experimental inflow and storage forecasts portal
DESCRIPTION
Experimental Inflow and Storage Forecasts Portal. Harminder Singh Department of Civil and Environmental Engineering State Climate Office of NC. Presentation Outline. Introduction and Objectives Inflow Forecasting Model Storage Forecasting Model Inflow and Storage Portal - Overview - PowerPoint PPT PresentationTRANSCRIPT
1
Experimental Inflow and Storage Forecasts Portal
Harminder SinghDepartment of Civil and Environmental Engineering
State Climate Office of NC
2
Presentation Outline1. Introduction and Objectives2. Inflow Forecasting Model 3. Storage Forecasting Model4. Inflow and Storage Portal - Overview5. Inflow Forecasts (Monthly and Seasonal)6. Storage Forecasts (Monthly and Seasonal)7. Conclusion and Future Work
Falls
Jordan
Philpott
Kerr Scott
Rocky Creek
SF Catawba
3
Presentation Outline1. Introduction and Objectives2. Inflow Forecasting Model 3. Storage Forecasting Model4. Inflow and Storage Portal 5. Inflow Forecasts (Monthly and Seasonal)6. Storage Forecasts (Monthly and Seasonal)7. Conclusion and Future Work
4
Need for Inflow and Storage Forecasts• Need of Inflow and Storage Forecasts
– Recent Increase in Demand – Urbanization– Demand induced droughts even under normal inflow variability
• Streamflow Forecasts and Water Management– Daily to Weekly time scale – Flooding, Peak power generation– Monthly to Seasonal Time Scales – Allocation and Firm Power
• Provide Inflow and Storage forecasts for Reservoirs in the Southeast US
• Inflow and Storage Forecasts Portal– Portal Overview and Skill Assessment
5
Inflow Forecasts - Challenges• Monthly to Seasonal Inflows
– Monthly to Seasonal Climate over the watershed– Current Basin Storage – Soil Moisture and Ground water
• Challenges in Seasonal Streamflow Forecasting– Climate Forecasts – Needs to be downscaled– Limited Basin Storage Data
• Streamflow Forecasts and Reservoir Management– Streamflow Forecasts needed at the reservoir site– Interest on net-inflows = Total streamflow - Evaporation– Inflow forecasts needs to be probabilistic
6
Need for Storage Forecasts• Reservoirs in the east are within-year
– Humid basins – Fill it up by April 1st
– Winter is the critical filling period – good skill
• Why we need storage forecasts?– Inflow forecasts related to storage projections– Initial conditions in the reservoirs also influence
• Issues in developing storage forecasts– Need to consider end of the month/season target storage– Varies depending on the user-defined releases– Probabilistic information in meeting the target storage
7
Presentation Outline1. Introduction and Objectives2. Inflow Forecasting Model 3. Storage Forecasting Model4. Inflow and Storage Portal 5. Inflow Forecasts (Monthly and Seasonal)6. Storage Forecasts (Monthly and Seasonal)7. Conclusion and Future Work
8
Inflow Forecasting Model - Overview
Precipitation Forecasts (Pt) from GCMs
IRI Data Library
PredictandModel
Observed Streamflow (Qt-1)
Statistical Downscaling Model (PCR)
Forecasted Streamflow (Qt)
Training Period : Data up to previous year?Archived Forecasts : 1990-till date
Predictors
State Climate Office of NC
Portal automatically downloads Updated Monthly/Seasonal Precipitation Forecasts from GCMs between 15-18 of each month
Use for Storage Forecast
(Reservoir Model)
• Climate Data (GCMs): ECHAM 4.5• Observed Streamflow: USACE Site
9
Inflow Forecasts – Statistical DownscalingStatistical Downscaling- Precipitation forecasts from ECHAM4.5
