I have completed the simulation through the SWAT model. However, as I don't have the measured flow data, I don't know the accuracy of model output, and I can't adjust the parameter based by measured data, which could improve the model output.
A professor told me that an uncalibrated, or limited calibrated model is still a useful management tool. In fact, he believes SWAT was developed in part to evaluate ungauged watersheds.
However, I believe that it is necessary to verify the accuracy of model output and adjust the parameter to improve the model output through comparing the simulated flow data and measured flow data.
I can't continue this work now, what should I do now? Can I apply the simulated data without verifying?
Hi Yi:
While SWAT is developed for ungauged watersheds, it should be calibrated and validated by some means in order to use it for answering applied questions in your watershed. Do you have monitored data within your watershed somewhere? If so you could do an internal validation. Another option is to look for data downstream and scale your watershed model. It is common in many countries to not have continuous data; in such cases you could rely on using the results for general policy analysis and not for predictive purposes. I need more information about your specific case to provide proper guidance.
Hi,Timothy O Randhir.
Thanks a lot!
We have some monitored data, but I think it is enough to verify the output of model. We have 9 days flow data for 6 years, and not continuous. Besides, we installed some transducers which could measure the pressure in the river at the beginning., we have enough pressure data.
I used want to apply these pressure data to estimate the flow data, and use the formula of the curve fit of the 9 calibration points to calculate the flow values.
However, a professor believe that in the datalogger flow calibration, the R2 value is very low i.e. 0.3238. This is the result when matched for only 9 days data. If I use such an erroneous method to estimate observed flows, the result will be misleading. I accept his opinion.
We also don't have flow data of downstream, and to get more monitored data seems like impossibile.
Dear Yu,
Modeling is basically invented to become alternative of measurement.
SWAT model already calibrated and validated. However, for accuracy your work if you have meassured data it would be great. In case you don't have measured data, your results will be acceptable.
Try checking at least evapotranspiration, use the data of precipitation and the number curve to estimate the flow, you can still generate the sediment yield by universal equation of soil loss. Apply in respresentativas sub-basins. I have helped.
Dear Liu:
I strongly suggested that for applying SWAT model to simulate the runoff and nutrient, the model builder should prepare enough observed data for calibration adn validation (R2 should be larger than 0.5 and also Nush-Sutclif Efficient should be larger than 0.5), daily observed runoff data and nutrient data are necessary.
If you do not have such data, first way is to obtained other "observed" data such like evapotranspiration (ET) data for model calibration, ET data may avaible for monthly step from the remote sensed data (MODIS MOD16 or other availble data source).
Anyway, without validation, the other research cannot trust your results and conclusions.
Hi Liu,
calibration and validation of models like SWAT is very important for most applications but not always necessary. For example in cases where you want to look at the change in flow or nutrients due to some kind climatic or land management shifts, you may not really need to calibrate and validate your model if you don't have data at all.
However, in your case it looks you may want to, some how, calibrate/validate your model. One thing you can do to 'by-pass' calibration/validation is to look around other SWAT models studies published for watersheds around your study area region or country or even continent, and make lists of parameters calibrated and ranges of their values. If you have too many of such studies, you can narrow your list by choosing studies which used watersheds that have similar landuse/landcover, soil, and/or topographic characteristics as your watershed. You can then use those parameters to at least estimate ranges of flow and nutrient loading for your watershed.
Good luck.
just a comment.........if you have complete data, then why you do need model and work on a big modelling tools? since we have limited data, we go for such huge and complicated task of modelling.
I also have a question, somehow like above!
I have calibrated a watershed with SWAT, and want to calibrate the nutrient loads, I have data for that but the flow is also needed.
Is there any way to use nutrient ex: NO3 (mg/lt) for calibration that needs kg/month?
- Badges
- Science topic
- Similar topics
- Agricultural Science
- Agronomy