Hi @xuehaixiaobai ,
It’s always important to be sure of your definitions. When you say, “potential evaporation”, do you mean “potential evapotranspiration” (PET)? The InVEST model takes the reference evapotranspiration (ET_0) as input and this is modified by the K_c coefficient in the biophysical parameters table to calculate PET. In other words, ET_0 is the potential evapotranspiration of a model grassland with grass 12 cm tall and an albedo of 0.23 etc. “Potential” here essentially meaning what you’d expect if the plant was growing healthily and not limited of water etc. So the K_c coefficient is a modifier to ET_0 that accounts for the actual vegetation present in a given LULC. Thus, InVEST’s PET would seem to be equivalent to FAO-56’s ET_c term, that is the potential evapotranspiration of a given LULC.
Thus, if your PET data source takes into account the LULC (e.g. via LAI, essentially via some knowledge of the characteristics of what’s growing in a given pixel), then I think you could plug in that PET in place of the input ET_0 but you should set all your K_c coefficients in your biophysical parameters table to unity to avoid double counting. I’m not familiar with the InVEST codebase, but from reading the user guide, the calculation of PET (from ET_0) is the only place I see K_c appear, so this approach may work. I’ve actually posed this question on a recent post myself, so hopefully will get a response from the experts in due course. I wonder if there are then constraints or caveats regarding the use of the same precipitation data as (may have been) used by the algorithm to calculate that PET…
However, whilst it might be tempting to just plug in a data product such as MODIS’ PET_500m and set all your K_c to 1, you still need precipitation as an input to the annual water yield model. Along with temperature data, which should be one of the easier datasets to obtain for a region, you’ve really got all you need there to estimate ET_0 anyway using the modified Hargreaves method. You’d still need to define the rooting depth for your LULC classes, so you can’t afford to be completely agnostic/ignorant of the LULC. You might then still argue that you could save yourself needing to obtain K_c values, but I understand FAO provides a bunch of them; I haven’t gotten that far yet, so don’t know how comprehensive that data is.
I hope this is useful. As this is the result of my own endeavours to understand this space, it’s subject to such caveats, but hopefully on this forum my understanding will be subject to validation/correction!