Convective precursors and predictability in the tropical western Pacific

Conditions leading to convective outbreak in the Tropics are investigated by multivariate analysis of sounding and satellite data from the tropical western Pacific area. Circumstances that make the prediction problem difficult are discussed and addressed by applying linear "error-in-variables" and nonlinear statistical simulation techniques to a large dataset. Low- to midtropospheric moisture is identified as the dominant factor regulating convective outbreak in this region. Based on the results it is argued that such moisture is particularly important in regulating spontaneous convective outbreak, but instability and near-surface wind speed probably play some role in allowing previous shallow or midtopped cumulus activity to deepen. Mesoscale-mean convective available potential energy sufficient for convection is found to exist almost 90% of the time. Quantitative estimates of noise in the data are obtained and accounted for in reaching these conclusions. The results imply that large-scale mean fields alone may not contain enough information to determine the behavior of convection except probabilistically. Both types of statistical model predict that even under favorable mesoscale-mean conditions, convection is typically only 20%-30% likely to break out during a given 3-h period.