Calibrating a SWAT+ Model in the Cloud
One real calibration run, start to finish: a SWAT+ model of a watershed in southwest Michigan with a USGS gauge near Sawyer. Choose the windows, the parameters and the optimizer, launch it on a cloud machine, follow each iteration on your dashboard, then read how the swarm converged and whether the held-out years hold the fit.
1:05 · captions on by default
Transcript
We start from a model we already built: a watershed in southwest Michigan, with a USGS gauge near Sawyer. Its ten-year run is compared with that gauge; calibration tunes the model, then tests it on years it never saw.
Calibration takes three choices. The gauge and its windows: twenty eighteen to twenty twenty-four to calibrate, two thousand to twenty ten to validate. The parameters: all forty-one in the bundle. And the optimizer: a multi-memory particle swarm, forty particles over fifty iterations.
The review shows the plan and its estimated time. Launch it, and the run goes to a cloud machine. Nothing to install, and you can close the page.
While it runs, your dashboard shows each iteration as it completes.
When it finishes, the page shows how the swarm converged, the calibrated flow against the gauge, and whether the held-out years hold the fit.
Calibrate your own model at swatgenx.com.
Credits
Data credit: USGS NWIS streamflow, gauge 04096015 (Galien River near Sawyer, MI); model inputs NHDPlus HR, USGS 3DEP, NLCD 2021, gSSURGO and PRISM; SWAT+ by USDA-ARS and Texas A&M AgriLife Research. Basemap: © OpenStreetMap contributors (openstreetmap.org/copyright).
