Observed and SST-forced seasonal rainfall variability across tropical America

Three experiments starting from different initial conditions have been made with the ECHAM-4 atmospheric General Circulation Model (GCM) integrated at T30 resolution forced with the observed sea-surface temperature (SST) over the period 1960-1994. The tropical America modes of seasonal rainfall anom alies whose time variation is most accurately simulated by the GCM have bee n searched for using Singular Value Decomposition Analyses (SVDA) and Canon ical Correlation Analysis (CCA) between observed and model fields. The lead ing modes revealed by SVDA and CCA are highly similar, even though the orde ring of the modes showed some fluctuation. A first skilful rainfall anomaly mode has weights of the same sign almost everywhere in tropical America. e xcept along the western coast and the sub-tropical margins. This mode appea rs in all of the four seasons assessed. A second major skilful mode is usua lly a bipolar north-south (N-S) rainfall anomaly pattern (clear in December -March, DJFM: March-May, MAM, and June-September. JJAS). A large portion of the skill of the first rainfall anomaly mode (same sign anomalies across tropical America except small patches along the western co ast) is through variance that is in common with the Southern Oscillation In dex (SOI). In addition to forcing from the central/eastern tropical Pacific SST. there also appears a contribution from contrasting SST anomalies in t he tropical Atlantic. This rainfall mode is usually a regional portion of a more large-scale mode encompassing at least the whole tropical zone (espec ially in DJFM, MAM and September-November. SON). Analysis of the relationsh ip of this mode with GCM circulation features reveals that a rainfall defic it (respectively excedent) over the main rainbelt of the tropical America r egion is associated with strengthening (respectively weakening) of the sub- tropical westerly jet streams, a global warming (respectively cooling) of t he tropical atmosphere. an anomalous divergence (respectively convergence) in the lower levels and an anomalous convergence (respectively divergence) in the upper levels over tropical America and in the region of the Atlantic Inter-tropical Convergence Zone (ITCZ). Such global features are not so ap parent for the dominant mode of JJAS. even though the correlations with El Nino-Southern Oscillation (ENSO) indicators (as SOI or NINO3 SST index) are as high as for the other seasons. The bipolar N-S rainfall anomaly mode in tropical America is mostly related to anomalous N-S gradient of SST anomalies in the tropical Atlantic. The a tmospheric circulation anomalies emphasize changes in 850 hPa meridional wi nds in the tropical Atlantic. However, there is also interannual variance o f this rainfall mode in both the model and observations that is unexplained by tropical Atlantic SSTs, but which is explained by central/eastern tropi cal Pacific SSTs and, potentially, SSTs from other tropical and extratropic al areas. This is especially true in MAM. Some differences in the details of the model and observed teleconnection pa tterns are noted. Such differences can be used to statistically adjust the model simulations using the CCA or SVDA modes as basis patterns. Both stati stical approaches have been applied and the results are consistent between the two, The increase of skill is stronger when temporal correlation (the p attern correlation) between the model and observed pattern is high (low) as for JJAS. The skill is moderate to high around the whole Amazon basin. but remains relatively low inside the Amazon basin, though reliability of the observations themselves may influence this result. Averaged over all the se asons, about 15-35% (35-55%) of the interannual grid-box (regional) seasona l rainfall variance is skilfully simulated from the observed SST forcing. C opyright (C) 2001 Royal Meteorological Society.