Skip to content

Louisiana Watershed Initiative EnDMC

LouisianaEndmc reads the Louisiana Watershed Initiative (LWI) Environmental Data and Model Catalog (EnDMC) and downloads its public model and data files. EnDMC is operated by The Water Institute. It catalogs HEC-RAS, HEC-HMS, HEC-FIA, and Go-Consequences model applications and supporting datasets for the eight LWI regions.

No account is required. RAS Commander never logs in to EnDMC and accepts no credentials for this source.

Need Public endpoint
Catalog GET https://lwi.endmc.org/search with Accept: application/json (the request the public search page makes)
Software per resource the same endpoint with software=<id>
Resource details public page https://lwi.endmc.org/view/<kind>/<id>
Files anonymous S3 ListObjectsV2 and range GET on bucket lwi-model-data (us-east-1)

Files live at endmc-managed/{model_application|dataset}/{24-hex id}/. A dated snapshot of the public catalog fields (id, kind, title, software, record date, keywords) ships with the package and is used automatically when the live catalog is unreachable.

The executed example notebook 932: Louisiana LWI EnDMC walks through each step below on live public data.

Browse the catalog

Python
from ras_commander.sources.state import LouisianaEndmc

catalog = LouisianaEndmc.list_resources()                 # model applications + datasets
ras = LouisianaEndmc.list_resources(software="HEC-RAS", region=2)
coastal = LouisianaEndmc.list_resources(search="coastal")  # site full-text search
offline = LouisianaEndmc.list_resources(source="snapshot")

details = LouisianaEndmc.get_resource_metadata("6a3d7158f74d9b1d5415493a")
details["fields"]["Coordinate system"], details["containing_packages"]

region is derived from titles and keywords ("Region 5"); a few statewide or pilot resources have none. date is the record date from the bundled snapshot; use get_resource_metadata() for resources published after the snapshot. list_s3_resource_ids() inventories the bucket directly, which can reveal prefixes that are not in the public catalog.

Inspect a resource's files

Python
objects = LouisianaEndmc.list_objects("685d772c2ecdbd05b0cb03d1")
objects.groupby("component")["size"].sum().sort_values()

Each object is classified into one component:

Component Contents
ras_project .prj project, .p##, .g##, .u##, .f##, .rasmap, .u##.hdf
geometry_hdf, plan_hdf .g##.hdf, .p##.hdf
ras_run_artifact .b##, .bco##, .c##, .x##, .o##, .rst, backups, snapshots, compute messages
terrain, land_cover, infiltration rasters and RAS Mapper layer HDFs in terrain, land-cover, or soils folders
output_depth, output_wse, output_velocity, output_raster_other flood output rasters and stored maps
hms, dss, gis, docs, metadata, other HEC-HMS files, DSS, GIS layers, documents, catalog metadata, everything else

Classification is a path heuristic tuned to the published deliveries. Raster sidecars (.aux.xml, .ovr, .tif.xml) follow their raster. Review the other rows before relying on a component filter for completeness.

Download selected components

Python
result = LouisianaEndmc.download_resource(
    "6a3d7158f74d9b1d5415493a",
    r"H:\Testing\endmc\abbeville",
    components=["ras_project", "geometry_hdf", "terrain", "land_cover", "dss"],
    max_workers=4,
)
result["status"].value_counts(), result.attrs["receipt_path"]
  • Resumable. Each file is assembled in <name>.part. Files larger than chunk_size (64 MiB) are fetched as parallel byte ranges, and completed ranges are recorded in <name>.part.json. Rerun the same call to resume. Ranges carry If-Match with the listed ETag, so a file replaced on the server restarts rather than mixing versions.
  • Verified. Every file is checked against the listed byte length. verify="checksum" also compares the full-object checksum S3 stored at upload, using the algorithm the listing reports. Published objects carry CRC32 (verified with the standard library) or CRC64NVME (requires the optional awscrt package, pip install awscrt). Objects without a stored checksum fall back to size and are counted in attrs["size_only"]. S3 ETags for multipart uploads are not MD5 digests.
  • Receipt. _endmc_<resource_id>.json in the destination records the key, size, ETag, modification time, status, and verification of each file.
  • Short paths. Resource paths are already long. The default max_path_length=240 refuses a download whose destination paths would leave no MAX_PATH headroom for HEC-RAS; choose a short destination such as H:\Testing\endmc\<short-name>.
  • Guards. dry_run=True returns the plan with sizes. max_bytes refuses oversized selections; download_model() (the ModelCatalog entry point) defaults to a 50 GiB limit because design-storm resources exceed 400 GiB.

Measured throughput from one workstation (November 2026) was about 60 MB/s on one stream and about 120 MB/s with four streams.

Read rasters without downloading

Python
rasters = LouisianaEndmc.get_raster_urls("69fc95b64c0345e94060a3bc")
path = rasters.loc[rasters["component"] == "output_depth", "vsicurl"].iloc[0]

import rasterio
with rasterio.open(path) as src:          # HTTP range reads through GDAL
    print(src.crs, src.nodata, src.overviews(1))

LouisianaEndmc.read_range(key, start, length) returns raw bytes for any object.

Data notes

These observations describe published files reviewed in November 2026. They are not library behavior; the library does not modify downloaded data.

  • Region 2 online output rasters declare EPSG:6477 (NAD83(2011) Louisiana North, US survey feet), which matches their coordinates.
  • Region 2 WSE and velocity rasters also carry a vertical CRS named "NAVD88 depth". That vertical component is not meaningful for velocity and should not be read as a datum statement for those values.
  • Region 7 mosaicked rasters use a nodata value of about 3.4e38 (the float32 maximum); mask it before computing statistics.
  • Product sets differ by resource. A resource may publish representative events (for example seven Atlas 14 return periods at one duration) rather than every modeled scenario.