Flood Exposure in Pakistan

Group project for CASA0025 (CASA, UCL), built with several other students on the course.

Goal

Identify which areas of Pakistan are most exposed to flood risk, on the assumption that areas flooded in the past carry a higher risk of flooding again, and weigh that against population, economic activity and agriculture.

Methodology

Built entirely in the GEE JavaScript API as a deployed Earth Engine App, with GAUL administrative boundaries (national/level-1/level-2) driving an interactive region picker.

Flood history comes from the Dartmouth Flood Observatory’s Global Flood Database (GLOBAL_FLOOD_DB/MODIS_EVENTS/V1), summed across all recorded events into a cumulative flood-frequency layer, with JRC permanent water masked out so lakes and rivers don’t register as “flooded”. Population comes from Meta’s High Resolution Settlement Layer, including age-banded layers (children under five, youth, elderly) for exposure by demographic group.

My part of the group build was the flood and population/vulnerability side:

• Cumulative flood-frequency layer from the Dartmouth database, with permanent-water masking.
• Population summed by flood-frequency zone, and youth/elderly population share as separate map layers.
• A vulnerability index (flood frequency × population), smoothed with a focal-mean kernel for a cleaner visualisation, mapped nationally and charted as a top-10 ranking across level-2 subregions.

Other students on the team built the land use, building footprints, SAR flood detection and the panel UI/region-selection logic that ties it all together into one app.

Results

Deployed as a public Earth Engine App: select a region (national down to district level) and the app returns population and area for that region, the proportion of it that’s flooded, and a ranked comparison of subregions by vulnerability index. Sindh, for instance - 51.4M people, 134,896 km² - shows which of its districts carry disproportionate flood-population overlap relative to the rest of the province.