Transcript PPT, 2.5 MB
Uncertainty in climate scenarios when downscaling with an RCM M. Tadross, B. Hewitson, W Gutowski & AF07 collaborators Water Research Commission of South Africa The 4 sources of uncertainty we will address: • General Circulation Models (GCMs); simulations of recent climate. • Regional Climate Models (RCMs); how well do they represent present climate ? • Land-surface exchange and influence on local climate • Observational network; how good/bad is the present observational network for assessing these models ? ERA ranalysis NCEP reanalysis ECHAM CSIRO Total moisture Sensitivity assessment alternative: First step is to assess system sensitivity to perturbations GCM derived anomaly provides credible perturbations of baseline climate HadCM3 ECHAM4 • GCMs simulate present climate within observational uncertainty - but uncertainty is large - how do these biases affect an RCM ? • They can be useful when used at the appropriate scales e.g. Hewitson, 2003 Station only rainfall Station + satellite rainfall Observed precipitation MM5 - Grell MM5 - Kain Fritsch Modelled precipitation MM5 inter-annual variability DJF 1988 (wet) - DJF 1991 (dry) Betts Miller convection Kain Fritsch convection CRU CMAP MM5 Precipitation change due to differences in representing the atmosphere NCEP boundary conditions ERA boundary conditions Large differences in the centre of the domain • RCM simulations strongly depend on the model physics - use multiple models - validate each over the region and variable of interest. • However, appropriately configured an RCM can give realistic simulations of local climate. Over to Bill for discussion of land-surface sources of uncertainty …………………….. Perturbation simulations…. Experiment: Desiccation of soil moisture – temperature increases up to 4°C Changing LAI …. Original: All vegetation have LAI 4 Set grassland/bare/shrubland Set evergreen broadleaf LAI 0.1 LAI 8 Satellite derived LAI Vegetation classes Precipitation change LAI 0.1 LAI 8 SDGVM re-classification of MM5 vegetation …. 30 year climatology assuming no human influence Equivalent LAI field What will be the effect on MM5 climate ? • Changes in land-surface characteristics can affect local climates as much as current climate change projections. • Correct representation of vegetation and/or land-use may improve model performance. Data example: Station presence between 19802000 Half the stations are present on any given day South African ~50-100km observational network. Urgent need for access to national data archives in other African nations! Rawinsonde count • Extreme sparsity of observations over most of Africa SEVERELY limits validation data. • ALL model testing and efforts to reduce uncertainty rely on this data. Climate change check list • Can you access data from multiple models for comparison ? • Is the model being assessed at the resolution it has skill ? • Has model been assessed over your region of interest and for the parameter your study is sensitive to ? • Was your future scenario created in the same way as your validation ? • Are the climate change differences important ?