Changelog

Source: NEWS.md

hydrofabric 0.1.0

First public release. Builds the NextGen hydrofabric for the National Water Model from the Lynker Reference Fabric and D8 flow-direction and flow-accumulation grids.

Pipeline

• refactor_fabric() splits and collapses reference reaches and reconciles them and their divides against the grids.
• aggregate_fabric() aggregates the reconciled network to distribution scale, conserving divide area.
• nextgen_fabric() builds the NextGen data model: flowpaths, divides, nexuses, flowlines, network, hydrolocations, POIs, lakes, and RiverML channel routing attributes.
• merge_vpus() merges per-VPU products into the SuperCONUS, Alaska, Hawaii and PRVI domain products with globally unique identifiers, and records an upstream index for range queries (add_upstream_index()).
• run_pipeline() runs the phases for one VPU; run_vpus() runs a set.

Validation

• Every phase is checked by check_invariants(): topology (acyclic, resolvable toids), divide coverage and area conservation, one divide per flowpath at most, and the POI placement contract. A VPU that fails does not ship.
• Inputs are checked before a build: check_reference_input(), check_reference_severed_terminals() and check_grid_coverage().

Provenance

• hf_input_manifest() and hf_output_manifest() record every input and product with its checksum; hf_verify_manifest() and hf_verify_output_manifest() check them.
• Every product carries its license, attribution and build metadata in its gpkg_metadata table.
• prepare_hydrolocations() snaps and assigns a raw hydrolocation set to a reference before a build.

License

Code is Apache-2.0 and data CC BY-SA 4.0, both © Lynker Corporation. See LICENSE.md, NOTICE and THIRD_PARTY_NOTICES.md.