Distance matrices, river-network crosswalks, water temperature observations, and river attribute data for twelve river basins in the United States

This data release contains input data used to test multi-scale modeling approaches for predicting water temperature in streams and rivers. The target application aims to predict stream temperature at relatively fine spatial resolution across a river basin by combining a machine learning (ML) model (optionally, with additional input data) configured at a relatively coarse spatial resolution with an ML model and additional input data at the fine resolution. The fine- and coarse- spatial resolutions were represented by two common river hydrography datasets. In each of 12 focal river basins, the coarse spatial resolution was represented by the Geospatial Fabric for Hydrologic Modeling (GFv1.1; Bock et al. 2020) and the fine spatial resolution was represented by the NHDPlusv2.1 flowline network. The model inputs include water temperature observations as well as information about the spatial relationships among river segments; stream and catchment characteristics; and daily meteorological drivers. Certain input datasets were prepared only for the fine spatial resolution (i.e., NHDPlusv2.1; flowline distance matrix, water temperature observations) to enable development of fine-resolution models for each basin, whereas other datasets were prepared at both spatial resolutions to provide additional data to inform the multi-scale modeling experiments (river attributes, meteorological data). The 12 focal river basins (436 - 958 km^2) are distributed across the conterminous United States and include Battle Creek, CA; Black Earth Creek, WI; Brandywine Creek, PA and DE; East River, CO; the Lower Delaware River, PA and NJ; the Lower West Branch Delaware River, PA and NY; Manhan River, MA; Neversink River, NY; Rancocas Creek, NJ; the South Fork McKenzie River, OR; Trinity River, TX; and the Upper South River, GA. This data release includes nine files that contain the model input data at one or both of the spatial resolutions described above for each river basin: 1. nhdv2_distance_matrix.npz: File includes a matrix indicating the upstream distances among flow-connected segments in the NHDPlusv2.1 river network. 2. nhdv2_nhgf_crosswalk.csv: File includes a crosswalk table that maps NHDPlusv2.1 flowlines to GFv1.1 flowline segments. 3. nhdv2_temp_observations.parquet: File includes water temperature observations summarized to daily values and aggregated to the NHDPlusv2.1 flowlines. 4. nhdv2_static_attributes.parquet: File includes attribute features that represent characteristics of the river segment, its catchment, or the upstream watershed area, summarized to the NHDPlusv2.1 flowlines. 5. nhgf_static_attributes.parquet: File includes attribute features that represent characteristics of the river segment, its catchments, or the upstream watershed area, summarized to the GFv1.1 segments. 6. nhdv2_inputs_io.zip: Zipped parquet file contains model input data including river segment characteristics and daily meteorological data, summarized to the NHDPlusv2.1 flowlines. 7. nhgf_inputs_io.zip: Zipped parquet file contains model input data including river segment characteristics and daily meteorological data, summarized to the GFv1.1 segments. 8. spatial_data.gpkg: File contains the spatial data representing the basin boundaries, the GFv1.1 segments and catchments, and the NHDPlusv2.1 flowlines and catchments. 9. source_code.zip: Compressed file contains R code used to generate the model input datasets.

[
  "010:12"
]

[
  "geospatial"
]