Discuss baseflow results and model architecture w/CBP team

[rburghol]

@COBrogan - outline for today's meeting with Gopal

• Methodology: Analyze gages in terms of HSPF paradigm ie Q = f(AGWS, AGWRC)
  • Also, in HSPF docs we see: $AGWO_{t} = f(AGWO_{t-1}) = AGWRC * AGWO_{t-1}$
• Is there hydraulic head varying baseflow/AGWRC?
  • AGWRC = f(AGWS) in gage flow we can analyze AGWRC = f(Q)
  • Evidence (Strasburg yes, Coote's maybe not)
    • 2002 case study
    • 2023 case study
  • Approaches to modeling (special action, Cheri's augmentation)
• Further validation of method (from meeting with Gopal):
  • Step 1: Analyze runoff from a single landseg/landuse
  • Step 2: Resegment model for Passage or Smith Creek (or Linville?) to find watershed with only 1 land segment overlapping. (note: may need to up the sophistication of our resegmentation routines to subdivide only a single land segment from the parent)
  • Step 3: Compare to larger watershed.
  • Step 4: Compare results to other "standard" baseflow separation techniques. Does our routine improve?
• Analysis of model baseflow inputs
  • Phase 7 runoff
    • Is the land hydro simulation streamlined in terms of number of LUs?
    • This could help, as reduces complexity for outputting baseflow into river segments (not currently)
    • Same WDM framework? Other?
    • Basically, this is ideal time for us to start running this
  • Draft modification to phase 5/6 WDM structure that separated local runoff inflow and upstream flow into DSN 10 & 11 (it worked sorta, for some reason my approach which just used the DSN basics double baseflow inputs to the model and I abandoned it). Add Local runoff inflows after ETM step meta_model#9
  • Gopal has replaced ETM I think?

[COBrogan]

Additional ideas from Gopal:

• Set model AGWR for all land segments to 0.95. Then check the results using the event parser. Where are we doing a good job of finding 0.95, where are we not?
  • Or, to simplify further, could we run our event parser on a single land use on a single land segment to check the performance of our isolation? This might be easier than requiring an a more extensive complete model rerun
• Maybe return to other "standard" baseflow separation techniques. How does a "standard" baseflow separation perform compare to our event isolation routine?