Identifying physio-climatic controls onwatershed vulnerability to climate and landuse change

SWAT 2018 IIT Madras, India

Ankit Deshmukh*1,Riddhi Singh2

January 12, 2018

1 Department of Civil Engineering, Indian Institute of Technology Hyderabad, Kandi, Telangana, India 502285 2 Department of Civil Engineering, Indian Institute of Technology Bombay, Powai, Maharashtra, India 400076

We need reliable projections of streamflow to manage freshwater

resources in the future

Figure 1: Global extent of water scarce regions in the last decade. (UNEP,

2008)

However, obtaining future projections are challenging! 1

Changing climate will lead to significant and uncertain changes

in future

1960 1980 2000 2020 2040 2060 2080 2100−1

0

1

2

3

4

5

6

7 Historical GCM uncertainty

Model uncertaintyInternal variabilityObservations (GPCP)

Scenario uncertainty

Projected change in global mean precipitation [%]

1960 1980 2000 2020 2040 2060 2080 2100−0.5

0

0.5

1

1.5

2

2.5

3

3.5

4

Projected change in global mean temperature [K]

Adler et al., 2003 Brohan et al., 2006

Decadal mean projections for the 21st century (the total

uncertainty in CMIP-3 global mean)2Hawkins and Sutton [2011]

Uncertainties are present in all the steps of obtaining a hydro-

logic projection

d y

nd

nEmission scenarios

(future CO2 concentration)Determine plausibility o

vulnerabilityGeneral circulation

models

Downscalingal fo

rwar

tion

Hydrologic model

Indicators of hydrologic

Tra

ditio

prop

aga Identify regions in

framework space that leato vulnerability

response

uncertainty about future economy

Initial conditionsParametric uncertainty

Downscaling uncertainty

Model structural uncertaintyParametric uncertainty

Uncertainty about critical thresholds

Climate change

Land use change

Determine if these ranges fall in the vulnerable region

Thus we need a way to provide useful information to stakeholders

in the presence of large uncertainties.

3

We provide an alternative approach to identify vulnerability to

climate change which is independent of future projections of

climate and land use change

y su

su

Determine plausibility ofvulnerability

p an

alys

i

Identify regions inframework space that lead

to vulnerability

Bot

tom

Determine vulnerability ranges forthe hydrologic indicator

Figure 2: Bottom up: A vulnerability based approach

Our vulnerability based approach is useful for decision makers and

water policy makers.4Singh et al. [2014]

We develop a framework which combines strength of bottom-up

approach and comparative hydrology

User should defined vulnerability: One can defined vulnerability as

”streamflow reduced by >50% in the watershed”

ThresholdsBottom-Up Framework Correlation

The framework provides a quantitative measurement (critical

thresholds) of watershed vulnerability to climate and land use

change.

5

We implement the bottom-up approach using exploratory mod-

elling framework

Deep-rooted

Vegetation

Climate and land use combinations

Exploring theprecipitation space

-40% +60%ΔP0oC

12oC

ΔT

0

1

Explo

ring

the

land

use

spa

ce

Exp

lorin

g th

ete

mpe

ratu

re s

pace

Hydrologic model

%Veg

GW store

ET

subsurfaceQ

Qsurface

baseflowQ Str

ea

mflo

w

Random sampling

Parameters θ

Screening (NSE & %Bias)

Mapping of classes to the sample space

-40% +60%ΔP0oC

12oC

ΔT

0

1

Deep-rooted

Vegetation CART

Indicators

Indicator classes

Simulations

C0 C1 C2 C3

6Deshmukh and Singh [2016]

We use a spatial lumped model to simulate runoff which account

land-use as fraction of deep rooted vegetation in the watershed

T

ET

Field Capacity

Shape parameter

LAImin/max

Surface flow

Subsurfaceflow

Baseflow

Max

imum

sto

rage

Ground water store

Recharge coefficient ASS

ABF

Legend

ParameterFluxes

Thresholdtemperature

Degree dayfactor

Precipitation

Rain/Snow

%veg:Fraction of deep rooted vegetation in watershed

Cei

Simulated runoff used to calculate indicator of vulnerability (i.e.

mean annual runoff).7Singh et al. [2014]

We classify an indicators based on class definitions

Indicator mean annual runoff (for water availability in the stream)

classified into 4 classes.

