Assessment of (sub-) seasonal prediction skill using a canonical event analysi

Abstract:

Hydrological extremes regularly occur in all regions of the world and as such are globally relevant phenomena with large impacts on society. Seasonal and sub-seasonal predictions could increase the preparedness to these extreme events. We investigated the skill of five seasonal forecast models from the NMME-II ensemble for the period 1982-2012 at a range of temporal and spatial scales. A canonical event analysis is used to enable a model validation beyond the ¨single¨ temporal and spatial scale. The model predictions are compared to two reference datasets on the seasonal and sub-seasonal scale. We evaluate their capability to reproduce observed daily precipitation and temperature. It is shown that the skill of the models is largely dependent on the temporal aggregation and the lead time. Longer temporal aggregation increases the forecast skill of both precipitation and temperature. Seasonal precipitation forecasts show no skill beyond lead time of 6 months, while seasonal temperature forecasts skill does extent beyond the 6 months. Overall the highest skill can be found over South-America and Australia, whereas the skill over Europe and North-America is relatively low for both variables. On the sub-seasonal scale (two week aggregation) we find a strong decrease in prediction skill after the first 2 weeks of initialization. However, the models retain skill up to 1-2 months for precipitation and 3-4 months for temperature. Their skill is highest in South-America, Asia and Oceania at the sub-seasonal level. The skill amongst models differs greatly for both the sub-seasonal and seasonal forecasts, indicating that a (weighted) multi-model ensemble is preferred over single model forecasts. This work shows that an analysis at multiple temporal and spatial scales can enhance our understanding of the added value of (sub-) seasonal forecast models and their applicability, which is important when these models are applied to forecasting of (hydrological) extremes.