26
MarkSim GCM: generating plausible weather data for future climates September 2011

MarkSim GCM: generating plausible weather data for future climates

Embed Size (px)

DESCRIPTION

CCAFS workshop titled "Using Climate Scenarios and Analogues for Designing Adaptation Strategies in Agriculture," 19-23 September in Kathmandu, Nepal.

Citation preview

Page 1: MarkSim GCM: generating plausible weather data for future climates

MarkSim GCM: generating plausible weatherdata for future climates

September 2011

Page 2: MarkSim GCM: generating plausible weather data for future climates

MarkSim GCM

A tool to generate daily data that are characteristic (to some extent) of future climatologies for any point on the globe, to drive agricultural impact models

Page 3: MarkSim GCM: generating plausible weather data for future climates

™A Markov weather simulator that generates simulated daily weather for any point in the tropics

• Runs off interpolated climate surfaces

• Model fitted to 10,000 weather stations

• World climates classified into 701 groups

• Climate group is known for any pixel on map

• Some model parameters estimated by regressions within climate groups based on the pixels’ climate from the interpolated surface

Page 4: MarkSim GCM: generating plausible weather data for future climates

Number of MarkSim rainfall stations per half-degree grid cell (as in early 2010)

• Still some major gaps in coverage, particularly in the tropical areas of Africa

• Version 2 of MarkSim is under development, and will be based on >50,000 sites globally with historical daily rainfall data

Page 5: MarkSim GCM: generating plausible weather data for future climates

Multi-model global averages of surface warming (relative to 1980-99) for the SRES scenarios

The future is uncertain: which GCM, which emissions scenario?

Page 6: MarkSim GCM: generating plausible weather data for future climates

Scenario 2011-2030

2046-2065

2080-2099

A2 (“high” emissions) 0.64 1.65 3.13

A1B (“medium” emissions) 0.69 1.75 2.65

B1 (“low” emissions) 0.66 1.29 1.79

Committed warming (emissions stabilised at 2000 levels)

0.37 0.47 0.56

Projected mean impacts on global temperatures of three different scenarios

Global mean warming from the IPCC multi-model ensemble mean for three periods relative to 1980–1999 (Wilby et al. 2009, data source IPCC 2007)

Page 7: MarkSim GCM: generating plausible weather data for future climates

Atmosphere-Ocean General Circulation Models used in MarkSim GCM4

Model Name Institution Code

BCCR_BCM2.0 Bjerknes Centre for Climate Research (University of Bergen, Norway)

BCC

CNRM-CM3 Météo-France/Centre National de Recherches Météorologiques, France

CNR

CSIRO-Mk3_5 Commonwealth Scientific and Industrial Research Organisation (CSIRO) Atmospheric Research, Melbourne, Australia

CSI

ECHam5 Max Planck Institute for Meteorology, Germany ECH

INMCM3_0 Institute for Numerical Mathematics, Moscow, Russia INM

MIROC3.2 (medres)

Center for Climate System Research (University of Tokyo), National Institute for Environmental Studies, and Frontier Research Center for Global Change (JAMSTEC), Japan

MIR

Ensemble average

Average climatology of the above 6 AOGCMs AVR

Page 8: MarkSim GCM: generating plausible weather data for future climates

How different are the projections of rainfall and temperature among the various GCMs?

One good place to find out: http://www.ipcc-data.org/maps/

Page 9: MarkSim GCM: generating plausible weather data for future climates

June precipitation anomalies (relative to the 20th century control 1961-1990 30-year normal) for 2046-2065 for SRES A2 and four GCMs www.ipcc-data.org

Page 10: MarkSim GCM: generating plausible weather data for future climates

Annual rainfall changes (mm) from 2000 to 2050, A2

CNR CSI

ECH MIR

Page 11: MarkSim GCM: generating plausible weather data for future climates

Where do the MarkSim model parameters come from?

From climate grids, or from the user directly:• Monthly rainfall amounts• Monthly average max and min temperatures

From the climate typing clusters:• Number of rain days per month• Monthly correlation matrix of raindays per month• Baseline probits of a wet day following three dry days and the “lag parameters”

Derived parameters:• Monthly solar radiation

Page 12: MarkSim GCM: generating plausible weather data for future climates

Where do the MarkSim model parameters come from?

