Interpreting AWY model outputs, and a note re. root restricting layer depth

I have run the AWY model but am having some difficulty interpreting the results.
The water yield and actual ET mean values appear to be a different magnitude than observed values. Wyield volume is 22,744,819,046 m3 vs observed 7,232,071,005 m3 (+/-20 year average), and for AET 508 vs 1800.
My study area is characterized by high water availability.

I am aware that the model results cannot really be compared with observed data until the sensitivity analysis and calibration has been done, which is what I intend to do in the next steps. But looking at the raster outputs and the range of values (wyield: 378 - 2023; mean 1974. AET: 436-1578, mean 507), it almost seems as if the wyield and AET outputs should be reversed (wyield output=closer to observed AET values and AET output = closer to wyield observed values in mm, which is 570).
Pending my further steps in calibratining, I was just wondering if it is at all possible that the model calculation somehow mixed up the wyield and AET?

And just a note regarding RRLD and PAWC, I am using the recommended soilgrids depth to bedrock R-horizon dataset, which is capped at 2 m (although the soil depth in the area is probably larger). However, when the RRLD raster provides values <2m and PAWC is calculated over 0 -2m, it is a bit strange to have Plant available water beyond the soil depth layer. So I set the the RRLD at a constant value of 2m for my entire study area. Does this make sense and could this somehow affect the outputs in a way as described above?

Apologies for the long text.


Attach the logfile here:

InVEST-natcap.invest.annual_water_yield-log-2023-10-06–09_24_07.txt (18.6 KB)

Hello @lisakb ,

As of InVEST 3.14.0, these outputs are being produced in the correct order … they are not mixed up.

As for the rest of your questions, maybe @swolny has some insight, especially into your question about input layers?


1 Like

Hi Lisa -

In the model, it adjusts PAWC using the minimum of the soil depth (RRLD) and rooting depth of the vegetation class on that pixel (here’s the methodology in the User Guide). I’m not sure if that helps ease your concern or not.

This is a very simple model, so it’s entirely possible that it’s not capturing a process that’s important to your water yield, which causes the overestimation. One other thing you could check is that the units of your inputs are what the model requires.

~ Stacie

1 Like

Thank you James and Stacie for responding. The inputs are in the correct units, so I don’t expect that’s where the issue lies. But I did have some difficulty in finding appropriate input for estimating the crop evaporation coefficient among others (data is scarce for my study area), so it might be caused by some of the input parameters. I will have a better look when calibrating the model and try as well to find some typical values for Z for similar basins and climates in literature.

Perhaps just one other thing. Since I don’t expect the strange output values are due to the ET, or P raster inputs, I’m thinking to look further into the Kc and Z parameters (also bringing me back to the sensitivity and calibration). But from the guidance documentations and literature suggestions, I read that one study found the AWY model treats natural vegetation the same as agriculture.
For my research focus, that would be an issue, since I’m looking into sensitivity to shifting cultivation compared to forests. Would you say it then makes less sense to use AWY model as it is not appropriate for that purpose?

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.