C0 C1 C2 C3

Class: Change in streamflowC0: > 0%C1: 0% to -25%C2: -25% to -50C3: < -50%

We use a classification method, CART to identify the combinations

of climate and parameters which leads to a specific class.

8

We select 77 United states watersheds for this study

Removed watersheds

WatershedsNorth0 1000

km

134 km2

8151 km2

We remove 8 watersheds based on model performance criteria and

low runoff ratio.9Deshmukh and Singh [2016]

We use CART to relate climate, land use and parameters sam-

ples into vulnerability classes

Node:1ΔP< 0.85

Samples: 28636

C2

Samples: 2631Samples: 26005

Samples: 2275

Node:2

Node:3

Node:5

C3

C1 C2

C0

C0

Node:4

C3

ΔP ≥ 0.75

M< 0.55

ΔP< 1.05

M ≥ 0.35

ΔP ≥ 0.95

if true if false

C0 C1 C2 C3Class: Change in streamflowC4: > 0%C1: 0% to -25% C2: -25% to -50C3: < -50%

Precipitation threshold = 24.3% (Reduction)

Landuse threshold = 0.550

Critical thresholds are computed by weighted averaging of thresholds for

all watersheds (for precipitation and land use change)10Deshmukh and Singh [2016]

We identity relationship between critical threshold and water-

shed physio-climatic characteristics

TO

PW

ET

1st o

rder

str

eam

4t h or

der

stre

am

Str

eam

den

sity

Ele

vatio

n

Asp

ect E

ast

Are

a

Asp

ect N

orth

Avg

. per

mea

bilit

y

HG

-B

Avg

. silt

Org

anic

mat

ter

Avg

. WT

dep

th

Urb

an

For

est

Agr

icul

ture

Pre

cipi

tatio

n

Eva

pora

tion

Arid

ity In

dex

M.A. runoff

IQdsc

Drought

Flood

-1 1

Correlation

Pre

cipi

tatio

n ch

ange

thre

shol

d La

ndus

e

chan

ge th

resh

old

Significant corrrelation Insignificant correlation (at P-value > 0.05)

Mean annual run

FPIFR

Drought

Flood

2nd o

rder

str

eam

a.

b.

1st o

rder

str

eam

4t h or

der

stre

am

Str

eam

den

sity

Ele

vatio

n

Asp

ect E

ast

Are

a

Asp

ect N

orth

Avg

. per

mea

bilit

y

HG

-B

Avg

. silt

Org

anic

mat

ter

Avg

. WT

dep

th

Urb

an

For

est

Agr

icul

ture

Arid

ity In

dex

Mean annual run

FPIFR

Drought

Flood

Drainage Topography Geology Landuse Climate

Land

use

ch

ange

thre

shol

d

-1 1CorrelationSignificant corrrelation (P-value < 0.05) Insignificant correlation

M.A. runoff

IQdsc

Drought

Flood

2nd o

rder

str

eam

a.

b.

Indi

cato

rsIn

dica

tors

TW

I

M. A

. PE

T

M. A

. pre

cipi

tatio

n

The correlation explain generalized relationship of watershed

physio climatic characteristic with watershed vulnerability.11Deshmukh and Singh [2016]

Thank you

Questions?

References

A. Deshmukh and R. Singh. Physio-climatic controls on vulnerability of watersheds to climate and

land use change across the united states. Water Resources Research, 2016.

E. Hawkins and R. Sutton. The potential to narrow uncertainty in projections of regional

precipitation change. Climate Dynamics, 37(1-2):407–418, 2011.

R. Singh, T. Wagener, R. Crane, M. Mann, and L. Ning. A vulnerability driven approach to

identify adverse climate and land use change combinations for critical hydrologic indicator

thresholds: Application to a watershed in pennsylvania, usa. Water Resources Research, 50(4):

3409–3427, 2014.