The potential value of seasonal forecasts of rainfall categories: Case studies from the wheatbelt in Western Australia's Mediterranean region

Hypothetical forecasts of categories of seasonal rainfall (May¿October) were evaluated in order to identify which requirements for seasonal climate forecasting are most useful in guiding decisions on both nitrogen (N) fertiliser rates, and whether or not to sow wheat, at different sites in the Mediterranean region of Western Australia. The analyses included the identification of the required number of forecast categories and their boundaries, and the quantification of effects of forecast skill (from zero to perfect) on the potential value of categorical forecasts. Simulated data from the APSIM-NWheat model were used to calculate gross margins for three soil types (clay, loam, sand) at four locations. The value of the hypothetical forecast information was calculated in terms of the additional gross margin achieved over strategies which ignore any forecasts (a ¿naïve forecast¿). In order for the conditional distribution (based on a forecast) of gross margins to differ from the all-years distribution (based on a naïve forecast) at p ¿ 0.1, a minimum of two categories (below/above median rain) were required at the low rainfall sites Merredin and Buntine, while at the medium rainfall sites Wongan Hills and Mingenew three categories (rainfall terciles) were required. This distribution shift was significant only on the clay soil, with exception to the loam at Merredin. With a conditional N management on the clay, the average gross margin increased by 31 AUD ha¿1 (+104%) at Merredin and 40 AUD ha¿1 (+40%) at Buntine when a perfect two-category forecast (below/above median) was assumed. With a perfect three-category forecast (terciles) the average increase was 62 AUD ha¿1 (+20%) at Wongan Hills and 60 AUD ha¿1 (+18%) at Mingenew. There was no additional benefit in adjusting N fertiliser rates based on forecasts of extreme categories (i.e. the 1st, 2nd, ¿ 8th, 9th rainfall decile) compared to forecasts of either above/below median or terciles. The minimum level of skill required for two- and three-category forecasts to be potentially valuable increased with seasonal rainfall, and was higher for a flexible than for a fixed sowing strategy. With fixed sowing on the clay soil, the threshold skill level was 13% at Merredin and 26% at Buntine in a two-category format; and 52% at Wongan Hills and 56% at Mingenew in a three-category format. At moderate levels of skill, the application of forecasts returned value at the low rainfall sites only. Given a three-category forecast with a hypothetical skill of 40%, the value was 15 AUD ha¿1 when used to modify N fertiliser rates on the clay at the dry site Merredin. The same forecast had no value at the wet site Mingenew