THE FORMATION AND TRANSFORMATION OF SPACE AND KNOWLEDGE IN ANCIENT CIVILIZATIONSTHE FORMATION AND TRANSFORMATION OF SPACE AND KNOWLEDGE IN ANCIENT CIVILIZATIONSThe Formation and Transformation of Space and Knowledge in Ancient Civilizations

Statistical downscaling of precipitation of two transient Holocene AOGCM simulations for central Sudan

Sebastian Wagner1,4, Janina Körper2,4, Jonas Berking3,4

1GKSS Research Center, Institute for Coastal Research, Geeshacht, Germany2Free University of Berlin, Department of Meteorology, Berlin, Germany3Free University of Berlin, Department of Physical Geography, Berlin, Germany4TOPOI Excellence Cluster, Free University of Berlin, Germany

11th International Meeting on Statistical ClimatologyJuly 12-16, Edinburgh, Scotland

Outline

• Motivation

• Climatic conditions in central Sudan

• Calibration and Validation of downscaling model

• Downscaled precipitation in central Sudan during the Holocene for July

• Summary and Outlook

Motivation

• Hypothesis I [ local ]:

Menoric culture around 2000 BP:

Hydrological changes responsible for collapse culture

• Hypothesis II [ large scale ]:

Decrease in Large-scale precipitation in the northern Sahel zone during the Holocene

→ Test with climate models:

Changes in hyrdological variables [ mean/extremes ]

BUT GCMs cannot reproduce hyrdological changes with good skill

→ Downscaling of large-scale circulation

= location of Naga

[ Location of Naga: ]

Climatic conditions in central Sudan: precipitation

Climatic conditions in central Sudan: Sea level pressure

Precipitation = circulation + other

Important note: ε(t) can be quite large

)()()(1

0log ttpcaatAPRECK

k

SLPk

SLPk

Calibration and validation of downscaling model: PCR method

Basis for Downscaling:

Precipitation time series for grid point of VASCLIMO[ DWD ] co-located with the location of Naga 1951-2000

SLP data from NCEP/NCAR re-analysis for period 1951-2000

rvalid=+0.43RE=+0.16

Calibration and validation of downscaling model: Validation

Atmosphere ECHAM4[ 3.75º x 3.75º,19 vert. layers ]

Ocean HOPE-G[ 2.8 x 2.8 , 20 vert. layers ]

[ Experimental Setup for Holocene simulations ]

ECHO-G T30

Holocene simulations 7 ka BP – present:

ORB: Transient orbital forcing ORBSG: Transient orbital, solar and GHG forcing

Precipitation in central Sudan during the Holocene:

Changes in mean precipitation

95% confidence

ORBORBSG

ORBORBSG

Precipitation in central Sudan during the Holocene:

Comparison with original precipitation output from climate model[ Reference: mean of period 1900-2000 AD ]

σSD_dt_IA=35.3 mmσmodel_dt_IA=25.9 mm

ORBSG

Summary

Methodological aspects:

• Statistical downscaling is potentially suited also for arid/semi-arid regions:However: model fit on inter-annual time scales shows only weak model skill

Test of the hypothesis:

• Based on changes in mean precipitation no change between 2000 BP and present-day

• Other reasons: changes in extreme precipitation and/or non-climatic influences

• Reproduction by decrease in precipitation during the Holocene in downscaled precipitation

Outlook :

Mean changes in sea level pressure during the Holocene

Thank you for your attention !

Outlook:

[ Methods for selection of proper geographical domain ]

• One-point correlation map:

Naga

Calibration and validation of downscaling model: PCR method

[ EOF patterns and their interpretation in the context of PCR ]

Precipitation = circulation + other

Important note: ε(t) can be quite large

)()()(1

0log ttpcaatAPRECK

k

SLPk

SLPk

a1=-0.337 a2=-0.23 a3=-0.104

Calibration and validation of downscaling model: PCR method

Calibration and validation of downscaling model: PCR method

[ EOF patterns and their interpretation in the context of PCR ]

Precipitation = circulation + other

Important note: ε(t) can be quite large

)()()(1

0log ttpcaatAPRECK

k

SLPk

SLPk

a1=+0.337 a2=+0.23

Calibration and validation of downscaling model: PCR method

[ EOF patterns and their interpretation in the context of PCR ]

Precipitation = circulation + other

Important note: ε(t) can be quite large

)()()(1

0log ttpcaatAPRECK

k

SLPk

SLPk

a1=+0.337 a2=+0.23 a3=+0.104

EOFs and their interpretation in the context of PCR cont.

a5=+0.328a4=+0.118

Precipitation = circulation + other

Important note: ε(t) can be quite large

)()()(1

0log ttpcaatAPRECK

k

SLPk

SLPk

Solar and CO2 changes:

Motivation: Why Downscaling at all ?

Comparison between

‘real world’ world of a climate model

at a global scale

= location of Naga

Linking scales in the ‘real world’ and apply the link to the GCM large scale

Local scale variable, e.g. precipitation

The solution:

Skill of statistical model calibrated with SLP predictors:

June July August September October

Calibration

correlation +0.41 +0.55 +0.45 +0.52 +0.46

RedinErr +0.16 +0.29 +0.2 +0.27 +0.21

Validation

correlation +0.25 +0.43 +0.27 +0.38 +0.16

RedinErr 0 +0.16 0 +0.1 -0.12

Mean wind conditions during July

jJ

jjecX

1

jjeXc T

Estimation of principal components by means of EOF analysis:

k j cc kj | 0

obsmoTmo

jjeXc

,

Estimation of GCM-modelled principal components cj

Principal Component Regression (PCR):

)()()(1

0 ttcaatPRECK

k

SLPk

SLPk

Changes in variability for downscaled large scale circulation:

Comparison with original precipitation output from climate model[ Reference: mean of period 1900-2000 AD ]

σmodel_dt_IA=26.6mm

σSD_dt_IA=35.3 mmσmodel_dt_IA=25.9 mm

σSD_dt_IA=34.9 mm

Mean differences in precipitation between different periods of the Holocene and present day [ PD, 1900-2000 AD ]

Naga [ 67.7mm ] ORB ORBSG

7k-PD +53.9 +39.5

6k-PD +37.1 +39.2

5k-PD +33.3 +23.8

4k-PD +21.7 +19.3

3k-PD +12.0 +8.7

2k-PD +1.4 -4.5

1k-PD +1.1 -6.2

Experimental Setup for downscaling:

• ECHO-G simulations starting 7000 BP forced with

I) only changes in orbital forcing [ ORB ]

II) additional changes in solar and GHG forcing [ ORBSG ]

Control simulation: pre-industrial with constant condition of 1750 AD

Mean differences in incoming solar radiation Mid-Holocene – PD