Using impact response surfaces to analyse the likelihood of impacts on crop yield under a changing climate.

Abstract

Most studies of future climate change impacts rely on estimates based on a limited set of projections of future climate. This way, it is not possible to determine whether one estimate is more or less likely than another. However, if future climate outcomes can be expressed probabilistically, this makes it possible to express impacts in terms of likelihoods, as demonstrated in this study.The approach involves overlaying joint probability density functions (pdfs) that describe uncertainties in projections of temperature and precipitation change over future time periods (using RCP-based climate model simulations) with impact response surfaces (IRSs). The IRS shows the modelled sensitivity of crop yield across a wide range of systematic changes in the same climate variables relative to the baseline (1981-2010). The likelihood of falling short of a target yield threshold is then calculated by integrating across the area of the pdf where yields are below the threshold. The WOFOST crop model was run for a locally grown cultivar of spring barley in south-west Finland assuming contrasting clay loam and sandy soils. IRSs were constructed for seven future CO2 concentrations representing time periods during the 21st century, so that the time-evolution of impact likelihoods with respect to mean yield levels and reliability can be presented. The effectiveness of adaptation options was demonstrated with simulations for cultivars with different development rates.The approach is an efficient way to summarise results and communicate them to a wider audience. Results indicate that the CO2 fertilisation effect counteracts the decline in yields with higher temperatures, and that a future switch to later maturing cultivars would lower the likelihood of a shortfall and produce higher yields.