Upload
barbara-barton
View
214
Download
0
Tags:
Embed Size (px)
Citation preview
A Probabilistic Model for Turbidity and Temperature in the
Schoharie Reservoir Withdrawal
Steven W. Effler and Rakesh K. GeldaUpstate Freshwater Institute, Syracuse, NY
Donald C. PiersonNew York City Department of Environmental Protection
2009 Watershed Science & Technical ConferenceSeptember 14th-15th,Thayer Hotel, West Point, New York
2
site 3
site 1.5
Schoharie Creek
dam
1 km
x
Schoharie ReservoirWater Supply Withdrawal and Esopus Creek
• water quality issues for withdrawal– temperature (T)– turbidity (Tn)
Ashokan ReservoirEsopus Creek
ShandakenTunnel ~ 29 km
withdrawal depthwhen full = 18 m
3
Variations in Water Quality of Withdrawal and Thresholds of Concern
1987-2004
J F M A M J J A S O N D
Tw
(°C
)
0
10
20
30
21.1 °C 1987-2004
J F M A M J J A S O N D
Tn
,w (
NT
U)
1
10
100
1000
15 NTU
threshold: 21.1 °C
drivers of variability:• meteorology• reservoir drawdown
threshold: ~ 15 NTU
drivers of variability:• runoff• reservoir drawdown• meteorology
related management modeling goal for Schoharie Reservoir:develop and implement a modeling strategy to represent this variability in model applications
4
Development of Modeling Strategy
• a “probabilistic” framework is desired to represent variability
• long-term records of environmental and operational drivers (model inputs), together with tested water quality models, offer opportunity to represent variability
• these historic conditions are inherently representative of the system
Withdrawal Temperature
% O
ccur
renc
e
21.1 °C
existing
Withdrawal Temperature
% O
ccur
renc
emanagementalternative
existing
21.1 °C
5
Design of Probabilistic Model Framework for Schoharie Reservoir
Water Quality Model• transport/hydrothermal sub model
• turbidity submodel with resuspension
Stream Temperature
(empirical model)
Stream Turbidity Loading
(empirical model)
Stream Flow(USGS)
Reservoir Operations(NYC DEP)
Met. Data (NOAA)
Withdrawal Temperature and
Turbidity
Wave Model
MLI Optimization Algorithm
long-term records independent emp. models multi-level opt. algo.
6
Water Quality Model (W2Tn)
• transport/hydrothermal submodel (W2/T)– mechanistic, dynamic, two-dimensional from CE-
QUAL-W2 (USACE)see Gelda and Effler 2007. J. Environ. Eng. Sci. 6:73-84
• turbidity submodel– three particle sizes of turbidity– sources – external loads (primarily Scoharie Creek),
resuspension (circulation and wave-driven)
– sinks – export and settlingsee Gelda and Effler 2007. J. Environ. Eng. Div. ASCE133:139-148
Water Quality Model• transport/hydrothermal sub model
• turbidity submodel with resuspension
Stream Temperature
(empirical model)
Stream Turbidity Loading
(empirical model)
Stream Flow(USGS)
Reservoir Operations(NYC DEP)
Met. Data (NOAA)
Withdrawal Temperature and
Turbidity
Wave Model
MLI Optimization
Algorithm
Predicted Tw (°C)
0 5 10 15 20 25
Obs
erve
d T
w (
°C)
0
5
10
15
20
25Tw,obs = 0.8676 Tw,prd + 1.2861
(r2 = 0.95; n = 1380)RMSE = 1.89 °C
2003
M J J A S Oc
660
(m
-1)
0
5
10
15
20(c) 10 m - bottom
7
Water Quality Model (W2Tn) Segmentation and a Simulation
Distance from dam (m)
0 2000 4000 6000 8000
Ele
vatio
n (m
)
300
310
320
330
340
350
site 3
site 1.5
Schoharie Creek
water supplyintake
dam
1 km
x
