Glacial Lake Growth Monitoring from Sentinel-2 Imagery for Outburst Flood Risk in Northern Pakistan
Glacial lakes in Gilgit-Baltistan and Chitral are growing as glaciers melt, and some can burst and flood villages downstream. Build a pipeline that maps lake outlines from free Sentinel-2 satellite images every month, tracks area change over several years, flags lakes growing fast, and lists the settlements downstream. Validate the outlines against published inventories and hand digitising.
Intermediate
A small team, or one strong student willing to learn something new.1 semester · 2 students
7 to ship
4 optional extrasSuggested stack
What you should ship
- A study area of one or two valleys with a list of glacial lakes from a published inventory
- Lake mapping from Sentinel-2 using a water index and thresholds, with cloud and shadow masks
- An improved classifier that handles frozen lakes, turbid water and mountain shadows
- A time series of lake area for each lake over several years
- Flags for lakes growing faster than a chosen rate, with uncertainty
- Downstream path tracing from a digital elevation model to list exposed settlements
- Validation against hand-digitised outlines on selected dates
If you have time left
- Using Sentinel-1 radar to see through clouds
- Monthly automated updates and an email summary
- Comparing growth with temperature records
- A simple public map with lake histories
The problem: lakes that grow quietly
Pakistan's north holds thousands of glaciers. As they retreat, meltwater collects in lakes held back by loose moraine or ice. When those dams fail, a glacial lake outburst flood (GLOF) can destroy bridges, roads, homes and farmland in minutes, as several events in Gilgit-Baltistan and Chitral have shown. Some lakes grow for years before a flood, but monitoring hundreds of remote lakes by field visits is not possible.
Free satellite images every few days make it possible to watch every lake, if someone builds and validates the pipeline.
What already exists and the gap
- ICIMOD, the International Centre for Integrated Mountain Development, has published glacial lake inventories for the Hindu Kush Himalaya.
- The UNDP-supported GLOF-II project with Pakistan's government installed early warning systems and other measures in vulnerable valleys.
- Research papers map glacial lakes from satellite images for single dates or regions.
Inventories are snapshots. The gap is an open, repeatable pipeline that updates lake areas regularly and flags fast growth for a chosen set of valleys.
Inspiration
ICIMOD's inventories show how glacial lakes are mapped consistently across a large region, and the GLOF-II project shows how that knowledge is used locally. Your project combines the two: inventory-style mapping, repeated over time, for valleys where communities and officials can use it.
What you build
- Disaster management officials and NGOs see each lake's area history and a list of lakes growing fastest.
- Researchers get code and data to extend to other valleys.
- Students learn practical remote sensing on a problem with real stakes.
The technically interesting part
- Hard pixels. Frozen lakes, turbid meltwater, shadows from steep slopes and clouds all confuse simple water indices. Start with an index such as NDWI, then improve with a classifier and masks.
- Consistency over time. Area changes from season to season. Compare like with like, for example only late summer images.
- Uncertainty. Each outline has an error of about one pixel along its edge. Estimate area uncertainty and only flag growth larger than it.
- Downstream exposure. Trace flow paths from a digital elevation model and intersect them with settlement data from OpenStreetMap.
How to evaluate it
- Outline accuracy: overlap with hand-digitised outlines on selected images.
- Agreement with the published inventory for lake count and area.
- Area uncertainty estimates and the share of flagged growth larger than that uncertainty.
- Run time to update the whole study area.
Suggested plan
- Weeks 1 to 3: read the inventories, choose valleys, set up data access.
- Weeks 4 to 7: water mapping and masks.
- Weeks 8 to 10: time series and growth flags.
- Weeks 11 to 13: downstream tracing, validation and map.
- Weeks 14 to 15: report and demo.
Path to a product
Disaster agencies, hydropower companies with assets in mountain valleys, insurers and NGOs need regular lake monitoring. A validated pipeline that produces monthly reports is a service a small geospatial team could offer. The project is strong evidence for geospatial data and remote sensing roles.
Scope warning
You are not issuing flood warnings. Do not publish "danger" labels for named lakes without expert review; describe growth rates. Lake area alone does not decide risk, dam type and slope matter, so say that clearly.
Learn what you need on Sefism
- Data science roadmap: classification and uncertainty.
- Data analytics roadmap: time series and maps.
- scikit-learn User Guide: pixel classifiers.
- Python Data Science Handbook: numerical work with arrays.
- Streamlit documentation: the lake history viewer.
Get help with this idea
Clouds and shadows can swallow weeks of effort. Tauseef can help you pick valleys and dates that make the project finishable. If you are new here, create a free account first.
Ideas and guidance, not finished projects
These are project ideas and scoping guidance, published free for students to use as a starting point. I do not build, write, or sell final-year projects, and I do not complete coursework for anyone. Take an idea, make it yours, and build it.
Not sure this one fits you?
I help students pick an idea that matches their skills and their deadline, then scope it down until it is finishable. Guidance only, never done for you.
Stuck on something specific?
Writing only gets you so far. If you want an answer to your situation rather than the general case, book a session and we will work through it together. Sessions are free for approved Sefism members, and a few slots open each week.
Follow along
New writing, resources and project ideas land here first.