TempEst-NEXT: Source Code, Automatic Validation Suite, and Test Data

This resource contains data and code related to the development of the TempEst-NEXT river temperature forecasting model. The provided data and automatic validation suite should be sufficient to fully reproduce the performance analysis contained in the associated manuscript. The code includes model implementations for TempEst-NEWT (watershed-specific model) and TempEst-NEXT (ungaged model) as well as some utility functions for data retrieval and analysis. Data files include the development dataset (DevData) and independent test dataset (TestData) with both estimated weather (...Buffers.csv) and forecasted weather (...HRRR3.csv) as weather inputs. The NEXT_validation notebook automatically analyzes NEXT performance (with some random variation, so results may not be identical to the published values but should be close).

NEXT_validation_20250917.pdf is a knitted version of the validation notebook, containing all key results, as of September 17, 2025 (model version used for manuscript submission).

To run the entire Notebook, requirements are:
• Python dependencies: tempest-next, matplotlib, seaborn, and scikit-learn. tempest-next is available (under that name) on the Python Package Index, so pip install tempest-next. This will also install, as a dependency, TempEst-NEWT, the calibrated version of the model.
• A directory (specify location in bp= in the first cell) containing DevDataBuffers.csv, TestDataBuffers.csv, DevDataHRRR3.csv, and TestDataHRRR3.csv. These should be provided with the notebook. It should also contain a results subdirectory.
• In the same directory as the notebook:
– An ecoregions directory, containing NA_CEC_Eco_Level1.shp (EPA Level I Ecoregions).
– A usa_states directory, containing usa_states/cb_2018_us_state_20m.shp (US Census state outlines shapefile).
– A val_figures directory
– A results directory (in the directory `bp`, which can be the same directory)

The above requirements, other than the Python dependencies, are included in this data release, so, with `bp=(current directory)`, it should be possible to run everything from this download.

These data and related items of information have not been formally disseminated by NOAA, and do not represent any agency determination, view, or policy.