Title: Regional Analysis of LowFlow in Western WA
1Regional Analysis of Low-Flow in Western WA
- Presented by Ken Eng1 and Chris Curran2
1USGS, NRP, WRD, BRR-ER Reston, VA2USGS, WAWSC,
Tacoma, WA May 23, 2008
2Regional Analysis of Low-flowin Western
Washington
- Decide on a list of candidate index sites.
- Select relevant basin characteristics.
- Choose a flow characteristic to estimate (Q7,10).
- Run correlation analysis (MOVE, etc) to estimate
Q7,10 at simulated misc. sites. - Test different scenarios for index-site
selection. - Develop regression model to estimate Q7,10 at
ungaged or future miscellaneous sites.
3Index site
4Index site
5Index-site selection
- Natural, unaltered flow.
- No significant storage (lakes, ponds).
- No significant contribution from snowmelt.
- Q7,10 cannot equal zero (correlation analysis).
- More than 10 years of continuous daily-streamflow
record. - Must have overlapping record (correlation only).
- Only USGS gages considered.
6USGS index gages used in the correlation analysis
7Basin characteristics
- Drainage area (StreamStats).
- Geographic proximity.
- Mean-annual precipitation (StreamStats).
- Surficial geology ( coarse material).
- Base-flow recession time constant, t.
8Base-flow Recession Time Constant, t
- Tau describes how fast base-flow recedes
Large t
Small t
Time
- Many ways to estimate t e.g., Brutsaert and
Nieber, 1977 Eng and Brutsaert, 1999 Eng and
Milly, 2007, Brutsaert, 2008.
9Base-flow Recession Time Constant, t
- Estimated t from pair of measurements
Single
Paired
Day 8
No t can be estimated for a single value.
Qt
QtDt
Dt5 days
10Regional low-flow correlation analysis
- Done using a Matlab script written by Ken
- Uses the daily-streamflow records for all index
gages. - Simulates a miscellaneous site by using an index
site. - Extracts the recession segments of hydrograph.
- Randomly picks a daily flow (or pairs) and treats
it as a miscellaneous measurement.
11Regional low-flow correlation analysis
- Runs each scenario with a user-defined
combination of basin characteristics and number
of miscellaneous measurements. - Determines best index site (similar A,G,P,t,S)
- Performs Q-ratio, MOVE, base-flow correlation
- Calculate estimates of Q7,10 and compares this
with the known value of Q7,10, records error. - Repeated 500 times.
12Low-flow correlation analysis
73
13Low-flow correlation analysis
14Summary for correlation analysis
- Minimum 4 measurements for Q-Ratio.
- Using t reduces RMSE by roughly 10.
- MOVE and base-flow correlation will underestimate
(more conservative). - Q-Ratio may overestimate (undesirable for low
flow).
15Recommendations based on correlation analysis in
western Washington
- Q-Ratio (lt8) and/or MOVE (gt8) methods.
- Minimum 4 measurements (paired for t).
- 3 day minimum wait between paired meas.
- Wait 6 days until after peaks, check a nearby
gage hydrograph to ensure recession - Make measurements during base-flow season
16Regional regression analysis
- Different from correlation analysis
- Uses basin characteristics.
- Uses index gages
- Can use discontinued sites.
- Miscellaneous measurements can be used.
- More accurate Q7,10 benefits regional regression
model.
17USGS index gages used in analyses
18Regional regression models using A, P, t
19Regional regression analysis
20Regional regression analysis
- Summary for regression models
- 8 pairs - 60 of potential reduction in RMSE.
- Diminishing returns after 8 pairs.
- Recommendations
- Use A, P, and t.
- Maximum of 8 paired measurements.
- Minimum of 4 paired measurements.
21Optimizing the Low-flow Networkin Western
Washington
What types of miscellaneous sites do we still
need?
22Optimizing the Low-flow Networkin Western
Washington
- Fill spatial gaps in the existing network.
- Use StreamStats to identify potential locations
for miscellaneous sites. - Balance location with operational concerns.
23Future Miscellaneous Sites
- Correlation analysis versus regional regression
to estimate Q7,10 for miscellaneous sites - Correlation
- Gambling, based on best single index site
- Lengthy, need measurements
- Regression
- Robust
- Simple
- Recommendation regional regression models to
estimate Q7,10 at future miscellaneous sites.
24Summary of What to do for a Study
- For estimating Q7,10 at current miscellaneous
sites - Run a correlation analysis to estimate Q7,10 at
current miscellaneous sites. - Use A, P and t as criteria to select index site.
- Use Q-ratio for lt 8 pairs of measurements or MOVE
(gt 8) to estimate Q7,10. - Need minimum of 4 singles/pairs.
25Summary of What to do for a Study
- Develop regional regression model to estimate
Q7,10 at future sites. - Use A, P, and t.
- Use all continuous and current miscellaneous
sites. - Use t (at least 4 pairs).
- Application of regional regression model.
- Make measurements at a stream of interest.
- Compute t from all pairs and average.
- Find A and P for the site (StreamStats).
- Place into regression model to calculate Q7,10.
26References
- Brutsaert, W. and J. L. Nieber (1977),
Regionalized drought flow hydrograph from a
mature glaciated plateau, Water Resour. Res.,
13(3), 637-643. - Brutsaert (2005), Hydrology An Introduction,
Cambridge University Press, 605. - Brutsaert (2008), Long-term groundwater storage
trends estimated from streamflow records
Climatic perspective, Water Resour. Res., 44,
W02410. - Eng K. and W. Brutsaert (1999), Generality of
drought flow characteristics within the Arkansas
River basin, J. Geophys. Res., 104(D16),
19,435-19,441. - Eng, K. and P.C. D. Milly (2007), Relating
low-flow characteristics to the base flow
recession time constant at partial record stream
gauges, Water Resour. Res.,43, W01201.