Identify Key Climate Change Scenarios

[rburghol]

Overview

Problem Statement Develop a method of screening the water resource effects from different future climate model options. We need to characterize the kinds of changes (temperature increase, precipitation increases, evaporation and storage) and what effect these changes have on the magnitude, timing, location, and duration of flows critical to water supply infrastructure.

Identify River Segment Outlets of Interest:

• Appomattox River at Mattoaca - USGS 02041650 - JA5_7480_0001 - Average Flow = 1262 cfs - Drainage Area = 1342 square miles
• Jennings Randolph - USGS 01595800 - PU3_4450_4440 - Average Flow = 551 cfs - Drainage Area = 266 square miles
• James River at Richmond VA - USGS 02037000 - JL7_6800_7070 - Average Flow = 7509 cfs - Drainage Area = 6,753 square miles
• Pamunkey River 0001 - USGS 01673000 - YP4_6720_6750 (or the 0001 segment is YP4_6750_0001, although this segment is downstream of any gage that it could be compared with) - Average Flow = 1046 cfs - Drainage Area = 1,078 square miles
• Rappahannock River at Fredericksburg - USGS 01668000 - RU5_6030_0001 - Average Flow = 1799 cfs Drainage Area = 1,595 square miles
• Potomac River at Little Falls - USGS 01646500 - PM7_4820_0001 - Average Flow = 12564 cfs - Drainage Area = 11,560 square miles
• Craig Creek at Parr, VA - USGS 02018000 - JU3_7490_7400

Model Scenarios

Below are the scenarios that we are currently examining.

• CFBASE30Y20180615 - Current conditions
• CBASE1808L55CY55R45P50R45P50Y - Climate change

Model vs. Emmissions Driven Changes

Different models may give different outputs, so too may different model Inputs give different outputs. For climate models, both things are at play. Our job is to try to understand them to the best of our ability.

Representative Concentration Pathways These are crucial to our understanding of what the different potential model INPUT differences. https://en.wikipedia.org/wiki/Representative_Concentration_Pathway

[hdaniel7]

After giving Emma a quick tutorial on last year's USGS gage vs. phase 5 model scripts, we split up to separately develop versions of the code which would utilize phase 6 data -- this gave us a bit of additional R practice and served as quality assurance, as we checked to make sure that our respective versions gave identical outputs, which they did. A spreadsheet showing all outputted metrics for the USGS gage, phase 5 model, and phase 6 model for the selected river segments was created by Emma here.

A quick proof-of-concept table for the Appomattox gage is:

[rburghol]

Good stuff folks -- I checked out the spreadsheet. Two observations/questions/request for extra attention:

• James River segment: See below, in a nutshell, flows are super low and gage needs to change.
• Potomac River: Mean flow looks like a decent match here (12,000 gage, 11,600 model), but the 7Q10 is nutso small for phase 5. We are onto Phase 6 now, but we should keep this one in our mind.

Richmond Gage and Segment the gage and segment for the James River at Richmond needs to change. It appears the gage is for the Kanawha canal (https://waterdata.usgs.gov/nwis/uv/?site_no=02037000&agency_cd=USGS) which is a very small diversion flow (like less than 100 cfs) -- it should be gage 02037500, which is on the James River proper (they are very close to one another). The River Segment is correct, though I noticed in VAHydro that the river segment doesn't actually contain the flow gage, the flow gage is located just downstream of where the next segment begins -- but it is still the appropriate segment. Don't recall if I was the one that listed the wrong gage in there, but if so, my apologies.

Also, and perhaps more importantly, the mean flow values for the model in that segment, both Phase 5 and Phase 6 are very small -- both roughly 76 cfs. At that location (right above the fall line in Richmond), the mean flow should be somewhere in the neighborhood of 7,000 cfs. Please check it out and make sure that all is copacetic. If you are very sure that the proper files have been analyzed, please take a look at the next downstream segment, JL7_7070_0001, and see if the flows are more in line (though bigger due to a little extra drainage area) -- you could also go one segment upstream to check. That is a funky bit of flow there.

