Page 95 - Climate Change and Food Systems

yields in Tanzania, finding that intra-seasonal variability has the dominant impact. To project future climate, Rowhani et al. apply projected temperature and rainfall outputs for SRES A1B from 22 GCMs for 2050. Controlling for intra- annual climate variability, Rowhani et al. show that climatic impacts by 2050 on crop yields
in Tanzania are underestimated by 4 percent,
9 percent, and 29 percent for maize, sorghum and rice, respectively, when only inter-annual variability is accounted for.
Lobell et al. (2012) also address intra-seasonal variability in their estimation of a statistical yield model of wheat production in India that describes the effects of extreme heat at the grain-filling stage, just prior to maturity. They compute that high heat truncates this process and lowers yields, with stronger effects as temperatures rise above 34C. Their finding is consistent with that of Schlenker and Roberts (2009), who carried out both cross- section (which assumes farmer adaptation) and time series (assumes no adaptation) studies of daily weather and yields in selected areas of
the United States. Results of both approaches describe asymmetric, non-linear effects of higher temperatures on corn, soybeans and cotton. Yields increase gradually until optimum temperatures are reached, but temperatures above that level result
in very steep yield declines. Schlenker and Roberts project that US area-weighted average yields will decline by 30-46 percent by the end of the century, compared to 1950-2005, under the slowest (B1) warming scenario and decrease by 63-82 percent under the most rapid warming scenario (A1FI), based on the Hadley III GCM model.
Ricardian models
Ricardian, also called hedonic, models are cross-section estimations of the effects of
climate, soils, geography, prices and farmer characteristics on land values or current net returns (Mendelsohn et al., 1994; Mendelsohn, 2009).
The models are based on the Ricardian view that land values reflect land productivity. The driving assumption of the model is that farmers seek to maximize their land value and to do so, will make
adaptations, such as changes in crop choices,
to maximize land productivity as environmental conditions change. Output of the models describe the marginal change in land rents or land values with respect to marginal changes in temperature and precipitation. These relationships can then
be combined with climate change projections to provide estimates of climate-change damages expressed in terms of changes in land values or net returns (Schlenker et al., 2006).
The key advantage of the Ricardian model’s cross-section approach is that its measure of climate-induced damage takes into account farmers’ observed, whole-farm adaptations. At the same time, however, the model always represent equilibrium situations and are not well-suited to describe adaptive transitions over time. Perhaps the model’s most important limitation is that prices are implicitly assumed to be constant. If agricultural prices rise as predicted in recent climate literature (e.g., Nelson et al., 2010), then land values also will rise (or not fall as much), so the Ricardian model may overestimate damages (Seo, et al. 2009).
Ricardian models also have been criticized because they do not describe the types and
costs of adaptations that have taken place. However, advances in Ricardian analysis have
led to important insights on adaptation. The
recent body of Ricardian analyses have largely stemmed from two major research programs supported by the Global Environment Facility
(GEF) and the World Bank. These generated a series of Ricardian analyses of crop and livestock production in 11 African countries and 7 countries in South America. Under these projects, surveys designed specifically to describe climate change were conducted of over 10,000 farms in Africa and almost 3000 farms in South America that represent a wide range of climates and farming systems.
In summarizing the African case studies, Dinar, et al. (2008) describe their methodological advances using the term “structural Ricardian” analysis. In this approach, farmer optimization
is described as a simultaneous, multi-stage procedure. A household model describes
the farmer’s adaptive choices, such as crop
chapter 3: economic modelling of climate impacts and adaptation in agriculture: a survey of methods, results and gaps
75