Testing the Effect of Domain Size on RegCM3’s Precipitation Output Over Southern Africa

Testing the Effect of Domain Testing the Effect of Domain Size on RegCM3’s Precipitation Size on RegCM3’s Precipitation

Output Over Southern AfricaOutput Over Southern Africa

Anne BarrableAnne Barrable¹ ¹

Mary-Jane Kgatuke² Mary-Jane Kgatuke²

Chris Lennard¹Chris Lennard¹

1- University of Cape Town ; 2- South African Weather Service

Aims and ObjectivesAims and Objectives

Establish the effect of different domain Establish the effect of different domain size and location on RegCM3 output.size and location on RegCM3 output.

Particularly examining precipitation in an Particularly examining precipitation in an agriculturally sensitive region.agriculturally sensitive region.

MethodsMethods

Define 3 different domains over southern Define 3 different domains over southern Africa including Madagascar. Africa including Madagascar.

Two domains chosen to overlap a Two domains chosen to overlap a common region, which formed the 3common region, which formed the 3rdrd domain… domain…

Grid points: 43x32

Grid points: 18x18

Grid points: 64x44

SA Domain

Common Domain

SA + Madagascar

MethodsMethods

Define 3 different domains over southern Define 3 different domains over southern Africa. Africa.

Examined February 1995, 1996, 1997. Examined February 1995, 1996, 1997.

1111thth February anomalously wet day in study February anomalously wet day in study region –compared model between domains.region –compared model between domains.

Model ParametersModel Parameters 60km grid.60km grid.

Lambert Conformal projection. Lambert Conformal projection.

NCEP reanalysis data. NCEP reanalysis data.

Grell convection scheme using Fritsch & Grell convection scheme using Fritsch & Chappell closure. Chappell closure.

ResultsResults Examined surface model precipitation Examined surface model precipitation

data.data.

Daily precipitation averages.Daily precipitation averages.Comparison over common domain.Comparison over common domain.

1111thth February 1996. February 1996.Comparison over common domain. Comparison over common domain.

1995 19971996

?

Tropical Cyclone Josie was present in the Mozambique Channel for 5 days (9-14 Feb 1997)

SA Domain

Common Domain

SA + Madagascar

Daily average Feb 1995

SA Domain

Common Domain

SA + Madagascar

Daily average Feb 1996

SA Domain

Common Domain

SA + Madagascar

Daily average Feb 1997

SA Domain

Common Domain

SA + Madagascar

11th February 1996

♠ Bloemfontein (11.5mm)

♣ Polokwane (10.4mm)

♦ Johannesburg (65.0mm)

♥ Van Reenen (74.2mm)

♠

♣

♦

♥

♥

♥

♦

♦

♣

♣

♠

♠

General ConclusionsGeneral Conclusions Summer and winter rainfall regions identified.Summer and winter rainfall regions identified. 1996 wetter year corresponds to known La 1996 wetter year corresponds to known La

Nina year. Nina year. Different domains display different spatial Different domains display different spatial

precipitation characteristics.precipitation characteristics. Although not presented here, temperature Although not presented here, temperature

(surface & 2m) showed a cold bias (3°).(surface & 2m) showed a cold bias (3°). Choice of domain location and size crucial Choice of domain location and size crucial

w.r.t. topography w.r.t. topography ANDAND synoptic systems. synoptic systems.

Caveats and Model “Hitches”Caveats and Model “Hitches”

Common domain too small; nudged by forcing Common domain too small; nudged by forcing data and did not allow the model to run freely.data and did not allow the model to run freely.

Running smaller grid sizes (30km & 20km) did Running smaller grid sizes (30km & 20km) did not work:not work: 2 segmentation faults (even with time step at 20 2 segmentation faults (even with time step at 20

seconds);seconds); 2 computer freezes;2 computer freezes; 1 floating point exception error….Gave up! 1 floating point exception error….Gave up!

Stretched grid for smaller domains?Stretched grid for smaller domains?

Caveats and Model “Hitches” Caveats and Model “Hitches” cont…cont…

Regrid for different grid size comparisons Regrid for different grid size comparisons (in GrADS; already available in Ferret).(in GrADS; already available in Ferret).

Map projection difficult to implement with Map projection difficult to implement with precipitation. precipitation. Map possibly shifted, particularly in SA-Map possibly shifted, particularly in SA-

Madagascar domain.Madagascar domain.

AcknowledgmentsAcknowledgments

Special thanks to Bill Gutowski for Special thanks to Bill Gutowski for meaningful discussions of model results. meaningful discussions of model results.

Thanks to Filippo Giorgi, his team, and the Thanks to Filippo Giorgi, his team, and the ICTP for allowing us the opportunity to ICTP for allowing us the opportunity to learn and run RegCM”3”. learn and run RegCM”3”.