Search

The main features of three atmospheric general circulation models forced with monthly-mean, observed sea surface temperature for the 1979-1995 period are analyzed and compared with observational data. The ensemble means and intra-ensemble standard deviations of several variables were investigated for the two seasons December-January-February and June-July-August. Correlations of the ensemble mean anomalies with observational data and with the indices Nino-3 and the Atlantic Dipole were also calculated. The probability distribution function of the precipitation at different regions was also examined, as well as the internal and external variance of the precipitation in the three models. No model has uniformly better characteristics than the others. On the contrary, each model has strengths and weaknesses that depend on the region and season.

The spatio-temporal variability of boreal summer monsoon onset over the Philippines is studied through the analysis of daily rainfall data across a network of 76 gauges for the period 1977 to 2004 and the pentad Merged Analysis of Precipitation from the US Climate Prediction Center from 1979 to 2006. The onset date is defined using a local agronomic definition, namely the first wet day of a 5-day period receiving at least 40 mm without any 15-day dry spell receiving less than 5 mm in the 30 days following the start of that period. The onset is found to occur rather abruptly across the western Philippines around mid-May on average and is associated with the set-up of a "classical" monsoonal circulation with low-level easterlies subsequently veering to southerly, and then southwesterly. The onset manifests itself merely as a seasonal increase of rainfall over the eastern Philippines, where rainfall occurs throughout most of the year. Interannual variability of the onset date is shown to consist of a spatially coherent large-scale component, rather similar over the western and eastern Philippines, with a moderate to high amount of local-scale (i.e. station scale) noise. In consequence, the large-scale signal can be easily retrieved from any sample of at least 5–6 stations across the network although the local-scale coherence and fingerprint of the large-scale signal of the onset date are found to be stronger over the central Philippines, roughly from Southern Luzon to Northern Mindanao. The seasonal predictability of local onset is analyzed through a cross-validated canonical correlation analysis using tropical Pacific and Indian Ocean sea surface temperature in March and the 850 hPa May wind field from dynamical forecast models as predictors. The regional-scale onset, defined as the average of standardized local-scale anomalies in onset date, shows good predictive skill (r ≈ 0.8). Moreover, most of the stations show weak to moderate skill (median skill = 0.28-0.43 depending on the scheme) with spatial averaging across stations typically increasing skill to > 0.6.