InVEST SWY Model Taking Over 14 Hours for Large Study Area

Software Support
1

Hey software team,

I’m running the InVEST Seasonal Water Yield (SWY) model for Uganda at 30 m resolution, and it has been running for over 14 hours.

The quickflow (QF) and local recharge (L) rasters were generated approximately 1.3 hours after the model started, and the baseflow (B) raster was generated after about 5.5 hours. However, the model is still running, apparently processing watershed-level statistics.
Since I’m mainly interested in the B, QF, and L raster outputs, I assume I can start using them for analysis while the model is still running, as long as those calculations are complete. Is that correct?
I’m also wondering if you have any thoughts on why it’s taking so long and whether this is expected for a study area of this size. The log is still showing progress, but I’m not sure how much longer it might take.

2

  1. Using completed outputs: Raster outputs that have been fully written can likely be used for analysis while the model continues running, since they are not expected to be overwritten during the remaining steps.

  2. Long runtime: Country-scale runs at 30 m resolution can take considerable time, particularly for models involving flow routing. Calculating watershed statistics and writing the .yml file may take several additional hours, potentially accounting for much of the total runtime.

  3. Testing at coarser resolution: For preliminary model testing and validation, consider using a 90 m DEM to reduce processing time, then return to 30 m for the final run.

  4. Dividing the study area: Running the model separately for larger subwatersheds may be another option