Actuaries Climate Index and Multivariate Wheat Yield Modelling: A Comparison of VAR and LSTM

Abstract:
The increasing frequency of extreme climatic events in Australia has been alarming and its implication towards the agriculture sector can’t be underestimated. Wheat – one of the most essential export commodities for Australia – has been experiencing a major yield damage due to the occurrences of extreme weather. Overcoming this issue requires a profound understanding of climate risk and the ability to foresee the impact of the risk to the wheat yield, which has been the core problem of this study. This study focused on using the Australian Actuaries Climate Index (AACI), as a holistic measure of climate risk, to forecast the quantity of wheat yield in Australia. By using multivariate forecasting methods such as Vector Autoregression (VAR) and Long ShortTerm Memory (LSTM) models, a forecasting result has been generated and the result showed that LSTM was the best performing model according to the Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) values. The result has also proven the usability of AACI as the independent variable to forecast wheat yield.