- forced with constructed analogue SST forecasts- Precipitation from the GCMs is spatially correlated - Principal component Regression
- Principal Component Analysis (PCA) is used to reduce the data
- PCA is applied on the predictors - Streamflow and ECHAM4.5 precipitation forecasts
- Principal component regression to obtain Inflow forecasts.- Inflow Forecasts are provides as Ensembles
10
Presentation Outline1. Introduction and Objectives2. Inflow Forecasting Model 3. Storage Forecasting Model4. Inflow and Storage Portal 5. Inflow Forecasts (Monthly and Seasonal)6. Storage Forecasts (Monthly and Seasonal)7. Conclusion and Future Work
11
Storage Forecasting Model• Net-Inflows Forecast: qtk; t=1…,T; k=1,…,N• Continuity Equation: t=1,2, …, T
• User will prescribe the releases or use observed releases• Critical variable: End of the season target storage
– Initial storage can provide water for entire forecasting period
• Simulation Model estimates P(STL < ST < ST
U)
– Probability of having the storage within the conservation pool
n
ititttt REqSS
11
12
Presentation Outline1. Introduction and Objectives2. Inflow Forecasting Model 3. Storage Forecasting Model 4. Inflow and Storage Portal – Individual Years5. Inflow Forecasts (Monthly and Seasonal)6. Storage Forecasts (Monthly and Seasonal)7. Conclusion and Future Work
13
Spatial and Temporal Extent• Inflow and Storage Forecasts – Spatial Extent
– Neuse - Falls Lake – Fully Automated– Cape Fear - Jordan Lake – Fully Automated– Yadkin – Scott Keer Reservoir – Roanoke – Philpott – Catawba – South Fork and Rocky Creek
• Inflow and Storage Forecasts – Temporal Extent– Monthly (at 1, 2 or 3 month lead time) and seasonal– Available from 1990 to present, updated month – Individual year forecasts or Retrospective forecasts
14
Inflow Forecasting Models• Inflow Forecasts Models
– Statistical Downscaling (PCR)• IRI Climate Forecasts – ECHAM4.5, Multimodel
– Monthly/Seasonal Climatology (No Forecasts)– Land Surface Models (Under Integration)
• NASA’s Land Information System – NOAH 3.2• Variable Infiltration Capacity Model
• Forecast Skills– Deterministic forecasts or as ensembles– Retrospective skill summary
15
Experimental Inflow and Storage Forecasts Portal(http://www.nc-climate.ncsu.edu/inflowforecast)
16
Inflow Forecasts (Individual Year)
17
Inflow Forecasts (Individual Year)
18
Forecast skill Evaluation
Month/SeasonObserved
Inflow (CFS)
Forecast 50th Percentile (CFS)
Relative RMSE MSSS RPSS
January '90 1310 754 0.425 -0.227 -0.313
Categorical Forecasts
Climatological PercentilePercentile Values (CFS) Model Probabilities
January '90 January '90<10% < 158 0.040
10-33% 158 - 700 0.42633-50% 368 - 700 0.09250-67% 857 - 1242 0.15867-90% 1242 - 1931 0.858>90% > 1931 0.142
Inflow Forecasts (Individual Year)
19
Storage Forecasts (Individual Year)
20
Storage Forecasts (Individual Year)
Stage Level(ft above MSL): Model Percentiles:
Below conservation pool: <236.5 0.00%Within limits of conservation pool: 236.5 - 251.5 79.40%Within limits of flood control pool: 251.5 - 264.8 19.00%Above flood control pool: >264.8 1.60%
21
Inflow Forecasts (Retrospective)
22
Inflow Forecasts (Retrospective)
January Forecast Skill EvaluationRelative RMSE RPSS Correlation MSSS
0.418 -0.148 0.663 0.156
January Forecast Probability Distribution
YearPercentiles
Observed Inflow (CFS)<33% 33-67% >67%
Percentile Flow Values < 700</td> 700 - 1242 > 1242 --
1990 0.466 0.250 0.284 1310
1991 0.342 0.256 0.402 2084
1992 0.290 0.250 0.460 1174
1993 0.322 0.256 0.422 1868
1994 0.404 0.258 0.338 784
1995 0.364 0.258 0.378 843
