Official providers, acquisition endpoints, access dates, licences, attribution requirements and known limitations are derived from the same release metadata used by the Data Catalog.
For exact project-document excerpts and dataset-version citations, open the evidence browser.
Atlas-derived analysis is listed separately from originating evidence. A derived product does not become an independent external source simply because it has its own release artifact.
Feature 42 contains original Atlas test cases under MIT, with no source readings or geographic locations. Fixed timestamps and workflow states are synthetic; they do not report current conditions. Inspect the checksum-verified cases and their validation method.
Feature 42 acquires no sources. Feature 43 adds manual USGS/NOAA acquisition. DHM is reference-only, and BIPAD/NDRRMA ingestion remains excluded from the public build. OpenAQ, Open-Meteo and IMERG remain disabled pending separate reviews. The fixture licence grants no rights to upstream datasets.
USGS preferred USGS-network earthquake summaries: public-domain guidance. NOAA/NCEP GFS 0.25° APCP: NWS data terms. Every imported snapshot has an exact-byte licence review, source hash, request and processing version. No warning or third-party contributor licence is inferred.
Boundaries are COD-AB v02, created by the source in 2024 and reviewed by OCHA/HDX in 2025; later legal or cartographic changes may not be represented.
Level 3 contains 753 local-government units plus 22 protected or special-area pieces encoded by the source; the interface distinguishes these categories.
Ward boundaries are not included in this source release.
English names are supplied; alternate-name fields are currently empty in the source.
Selected district display names follow the public reference list (https://en.wikipedia.org/wiki/List_of_districts_of_Nepal); original COD-AB spellings remain in aliases.
Web geometry is topology-preserving display simplification and must not be used for cadastral, legal, distance, area, exposure, or aggregation calculations.
International-boundary representation follows this source and does not resolve disputed boundary claims.
The network is derived from HydroRIVERS/HydroSHEDS terrain-based hydrography; displayed lines are mapped/modelled river reaches, not a guarantee of present-day wetted channels.
HydroRIVERS includes rivers meeting its catchment-area or average-flow thresholds, so smaller channels are absent.
The FAO edition provides no river-name field for these reaches; the atlas therefore shows UNKNOWN for river name and uses HYRIV_ID as the stable identifier.
Average discharge and flow-regime attributes are source dataset descriptors and are not live gauge observations.
Reaches are selected by intersection with Nepal and are not clipped, preserving source geometry and downstream continuity at national boundaries where available.
RGI 7.0 is a dated inventory targeting approximately the year 2000; individual source outline dates in Nepal vary and are shown per glacier.
Glacier outlines are inventory observations derived from remote-sensing source material and do not represent the current glacier margin in 2026.
The GLIMS WFS display export rounds/reprojects coordinates; invalid presentation geometries are explicitly repaired and simplified. Source RGI area is retained as the measurement and display polygons must not be used to recompute area.
Most RGI glacier records in this region have no published glacier name; the atlas shows UNKNOWN rather than inventing a name.
RGI/GLIMS coverage and outline uncertainty varies with source imagery, mapping method and acquisition date.
The browser layer uses published EPSG:4326 unique-lake centroids, not the equal-area lake polygons. The polygon dataset is used only for its source area/perimeter attributes in this release.
AREA_DISSOLVED is the dissolved maximum mapped extent across the 2017–2024 unique-lake record and must not be interpreted as a lake area measured on 2026-09-07.
Snow, icebergs, seasonal freezing, observation timing, water-level variability and mosaicking can affect classification and mapped area.
Connectivity is the source GLO glacier-fed/non-glacier-fed classification. It does not identify a specific RGI glacier and must not be converted into glacier-lake causality.
No GLO record is labelled dangerous or safe by this atlas. Expansion statistics describe mapped change and are not a GLOF hazard assessment or forecast.
The source does not publish lake names in this layer; names remain UNKNOWN rather than being inferred from nearby features.
Phewa Lake observation window only; not Nepal-wide water coverage.
