Error growth in Climate Forecast System daily retrospective forecasts of South Asian monsoon

Abstract

[1] The forecast skill and predictability of the Climate Forecast System, the operational model of the National Centers for Environmental Prediction, have been assessed on daily and intraseasonal time scales for the South Asian monsoon region. Using retrospective forecasts for the period 1981–2005, the growth of forecast errors and predictability errors in rainfall has been studied in detail for forecasts initiated in different months of the monsoon season. The growth rates of small predictability errors have been estimated by applying the Lorenz method. The doubling time of small errors is found to be in the range of 4–7 days for forecasts initiated in July, whereas it is in the range of 5–14 days for forecasts starting in May. The growth of errors associated with forecasts which start from different phases of the monsoon intraseasonal variability has also been examined. For the rainfall averaged over the land points of India, it is found that the errors grow faster in the case of forecasts initiated during active (above normal) or break (below normal) phase while they grow slower with forecasts starting from normal phase.