Outstanding Issues in Seasonal to Interannual Climate Prediction IV
Presiding: M Ji, NOAA Office of Global Programs; P Chang, Texas A&M University
A44A-01 15:30h
Diagnosing Sources for the Contiguous US Seasonal Forecast Skill
The skill in U.S. seasonal forecasts is widely held to arise first and foremost from the atmosphere's sensitivity to anomalous sea surface temperatures. Much progress has been made in recent years in predicting the latter, especially at one-season leads. General circulation models used in modern prediction methodologies also exhibit superior fidelity relative to earlier generations. Notwithstanding these advances, a situation of modest U.S. seasonal forecast skill remains. To interpret the current state of U.S. skill, our study diagnoses the nature of the SST-related skill and its origins. The simulation skill of a multiple model, large ensemble suite of atmospheric GCM's using the monthly observed specified global SSTs is first presented. Skill is stratified by season, region, and for the variables surface temperature (T) and precipitation (P). The skill sources are diagnosed by constructing a multi-variant CCA whose co-variance matrix relates tropical SST "predictors' to the GCM simulated US T and P "predictands". This diagnostic model confirms that the bulk of skill is of tropical SST origin, and is furthermore linearly related to those SSTs. The CCA model is subjected to various truncations, from which it is discovered that the bulk of skill originates from a single ENSO pattern of SST predictor. A second, additional source of US skill is discovered. This source is largely non-ENSO. The predictor consists of SST variations in the subtropical west Pacific including the South China Sea, and explains much of the GCM's Fall seasonal temperature skill in northern US. A seasonal hindcast model is then constructed based on lagged CCA. Hindcast skill for 1950-99 reveals the majority of simulation skill of the same period is realized using the above-mentioned two primary tropical SST predictors.
A44A-02 15:45h
On the Interpretation of Skill Information for Seasonal Climate Predictions
Based on the analysis of atmospheric general circulation model (AGCM) simulations, certain caveats inherent in the estimates of seasonal prediction skill from the past performance of seasonal forecasts will be highlighted. These include: (1) the stability of estimated skill depends on the length of the time-series over which seasonal forecasts are verified, leading to possible scenarios where error bars on the estimated skill could be as large as skill itself, particularly in the extratropical latitudes. (2) A single estimate of skill obtained from the verification over a given forecast time-series, due to variation in signal-to-noise ratio from one year to another, may not be representative of seasonal prediction skill on a case-by-case basis. It is hoped that the analysis presented will lead to further dialogue on the interpretation, presentation, and utilization of skill information for seasonal prediction efforts.
A44A-03 16:00h
Storm track predictability on seasonal to decadal scales
This talk is concerned with estimating the predictable variation of extratropical daily weather statistics ("storm tracks") associated with global sea surface temperature (SST) changes on interannual to interdecadal scales, and its magnitude relative to the unpredictable noise. The SST-forced stormtrack signal in each northern winter in 1950-2004 is estimated as the mean stormtrack anomaly in an ensemble of atmospheric general circulation model (AGCM) integrations for that winter with prescribed observed SSTs. Since the stormtrack signals cannot be derived directly from the archived monthly AGCM output, they are diagnosed from the SST-forced winter-mean 200 mb height signals using an empirical linear stormtrack model (STM). For two particular winters, the El Nino of JFM 1987 and the La Nina of JFM 1989, the stormtrack signals and noise are estimated directly, and more accurately, from additional large ensembles of AGCM integrations. The linear STM is remarkably successful at capturing the AGCM's stormtrack signal in these two winters, and is thus also suitable for estimating the signal in other winters. Our principal conclusion is that a predictable SST-forced stormtrack signal exists in many winters, but its strength and pattern can change substantially from winter to winter. The pattern correlation of the SST-forced and observed stormtrack anomalies is high enough in the Pacific-North American sector to be of practical use. In the Euro-Atlantic region, we find much lower correlations, which we argue arise from substantial AGCM error in representing the regional response to tropical SST forcing, rather than intrinsically low stormtrack predictability.
A44A-04 16:15h
Measuring the potential utility of seasonal climate predictions
Variation of sea surface temperature (SST) on seasonal-to-interannual time-scales leads to changes in the distribution of seasonal climate anomalies. Relative entropy, an information theory measure of utility, is used to quantify the impact of SST variation on seasonal precipitation compared to natural variability. Ensemble simulations from two general circulation model (GCMs) are used to estimate relative entropy in three regions where tropical SST has a large impact on precipitation: South Florida, the Nordeste of Brazil and Kenya. The impact is statistically significant about half of the years. Yearly variation of relative entropy is strongly correlated with shifts in ensemble mean precipitation and weakly correlated with ensemble variance. Further analysis using relative entropy as a metric indicates only modest useful and detectable year-to-year variation of higher order distribution moments. Relative entropy is also found to be related to measures of the ability of the GCMs to reproduce observations.
A44A-05 16:30h
Extratropical Sensitivity to Tropical Sea Surface Temperatures
Despite numerous investigations of the extratropical response to prescribed observed SSTs in atmospheric general circulation models, a general understanding of the sensitivity of that response to SST anomalies in different parts of the tropical oceans is lacking. In this study, such a general sensitivity analysis has been conducted using the NCAR atmospheric general circulation model (CCM3.10). Ensemble-mean model responses were determined for an array of 43 regularly spaced localized SST anomaly patches over the tropical Indian, Pacific, and Atlantic oceans. A singular vector analysis of these 43 responses shows that the extratropical response is much more sensitive to SST forcing in the western half of the Indo-Pacific basin than in the eastern half, the region of largest observed SST variability. This larger sensitivity is only partly due to the larger thermodynamic sensitivity of the precipitation response to SST anomalies over the warm pool. The dynamic sensitivity of the extratropical response to forcing south of the Asian-Pacific jet is even more important, especially in winter. Remarkably, this sensitivity is of opposite sign for SST anomalies in the Indian and western Pacific oceans. Thus a warmer western Pacific tends to force the positive phase of the PNA pattern, but a warmer Indian ocean the negative phase. Our results strongly indicate the need to improve SST predictions in this sensitive western half of the Indo-Pacific basin to improve predictions of global climate variability from seasonal to centennial scales.
A44A-06 16:45h
The Roles of Surface Heat and Freshwater Fluxes in the Context of ENSO Prediction
An outstanding problem with all present ENSO forecast models is their lack of skill in predicting the SSTA in the western tropical Pacific, a region with a strong impact on extra-tropical climate variations. In part, this is due to the relatively weaker ENSO signal and thus the larger noise-to-signal ratio within the western Pacific warm pool. But more importantly, this is because of the dominant roles of surface heat and fresh water fluxes in determining the SSTA there, and the mistreatment of these fluxes in present ENSO models. The interannual variabilities of surface heat and freshwater fluxes are completely ignored in intermediate coupled models and are presently not well simulated in coupled GCMs. The moderate success of these models is mostly based on their ability to simulate large-scale ocean dynamics, which has a dominant influence on the SSTA in the central and eastern equatorial Pacific. In the west, however, the SSTA is decoupled from the deep thermocline and is mostly controlled by the surface layer thermodynamic processes. Here we present some results from a systematic study on the effects of surface heat and freshwater fluxes in the context of ENSO prediction. Through a combination of data analyses and model experiments, it is found that these fluxes play significant roles in the ocean-atmosphere interaction on interannual timescales; they are fundamentally important in controlling the upper-ocean thermohaline structure in the western tropical Pacific; and they are quantitatively not negligible even in the central and eastern tropical Pacific.