Over the past almost four years, much of my research has been done as part of teams that have been using crop models together with climate models to investigate the direct impact of climate change on crop productivity. Much of the work has been completed and published. Examples include the recently released 3-volume set consisting of West African Agriculture and Climate Change, Southern Africa Agriculture and Climate Change, and East Africa Agriculture and Climate Change. Much of the groundwork for these publications were completed by the end of 2010.
In some of the more recent work, we expanded the analysis by also using the crop models to explore the effectiveness of a variety of adaptation options. Results from two of these analyses are out as discussion papers: Agriculture and Adaptation in Bangladesh and Cambodian Agriculture: Adaptation to Climate Change Impact. Another, focusing on the Pacific island nations of Fiji, Papua New Guinea, and the Solomon Islands is expected to be published by the Asian Development Bank sometime in 2014.
In using crop models with climate models, we divide up a country or region into a grid (set of squares) that span the study area. Usually these squares are 5 arc-minutes (around 9 or 10 kilometers) on edge, though in the case of West Africa, many of the squares are 15 arc-minutes (around 30 kilometers) on edge. Then, we do analysis on each square, first for the “baseline” climate (that of the 1950 to 2000 period), and then on the climate of the future (most of the work focused on 2050).
The crop model uses mathematical equations to “grow” each crop a day at a time. The DSSAT software, which we used in these studies, allows the user to specify a variety of the crop that is being analyzed (each variety has its own set of parameters to be used in the model, and generally each crop has its own model or set of equations). The model uses soil characteristics together with user-specified crop management methods. In addition, daily weather data is used. Since we usually do not have daily weather, we usually use weather data generated by a simulator that uses the monthly climate parameters to determine what kind of weather would be feasible. Since weather can vary from year to year, we usually simulate between 30 and 90 years worth of weather data, computing yields for each annual set of data.
In the next post (part 2 of this article), we will look at a case study for rainfed maize in Kenya that will illustrate the incredible usefulness of this approach. We will help the reader better understand what the maps are telling us in terms of climate impacts. We also hope to help the reader understand the kind of policies and projects that would be appropriate to use in response to the information they contain.