Water Resources in a Changing Climate: NSF ESPCoR VI Hydroclimatology V. Sridhar, Xin Jin, David...

Post on 22-Dec-2015

214 views 0 download

Tags:

Transcript of Water Resources in a Changing Climate: NSF ESPCoR VI Hydroclimatology V. Sridhar, Xin Jin, David...

Water Resources in a Changing Climate: NSF ESPCoR VI Hydroclimatology

V. Sridhar, Xin Jin, David Hoekema, Sumathy Sinnathamby, Muluken Muche

R. Allen, Wenguang ZhaoM. Germino

Background and Context

Region-wide warming Precipitation change Decline of snowpack Earlier spring runoff and Decline in summer streamflow trends

Research Questions

How will future climate change impact water resources?

Hydro - Climate

Hydro - Economic / Policy

Hydro - Ecology

Research Questions

Hydro-climate interactions

What are the relationships between climate change, vegetation, snow pack, and the resulting stream flows in managed and unmanaged river systems?

How will aquifer systems exchange with surface and

groundwater under various climate change scenarios?

What will both the supply and demand on water be in these systems under various climate change scenarios?

How will fire and invasive species (cheat grass, some bunchgrasses) impact:

Rates and durations of ET fluxes from desert systems?

Changes in infiltration patterns for precipitation?

Interactions of ET, infiltration, thermal profiles and microbial populations and feedbacks?

Erosion and Sedimentation in Tributaries of the Snake and Salmon basins?

Hydroclimate Bio Interactions

Focus Area 2009-2011

Research 1: Advance our ability to model surface energy balance processes

Tasks: Use of scintillometer and eddy covariance (EC) systems to measure sensible (H) and latent (LE) heat fluxes in desert and timber

Research 2:Advancement in Basin scale Hydrologic Forecasting – verification and operation under climate change scenarios

Task1 : Operate, test and calibrate the CIG Variable Infiltration Capacity (VIC) model and combine with the groundwater model

Task2: Implement VIC –groundwater model and evaluate the reservoir optimization techniques

Scintillometry

Use to Retrieve Sensible Heat Flux (H) over large, integrated transects

Use to improve components in METRIC and VIC Reduce variances

Use to derive ET experimentally Determine how soil water meters out from desert

and lodgepole

Transmitter

Receiver

Surface Energy Balance Processes----Large aperture scintillometer--transmitter (left) and 3 intercompared receivers (right) purchased by Idaho EPSCoR RII

Scintillometers to measure surface heat flux densities

Scintillometer – Idaho RII Deployments

Three systems (BSU, ISU, UI)

Three deployments Sage brush -- Snake River Plain

Cheat grass or recent burn – Snake River Plain

Timber – Upper Reaches

Idaho RII: Sagebrush Deployment

Sage brush ecosystem located west of Hollister, Idaho.

Raft River Site Cheat grass

LodgePole Pine Site --- Macks Inn Area

Scintillometers measure only Sensible Heat flux (H)

ET is calculated as a residual of the energy balance

ET = Rn – G – H

Net radiation, Rn, and soil heat flux, G, must be measured

All error in Rn, G and H transfer into ET

Preliminary VIC model calibration & Flowchart for sensitivity analysis

Choose parameters to be calibrated. In this study, 7 parameters were chosen recommendation

Construct nxp Jacobian matrix and

calculate CSSj

Identify the most sensitive parameters (with the biggest CSSj) end

Choose observation locations to be used as reference. In this study, 6 locations were chosen

Calculate sensitivity for each observation yi (i=1,…,n) over each

parameter bj (j=1,…,p).

Divide the range of each parameter into equal space

Run VIC and routing with one parameter changed and the others unchanged. Obtaining new RMSE.

Divide the Snake River Basin into 6 sub-basins and select the new set of parameters that make the RMSE minimum and apply it to the VIC

Ws Ts3 binf Ts2 Ds Dsmax Ts1

2.54×105 2.00×105 1.67×105 1.48×105 1.11×105 1.08×105 4.87×104

Parameter Dsmax Ds binf Ws Ts1, Ts2, Ts3

Unit mm/day % NA % mRange >0 to ~30 >0 to 1 >0 to ~0.4 >0 to 1 0.1 – 1.5

• Ws fraction of maximum soil moisture where non-linear baseflow occurs

VIC model calibration results (all 6 locations, 1928 - 1978)

RMSE (cfs) r2

uncalibrated calibrated uncalibrated calibrated

Heise 3582 2427 0.88 0.90

Rexburg 2168 1683 0.87 0.89

Milner 5649 4708 0.85 0.86

Oxbow 15232 12099 0.90 0.85

Parma 3127 1837 0.75 0.81

Payette 2603 1532 0.86 0.88

VIC model validation results (all 6 locations, 1979 - 2005)

RMSE (cfs) r2

uncalibrated calibrated uncalibrated calibrated

Heise 3462 2556 0.92 0.90

Rexburg 2158 1844 0.87 0.83

Milner 5400 4871 0.87 0.86

Oxbow 14026 12796 0.92 0.86

Parma 2639 1687 0.77 0.83

Payette 1988 1859 0.88 0.88

VIC model calibration results (at Heise)

Default calibrated from University of Washington

calibrated

Preliminary VIC results (1979-2004)

