Quantifying the effects of climate change on crop yields.

摘要

Climate is one of the primary determinants of agricultural productivity. For any given location and crop, climate determines the length of the growing season, the amount of heat that can be used in photosynthesis and carbon fixing, and the expected amount of temperature and water stress. Anthropogenic warming of the climate is therefore of potentially great consequence to agriculture and food security on a planet under increasing environmental and population pressures. Technology in the form of improved seed varieties, chemical inputs, and mechanization all serve to mitigate to varying degrees the deleterious effects of climate-induced stress on crops. In the U.S. this 'technology signal' is evident in the most widely grown cereal crops, particularly corn ( Zea mays), where yields have increased by two bushels per year on average since 1950. Despite the overall increasing yield trend, crops are still vulnerable to climatic variability and extreme events such as drought and heat stress, which will reduce yields below the expected amount. The response of crops to an increase in these extreme events is uncertain because of the co-mingling effects of technology and climate that is manifest in the observed yield trend. In addition, while there is high confidence that mean temperatures will increase in response to rising greenhouse gasses, our limited knowledge and imperfect ability to model the climate system results in large uncertainties about how the more agriculturally-relevant tails of the temperature distribution will change. This dissertation develops probabilistic projections of the potential effects of anthropogenic climate change on corn yields. Three research objectives are developed that improve our understanding of the climate change-yield relationship: (1) identify and evaluate the portions of the temperature distribution that drive yield variability notwithstanding the technological signal, (2) develop projections of how these agroclimate variables could change in the future while accounting for uncertainty, and (3) develop projections of the expected effects on crop yields given these changes in the agro-climate. These objectives are addressed in three research articles. The first article is an evaluation of recent changes in three agro-climate indices (frost days, thermal time, and heat stress index) in North America and the ability of general circulation models (GCMs) to reproduce the observed patterns and trends. The next article uses bootstrapping and a statistical method known as Bayesian Model Averaging (BMA) to develop probabilistic projections of the three agro-climate indices. The method accounts for important structural uncertainties between climate models that are often neglected or minimized in many impact studies. The final article uses the bootstrapping and BMA methods described in the second article, as well as an empirical climate-yield damage function, to develop probabilistic projections of corn yields in the 21 st century in the eastern U.S. for a high greenhouse gas emissions scenario. Two different sources of uncertainty in the projections are quantified: structural uncertainty between GCMs, and parametric uncertainty in the damage function due to spatial dependence between observations. The results show that severe corn yield damages are possible by the end of the 21st century if the current emissions trajectory is maintained over the entire interval. After accounting for both GCM and damage function uncertainty, the full 95% projection interval shows potential yield declines of between 16% and 77% by 2100, relative to current average yields. These results suggest that substantial adaptation and mitigation will have to occur if the concomitant impacts on global food security are to be avoided without also causing further environmental degradation.