streamflow nudging with wrf-hydro

[Bongchul Seo]

Hello Everyone,

I am running wrf-hydro (close to the NWM 1.2) with streamflow nudging for the entire State of Iowa.

Observations from about 140 USGS stations in Iowa have been included in my modeling framework and I have run the model for three years (2016-2018).

What I am wondering when looking at streamflow output files (e.g. CHRTOUT) is where I can find "model assimilated streamflow" at the links where the USGS stations are located.

It seems that the "streamflow" variable in CHRTOUT was replaced with USGS observations I provided.

My current approach to see the nudging effect (propagated from upstream) at the USGS stations is to read streamflow at the closest upstream links to corresponding USGS stations.

Is there any way to directly read model assimilated streamflow at the USGS stations? I'd like to compare open loop simulated streamflow (not associated with USGS observations) vs. assimilated streamflow by the current nudging scheme.

Thanks,

Bongchul Seo

[jamesmcc]

Hi Bongchul, 

Thanks for using nudging! Please explain what you mean by or expect for "model assimilated streamflow" in detail. I do not understand what you are expecting. I will attempt to answer and provide guidance based on what I think you are looking for/expecting.

At least the method appears to be functioning as intended (from my perspective). Indeed nudging, depending on the quality flag associated with the obs, "replaces" the prognostic streamflow solution at the observed location with the observed value at the time of observation. This is a nudging, time-continuous solution and there is no "prior" or "posterior", if this is what you are looking for, because the increment is added and adjusted over time. There is a "nudge" amount that is output at the output frequency, but one can not simply subtract if from the streamflow to get "open loop" because the solution is non-linear and is performed on the routing timestep (dtrt_ch). 

There are also difficulties in comparing with the directly upstream value because the current solution of the nudging equation applies the nudge to the upstream link as well. Please see page 43-44 https://ral.ucar.edu/sites/default/files/public/projects/Technical%20Description%20%26amp%3B%20User%20Guides/wrfhydrov511technicaldescription_0.pdf. This method significantly reduced the size of the increment/nudge at the downstream reach, leading to a smoother solution.

If you want to understand the simulation without nudging, you have to run a separate simulation without nudging. The additional cost of this can be mitigated by running the model in "channel-only" mode (namelist.hrldas FORC_TYP=9) where the CHRTOUT files from a full-model run (e.g. the nudging run above) can force the channel-model (in this case open-loop). Note that the boundary conditions for the channel model are not affected by nudging.

[Bongchul Seo]

Hi James,

Thank you for your response and comments.

My goal is to evaluate the nudging effect at all USGS locations in Iowa by comparing the results between without nudging and with nudging (I already ran the model without nudging) .

If I have only one location to evaluate, I can run the model with nudging all upstream observations and see the difference (vs. without nudging) at my target location.

Since my target locations are about 140 USGS stations in Iowa, I included observations from all locations and ran the model with nudging (so each target location can be upstream and downstream of other target locations in the stream network).

Then, model simulation results at the target locations are all replaced with observations and there is no way to get the streamflow value propagated from upstream nudging at the target locations.

Can I just add the amount of nudge to the simulation result without nudging to get the streamflow value after nudging?

Thanks again for your help!

Bongchul

[zed li]

Hi Bongchul ,

  I'm currently conducting nudge research,but I've run into some problems. I've created the input files and checked each of them. The stationId in all the input files are corresponding. In hydro.namelist, I commented out the Unsupported Options, but the nudge value in the generated nudgingLastObs file is still 0. I don't know where the problem lies. Could you offer some suggestions? Thank you very much. I've attached the diag file, the output nudgingLastObs file, and a sample of the input files.