Matthew Harrison1, Richard Rawnsley1, David Phelan2, Ross Corkrey2 and David Henry3
2 School of Land and Food, University of Tasmania, Private Bag 78, Hobart Tasmania 7001
3 Agriculture Flagship, CSIRO, Sneydes Road, Werribee Victoria 3030
Knowledge of forthcoming seasonal weather conditions may help end-users make complex farm decisions, such as purchasing livestock feed or pasture fertiliser. We assessed differences between seasonal pasture growth rates collected at two sites over a two year period with 30-90 day growth rate forecasts in southeastern Australia. We contrasted forecasts generated using either climate data from historical records, subsets of historical records that aligned with the current phase of the Southern Oscillation Index (SOI), or from the global circulation model POAMA. Growth rate predictions in late winter/early spring (Aug-Sep) were least reliable overall, mainly because the long-term variation in forecast growth rates was low due to consistent cool, wet climes typical of north-western Tasmania. In contrast, our nonparametric statistical analyses indicated that early summer (Dec) forecasts from all methods and for all durations were most reliable, but this result was driven by larger long-term variation in this season. Across both sites and all months, the 30-90 d POAMA forecasts and the 60-90 d forecasts from the Historical and SOI methods were the most and least reliable, respectively. However, the better performing method varied considerably across months, indicating that multiple seasonal climate forecasts produced using different means should be used to forecast growth rates, depending on time of year of the forecast. At present, seasonal climate forecasts need to improve considerably to be of value in complex farm-decision making. As GCM forecasts advance in future, we expect that seasonal growth rates forecast using dynamical climate models will provide the lowest uncertainty, and may become more useful to growers and policy-makers than statistical methods developed using historical data.