Calibration of ACCESS-S forecasts on a daily timescale — Australian Meteorological and Oceanographic Society

Calibration of ACCESS-S forecasts on a daily timescale (#277)

Morwenna Griffiths 1 , Li Shi 1 , Debbe Hudson 1 , Oscar Alves 1
  1. Australian Bureau of Meteorology, Docklands, VIC, Australia

The Bureau of Meteorology, with support from the Federal Government under the Agricultural Competiveness White Paper and from the Managing Climate Variability Program, has commenced a four year project to develop a significantly improved seasonal climate forecast service for Australia. A key component of this project is a new forecast system, ACCESS-S.  ACCESS-S has enhanced features compared with POAMA, the most obvious being an increase in the spatial resolution from 250 km to 60 km in the Australian region which contributes to an improved depiction of the mean climate.  

Despite improvements to seasonal prediction systems, model errors and uncertainties remain. Post-processing seasonal forecasts based on their historical error characteristics, i.e. calibration, is common practice and can enhance the utility for use in application models such as for streamflow and crop yield.  Calibration is based on the model forecast characteristics obtained from a large set of hindcasts.  The calibration is a function of start-date, lead time and location.   We have implemented a quantile-quantile matching approach to produce daily calibrated maximum and minimum temperature, rainfall, vapour pressure, wind speed, solar radiation and evaporation forecasts over Australia.   We will show that we improve the distribution whilst not degrading the error, spread or forecast skill.  The calibration is done on each day of the forecast independently.  We investigate the effect that the calibration has on the monthly and seasonal time-scales, as well as other aspects of the forecasts such as the correlation between variables

Provision of this calibrated data will support a range of existing and new applications, including crop, pasture-growth and streamflow modelling, as well as a variety of forecast products (such as probability of exceedance products and heatwave forecasts).

#AMOS2019