An analysis of prediction skill of monthly mean climate variability

In this paper, lead-time and spatial dependence in skill for prediction of monthly mean climate variability is analyzed. The analysis is based on a set of extensive hindcasts from the Climate Forecast System at the National Centers for Environmental Prediction. The skill characteristics of initialized predictions is also compared with the AMIP simulations forced with the observed sea surface temperature (SST) to quantify the role of initial versus boundary conditions in the prediction of monthly means. The analysis is for prediction of monthly mean SST, precipitation, and 200-hPa height. The results show a rapid decay in skill with lead time for the atmospheric variables in the extratropical latitudes. Further, after a lead-time of approximately 30-40 days, the skill of monthly mean prediction is essentially a boundary forced problem, with SST anomalies in the tropical central/eastern Pacific playing a dominant role. Because of the larger contribution from the atmospheric internal variability to monthly time-averages (compared to seasonal averages), skill for monthly mean prediction associated with boundary forcing is also lower. The analysis indicates that the prospects of skillful prediction of monthly means may remain a challenging problem, and may be limited by inherent limits in predictability.