How the screening works

1. What this application measures

The core question is: does a hydropower project point lie downstream, on the same mapped HydroRIVERS network, of a glacial lake? When it does, the tool records the connection, estimates the downstream river-network distance, reports the lake's hazard evidence (ICIMOD expert rank where one exists, and observable GLO lake characteristics), and calculates transparent screening scores.

2. Analysis modes: PDGLs Only and All GLOs

All projects, all GLO lakes and the ICIMOD PDGLs are always visible on the map. The analysis mode decides which lakes contribute to the exposure calculation and statistics:

An ordinary GLO lake is never presented as if ICIMOD had classified it: its panel reads “ICIMOD PDGL: NO · ICIMOD hazard rank: Not available”. In PDGLs Only mode, ordinary GLO lakes remain visible (dimmed) and can still be searched and opened; their panel explains that they are not part of the current calculation.

3. Data sources

Published build: 2026-10-06T20:50:45+00:00. Geographic dataset built: 2026-10-06T20:50:45+00:00. Project sheet snapshot fetched: 2026-10-06T20:50:42+00:00. Hydropower projects keep stable internal IDs (HP_000001, …) across sheet edits.

4. Building one lake inventory

1. GLO lakes. Every GLO unique-lake record inside the analysis window is imported with its attributes (area, 2017–2024 expansion rate, uncertainty and significance, connectivity, elevation) and its outline. Outlines stored in a projected CRS (e.g. Albers Equal Area) are converted to longitude/latitude; the conversion is checked against the record's own coordinates.
2. PDGL matching. Each ICIMOD PDGL is compared with nearby GLO lakes using distance, outline containment, lake-area similarity and country consistency. Matches are one-to-one. High = the PDGL point lies inside the GLO outline (or ≤ 1.5 km with similar area); Medium = ≤ 3.5 km with broadly similar area; otherwise Review.
3. Merging. A High/Medium match merges the PDGL into its GLO lake, so one physical lake is one map object carrying both the ICIMOD classification and the GLO change evidence. A PDGL with only a Review candidate (or none) stays a separate ICIMOD record, and its GLO evidence is not used.

5. Lake → river → project method

1. Regional river extraction. The Asia-wide HydroRIVERS shapefile is scanned; only reaches intersecting the Nepal/transboundary analysis window are kept.
2. Lake connection hierarchy. Each lake is matched to its nearest HydroRIVERS reach. The distance is measured from the lake centroid and reduced by the lake's equivalent radius (√(area/π)), so large lakes are not penalised for their size.
   • Direct HydroRIVERS connection — reach within 8 km: High (≤ 3 km) or Medium (≤ 8 km). If the lake and the reach fall in different HydroBASINS Level-6 basins, the grade is lowered one step.
   • Catchment-supported connection — reach 8–20 km away and inside the same HydroBASINS Level-6 basin as the lake: graded Review and labelled as catchment-supported.
   • Unresolved — no defensible connection. No result is forced: unresolved lakes are shown but never contribute to exposure.
3. Project snapping. Each project point is matched to its nearest reach (≤ 15 km) and the snap distance is stored and disclosed.
4. Downstream tracing. From each connected lake's reach the engine follows HydroRIVERS NEXT_DOWN reach by reach. A project is connected only if its reach lies on that downstream chain. Links are stored once per build, so the public map only performs database look-ups.
5. Flow-path distance. Estimated from HydroRIVERS downstream-distance attributes. It is a screening distance along the mapped network, not a simulated GLOF travel distance or arrival time.

6. Connection confidence

A score says how strong the screening signal is; connection confidence says how certain the river connection itself is. They are reported separately. A link's confidence is the weakest of the lake connection, the project snap and the same-reach check.

Links where lake and project fall on the same reach are also marked Review, because a generalized reach cannot prove their order within it.

7. Screening scores

All scores are transparent, analyst-designed prioritization indices. They are not calibrated probabilities.

PDGLs Only — link score (0–95)

All GLOs — lake hazard proxy (0–100)

Project scores (both modes)

A project's GLOF exposure screening score is its strongest link in the current mode plus a multiple-lake modifier (0 for one link, 2 for two, 3.5 for three, 5 for four or more), capped at 100. Classes: Low < 35, Moderate 35–54.9, High 55–74.9, Very High ≥ 75.

MW is a materiality proxy only; it is not project value, loan exposure or a loss estimate. National statistics (lakes analysed, linked lakes, connected projects and capacity) are computed separately for each mode.

8. Reading the hydrological network overlay

When a project or lake is selected, Network overlay draws the relevant part of the river network around it — not every tributary, only reaches that connect the selection to mapped lakes and projects.

• Project selected: the upstream network linking it to every upstream lake (in the current mode) and upstream project, plus the downstream path to the last downstream project.
• Lake selected: the downstream path (the potential exposure pathway) to the last downstream project, with projects and lakes on it, plus the lake's immediate upstream catchment drawn dashed as context only. A lake's upstream catchment is not part of a GLOF propagation path.
• The faint lines show river topology; they do not claim that every drawn object affects every other. Hovering a line labels it and shows small flow-direction arrows.
• For a selected project, the route to its strongest upstream link in the current mode is highlighted automatically (with flow-direction chevrons). Any other link can be highlighted with its Route button. Clicking a connected lake or project re-centres the overlay on it.

Map symbols: hydropower projects are circles sized by MW and coloured by exposure class; GLO lakes are blue diamonds; ICIMOD PDGLs are magenta diamonds (darker and larger for Rank I).

9. Hydropower project data

Projects come from an administrator-maintained Google Sheet (survey, generation and operation licences), read through its public CSV export. The administrator refreshes it on demand; a refresh never re-processes the GIS sources.

• Validation. Rows without a project name or coordinates, or with coordinates outside Nepal, are skipped and listed for the administrator. Latitude/longitude entered in swapped columns are corrected and flagged.
• Review before publishing. Each refresh shows added, modified and removed projects and skipped rows before anything is published, with a warning if many rows disappear. The previous snapshots are kept for rollback.
• Stable IDs. IDs follow the licence number within its licence series when it is unique, otherwise the normalized name and location. Re-ordering the sheet does not change IDs; a project whose licence type changes keeps its ID via name + location matching; removed IDs are never reused.
• Incremental processing. Only new or moved projects are re-snapped; lake → project links and scores are then recalculated for everything.
• Licence coordinates are treated as representative project points, not exact intake, dam or powerhouse locations.

10. What the current data cannot establish

• No GLOF occurrence probability is calculated, and the All-Lake hazard proxy is not a substitute for an expert hazard assessment.
• No breach hydrograph, flood depth, velocity, arrival time or inundation extent is simulated.
• No dam/intake/powerhouse vulnerability function is applied; no monetary damage, bank loss or probability of default is calculated.
• HydroRIVERS may omit small outlet channels and generalizes geometry; some small high-altitude lakes therefore remain Unresolved or only catchment-supported.
• “No identified connection” means no connection was identified under this network method in the current mode; it does not mean zero GLOF hazard.
• GLO expansion measurements are remote-sensing estimates and retain the source dataset's uncertainty and significance fields.

11. How to use the result responsibly

Use this tool to screen, compare and prioritize further investigation. For project-level decisions, review the mapped route, snap distances, connection confidence, actual infrastructure coordinates, terrain and project protection/design information, and commission hydraulic modelling where warranted. Compare both modes: PDGLs Only reflects expert-classified hazard; All GLOs reveals exposure to lakes that have not been expert-assessed.