Image 1: James River Gage 02037500 (located at the upstream edge of the next downstream model segment).

[emmaa4]

Rob - We reanalyzed the James River segment to Gage 02037500 and segment JL7_6800_7070. We updated the Google sheets document to reflect this - looks like the numbers are much larger for this gage than the previous one.

[hdaniel7]

Two quick issues I noticed while generating .csvs from the .wdms for phase 6 data --

1) For the initial gage vs phase 5 vs phase 6 CFBASE30Y20180615 comparison, we only used data from 1984-2005. While re-generating .csvs for phase 6 scenario CFBASE30Y20180615 to last from 1984-2015, I got "Fortran runtime error: End of record" -- from what I can tell, the .wdms in the phase 6 CFBASE30Y20180615 scenario do not contain any data for years after 2005. [Edit] Issue is detailed here -- resolved by creating a new quick_wdm_2_txt script with a higher maximum number of days.

2) I was generating .csvs from the .wdms for the new phase 6 CBASE1808L55CY55R45P50R45P50Y scenario but ran into a permissions issue once again -- I don't believe we have permission to write files within the CBASE1808L55CY55R45P50R45P50Y directory. [Edit] Access has been granted to write files for this scenario -- however, data only stretches from 1984-2000.

[emmaa4]

Rob and Joey - we have been knitting our markdown files for the comparisons of these segments of interest into this pdf document. We were unable to do analysis for Craig Creek though, as there is no .csv file available for us to download data from

[hdaniel7]

My bad -- I forgot to generate those .csvs from the .wdms. I'll get them generated in the next few minutes so you can knit the markdown.

[rburghol]

Thanks for getting on the Craig Creek segment(s) folks. Just updated the top issue description to add a scenario key -- here's to knowing those codes by heart!! :)

[rburghol]

Hey all - let's do a briefing this morning on where we are with this. I also want to share some of the relevant next steps if we are at a good place to do so. I propose 11:00 AM. @hdaniel7 @jdkleiner @emmaa4 @kevindandrea @durelles

[hdaniel7]

Alright, 11am works for us -- we'll throw it on the HARP calendar.

[hdaniel7]

While reading through some documentation of R libraries I came across a few interesting functions pertaining to low flows which might lead to a few new metrics that might be helpful -- in the hydrostats package, these two functions in particular stood out to me:

• CTF (Calculates summary statistics describing cease-to-flow spell characteristics) -- in particular, it outputs:
  • p.CTF (proportion of study period when cease to flows occur)
  • avg.CTF (average cease-to-flow spell duration)
  • med.CTF (median cease-to-flow spell duration)
  • min.CTF (minimum cease-to-flow spell duration)
  • max.CTF (maximum cease-to-flow spell duration)
• Low.spell.lengths -- returns the length and start date of all flow spells above or below a given percentile or user defined threshold

I was thinking these functions might be useful in the generation of a few metrics to add to our analysis, such as:

• proportion of study period when cease to flows occur
• either average or median cease-to-flow spell duration
• maximum cease-to-flow spell duration
• length of longest flow spell below a certain a certain percentile (maybe 5%?)
• the start date of the longest flow spell below a certain percentile (maybe 5%?)

Do any of you have any opinions on how useful these metrics might be and if you think they would be worth adding to our analysis?

[kevindandrea]

Rob and Joey, Emma and I have been working on QA checks between the metrics calculated from last years dashboard creator versus this years, here is a pdf document I put together showing that we end up with the same values for each metric as last year. We did not utilize the gage data because we are not utilizing it in our current analysis. One important thing to note, the September 10% flow appears to be wrong, as it matches the gage value and not the model value, but that is the correct value that is displayed. Last year's dashboard creator accidentally called the September 10% values in the wrong order, so it displays the model value where the gage value should be and vice versa.