Parameter Selection for the SWAT ModelSnowmelt and snow formation parameter

Ground water parameter

Soil parameter

Surface Runoff parameter

SWAT Calibration and validation

Salmon River Snake River

Optimization objective functions

White bird Krassel Ranger

Yellowpine Millner Oxbow

Nash -Sutcliffe model efficiency (Ens) ∑∑==

−−−=n

iobsobssim

n

iobs QmeanQQQ

1

22

1

))((/)(1

Calibration Validation

0.57 0.49

0.25 0.23

0.47 0.49

0.52 0.46

0.39 0.40

Percent bias (PBIAS)

=

∑ ∑= =

−n

i

n

iQobssimobs QQ

1 1

/)(

Calibration Validation

0.04 0.03

0.11 0.07

0.2 0.14

0.06 0.05

0.07 0.08

Regression coefficient R 2 Calibration Validation

0.57 0.60

0.53 0.55

0.63 0.51

0.53 0.46

0.42 0.42

Where Qobs is the measured monthly s treamflow, Qsim is the simulated monthly streamflow, mean (Qobs) is the mean of the measured monthly streamflow, and n is the number of measurement.

Results

PrecipitationMaximum

TemperatureMinimum

Temperature PrecipitationMaximum

TemperatureMinimum

TemperatureA1B + + + + + +

ECHAM A2 - + + - + +B1 + - - + + +

A1B + + + + + +GISS A2 - + + + + +

B1 - + + + + +A1B + + + + + +

IPSL A2 + + + + + +B1 + + + + + +

Model Scenario

Future trendSnake River watershedSalmon River Watershed

Decreasing trend in monthly discharges

Salmon River watershed Snake River watershed

White bird Krassel Ranger Yellowpine

ECHAM GISS IPSL

A1b

A2

B1

A1b

A2

B1

A1b

A2

B1

Oxbow Milner

ECHAM

GISS

IPSL

A1b

A2

B1

A1b

A2

B1

Oxbow Milner

Snake River watershedA1b

A2

B1

A1b

A2

B1

GUI development of ESPAMGUI development of ESPAM

GUI development of ESPAMGUI development of ESPAM

Flow Distribution & Points of Interest (POI)Six Points of Interest

1) Heise (Snake River)

2) Rexburg (Henry’s Fork)

3) Milner (Snake River)

4) Parma (Boise River)

5) Payette (Payette River)

6) Oxbow (Snake River)

58%

+32%

90 % Total

Points of interest were chosen from which projected flows could be distributed to simulate upstream reach gain contributions. As represented in the chart below, we selected six points of interest that cover 90% of the flow in the upper SRB.

1) Monthly Natural flow (sum of upstream reaches)

Where,

NFm = monthly natural flow at reach d

d = downstream reach (or point of interest)

u = furthest upstream reach

xi = any given reach between u and d

• Annual Natural Flow (sum of monthly Natural Flows)

Where, NFm,1 = natural monthly flow in October, NFM,2 = natural monthly flow in November….

Reach Gain Simulation Calculations

∑=

+++=u

diu

xi

xd

x )....(m

NF

∑=

=12

1,y

NF

iim

NF

The first step of the reach gain simulation method is to categorize flow based on a range of historic annual natural flows. The equations for calculating natural flow from IDWR historic reach gains are presented here.

Flow Categorization Henry’s Fork

Flow Range per Category: 3000 (100 acre-feet) Minimum: 13678Maximum: 40697Mean: 24768

Dry < 180001 18000 210002 21000 240003 24000 270004 27000 300005 30000 33000

Wet > 33000Flow categorization is based on annual flows while simulation of these flows are based on monthly distributions of the projected flow. Along the Henry’s Fork flows are categorized with a range of 300,000 acre-feet per category.

Predicting Minor Flows—Linear Model

Flow W. = (2.18*(%Avg. Flow Ox.) - 1.18)*W. Avg.

Model Validation

Irrigation Shortage Comparison: Historic vs. Simulated (1980-2005)

A comparison between SRPM calculated irrigation shortages as represented by historic and simulated reach gains reveals that the reach gain simulation method was able to provide perfect replication of historic irrigation shortages in the river between the years 1980 and 2005.

Falls Teton Henry's Above Lorenzo Willow Blkft Blkft Milner Boise NewY Payett

River River Fork Lornezo Blkft Creek Prtnf Milner Murphy River Canal River TOTAL

His. 0.3% 0.9% 0.6% 0.6% 15.0% 0.4% 49.0% 36.2% 0.0% 27.8% 74.8% 14.6% 20.7%

Sim. 3.8% 6.0% 8.3% 0.5% 21.8% 3.3% 47.6% 56.8% 0.0% 43.3% 87.1% 15.3% 30.3%∆Shortage 3.5% 5.1% 7.8% 0.1% 6.8% 2.9% 1.4% 20.6% 0.0% 15.5% 12.3% 0.7% 9.5%

Payette Watershed-Future Climate: Echam-5

Deadwood Dam

Cascade Dam

Black Canyon Dam

Future Climate Payette River

Future Climate Payette River

Future Shortages Payette River

VIC+MODFLOW flowchartVIC+MODFLOW flowchart

Run VIC model to generate the infiltration, evapotransporation

(ET), runoff and baseflow at each cell of unsaturated zone

Run MODFLOW and generate recharge, water content

Infiltration, ET

Add a fraction of recharge from MODFLOW the baseflow in VIC output

Run VIC routing model

No

Yes

Stop

Reaching time step limit

Water content

END