability maps), depending on environmental conditions and
• calculation of spillover potential, depending on the suitability maps and livestock and animal data.

Calculation of potential risk areas

• Tool for Analysis: Maxent
  • Used within GRASS GIS
  • Installation with r.maxent.setup:
    • Note: need to have java installed
• Analysis
  • Define configuration file: potential_risk_areas_config.cfg
  • Train model: potential_risk_areas_train.sh
  • Apply model: potential_risk_areas_predict.sh

Analysis procedure:

Input

• Prepared disease data and covariates
• Defined within config
• For different data versions see table potential_risk_areas_data_versions

Train model:

• Script: potential_risk_areas_train.sh
• AOI/Region setting
  • Use all given positive and negative samples -> AOI defined by them (Covariates only sampled at negative/positive result coordinates)
• Model options: Usage of monthly disease data
  • Option 1: single model
    • Combine all given positive and negative samples from all months/years
    • Sample covariates for all given positive and negative results (i.e. create monhtly SWD files) and combine them --> use this as input to train one single model
  • Option 2: monthly models
    • Train monthly models (for which enough positive (and negative) samples given)
  • For different model versions see table potential_risk_areas_model_versions

Apply model

• Script: potential_risk_areas_predict.sh
• Apply model to all monthly data
  • From 2019-2023, see also PREDICTION section within config

Calculation of spillover potential

The spillover potential calculation follows Hardcastle et al. 2020 with two differences:

• the geographic units are not administrative areas but pixels
• livestock movement from Jahel et al. 2020 is included

The main script is spillover_risk.sh which includes livestock movement: the number of animals per pixel is increased with increasing movement through a pixel to simulate more contacts between animals.

Input data for the spillover risk calculation are

• potential risk calculated with maxent for each month separately (see above)
• human population per pixel
• livestock population per pixel
• livestock movement, separately for the dry and wet season