Complex Models regarding general atmospheric circulation (GCM) have been developed to predict the future earth climate. Those models are able to simulate the energy and mass exchanges between the atmosphere and the earth surface, according to several man-due scenarios of greenhouse gases emissions (IPCC, 2007). The HadCM3 model, developed by the United-Kingdom Meteorological Office, and the German ECHAM4 model were considered in the IPCC (2007) report, among other non-European GCM's.
On the other hand, seasonal time-scale climate predictions are now made routinely at a number of operational meteorological centers around the world, using comprehensive coupled models of the atmosphere, oceans, and land surface (Stockdale et al. 1998; Mason et al. 1999; Kanamitsu et al. 2002; Alves et al. 2002; Palmer et al., 2004). Particularly, GCM were integrated over 4-month time scales with prescribed observed sea surface temperatures (SSTs) within the PROVOST project (Palmer et al., 2004). Single model and multi-model ensembles were treated as potential forecasts. A key result was that probability scores based on the full multi-model ensemble were generally higher than those from any of the singlemodel ensembles (Palmer et al., 2004).
Based on PROVOST results, the Development of a European Multi-model Ensemble System for Seasonal to Inter-annual Prediction project (DEMETER) was conceived, and funded under the European Union 5 th Framework Environment Programme (Palmer et al., 2004). The principal aim of DEMETER was to advance the concept of multi-model ensemble prediction by installing a number of state-of-the-art global coupled ocean-atmosphere models on a single supercomputer, and to produce a series of 6-month multi-model ensemble hindcasts with common archiving and common diagnostic software. As a result of
DEMETER, real-time multi-model ensemble seasonal global predictions are now routinely made at the European Centre for Medium-Range Weather Forecasts (ECMWF).
Palmer et al. (2004) showed some DEMETER applications. Results indicate that the multi-model ensemble is a viable pragmatic approach to the problem of representing model uncertainty in seasonal-to-inter-annual prediction, and will lead to a more reliable forecasting system than that based on any one single model (Palmer et al., 2004). On the other hand, Doblas-Reyes et al. (2006), pointed out the potential of DEMETER predictions of seasonal climate fluctuations to crop yield forecasting and other agricultural applications. They recommend a probabilistic approach at all stages of the forecasting process.
The ENSEMBLES EU-funded proposal (Hewitt, 2005) is an important recent effort to improve the skill of seasonal forecasts and to make them available to stakeholders. The ENSEMBLES proposal uses the collective expertise of 66 institutes to produce a reliable quantitative risk assessment of long-term climate change and its impacts. Particular emphasis is given to probable future changes in climate extremes, including storminess, intense rainfall, prolonged drought, and potential climate 'shocks' such as failure of the Gulf Stream. To focus on the practical concerns of stakeholders and policy makers, ENSEMBLES considers impacts on timeframes ranging from seasonal to decadal and longer, at global, regional, and local spatial scales.
Several useful tools to assess climate-change impacts on agriculture have been developed during the last years. GCM are among such tools. However, GCM estimations of temperature, precipitation and other meteorological variables are usually made for large areas. For instance, Guereña et al. (2001) showed that those estimations are not very useful to Spanish agricultural climate-change impact assessments, due to the notable topographical changes within the Peninsula for relative small distances. Therefore, a "downscaling" of GCM outputs is absolutely needed before using their estimations for agricultural applications.
Wilby and Wigley (2001) summarized the available downscaling techniques; which can be classified as statistical, dynamical and weather generators. A dynamical downscaling method is to apply numerical regional climate models at high resolution over the region of interest. Regional models have been used in several climate impact studies for many regions of the world, including parts of North America, Asia, Europe, Australia and Southern Africa (e.g Giorgi and Mearns, 1999; Kattenberg et al., 1996; Mearns et al., 1997). The regional climate models obtain sub-grid scale estimates (sometimes down to 25 km resolution) and are able to account for important local forcing factors, such as surface type and elevation. Particularly, the regional climate model RegCM was originally developed at the National Center for Atmospheric Research (NCAR), USA and has been mostly applied to studies of regional climate and seasonal predictability around the world. It is further developed by the Physics of Weather and Climate group at the Abdus Salam International Centre for Theoretical Physics (ICTP) in Trieste, Italy. The PRUDENCE Regional Models Experiment has been developed in Europe under the EU Framework Research Program (Christensen and Christensen, 2007). PRUDENCE project provides a series of high-resolution regional climate change scenarios for a large range of climatic variables for Europe for the period 2071-2100 using four high resolution GCMs and eight RCMs.
Wilby and Wigley (2001) classified statistical downscaling in regression methods and weather-pattern approaches. The regression method uses statistical linear or non-linear relationships between sub-grid scale parameters and coarse resolution predictor variables. Wilby and Wigley (2001) included Artificial Neural Network within the regression-type statistical downscaling. On the other hand, weather-pattern based approaches involve grouping meteorological data according to a given classification scheme. Classification procedures include principal components, canonical correlation analyses, fuzzy rules, correlation-based pattern recognition techniques and analogue procedures; among others (Wilby and Wigley, 2001). Theoretically, dynamical downscaling methods are better than simple statistical methods since they are based on physical laws. However, statistical downscaling method are less computational exigent and can give good results if the relationships between predictand and predictors are stationary (Wilby and Wigley, 2001).
Was this article helpful?
You Might Just End Up Spending More Time In Planning Your Greenhouse Than Your Home Don’t Blame Us If Your Wife Gets Mad. Don't Be A Conventional Greenhouse Dreamer! Come Out Of The Mould, Build Your Own And Let Your Greenhouse Give A Better Yield Than Any Other In Town! Discover How You Can Start Your Own Greenhouse With Healthier Plants… Anytime Of The Year!