Hi,
Please let me know is this the right approach for the calibration of flow in SUFI2?
1. I should first check the performance of initial model with observed data :
i. Should I compare the simulated results given as "flow_out" in the "rch" file in SWATOUTPUT.mdb with the observed data?
ii. Should this be performed for the whole discharge period or only the period for which I intend to perform calibration?
iii. Should I compute the waterbalance before doing calibration using my first model and pleae let me know what else i should check apart from comparing the observed and simulated discharge in the first step.
2. Based upon the comparsion of simulated and observed flow I should do the regionalisation of parameters.
2. Sensitivity Analysis to find out the parameters most sentive for the watershed with regards to flow.
3. I should set the range for the parameters identified in step 3.
Can i first use the ranges provided as default in SWAT-CUP or I should set the ranges based on literature review (in case I find any). For this step I should run the simulations from 500-1000 times and check the values of Objective funcion and 95PPU and r-factor. If the ranges are satisfactory then those ranges set for the parametsr will be final otherwise i should repeat the model for another 500-1000 simulations by changing the ranges.
please correct me if my approach is wrong.
thanks
My suggestions as per your mentioned points wise
1. Compare the model simulation with the observed data graphically and statistically to check whether the model is properly set up or not. If there is huge difference in the comparison, check the quality and errors in the data inputs. For calibration, you can run only the intended period.
2. yes you can do regionalisation of parameters
3. after every iteration process during calibration, sensitivity of parameters can be computed and accordingly identify the non-sensitive parameters (which can be removed from the calibration set parameters)
4. Set the range of parameters with accordance to the characteristics of the catchment. or set the maximum range possible reasonably and try to iterate the processes as per your results.
Dear Saima,
In addition to the very good points of Shanbor Kurbah, two additional points:
- It would be good to calculate the water balance for a long time period. Then, the change in storage is relatively small compared to the incoming (precipitation) and outgoing (discharge, evapotranspiration) fluxes and you can determine the long-term actual evapotranspiration based on precipitation and discharge data. Obviously, the actual evapotranspiration should be smaller than the long-term potential evapotranspiration (which can be calculated using meteorological data only).
- I am not sure what you mean with regionalisation and how you would like to do that, but generallly it is quite difficult to establish reasonable regionalisation relationships (relations between catchment characteristics and model parameters). This is because most model parameters (e.g. in SWAT) cannot directly be observed in a real catchment, but rather are part of the conceptualization of the hydrological reality in a model. Furthermore, in the calibation different uncertainty sources (data, model structure, model parameters) will influence the calibration results and estimated parameters and therefore also obscure the regionalization relationships.
Good luck,
Martijn
- Badges
- Science topic
- Similar topics
- Geoscience
- Hydrology