Climate change is expected to substantially reduce agricultural yields, as reported in the by the Intergovernmental Panel on Climate Change (IPCC). In Sub-Saharan Africa and (to a lesser extent) in South Asia, limited data availability and institutional networking constrain agricultural research and development. Here we performed a review of relevant aspects in relation to coupling agriculture–climate predictions, and a three-step analysis of the importance of climate data for agricultural impact assessment. First, using meta-data from the scientific literature we examined trends in the use of climate and weather data in agricultural research, and we found that despite agricultural researchers’ preference for field-scale weather data (50.4% of cases in the assembled literature), large-scale datasets coupled with weather generators can be useful in the agricultural context. Using well-known interpolation techniques, we then assessed the sensitivities of the weather station network to the lack of data and found high sensitivities to data loss only over mountainous areas in Nepal and Ethiopia (random removal of data impacted precipitation estimates by ±1300 mm/year and temperature estimates by ±3 °C). Finally, we numerically compared IPCC Fourth Assessment Report (4AR) climate models’ representation of mean climates and interannual variability with different observational datasets. Climate models were found inadequate for field-scale agricultural studies in West Africa and South Asia, as their ability to represent mean climates and climate variability was limited: more than 50% of the country-model combinations showed less than 50% adjustment for annual mean rainfall (mean climates), and there were large rainfall biases in GCM outputs (1000–2500 mm/year), although this varied on a GCM basis (climate variability). Temperature biases were also large for certain areas (5–10 °C in the Himalayas and Sahel). All this is expected to improve with IPCC's Fifth Assessment Report; hence, appropriate usage of even these new climate models is still required. This improved usage entails bias reduction (weighting of climate models or bias-correcting the climate change signals), the implementation of methods to match the spatial scales, and the quantification of uncertainties to the maximum extent possible.
Ramirez-Villegas, J.; Challinor, A. Assessing relevant climate data for agricultural applications. Agricultural and Forest Meteorology (2012) 161: 26-45. [DOI: 10.1016/j.agrformet.2012.03.015]