File size: 4,962 Bytes
110026e 8c34c86 110026e 8c34c86 110026e 8c34c86 110026e 8c34c86 110026e | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 | ---
license: cc-by-4.0
pretty_name: h3-lowpoints explorer data
tags:
- geospatial
- h3
- hydrology
- flooding
- united-states
---
# h3-lowpoints explorer data
The data the [h3-lowpoints](https://github.com/UJJAWAL-01/h3-lowpoints) explorer web app reads in
the browser: every closed depression on land in the lower 48 US states, packed by area so a page
can load one place at a time, plus the NOAA Atlas 14 tables that say how rare a storm is.
Built from the national depression dataset, version 0.3
([doi.org/10.5281/zenodo.23731345](https://doi.org/10.5281/zenodo.23731345)), with the h3-lowpoints
library 0.5.0. Nothing here is new analysis: it is the published dataset rearranged for browsers.
The same files are archived on Zenodo: [doi.org/10.5281/zenodo.23824290](https://doi.org/10.5281/zenodo.23824290).
## Files
| File | Size | What |
|---|---|---|
| `packs.bin` | 1.2 GB | 4,576 packs (format 2), one per H3 resolution-4 region, each gzipped: the region's depressions (catchment area, curve number, the fill curve from their measured levels, spill link, drained or landlocked), their H3 resolution-10 cells, and the water reaching them from neighbouring regions for storms of 0 to 1,000 mm |
| `index.json` | 0.2 MB | where each pack starts in `packs.bin` and how long it is |
**Format 2 (November 2026)** adds, for every depression, `target_id` (the dataset id of the depression it overflows into, 0 when the water leaves the data) and `target_region` (that depression's H3 resolution-4 region), so overflow paths can be followed across packs. Every format-1 field is unchanged (checked byte for byte in all 4,576 packs).
| `rarity.bin` | 0.25 GB | NOAA Atlas 14 rainfall depths for storms of 1 to 48 hours and return periods of 1 to 1,000 years, per H3 resolution-7 cell, one gzipped block per region |
| `rarity_index.json` | 0.2 MB | where each block starts |
| `extras.bin` | 40 MB | per pack, in the same regions and depression order as `packs.bin`, gzipped: the rim (spill level) and the lowest ground of every depression, in metres above sea level (NAVD 88, from USGS 3DEP via the national dataset), the distance from its lowest point to the nearest major road (metres, empty beyond 100 m), and a road class |
| `extras_index.json` | 0.2 MB | where each extras block starts, and the road-class rules |
**Extras (version 2.1, November 2026)** add a road-feature flag from the US Census Bureau's TIGER/Line 2024 primary and secondary roads (feature classes S1100 and S1200). Class 0: no major road within 30 m of the lowest point. Class 1: the lowest point is within 30 m of a major road. Class 2: class 1, and the depression is at least 2.5 times longer than wide with its long side within 25 degrees of the road: a road cut or underpass. Such hollows are usually emptied by storm drains the terrain does not show. 114,536 depressions are class 1 and 36,002 class 2, of 10,570,791. The packs and rarity files are unchanged from version 2.0.
Read a pack with one HTTP range request (`bytes=offset-(offset+length-1)` from `index.json`),
gunzip it, and parse it as described in
[`app/src/pack.js`](https://github.com/UJJAWAL-01/h3-lowpoints/blob/main/app/src/pack.js). The
tools that build these files are in
[`app/tools/`](https://github.com/UJJAWAL-01/h3-lowpoints/tree/main/app/tools).
## Checks
The app's simulator, run on these packs in Node and in Chrome, equals the h3-lowpoints library to
rounding (largest relative difference 1e-15) in 36 test cases. Storing the data compactly and
precomputing the water from neighbouring regions changes the water held per depression by a
median of under 1e-5 relative; a few depressions per region, below a region boundary, can be off by
more at some storm depths (the worst seen, 56%). Rarity read from these tables is within 7% of
Atlas 14 at the exact point for 95% of places. Atlas 14 does not cover Washington and Oregon.
The rim and lowest ground in `extras.bin` equal the national dataset (`water_level_m`, `ground_m`) to under 0.02 mm for all 10,570,791 depressions. The road classes were recomputed independently in three boxes (Austin, Houston, Minneapolis; 2,954 depressions) with no mismatch, and the app shows the same values as the dataset for 30 depressions checked in Chrome.
## Limits
Everything in the national dataset's limits applies: 30 m terrain, storm drains and pumps not
included (the road flag marks only the depressions on major roads; drains under streets and car parks are not known), storms treated as totals. These data describe where water can collect, not a flood
forecast or warning.
## Licence and credit
CC BY 4.0. Creator: Ujjawal Patel. Sources: USGS 3DEP elevation, ESA WorldCover, Annual NLCD 2024,
USDA gNATSGO, NOAA Atlas 14 (via the national dataset and the library), and US Census Bureau TIGER/Line 2024 roads (public domain, used directly). Please cite the national dataset
(10.5281/zenodo.23731345) or these files (10.5281/zenodo.23824290).
|