1996 0.506 0.242 0.252 1682
1997 0.450 0.254 0.296 1151
1998 0.172 0.194 0.634 3006
1999 0.586 0.218 0.196 1598
2000 0.552 0.228 0.220 1127
Inflow Forecasts (Retrospective)
This is the percentile range which was predicted to be the
most likely to occur
The observed flow falls between the percentiles
indicated by the column for each year
The observed flow was forecasted correctly by the
model for this yea23
24
Storage Forecasts (Retrospective)
25
Storage Forecasts (Retrospective)
Year Observed Outflow (cfs)
Observed Inflow (cfs)
Median Forecasted Inflow (cfs)
Start-of-Month Elevation
End-of-Month
Elevation
1990 1291 1310 756 250.54 250.21991 2125 2084 999 245.57 250.111992 501 1173 1138 250.23 250.051993 1873 1867 1048 246.42 250.21994 121 783 864 247.31 250.621995 367 843 946 250.39 250.171996 1616 1682 690 250.64 250.451997 1126 1150 780 249.88 250.731998 1165 3006 1709 247.36 257.651999 886 1598 577 251.98 251.092000 693 1127 625 249.86 254.44
Reservoir Information for January:Storage Range Probabilities:
YearBelow
ConservationPool
Within Conservation
PoolWithin Flood Control Pool
Spilling Over Flood Control
PoolObserved
Storage (acre-ft)Median
Forecasted Storage (acre-ft)
1990 0.00% 77.80% 20.40% 1.80% 116020 87057
1991 64.00% 26.00% 8.20% 1.80% 114975 2370
1992 0.00% 31.60% 61.60% 6.80% 114278 155522
1993 48.00% 38.20% 11.60% 2.20% 116020 27816
1994 0.00% 49.40% 47.40% 3.20% 120895 132051
1995 0.00% 28.80% 66.40% 4.80% 115671 153849
1996 0.20% 85.60% 13.00% 1.20% 118922 64217
1997 0.00% 75.80% 22.40% 1.80% 122172 91101
1998 0.80% 53.40% 34.80% 11.00% 219366 120241
1999 0.00% 63.60% 35.00% 1.40% 126431 118481
2000 0.00% 70.40% 28.20% 1.40% 170804 108012
26
Presentation Outline1. Introduction and Objectives2. Inflow Forecasting Model 3. Storage Forecasting Model4. Inflow and Storage Portal 5. Inflow Forecasts (Monthly and Seasonal)6. Storage Forecasts (Monthly and Seasonal)7. Conclusion and Future Work
Seasonal Inflow Forecasts - JFM
27
Seasonal Inflow Forecasts - AMJ
28
Seasonal Inflow Forecasts - JAS
29
Seasonal Inflow Forecasts - OND
30
Seasonal Inflow (Climatology)
31
Seasonal Inflow Forecasts – Skill Summary
Falls Lake - SeasonJan - Mar Apr - Jun Jul - Sep Oct - Dec
R-RMSE 0.703 0.959 1.563 2.465RPSS 0.23 0.1 -0.13 0.1Correlation 0.861 0.547 0.42 0.413MSSS 0.408 0.146 0.159 0.076
• Relative-RMSE : A good forecast is expected to have R-RMSE closer to zero
• RPSS : If RPSS is positive, then the forecast skill exceeds that of the climatological probabilities.
• Correlation: A good forecast is expected to have a correlation around one.
• MSSS : A good forecast is expected to have MSSS be closer to one.
32
33
Seasonal Inflow Forecasts - PCR vs. Climatology Falls Lake
YearPercentiles (Climatology)
<33% 33-67% >67%1990 0.314 0.388 0.2981991 0.314 0.388 0.2981992 0.314 0.388 0.2981993 0.314 0.388 0.2981994 0.314 0.388 0.2981995 0.314 0.388 0.2981996 0.314 0.388 0.2981997 0.314 0.388 0.2981998 0.314 0.388 0.2981999 0.314 0.388 0.2982000 0.314 0.388 0.2982001 0.314 0.388 0.2982002 0.314 0.388 0.2982003 0.314 0.388 0.2982004 0.314 0.388 0.2982005 0.314 0.388 0.2982006 0.314 0.388 0.2982007 0.314 0.388 0.298
Percentiles (PCR Model)<33% 33-67% >67%0.364 0.322 0.3140.160 0.270 0.5700.260 0.328 0.4120.412 0.322 0.2660.306 0.328 0.3660.226 0.300 0.4740.204 0.266 0.5300.458 0.284 0.2580.368 0.322 0.3100.210 0.248 0.5420.574 0.264 0.1620.624 0.244 0.1320.324 0.332 0.3440.334 0.332 0.3340.288 0.318 0.3940.532 0.290 0.1780.370 0.324 0.3060.402 0.328 0.270
Model Predicted Percentile
