Urban Flood Risk Mitigation Calibration

Hi Everyone,

I am working with the urban flood risk mitigation model and I didn’t find any guidelines on how to calibrate (or validate) the model. I used other InVEST models (i.e. UHI or nutrient retention) and I always calibrated those model with empirical observations.

Do you have any suggestions on how to “verify” the results obtained from the flood risk mitigation model?
Thank you so much!

1 Like

Hi @laura78 ,

While I don’t have personal experience applying this model, from looking at the user’s guide chapter it seems like the most readily-comparable output to calibrate against would be to use the runoff volume raster (Q_m3.tif, found in the intermediate folder in your workspace). Particularly if you aggregated this to a relevant watershed/sewershed, you might be able to compare this against any available water volume gauges in your area of interest. That’s what seems reasonable anyways!

But just to make sure we have a more informed opinion weigh in here, I’ve reached out to our urban team to see if they have anything to add.


Hi Laura, James -

The suggestion by James is probably your best option to calibrate the flood mitigation model. You would need to have a known watershed delineation within (or maybe containing) your area of interest, and this known watershed would need to have associated hydrologic data collection (stream gauge, stormwater monitoring program, etc.) that includes runoff volumes for individual storm events along with the measured rainfall that produced the event(s). If the delineated watershed has rainfall-runoff results from a more dedicated stormwater model (such as SWMM), then those outputs could also be used in place of monitored runoff, acknowledging that it might not be as accurate as observations.

You could then run the flood mitigation model for that known watershed using the measured rainfall as input, and compare that “intermediate” runoff volume raster to the observed runoff volume for the watershed for that particular storm event. I think the raster provides a volume of runoff on each pixel, so you would need to sum all of the pixels within that known watershed to compute total runoff volume and make the comparison to observed runoff volume from the external data. Since the model is intended for larger (flood-type) events, it might be most useful to pick a larger storm event within the data record, if you have the choice!

Once those volumes match pretty well for one event or a set of events – you might have to tweak curve numbers and/or soil or land cover types to get a better match – you could then consider the model “calibrated” and begin analyzing in more detail your area of interest (if it doesn’t correspond to the known/delineated watershed boundary).

If there are not existing runoff and rainfall data for your target watershed, then it may not be possible to precisely calibrate the model… but you may still be able to use it in a comparative framework, such as looking at the change in flood mitigation among (for example) several land use scenarios, acknowledging that the absolute outputs (e.g., cubic meters of retained runoff) should probably not be interpreted literally.

Other folks might have some better suggestions on how to use the model in uncalibrated form, but I hope this answers your question.

Ben J.


Hi Ben,

Thank you so much for your response…it was a great help!!

This topic was automatically closed 7 days after the last reply. New replies are no longer allowed.