Towards a workflow for operational mapping of Aedes aegypti at urban scale based on remote sensing

This repo holds the scripts that we have developed towards an operational workflow for mapping dengue vectors in urban areas.

The manuscript was published on August 2021 in Remote Sensing Applications: Society and Environment and can be found at: https://doi.org/10.1016/j.rsase.2021.100554

The scripts were originally developed using SPOT as input imagery, but these can be replaced by Sentinel 2 level 2A data in order to avoid the atmospheric correction step and automate the ingest of satellite data.

All image processing was done with GRASS GIS 7.8+. The scripts assume that the users have the software, all dependencies and extensions installed. Moreover, it requires that the GRASS DATABASE is properly set with locations and mapsets created accordingly. For extra info, visit the first time users and tutorials pages in the GRASS GIS website.

The order in which scripts should be used is as follows:

1. grass_scripts/data_import.sh: This script imports ancillary data like DEM from SRTM and vectorial data like water bodies, railway lines, neighborhood polygons.
2. grass_scripts/img_processing.sh: This script imports SPOT data and does all the image processing for one scene. It requires some variables to be set by the user. The atmospheric correction step requires also intervention to create simple files with specific parameters. In the end, it exports all relevant raster maps and removes them from the mapset.
3. r_scripts/modeling.r: This is the core script that performs calibration, variable selection, model selection and threshold independent evaluation. It requires intervention in identifying the best models and copying relevant files to produce the final predictions.
4. r_scripts/thres_depen_validation.r: This script uses independent data (in this case, from larval samplings) to perform a threshold dependent validation. It requires intervention to determine the best threshold that is then used to create presence- absence maps.
5. r_scripts/output.r: This script creates the outputs of the workflow: average probability and standard deviation; average probability per neighborhood and presence-absence binary maps.

TO DO:

• Replace SPOT by Sentinel 2 data. Use GRASS GIS modules i.sentinel.* to search for available scenes, filter by location and cloud cover, download, import and mask clouds and clouds' shadows.
• Improve mosquito data flow with Health Authorities so data can be ingested automatically.
• Improve modeling and validation steps that still require intervention, i.e., copying of environmental layers, threshold selection.
• Update threshold dependent validation to use a package available in R 4.0 (SDMTools runs up to R 3.6)
• Automatically generate a report from output maps to be sent to Health Authorities.
• Upload output raster and vector maps to a webGIS.