Observed Flow Percentile
Model Predicted and Observed Flow
Percentile
34
Seasonal Storage Forecasts - JFM
35
Seasonal Storage Forecasts – JFMFalls Lake
YearBelow
ConservationPool
Within Conservation
PoolWithin Flood Control Pool
Spilling Over Flood Control Pool
2000 30.90% 33.50% 23.20% 12.40%
2001 0.00% 17.30% 62.20% 20.50%
2002 0.00% 18.10% 62.50% 19.40%
2003 18.60% 26.00% 30.50% 24.90%
2004 0.00% 18.60% 57.10% 24.30%
2005 7.50% 34.80% 40.20% 17.50%
2006 0.00% 12.40% 66.60% 21.00%
2007 4.80% 28.20% 43.20% 23.80%
2008 0.00% 34.90% 52.60% 12.50%
2009 6.10% 36.80% 39.00% 18.10%
2010 20.50% 24.20% 30.30% 25.00%
2011 0.00% 13.30% 68.90% 17.80%
2012 0.00% 7.50% 64.10% 28.40%
Model Predicted Percentile
Observed Storage Percentile
Model Predicted and Observed
Storage Percentile
36
Seasonal Storage Forecasts: Climatology – JFM
Monthly Inflow Forecasts - 6/2007 (3 Month Lead)
Climatological Percentile Percentile Values (CFS) Model Probabilities June '07 June '07
<10% < 0 0.51810-33% 0 - 67 0.03633-50% 55 - 67 0.03250-67% 130 - 325 0.09867-90% 325 - 1093 0.93>90% > 1093 0.07
37
Monthly Inflow Forecasts - 7/2007 (3-Month Lead)
Climatological Percentile Percentile Values (CFS) Model Probabilities July '07 July '07
<10% < 2 0.26210-33% 2 - 34 0.03433-50% 31 - 34 0.05850-67% 83 - 232 0.1967-90% 232 - 756 0.962>90% > 756 0.038
38
Monthly Inflow Forecasts - 8/2007 (3-month Lead)
Climatological Percentile Percentile Values (CFS) Model Probabilities August '07 August '07
<10% < 0 0.40010-33% 0 - 28 0.03233-50% 21 - 28 0.1350-67% 139 - 291 0.16667-90% 291 - 795 0.978>90% > 795 0.022
39
40
Monthly Storage Forecasts – JJA 2007- 3-Month Lead
41
Monthly Inflow Forecasts: Sep 1996
42
Inflow Forecasts- Sep 2012 – 1-Month lead
43
Inflow Forecasts - Oct 2012 – 2 Month Lead
44
Inflow Forecasts - Nov 2012 – 3 Month Lead
45
Storage Forecasts - Nov 2012 – 3 Month Lead
Storage Forecasts - Nov 2012 – 3-Month Lead
Stage Level (ft above MSL): Model Percentiles:Below conservation pool: <236.5 10.60%Within limits of conservation pool: 236.5 - 251.5 24.80%Within limits of flood control pool: 251.5 - 264.8 57.80%Above flood control pool: >264.8 6.80%
46
47
Overview of Forecasting• Individual Year or Retrospective Forecasts• Seasonal and Monthly Forecast
– Inflow forecast with skills summary– Storage forecast
• User defined outflow or observed outflow
• New Inflow and Storage Monthly Forecast – Available at the middle of month – Various lead times– Requires user defined outflows
48
Presentation Outline1. Introduction and Objectives2. Inflow Forecasting Model 3. Storage Forecasting Model4. Inflow and Storage Portal 5. Inflow Forecasts (Monthly and Seasonal)6. Storage Forecasts (Monthly and Seasonal)7. Conclusion and Future Work
49
Conclusion, Future Work
• Add Land Surface Models for Inflow Forecasts• Develop Multimodel Inflow Forecasts• Provide Inflow and Storage Forecasts for other
Reservoirs in the Southeast US• Available at the State Climate Office Website
(http://www.nc-climate.ncsu.edu/inflowforecast)
50
Acknowledgements• Project Funded by: Water Resources Research Institute (WRRI)
and NC Urban Water Consortium (NC UWC)• Dr. Sankar Arumugam – Associate Professor - NC State
University • Dr. Ryan Boyles - State Climatologist and Director - State
Climate Office of North Carolina• Dr. Tushar Sinha - Postdoctoral Research Scientist - NC State
University • Simon Mason - Research Scientist - IRI • Andrew McNamara - Graduate Student - NC State University• Thomas Petersen - Prospective Graduate Student