Mapped open-water candidates require NDWI > 0.05 and SCL water. Land requires NDWI < -0.05 and non-water SCL. All other pixels are UNKNOWN.
Cloud, shadow, snow, defective, unclassified and NoData pixels plus a 20 m buffer are excluded. SCL and index errors can remain.
Minimum mapped water patch is 9 eight-connected 20 m cells (0.0036 km²). Narrow rivers and small ponds may be absent.
April observations differ in day, year, sensor and conditions. Gain/loss is a classification difference, not a causal claim, flood footprint, trend or forecast.
Only pixels valid on both dates contribute. Area is partial mapped area, not total lake area; comparisons below 80% jointly valid coverage of either the full window or the union of mapped water are unavailable.
No independent ground-truth accuracy estimate or confidence interval is available.
Data obtained from the NASA Langley Research Center (LaRC) POWER Project funded through the NASA Earth Science/Applied Science Program (POWER v10.0.0, MERRA-2). Licensed CC BY 4.0.
MERRA-2 is a reanalysis: modelled fields constrained by observations, not station measurements. Each cell is about 55 x 60 km and cannot resolve valleys, slopes or individual settlements.
PRECTOTCORR is a monthly mean daily rate (mm/day). Monthly totals use the number of days in the month; daily extremes and cloudbursts are not represented.
SPI-3 compares a 3-month total with the 1991-2020 distribution for the same months in that cell. It is a relative dryness/wetness index, not a drought declaration, crop or water-supply impact, or forecast.
In the dry season (about November-April) 3-month totals are small, so modest absolute differences can give large SPI magnitudes; read SPI together with the millimetre totals.
T2M_MAX is the highest hourly 2 m temperature in the month. A positive anomaly is not a heatwave classification; heat-health impacts are UNKNOWN.
Snow depth and snow-cover fraction are model land-surface outputs and are known to be uncertain over high mountains and glaciers.
Cells partly outside Nepal are kept whole and labelled with their Nepal area share; values are not reallocated.
This is a periodically refreshed snapshot (latest month shown), not a monitoring service. Official climate information comes from DHM.
This is terrain context, not landslide susceptibility, probability or hazard. Many slope failures occur on moderate slopes and steep rock can be stable; no inventory-calibrated model is applied.
GLO-90 is a digital surface model including vegetation and buildings; about 90 m cells smooth ridges and gullies, so local slopes are underestimated.
Slope classes are declared reporting intervals, not hazard thresholds.
Districts are assigned by cell centre; boundary cells are not split.
Geology, soils, land cover, rainfall triggers, earthquakes and drainage are not included. Reported landslide records are a separate dataset excluded from the public build pending licence review.
Public evidence corpus
Himalayan Disaster Atlas project documents and public model-release metadata
Source link
UNKNOWN — no public source URL is declared in the release metadata
OpenStreetMap completeness and tagging vary geographically; absence in this layer does not prove an asset is absent.
The acquisition bounding box includes neighboring territory, so every published record is filtered/clipped against pinned Nepal COD-AB v02.
This release deliberately limits roads to motorway/trunk/primary classes and links, and bridges to motorway through tertiary road classes.
School, health, emergency and bridge ways/relations are browser point representations using source-provided Overpass centres; source OSM geometry remains the analytical reference.
Major roads are topology-preserving display simplifications at 0.0002 degree tolerance and must not replace source geometry for precise analysis.
Buildings are not included because national OSM building completeness is too uneven for defensible exposure counting in this feature.
No heuristic geographic deduplication is applied. Only repeated occurrences of the same OSM element/class or road way ID are removed, preserving stable source identity for future exposure deduplication.
The Atlas preserves the licence/attribution text supplied by each release rather than substituting one project-wide licence for upstream data. Where a derived product combines multiple parents, its catalog lineage remains the authoritative path to each parent licence.
Source links are not credentials. Private account secrets, OAuth tokens, API keys and local acquisition configuration are not published in this directory.