using bias-corrected regional climate modelling outputs for the...

11
09/06/2015 1 LAS RICCAR Using Bias-Corrected Regional Climate Modelling Outputs for the Arab Domain to support Hydrological Modelling Joel Dahné Swedish Meteorological and Hydrological Institute (SMHI) LAS RICCAR Future Climate Projections A core activity within RICCAR is to produce regionally downscaled regionally downscaled regionally downscaled regionally downscaled future climate projections for the Arab Region We need an ensemble of projections ensemble of projections ensemble of projections ensemble of projections because there is no single answer ! Arab RCM Domain ensemble” > reproduce results many times using variations in how we go about it

Upload: others

Post on 12-Sep-2020

6 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

1

LAS

RICCAR

Using Bias-Corrected Regional Climate

Modelling Outputs for the Arab

Domain to support Hydrological

Modelling

Joel Dahné

Swedish Meteorological and Hydrological Institute (SMHI)

LAS

RICCAR

Future Climate Projections

A core activity within RICCAR is to produce

regionally downscaledregionally downscaledregionally downscaledregionally downscaled future climate

projections for the Arab Region

We need an ensemble of projectionsensemble of projectionsensemble of projectionsensemble of projections because

there is no single answer!

Arab RCM Domain“ensemble” > reproduce results many times using variations in how we go about it

Page 2: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

2

LAS

RICCAR

Future Hydrological Projections

Regional Hydrological Modelling over the Arab

Region is a key component of RICCAR

• (RCP2.6)

• RCP4.5

• RCP8.5

• EC-EARTH

• CNRM-CM5

• GFDL_ESM2M

• RCA4 • VIC

• HYPE

• Python

• ArcGIS

LAS

RICCAR

RCM

(Institute)

GCM Historical

1950-

2005

RCP4.5

2006-

2100

RCP8.5

2006-

2100

RCA4

(SMHI)

EC-Earth

50km✔ ✔ ✔

RCA4

(SMHI)

EC-Earth

25km✔ ✔

RCA4

(SMHI)

CNRM

50km✔ ✔ ✔

RCA4

(SMHI)

GFDL-ESM

50km✔ ✔ ✔

RCA4

(SMHI)

GFDL-ESM

25km✔ ✔

CORDEX-MENA/Arab Ensemble Matrix

Used for Hydrological analysis

Page 3: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

3

LAS

RICCAR

Requires an interface to overcome RCM biases

Downscaling of RCMs

LAS

RICCAR

Temperature

Precipitation

Example: Bias in EC-EARTH as compared to reference data

Page 4: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

4

LAS

RICCAR

Obs

Direct

RCA3_E5/Ctrl

overestimated

dischange

Example from Torpshammer River Basin in Sweden

(Observations & RCA-ECHAM5) - RCM overestimates precipitation

Direct input of Prec. & Temp. from RCM

control period: 1961-1990

Hydrological model simulations (HBV Model)

LAS

RICCAR

Obs

Direct RCA3

E5/Ctrl

Bias corrected

E5/Ctrl

Bias-corrected input of Prec. & Temp. from

RCM control period: 1961-1990

improved

dischange

Example from Torpshammer River Basin in Sweden

(Observations & RCA-ECHAM5) - RCM overestimates precipitation

Hydrological model simulations (HBV Model)

Page 5: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

5

LAS

RICCAR

Some 30 000 subbasinsaverage size 650 km2

Gridded 50 x 50 kmGrid box area 2500 km2

Hydrological Models used for analysis

Main differences

More conceptual �Mode physical

Division into basins � Gridded

Integrated routing, lake/dam routines � routing, lake/dam as modules

Parameterization and calibration

Hype Model VIC Model

LAS

RICCAR

Regionalization, from local to regional model setup

• Calibrate gauged catchments

• Regionalization by nearest neighbor

SoilLandcover

• Calibrated gauged basins (pilot sites)

• Regionalization by linking parameters

to land cover or soil.

Characterization

(VIC Global Input Parameters)

• Hydraulic conductivity

• Elevation

• Soil density

• Water content at wilting point

• Water content at field capacity

Globcover Digital Soil Map of the World

Hype Model VIC Model

Characterization

Landuse Soil type

Arable Coarse

Forest Medium

OpenLand Fine

Wetland Organic

Urban Bare/Shallow

Bare land Submerged/Irrigated

Water

Page 6: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

6

LAS

RICCAR

River discharge

Runoff from land

Evapotranspiration

Soil water storage

Precipitation

Temperature

Hydrological modelling input and output

HYPE

RCP 4.5 RCP 8.5

Hydro Ensemble

Mean, Max, Min

Creating Future Hydro Projection Ensembles

VIC

HYPEVIC

Hydro Ensemble

Mean, Max, Min

VIC

HYPE

Modelrun 1980 – 2100 Modelrun 1980 – 2100

HYPE

VIC

2*high resolution

VIC

HYPE

Page 7: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

7

LAS

RICCAR

Future hydro climate projectionsSub-regions

LAS

RICCAR

Maps

Timeserie graphs (to ensembles)

