Calibration and bias correction of POAMA forecasts for use in Agricultural models.

[摘要] Seasonal climate forecasts from global circulation models have now reached a level of skill where it is becoming appropriate to incorporate them directly into agricultural models and tools. The output from such models arrives on a large spatial grid and can contain biases. Examples of these biases include temperatures being too warm, or rain being the correct amount but falling as drizzle over many days instead of distinct large rain events. Here we present and evaluate a method to downscale and calibrate the output from a seasonal climate model to locations where weather stations exist. The technique is designed to be effective but with low-computational cost so it can be delivered quickly to websites and Apps.A frequency correction is first applied to the rainfall in the model output, particularly at adjust for the number of dry days. Weather station records are then used to adjust the statistical properties of the model output through a quantile mapping approach.In eight case study sites the adjustment is effective with the MSE being close to the ensemble variance.Presently this downscaled and calibrated data is obtainable from http://www.agforecast.com.au/ and has been used in the development of yield forecasts in Yield Prophet (http://www.yieldprophet.com.au/YP) and Yield Prophet Lite (http://www.yieldprophet.com.au/yplite)

[发布日期] 2017-05-08 [发布机构] CSIRO

[学科分类] 地球科学(综合)