posted on 2022-11-10, 01:50authored byNatalia Bailey, Zvi Hochman, Yufeng Mao, Mervyn J. Silvapulle, Param Silvapulle
This paper introduces a statistical model to estimate and evaluate the predictability of the response of wheat yield to extreme temperature exposures and rainfall during the three phases of wheat grain production (vegetative, reproductive and grain filling) in northwestern (NW) Victoria, Australia. Unlike crop models which rely on functions developed from field experiments, we use observed data on annual wheat yields from 44 farms in the region over a period of 26 years (1993-2018). We find that the one-way fixed effects panel data model tends to outperform competing models in the out-of-sample prediction of future yields. We detect as positive drivers of NW Victorian wheat yield growth, exposure to moderate temperatures in all the three phases of the wheat production and total rainfall in the first two phases of the growing season. Providing adequate soil moisture, January-March rainfall also was found to be a positive driver of yields. Conversely, exposure to freezing temperatures during the vegetative and reproductive phases as well as to extreme high temperatures in all three phases of wheat production constitute negative drivers of NW Victorian wheat yields. The reproductive phase appears to be the most sensitive to climate variability, with adverse extreme heat and frost having sizeable negative impacts on yields. These negative effects are partially offset by increased rainfall in the same phase of wheat production. Moreover, we compare yield predictions by our statistical model to yield potentials calculated by APSIM. The gaps can be used to make recommendations on some adaptation opportunities available to farmers in the NW Victoria region.