I’m using the SWY model in Uganda across areas with some large waterbodies (including Lake Victoria) and wetlands and I’m trying to work out the best way to handle runoff and evaporation processes in these systems.
If I set a very high curve number (say 98-100) for waterbodies, then surely the quickflow yield is over-estimated? I say this because my understanding is that the model only calculates evapotranspiration losses on the amount of water remaining after quickflow has been calculated. Even with a high Kc value, the evapotranspiration losses will only apply to the small portion of water that is not considered quickflow (which would be zero if using a CN of 100), so this does not really solve the problem. So using this approach doesn’t seem to account for the fact that a large portion of the precipitation falling on the lake is lost to evapotranspiration, and will not actually reach the watershed outlet.
I’m curious to hear how other modellers have handled this issue. Do you perhaps mask out large waterbodies and ignore their contribution to quickflow and/or local recharge, thus focusing on the contribution of the terrestrial landscape instead? Another option I see would be to give waterbodies a lower CN in combination with high Kc values. This will then mean a larger portion of the incoming rainfall is evaporated. However, I see no guidance for this, as studies always use high CN for waterbodies since little to no infiltration of rainfall is occurring.
Wetlands raise a similar issue. For wetlands, I have been using a curve number of around 80 and high Kc values. However, I am still finding that my quickflow estimates alone are much higher than the flow estimates from gauging stations in catchments with high wetland and open water coverage (this is before even considering any contribution of baseflow to streamflow). Notably, this is not the case when I compare my modelled flow estimates to gauged catchments which do not have many wetlands and lakes. Again, it seems I need to somehow incorporate greater evaporation losses from wetlands, though a CN of 80 for wetlands is already low based on what I’ve seen in other studies. Interestingly, I have seen that measured flows from gauging stations in Uganda often decline as one moves from the upstream to downstream end of a wetland complex, suggesting that high evapotranspiration results in a negative water balance in many of these systems. Again, I can’t see how to capture this other than using an unusually low CN and high Kc values.
Interested to hear any suggestions for handling these systems within the model, and sorry for this long post!
Hi Luke, I am glad to see that you are digging into the SWY model! Here are some thoughts in reaction to your post:
The SWY model does not directly model many of the dynamics in wetland systems, so it is probably best to focus on getting the terrestrial quickflow/baseflow portion correct, and then handle the extra evaporation from wetlands in post-processing. For example, you could set the CN for wetlands and water bodies high, as recommended in the model settings, and then subtract evapotranspiration losses from the wetland/lake pixels before summing and comparing to your gauge data.
Are you sure that the evaporative losses are causing the reduction in measured streamflow as you move downstream in Uganda? I agree that this is a valid hypothesis and might very well be correct, but I would also check if there are some other avenues for losses that might be affecting the streamflow. For example, are these wetlands possibly perched above a permeable layer that could be recharging an aquifer below? Are there direct extractions of water for irrigation or urban water supplies that could be reducing the flows? If any of these conditions exist, they could also be handled in post-processing. Possibly you already excluded these but I wanted to mention for completeness.
Thanks Doug and Adrian, and great to connect with you again so soon Adrian.
My apologies for taking so long to respond as I had some other pressing work to attend to.
I am proceeding with your suggestion to use a high CN for wetlands and lakes and then subtracting measured actual evapotranspiration after that. Would it make sense to use a CN of 100 for open water, so that everything can be assigned to quickflow and then evapotranspiration losses subtracted from that? Or would you recommend allowing a portion of rainfall to infiltrate even under open waterbodies? My understanding is that wetlands could be treated the same way, but potentially with a slightly lower curve number than for open water?
Regarding the nature of the wetlands, the wetlands reflecting in the land cover I am using are mostly permanently flooded papyrus swamps, which are extensive in Uganda. While I haven’t found many detailed studies of their hydrology, most reports suggest that the papyrus wetlands reduce overall water yield in Uganda due to high rates of evapotranspiration. Localised studies have also shown that evapotranspiration rates are higher in the papyrus swamps than adjacent areas of open water. So overall my sense is that they are causing a net decline in the water balance rather than net recharge of groundwater.