Monitoring environmental damage in conflict zones

1. Step 1.Define the environmental harm you are documenting

   Name the specific damage in the environment you are testing for: cratered farmland, oil-facility fires, contaminated aquifers, deforested buffer zones around IDP settlements. A concrete claim keeps the evidence chain auditable.

2. Step 2.Anchor the claim to a place and a time

   Draw the smallest bounding box that covers the incident and pick a pre-event and post-event window. Same-season pairs (dry-to-dry, wet-to-wet) prevent seasonal shifts from being misread as war damage.

3. Step 3.Pull free imagery and conflict events for that window

   Copernicus Sentinel-2 (10 m optical) and Sentinel-1 (SAR) cover most conflict zones on a 5–12 day revisit. Pair every pull with the ACLED events that fall inside the bounding box and window so damage and incidents are already co-located.

4. Step 4.Compute a change index, not a screenshot

   Use dNBR for burn scars, NDVI difference for vegetation loss, and Sentinel-1 VV backscatter change for flooding or built-up damage. Change indices survive haze, shadow, and phenology in a way true-colour composites do not.

5. Step 5.Cross-check with independent evidence

   Confirm each anomaly against NASA FIRMS active fires, Global Forest Watch loss alerts, ReliefWeb reporting, and named local media. A single-sensor anomaly is a lead, not a finding, until at least one independent source agrees.

6. Step 6.Publish with methodology, limits, and protections

   Every finding needs the sensor, dates, formula, bounding box, and cloud mask. Redact precise coordinates of IDP camps, informal settlements, or humanitarian convoys before publication, following the Berkeley Protocol.

This is the applied version of the six-step workflow, wired to the data infrastructure Humanity Centered Data already runs: country pages for scoping, ACLED for events, HDX for boundaries and populations, Sentinel Hub and rasterio for change detection, and the shared CitationsMethodology block for publication.

1. Step 1.Scope the incident against site infrastructure

   Humanity Centered Data country + ACLED

   Start on the relevant country page (for example /sudan or /ukraine) to pull the current displacement figures, active thematic briefs, and the list of ACLED-flagged hotspots the site already tracks. Copy the ISO3 code, the admin1 name, and the incident date range into a scoping note. This ties every downstream raster back to a population and a governance unit the platform already covers.

2. Step 2.Pull ACLED events for the bounding box and window

   ACLED API

   Query ACLED for the same bounding box and dates. Filter to event types that plausibly cause environmental harm: explosions/remote violence, battles, and strategic developments involving infrastructure. Save the response as GeoJSON so it drops straight into QGIS or Leaflet.

   curl -G 'https://api.acleddata.com/acled/read' \
     --data-urlencode 'key=YOUR_KEY' \
     --data-urlencode 'email=you@org' \
     --data-urlencode 'iso=729' \
     --data-urlencode 'event_type=Explosions/Remote violence|Battles' \
     --data-urlencode 'event_date=2025-01-01|2026-06-30' \
     --data-urlencode 'event_date_where=BETWEEN'

3. Step 3.Grab HDX admin boundaries and settlement layers

   Humanitarian Data Exchange (HDX)

   Download COD-AB admin boundaries (admin1/admin2) and, where available, IOM DTM site coordinates and WorldPop constrained population rasters for the country. Clip everything to the bounding box. These layers turn a raw change pixel into an answer to "which district, which settlement, how many people".

   # HDX CKAN API - list Sudan COD boundaries
   curl 'https://data.humdata.org/api/3/action/package_search?q=cod-ab+sudan&rows=5' | jq '.result.results[].name'

4. Step 4.Compute a change index in Sentinel Hub or rasterio

   Copernicus Sentinel-2 / Sentinel-1

   Use the Sentinel Hub Custom Script below for dNBR (burn severity). For deforestation swap to NDVI difference; for flooded built-up areas run Sentinel-1 VV backscatter thresholding. Export as a GeoTIFF with the sensor, dates, and formula written into the filename so provenance survives.

   // Sentinel Hub evalscript - dNBR (Normalized Burn Ratio difference)
   //VERSION=3
   function setup() {
     return {
       input: [{ bands: ['B08', 'B12', 'dataMask'] }],
       output: { bands: 1, sampleType: 'FLOAT32' },
       mosaicking: 'ORBIT',
     };
   }
   function evaluatePixel(samples) {
     const pre = samples[samples.length - 1];
     const post = samples[0];
     const nbrPre  = (pre.B08  - pre.B12)  / (pre.B08  + pre.B12);
     const nbrPost = (post.B08 - post.B12) / (post.B08 + post.B12);
     return [nbrPre - nbrPost]; // dNBR > 0.27 = moderate/high severity
   }

5. Step 5.Join the change raster to ACLED and HDX layers

   QGIS / GeoPandas

   Overlay the dNBR (or NDVI difference) raster with the ACLED GeoJSON and HDX admin/settlement layers. Keep only change pixels that fall within a defined buffer (typically 2 km) of an ACLED event in the same window. This is the step that converts "something changed" into "damage in the environment co-located with a documented incident".

   import geopandas as gpd, rasterio
   from rasterio.features import geometry_mask
   
   acled  = gpd.read_file('acled_sudan_2025_2026.geojson').to_crs(3857)
   buffer = acled.buffer(2000)  # 2 km buffer around each event
   with rasterio.open('dnbr_sudan_2025_2026.tif') as src:
       mask = geometry_mask(buffer, out_shape=src.shape, transform=src.transform, invert=True)
       dnbr = src.read(1)
       attributed = (dnbr > 0.27) & mask   # burn severity co-located with events

6. Step 6.Corroborate anomalies with independent sources

   NASA FIRMS + GFW + ReliefWeb

   For every attributed cluster, pull FIRMS active fires (VIIRS/MODIS), Global Forest Watch GLAD/RADD alerts, and ReliefWeb situation reports for the same admin unit and window. Log at least one independent match per finding. Single-sensor anomalies stay in the working file, not in the publication.

   # NASA FIRMS - last 10 days of VIIRS active fires for Sudan
   curl 'https://firms.modaps.eosdis.nasa.gov/api/country/csv/YOUR_MAP_KEY/VIIRS_SNPP_NRT/SDN/10'

7. Step 7.Publish through the site methodology block

   Humanity Centered Data CitationsMethodology

   Wrap every published chart or map with the shared CitationsMethodology component so readers see the sensor, dates, formula, bounding box, ACLED query, and independent sources. Redact coordinates of IDP sites and informal settlements per the Berkeley Protocol before pushing.