Rethinking the Prediction of Intraseasonal Variability

Current statistical techniques used for the prediction of intraseasonal variability usually extrapolate principal components of the canonical Madden-Julian as a base. Yet, diagnostic studies show that the canonical phenomena accounts for less than half of the variance in the 20-60 day intraseasonal band. Convective maxima, often (although not ubiquitously) form in the eastern equatorial Indian Ocean. Some remain stationary, some move northeastwards and others propagate towards the east. Each has differing teleconnection patterns influencing the middle latitudes in different ways depending on the state of the higher latidtudes. They also exist over a wide periods within the 20-60 day period range. Furthermore, hindcasts with the ECMWF System 4 and the CFSV2 show that there is a great variability of predictability depending on the strength of the MJO signal. First we show that the wide variability of the MJO period observed in nature is broad, depending on stochastic ocean-atmosphere interactions. Second, we examine the teleconnection patterns related to the MJO as a function of MJO period, intensity and type. Third, we examine the predictability of the different spatial and temporal forms of intraseasonal variability using hindcasts of the two coupled climate models. Finally, we attempt to assess the biases in the coupled models for the wide range of intraseasonal variability and assess if there are a priori expectations of prediction that can be applied in real time.