Global climate change models are of limited use to agricultural policymakers in some regions of the developing world, according to a report by the CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS).
The report was launched at the Climate Models and Farm Crop Forecasting in South Asia and Africa meeting last month (21 February).
The researchers, from Oxford University, United Kingdom, and the University of Cape Town, South Africa, studied the ability of global climate models to predict regional climate events such as monsoon rains and temperatures — and found mixed results.
"The models have a reasonable capability in terms of reproducing [trends in the] East African climate," said Richard Washington, professor of climate science at the University of Oxford.
But in West Africa, particularly in the Sahel region, the models predicted more monsoon rains, of different duration, to those that were actually observed.
Similar difficulties were encountered with India's monsoons, the authors said. Global models were generally accurate in predicting large-scale horizontal movement of air (atmospheric flow), but not the specific timing or patterns of monsoon rainfall, according to Mark New, professor of climate science at the University of Cape Town.
The authors said global models often failed to take account of complex regional climatic factors — making them less useful for policymakers.
For example, Asia's monsoons are affected by many region-specific factors, such as El Niño events, atmospheric pressure over the North Atlantic Ocean, and Asia's so-called 'brown cloud' air pollution.
"If you want a climate model that predicts monsoon rainfall variability you need one that gets each of these factors right," New told SciDev.Net.
The authors suggested greater use of 'ensemble' models which combine results from several models to generate averages; and also the use of global models in conjunction with regional ones, to enable regional information to be factored in alongside larger-scale processes.
Washington said it was also important to improve field data on which models are based, and remove any existing biases.
"Otherwise we are left to choose between models that are different [without knowing] which one is better," he said.
Philip Thornton, a senior scientist at the International Livestock Research Institute in Kenya and a modelling tools leader at CCAFS told SciDev.Net: "The more we understand [uncertainty in models], the better we can deal with it".
Link to full report [265kB]
See below for a video of event: