Green River Basin

1
Green River Basin

• New Verification Tool
• 7 Forecast Points Selected
• Five Major Inquires

William B. Reed Green River Basin November 2008
2
Five Major Inquires

• Climate Indices
• ESP Distributions
• Anomalous Years
• Basic Hydrology
• ESP vs. SWS
• Joes Valley

3
Green River Basin

• WBRW4 (Green Warren Bridge
• GBRW4 (Green Fontenelle Res)

4
Green River Basin
GBRW4

• GBRW4 (Green Fontenelle Res)
• MBLC2 (Yampa nr Maybell)
• LILC2 (Little Snake nr Lily
• GRVU1 ( Green at Green River)

5
Green River Basin

• USTU1 (Upper Stillwater Res)
• JOVU1 (Joes Valley Res)
• GRVU1 ( Green at Green River)

6
Climate Indices

• Is there a Signal?
• Answer 1 Looking at 3 El Nino indices there were
  no significant correlations.
• Answer 2 Best correlation was with PDO but still
  likely not significant.

7
Climate Indices Green Fontenelle Reservoir.
Looked at SST, MEI and ONI with essentially the
same results no significant correlation.
8
Climate Indices Green Fontenelle Reservoir.
Best R2 0.0679 but still likely not
significant.
9
Climate Indices Green - Warren Bridge
No Improvement Further North, essentially the
same results Best R2 0.0229.
10
ESP Distributions

• Is there a major difference between using one
  ESPADP distribution over the others.
• Answer For the Green River at Green River the
  Normal and Wakeby distributions may be best to
  use in January.
• However, on average there is really no difference
  between the top six.
• Never use the Loglogistic distribution.

11
Looking at CategoriesContingency Table for
JanuaryGreen River Green River, UT
Bad News
Normal Wakeby - slightly better?
12
Contingency Table for April(All Distributions
OK) Green River Green River, UT
In April these five all look okay categorically.
13
RMSE for Distributions
On the average, all distributions are OK with the
exception of Loglogistic (not used in previous
plots). This is true regardless of the lead time.
14
Anomalous Years

• Contingency Table
• Historical Plot
• Cause of Forecast Error Due To Seasonal Change
• Basic Hydrology
• Low to High/High to Low
• Above Record

15
Looking at CategoriesContingency Table for
JanuaryGreen River Green River, UT
1993, 1995, 1998
16
Historical GRVU1
1993 under then over forecasted 1995 under
forecasted then caught up
1998 never caught up
17
1993 under then over forecasted
Late season snow starting in mid-March and
continuing thru May.
18
1995 under forecasted then caught up
Late season snow staring in mid-April with late
seasonal peak. 1993 happened two years prior.
19
1998 never caught up
Then peaks slightly above average.
Dry thru late January.
20
Time series
Basic Basin Hydrology
Can go from below normal to above normal
Can go from above normal to below normal
21
Time series
Can go from below normal to above normal
But I was still surprised by the size of this
jump.
22
ESP vs. SWS

• Is One Better?
• Answer In the Green Basin for the 6 Forecast
  Points investigated, on average ESP was better.
• Is it Always Better?
• Answer Yes, on average, for the sites
  investigated.

23
RMSE (lead time) 1991-2008
On average ESP is better than SWS.
24
RMSE (lead time) 1991-2008
On average ESP is better than SWS.
25
RMSE (lead time) 1991-2008
On average ESP is better than SWS.
26
Joes Valley

• Joes Valley is consistently under-forecasted.
• Low flows are forecasted okay but high flows are
  a problem.

27
Rank histogram (1993-2008)
General Tendency To Be Under-Forecasted.
28
Rank histogram by LT (1993-2008)
General Tendency To Be Under-Forecasted
regardless of Month.
29
Historical JOVU1
Forecast have tended to be below blue lines
(Observed). Have ESP for only one year.
30
POD below mean
Good Job of Detecting Volumes Below Mean.
31
POD above mean
Poor Job of Detecting Volumes Above Mean. Very
Little Improvement Thru Season.
32
Error Doesnt Improve
33
Time series (short time series 2002 missing)
Further Investigation Required.
34
SUMMARY

• With regard to climate indices, no significant
  signals were found.
• With regard to ESPADP distributions, on average
  there is really no difference between the top
  six. However, never use the Loglogistic
  distribution.
• Anomalous forecasts can often be explained by
  looking at what climate occurred within the basin
  during the anomalous year.
• For the sites investigated within the Green
  Basin, on average ESP is better than SWS as a
  tool.
• Joes Valley requires further investigation.