Simulating Climate Change impact

[Zé Pedro Ramião]

Dear all,

I’m using SWAT-CUP to calibrate discharge, sediments, and phosphorous, and then I want to assess the impact of climate change

In SWAT-CUP guidelines (Karim et al., 2015, http://dx.doi.org/10.1016/j.jhydrol.2015.03.027) and video tutorials (2w2e YouTube channel), it is suggested that the parameters selected depend on the behavior of the initial model. The simulated discharge is quite similar to the observed discharge (NSE>0.7) in my initial model, but I still want to calculate P-factor and R-factor. Since my initial model does not mirror any of the behaviors in Table 2 of Karim et al. (2015), should I select all discharge related parameters and perform a global sensitivity analysis?

I’ve notice that some SWAT papers on climate change report the 95PPU as well as the P and the R-factors, but then the climate change simulations are reported as single values. However, in the video tutorials, it is suggested to make an iteration with the calibrated parameter ranges with as many simulations as before, for simulating climate change (after updating pcp1.pcp & tmp1.tmp). Since I’m using 4 Regional Climate Models, does it mean I will have one 95PPU for each climate model, and then four 95PPU for each climate scenario? In that case, the 95PPU for each climate scenario is the lowest and the highest values of the 95PPUs of the four RCMs?

Thanks in advance

Best regards,

José Pedro Ramião

[Natalja C.]

Hello,

1. You can do that, if you want. Or you can select a smaller sample of parameters based on the conditions of your modelled region. Anyway, you should do a sensitivity analysis before calibration.

2. Yes and no. You see, papers often have "what the reviewer asked for", hence some report  95PPU, some don't. Some use autocalibration, some don't. If your want to quantify the uncertainty related to the model internal functioning/process representation - then you should do as you described. If you want to quantify the average effects of climate change, then that might not be necessary. Still, I would consider the uncertainty analysis as a good addition to any modeling study.

Best of luck,

Natalja 

[Zé Pedro Ramião]

Dear Natalja,

Many thanks for your reply, it was really helpful

Best,

José Pedro Ramião