hi
i have calibrated a snow/glacier-fed mountainous watershed.
The calibration results are good with R2 and NSE >0.80 and p factor 0.77 and r factor 0.71, PBIAS 7.5
I have got the following values.
v__SFTMP.bsn -1.750000 v__SMTMP.bsn -5.850000 v__SMFMX.bsn 1.100000 v__SMFMN.bsn 2.100000 v__TIMP.bsn 0.083333
Is it Ok to have SFTMP < SMTMP and similarly SMFMX < SMFMN for a watershed in the Northern Hemisphere?
It really depends on how you wish to define "ok".
Regarding the model's functionality, yes, it does not matter very much: even if the calibration process suggested you some parameter values that are physically ridiculous, as long as each value is within its own "functionally acceptable" range, the model can still work and yield discharge result for you.
You can find the reference to this range in the SWAT-CUP interface, "Calibration inputs" >> "Absolute_SWAT_values.txt" (and some of the parameter values can even be outside the given range there).
On the other hand, if you are wondering whether the physical meaning of these calibrated parameters are accetpable, then it is a much more complicated question.
The quick answer to your question would be, no it is not ok.
A suggestion would be to specify the numerical relationship in the initial calibration setting so that this situation can be prevented from the beginning. Assuming you are using SWAT-CUP for calibration, when you add SFTMP and SMTMP in the "Par_inf.txt" of the interface, the third larger column is "Filter condition (optional)", and the last sub-column there is "Conditional filter". Select you desired parameter, and you will see a "..." under this sub-column. Click it and you will open the Conditonal Filter Editor, then include both SFTMP and SMTMP via the "New Parameter", then define their relationship on the right pannel. It is a very simple process, I trust you can understand how to use it once you see it. If not, you may ask me again.
By this way, I think the problem itself can be solved.
However, as I said, it is actually a more complicated question. So I have a longer explanation for you on this issue, you can keep reading if you are interested.
--------------------------------------
Before explaining any further, one essential point you need to understand is that an optimization process such as the one that SWAT-CUP use is highly statistics-based. In other words, the calibration process does NOT guarantee the physical meaning of any parameter at all during the calibration process. As a result, even when the NSE, P-bias, or r^2, are all very good after the calibration, it does not necessarily mean the numerical values of the calibrated parameters are physically true to the local reality in your target catchment.
If this notion is the definition of "ok", then theoretically you need to make sure the physical meaning of every calibrated parmater is reasonable and somewhat true to the local reality of your target catchment.
However, you may notice that this notion is rarely realized in most publsihed papers (even from highly ranked journals). This is because without the validation data (not the one to validate discharge result, but to validate the physical meaning of parameters), no one can really prove the numerical values of the calibrated parameters are physically wrong (although at the same time, you cannot prove them correct either). As an alternative solution, the commonly accepted way is to ensure two points:
1. The initial calibration range(s) is physically meaningful (but only in a general sense, not necessarily true to the local reality), so that after the calibration process, the calirated values will still be physically meaningful.
2. The calibrated parmater(s) shows acceptable sensitivity.
Now back to your initial question, if the term "ok" refers to whether the consequent result would be good enough for a thesis or a journal paper, then in most cases, it will be ok as long as you fulfilled these two points and show the evidence in your work.
Hope this helps.
Muqing Shi thanks a lot. your way of answering my query is awesome.
- Badges
- Science topic
- Similar topics
- Phytochemistry
- Extracts