Analyzing Seasonal Water Yield Results: very low values of Qf

Hi, Nat Cap!

Can you help me please with some calibration in Seasonal Water Yield model? I came trough the results of the SWY model for some watersheds in south Brazil, and while some results seem very fair to me (as annual AET), others are difficult to me to interpret. I managed to have actual evapotranspiration rates very similar to observed data (about 900 mm), using CN and Kc parameters based in local literature. But the baseflow (Qb) is relatively higher (557 mm/yr) that other studies in the same region (the areas with bigger baseflow rarely overpass 300 mm/yr, according literature). What caught my attention was the low average value of quickflow (68mm/yr); I'm not sure if quickflow in InVEST's context means streamflow (if so, the actual values are much larger than SWY output). Do you have a clue on how to balance the Qb and Qf values to have more realistic outputs?

I have created a monthly table for the alpha parameter, and have amde few tests with beta and gamma, which seem to have not much influence in the results.

Thank you, cheers,


  • PerrinePerrine Moderator, NatCap Staff
    Hi João,

    Thanks for sharing this application of the Seasonal Water Yield model.
    Here are a few comments that I hope can help interpreting results.

    1. If the annual AET seems right, the total flow (baseflow: Qb + quickflow: Qf) must also match the observed streamflow. This is because at the watershed scale, the water balance is closed (ie. P=QF+B+AET).

    2. Re: the quickflow: it represents the surface runoff from a pixel, as predicted by the SCS Curve Number approach. It is a very simple approach and I have found also that in many cases the direct runoff is very small (which is fair for most vegetated LULC, since only large events produce surface runoff).

    3. The model is not a traditional rainfall-runoff model in the sense that the routing is very simplified. The indices (B, QF) have units but they are interpreted as relative values to rank pixels. Ultimately, our goal is to provide better guidance on how to use the quantitative values too (because it helps with model verification) but things are going more slowly than expected. I can share with you the analyses that I'm doing now in Peru to go more in-depth on these questions.

    4. I'm surprised that beta did not have a large effect on the results. You may want to change alpha too, which affects the proportion of the annual contribution from above areas that is available to a given pixel.



  • Thanks, Perrine!
    I 've just got back from a short leave and I will try to apply your guidance on it, hope it works! I'd like to take a look in these analyses on Peru, thanks, it's always useful to see results of similar applications. I'll be back soon with some news after further tests.
    Thanks again
  • Perrine, hope you are well!
    Would you mind if I send to you my datastack for you to take a look, run it, and see if the results are plausible? 

    Thanks, cheers
Sign In or Register to comment.