Bear Kill
Manor Kill
intake
8
Long-Term Records to SpecifyInputs for Probabilistic Model
Water Quality Model• transport/hydrothermal sub model
• turbidity submodel with resuspension
Stream Temperature
(empirical model)
Stream Turbidity Loading
(empirical model)
Stream Flow(USGS)
Reservoir Operations(NYC DEP)
Met. Data (NOAA)
Withdrawal Temperature and
Turbidity
Wave Model
MLI Optimization
Algorithm
Model Driver/Input Specifications
meteorology1948-2004 (57 years); off-site Albany (NOAA) since 1948, on-site since 1997
inflows (gaged)Schoharie Creek since 1948; others more recent (USGS)
outflows (operations)1948-1996, NYCDEP; 1997-2004, USGS
9
Independent Empirical Models to Specify Inputs for Probabilistic Model
Water Quality Model• transport/hydrothermal sub model
• turbidity submodel with resuspension
Stream Temperature
(empirical model)
Stream Turbidity Loading
(empirical model)
Stream Flow(USGS)
Reservoir Operations(NYC DEP)
Met. Data (NOAA)
Withdrawal Temperature and
Turbidity
Wave Model
MLI Optimization
Algorithm
Stream temperature (plunging)Ts,i = a0 + a1 Tair,i-3 + a2 log (Qi)
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31
Ts
(°C
)
16
20
24
28
observedpredicted - long-term stream T predicted
from Tair and Q records
Turbidity-Flow Relationship (external loads)Tn = 2.5 C660
- long-term stream Tn loads predicted from Q records
9/18/2004
Distance from Creek Mouth (km)
3 4 5 6 7 8
De
pth
(m
)
10
20
30
0
20
40
60
80
100
120
Flow (m3·s-1)
0.1 1 10 100 1000
c 660
(m
-1)
0.1
1
10
100
10001996-2001best fit (r2=0.70)
10
Performance of Probabilistic Model in Representing Variability of Withdrawal T
• observations: 1987-2004• prediction bounds: for driving conditions of 1987-2004
J F M A M J J A S O N D
Tw (
°C)
0
5
10
15
20
25
30
• probabilistic model succeeds in representing range of observations
11
Performance of Probabilistic Model in Representing Variability of Withdrawal Turbidity
• observations: 1987-2004• prediction bounds: for driving conditions of 1987-2004
• probabilistic model succeeds in representing range of observations
J F M A M J J A S O N D
Tn
,w (
NT
U)
1
10
100
1000
12
Performance of Probabilistic Model in Simulating Water Quality in the Withdrawal
Withdrawal Temperature (°C)0 5 10 15 20 25 30
Cu
mu
lativ
e O
ccu
rre
nce
(%
)
0
20
40
60
80
100
predictedobserved
21
.1 °
C (
70
°F
)
Withdrawal Turbidity (NTU)
1 10 100
Cu
mu
lativ
e O
ccu
rre
nce
(%
)
0
20
40
60
80
100
predictedobserved
15 NTU
(a)
(b)
generally good performance
13
Example Application of the Probabilistic Model: Scenario Description
• potential benefits of multi-level intakes (MLI) and location in the reservoir
• is there a benefit to “spatial avoidance” of turbid plumes?E
leva
tion
(m)
300
310
320
330
340
350
Col 17 vs Col 18 Col 20 vs Col 19 Plot 2 Mean
site 3
site 1.5
spillway elevation
multi-levelintakes
14
Projections for MLI Scenario with Probabilistic Model: Site 3 versus Site 1.5
• for 57 years of historic conditions• summary statistic of number of days withdrawal Tn > 15
NTU, for individual years of record
- no noteworthy benefit for MLI at site 1.5 versus site 3E
leva
tion
(m)
300
310
320
330
340
350
Col 17 vs Col 18 Col 20 vs Col 19 Plot 2 Mean
site 3
site 1.5
spillway elevation
multi-levelintakes
Schoharie Cr.
