Could you please define how complex the terrain is ? I have experienced to verify WRF surface precipitation (apccaccsfc) output over Java island (Indonesia) which has quite complex mountainous terrain with TRMM/GSMaP dataset using dichotomous grid-to-grid categorical verification method by calculating the CSI, POD, FAR etc of the forecasted rainfall.
I am using a 10 and 30km WRF derived dataset for the Himalayas and Pamirs (TU Berlin --> High Asia Reanalysis). Comparison with rain gauge data shows a resoanably well capture of spatio-temporal distribution albeit overestimating amounts/intensities.
Thanks for replying. I am working over Himalaya region. Well I have also look TRMM and CMORPH datasets and they show good agreement for heavy rainfall events. well I will look High Asia Reanalysis data as well and let you know my experiences.
Hi! It will be very interesting having two or more precipitation analysis to perform the verification of the WRF output. If you manage to convert your pcp analyses in netcdf format and the WRF output in grib format, you could use the MET software to compute the verification scores: http://www.dtcenter.org/met/users/
However, it is always hard to conduct such a verification over complex terrain. I have experience doing so over eastern Pyrenees, using an analysis from a dense network of 4 radars operating in Catalonia improved by ingestion of rain gauge data. Despite this high resolution analysis (~1 km), I've seen large discrepancies between rain gauge records (not used in the analysis) and the final analysis product. That's the reason why I recommend using two or more datasets of "truth" to verify QPF. I hope my experience helps. Good luck!
Thanks jordi I am struggling to convert data readable by METV5 yet I couldn't I have chosen term cmorph and observed data I have successfully did it usin arctics but that method is terrible as wrk output is not complaint to arctic so I have to convert to ascii then create hrd file and bla bla. Could you share some data conversion experiences for MET? Thanks in advance
For data conversion I use the CDO tools: https://code.zmaw.de/projects/cdo
I've been able to convert the netcdf WRF output to grib using some tricks (such as extracting only the pcp fields and then converting them into grib). TRMM is available in netcdf format... As for cmorph data, I don't have experience, unfortunately.
I did some simulations not with WRF but with MM5 at 1 km resolution over the island of Kauai back in 2002. Under a trade wind case the model successfully simulated something like 10 times more rain over the highest point Mt Waialeale which is in the league of wettest spots on Earth although unlikely to be the wettest. Fit roughly with rain gauge observations. I tried to publish but was rejected for valid reasons but if you want to check out the manuscript I have attached it.
Thanks Christopher for sharing your valuable work. I will go through the manuscript and will let you know my observations as well. but now I am little busy in writing my two papers :(