Seasonal graph

(to ensembles)

Graphs relating rcp8.5 ensemble to

High Resolution projections

Hyd-results

• Evapotranspiration

• Soil water storage

• Runoff from land

• River discharge

Climate projections

• 3 RCP 8.5

• 3 RCP 4.5

• 2 RCP8.5, high res

Hyd-models:

• VIC

• HYPE

Time periods:

• 1986-2005

• 2016-2035

• 2046-2065

• 2086-2100

Future hydro climate projectionstype of outcome

Tables

Value Change Agreement

Page 8: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

8

LAS

RICCAR

Model forcing - RCP 8.5Climate Models: 3-member ensemble

Temperature

Precipitation

Reference

1986-2005

Ensemble

agreement

Change to

2081-2100

> 16 <-16[mm/month]<0.1 >150[mm/month] All - All +

0 35[°C] > 4.0 <-4.0[°C] All - All +

LAS

RICCAR

Evapotranspiration - RCP 8.5Hydro Models: 3-member ensemble

HYPE-m

odel

VIC-m

odel

Reference

1986-2005

Ensemble

agreement

Change to

2081-2100

> 16 <-16[mm/month]<0.1 >150[mm/month] All - All +

Page 9: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

9

LAS

RICCAR

Runoff - RCP 8.5Hydro Models: 3-member ensemble

HYPE-m

odel

VIC-m

odel

Reference

1986-2005

Ensemble

agreement

Change to

2081-2100

> 16 <-16[mm/month]<0.1 >150[mm/month] All - All +

LAS

RICCAR

Mediterranean

Annual HYPE VIC HYPE VIC

Period RCP Runoff Evap

2016-2035 4.5 3% 10% -1% -1%

8.5 9% 17% 0% 1%

2046-2065 4.5 2% 8% -3% -2%

8.5 -2% 2% -7% -7%

2081-2100 4.5 8% 17% -2% -1%

8.5 -13% -10% -14% -15%

Changes to 1986-2005

Pre

cip

ita

tio

n

[%]

Tem

pe

ratu

re

[⁰C

]

Ru

no

ff (

HY

PE

)

[%]

Ev

ap

ora

tio

n (

HY

PE

)

[%]

Seasonal changes 2071-2100 to 1986-2005

RCP 8.5

RCP 4.5

mm

/da

y]

[⁰C

]m

m/d

ay

]

Summary table (ensemble mean)

Page 10: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

10

LAS

RICCAR

Moroccan Highlands

Annual HYPE VIC HYPE VIC

Period RCP Runoff Evap

2016-2035 4.5 3% 2% -2% -2%

8.5 -7% -9% -3% -2%

2046-2065 4.5 -17% -22% -8% -7%

8.5 -32% -40% -15% -12%

2081-2100 4.5 -23% -28% -11% -10%

8.5 -48% -59% -26% -22%

Pre

cip

ita

tio

n

[%]

Tem

pe

ratu

re

[⁰C

]

Ru

no

ff (

HY

PE

)

[%]

Ev

ap

ora

tio

n (

HY

PE

)

[%]

Changes to 1986-2005 Seasonal changes 2071-2100 to 1986-2005

RCP 8.5

RCP 4.5

mm

/da

y]

[⁰C

]m

m/d

ay

]

Summary table (ensemble mean)

LAS

RICCAR

Ethiopian highlands

Pre

cip

ita

tio

n

[%]

Tem

pe

ratu

re

[⁰C

]

Ru

no

ff (

HY

PE

)

[%]

Ev

ap

ora

tio

n (

HY

PE

)

[%]

Changes to 1986-2005 Seasonal changes 2071-2100 to 1986-2005

Annual HYPE VIC HYPE VIC

Period RCP Runoff Evap

2016-2035 4.5 0% 0% -2% 0%

8.5 -2% 0% -4% 0%

2046-2065 4.5 -5% -1% -7% 0%

8.5 -3% 0% -3% 0%

2081-2100 4.5 -8% -1% -13% -1%

8.5 -7% -1% -10% -1%

Summary table (ensemble mean)

Page 11: Using Bias-Corrected Regional Climate Modelling Outputs for the …css.escwa.org.lb/SDPD/3639/PR-S2-1.pdf · 2015. 6. 9. · • HYPE • Python • ArcGIS LAS RICCAR RCM (Institute)

09/06/2015

11

LAS

RICCAR

Final remarks:

• Bias-correction proved to be more demanding for

the highly variable climate of the region than

expected

• The volume of data processed required a lot of

time, both computational and in man hour

• We are still in the process of identifying what the

main outcomes are