From climate grids, or from the user directly:• Monthly rainfall amounts• Monthly average max and min temperatures

From the climate typing clusters:• Number of rain days per month• Monthly correlation matrix of raindays per month• Baseline probits of a wet day following three dry days and the “lag parameters”

Derived parameters:• Monthly solar radiation

Grids of possible future climates from GCMs, from RCMs

Page 13: MarkSim GCM: generating plausible weather data for future climates

MarkSim GCM

Climate model outputs:

• Calculate “long-term” monthly means for rainfall and max & min temperatures from daily output:

20-or 30-year monthly averages, say for 2041-2060 (the “2050s”)

• Downscale spatially using “unintelligent” downscaling (e.g. the “delta” method):

interpolate from the coarse (often 200-300 km grids) spatial resolution of the climate models to a higher resolution and add differences to a baseline climatology such as WorldClim (www.worldclim.org)

Page 14: MarkSim GCM: generating plausible weather data for future climates

What to use for “observations”?Could use www.worldclim.org

Page 15: MarkSim GCM: generating plausible weather data for future climates

Where do the MarkSim model parameters come from?

From climate grids:• Monthly rainfall amounts• Monthly average max and min temperatures

From the climate typing clusters:• Number of rain days per month• Monthly correlation matrix of raindays per month• Baseline probits of a wet day following three dry days and the “lag parameters”

Derived parameters:• Monthly solar radiation

With future climatologies of rainfall and max & min temps, we could then generate the remaining parameters and simulate “plausible” daily data for these climatologies

Page 16: MarkSim GCM: generating plausible weather data for future climates

Where do the MarkSim model parameters come from?

From climate grids:• Monthly rainfall amounts• Monthly average max and min temperatures

From the climate typing clusters:• Number of rain days per month• Monthly correlation matrix of raindays per month• Baseline probits of a wet day following three dry days and the “lag parameters”

Derived parameters:• Monthly solar radiation

But: remember Cape Town?

Differences between coarse-grid statistical downscaling and RCM-based downscaling

We could do much better using future climatologies derived in different / better ways

Page 17: MarkSim GCM: generating plausible weather data for future climates

MarkSim GCM

• “All” that is needed from a climate model is a set of long-term mean monthly data for rainfall and max & min temperatures for a specific time slice and GHG emissions scenario

• Climate typing is then used to generate the remaining parameters

• MarkSim’s climate clustering is based on current climate types: climates are changing through time, and the climate cluster to which any point belongs often changes into the future

Page 18: MarkSim GCM: generating plausible weather data for future climates

MarkSim GCM

• MarkSim asymptotically matches monthly means of rainfall (whether current or future); but concerning changes in climate variability, the only information in MarkSimGCM relates to the current variability of the fitted cluster – there is no other information available

• What this means: GCMs are climate models, not weather models; the methods used in MarkSimGCM cannot capture future (unknown) variability in weather (although they might do better with RCM-derived climatologies)

Much care is needed in how we use the outputs of MarkSimGCM and similar tools

Page 19: MarkSim GCM: generating plausible weather data for future climates

• MarkSim is a climate typer – so shifts in climate cluster between now and 2050, say, may result in shifts in weather variability (associated with the new cluster) – but not much if any reason to suppose that this may be realistic

• Use ensembles of GCM-scenario combinations, look at the variation in mean response, and present this variability (uncertainty)

• The general approach may be OK; the weakest link is the future climatologies. Much more advisable to use a better climate downscaling approaches + weather generation?

MarkSim GCM

Page 20: MarkSim GCM: generating plausible weather data for future climates
Page 21: MarkSim GCM: generating plausible weather data for future climates

The tool allows you to select one of the three scenarios, and one of 6 climate models (or their average)

Page 22: MarkSim GCM: generating plausible weather data for future climates

Select climate model

Select emissions scenario

Select the centre year of the time slice and number of years of data wanted

Select location (the ILRI cafeteria in Nairobi)

Page 23: MarkSim GCM: generating plausible weather data for future climates

… graphed …

After running the model, the daily data can be viewed directly …

… or downloaded as a ZIP file to the user’s computer

Page 24: MarkSim GCM: generating plausible weather data for future climates

Region Dec-Jan Jun-Aug

Sahara Small decrease(5-20%)

Inconsistent

West Africa Inconsistent Inconsistent

East Africa Small increase (5-20%)

Inconsistent

Southern Africa Inconsistent Large decrease (>20%)

Limitations and uncertainties associated with these data

GCM consistency in regional precipitation projections for 2090-2099 (SRES A1B). IPCC, 2007

Page 25: MarkSim GCM: generating plausible weather data for future climates

http://gismap.ciat.cgiar.org/MarkSimGCM/

Demo

Page 26: MarkSim GCM: generating plausible weather data for future climates