Exceedences of Tn,w for site 1.5 MLI
0 50 100 150 200
Exc
eed
en
ces
of
Tn
,w
for
site
3 M
LI
0
50
100
150
200hypotheticaloutcome:site 1.5 betterthan site 3
equivalence; i.e.,no benefit from location change
Exceedences of Tn,w for site 1.5 MLI
0 50 100 150 200
Exc
eed
ence
s o
f T
n,w
fo
r si
te 3
MLI
0
50
100
150
200
hypotheticaloutcome:site 1.5 betterthan site 3
equivalence; i.e.,no benefit from location change
15
Projections for MLI Scenarios with Probabilistic Model,
Comparisons to existing Withdrawal Case
• for 57 years of historic conditions• cumulative distribution format for presentation of results
Withdrawal Turbidity (NTU)
1 10 100
Cum
ulat
ive
Occ
urre
nce
(%)
0
20
40
60
80
100
site 3, 3-level
site 1.5, 3-level
15 NTU
existing
MLI scenarios
- modest benefit of MLI; exceedences decrease from 27 to 16% of days
16
Summary
• probabilistic modeling framework for temperature and turbidity for Schoharie Reservoir developed, tested and preliminarily applied– key components: tested mechanistic water quality models, long-
term records for drivers, and empirical models
• insights from preliminary applications concerning multi-level intake alternatives
• broad utility of approach– other issues and systems (Ashokan, Kensico)– flexibility to accept upgrades/updates– coupling with hydrologic model (OASIS)
• to integrate water quantity needs of overall system
17
Related Professional Journal Citation
• a more complete treatment of material addressed in this presentation can be found in the following peer-reviewed journal paper
Gelda, R. K. and S. W. Effler, 2008. Probabilistic model for temperature and turbidity in a reservoir withdrawal. Lake and Reserv. Manage. 24: 219-230.
18
Investigation of Model and Input Updates/Upgrades (2009)
• turbidity submodel and stream turbidity loading model
Water Quality Model• transport/hydrothermal sub model
• turbidity submodel with resuspension
Stream Temperature
(empirical model)
Stream Turbidity Loading
(empirical model)
Stream Flow(USGS)
Reservoir Operations(NYC DEP)
Met. Data (NOAA)
Withdrawal Temperature and
Turbidity
Wave Model
MLI Optimization
Algorithm
19
Investigation of Model and Input Updates/Upgrades (2009)
• turbidity submodel and stream turbidity loading model
Updates based on1. new particle characterizations
(Peng et al. 2009)2. resuspension studies (Cornell)
and modeling (Owens et al. 2009)3. expansion of model testing for
additional years of detailed monitoring (Owens et al. 2009)
4. correction of coding error for resuspension
Water Quality Model• transport/hydrothermal sub model
• turbidity submodel with resuspension
Stream Temperature
(empirical model)
Stream Turbidity Loading
(empirical model)
Stream Flow(USGS)
Reservoir Operations(NYC DEP)
Met. Data (NOAA)
Withdrawal Temperature and
Turbidity
Wave Model
MLI Optimization
Algorithm
based on additional stream monitoring data
Schoharie Creek Flow (m3·s-1)
1 10 100 1000
Tu
rbid
ity (
NT
U)
1
10
100
1000
10000
Phase II
Schoharie Creek Flow (m3·s-1)
1 10 100 1000
Tu
rbid
ity (
NT
U)
1
10
100
1000
10000Upgraded - multipleUpgraded - singlePhase II
20
Effects of Updates/Upgrades on Probabilistic Model Projections
• an example
Exceedences of Tn,w for site 1.5 MLI
0 50 100 150 200
Exc
ee
de
nce
s o
f T
n,w
fo
r si
te 3
ML
I
0
50
100
150
200Phase II
equivalence
Exceedences of Tn,w for site 1.5 MLI
0 50 100 150 200
Exc
ee
de
nce
s o
f T
n,w
fo
r si
te 3
ML
I
0
50
100
150
200
equivalence
Updated/Upgraded with errorPhase II
Exceedences of Tn,w for site 1.5 MLI
0 50 100 150 200
Exc
ee
de
nce
s o
f T
n,w
fo
r si
te 3
ML
I
0
50
100
150
200
equivalenceUpdated/Upgraded with error corrected
Phase IIUpdated/Upgraded with error
Ele
vatio
n (m
)
300
310
320
330
340
350
Col 17 vs Col 18 Col 20 vs Col 19 Plot 2 Mean
site 3
site 1.5
spillway elevation
multi-levelintakes
Schoharie Cr.
• management perspectives on MLI/location alternatives remain unchanged
21
Summary
• probabilistic modeling framework for temperature and turbidity for Schoharie Reservoir developed, tested and preliminarily applied– key components: tested mechanistic water quality models, long-
term records for drivers, and empirical models
• insights from preliminary applications concerning multi-level intake alternatives
• broad utility of approach– other issues and systems (Ashokan, Kensico)– flexibility to accept upgrades/updates– coupling with hydrologic model (OASIS)
• to integrate water quantity needs of overall system