A method for improving weather and climate forecast skill has been developed. It is called a superensemble, and it arose from a study of the statistical properties of a low-order spectral model. Multiple regression was used to determine coefficients from multimodel forecasts and observations. The coefficients were then used in the superensemble technique. The superensemble was shown to outperform all model forecasts for multiseasonal, medium-range weather and hurricane forecasts. In addition, the superensemble was shown to have higher skill than forecasts based solely on ensemble averaging.

In this paper the performance of a multimodel ensemble forecast analysis that shows superior forecast skills is illustrated and compared to all individual models used. The model comparisons include global weather, hurricane track and intensity forecasts, and seasonal climate simulations. The performance improvements are completely attributed to the collective information of all models used in the statistical algorithm.The proposed concept is first illustrated for a low-order spectral model from which the multimodels and a ''nature run'' were constructed. Two hundred time units are divided into a training period (70 time units) and a forecast period (130 time units). The multimodel forecasts and the observed fields (the nature run) during the training period are subjected to a simple linear multiple regression to derive the statistical weights for the member models. The multimodel forecasts, generated for the next 130 forecast units, outperform all the individual models. This procedure was deployed for the multimodel forecasts of global weather, multiseasonal climate simulations, and hurricane track and intensity forecasts. For each type an improvement of the multimodel analysis is demonstrated and compared to the performance of the individual models. Seasonal and multiseasonal simulations demonstrate a major success of this approach for the atmospheric general circulation models where the sea surface temperatures and the sea ice are prescribed. In many instances, a major improvement in skill over the best models is noted.

Abstract. The Pacific Exploratory Mission to the Tropics phase B (PEM-Tropics

[1] We examine the sources and chemistry affecting nitrogen oxides (NO x = NO + NO 2 ) over the tropical Pacific (30°S-20°N) using observations from the Pacific Exploratory Mission to the Tropics B (PEM-Tropics B) aircraft mission conducted in March-April 1999. A global model of tropospheric chemistry driven by assimilated meteorological data is used to interpret the observations. Median concentrations observed over the South Pacific during PEM-Tropics B were 7 pptv NO, 16 pptv peroxyacetyl nitrate (PAN), and 34 pptv nitric acid (HNO 3 ); the model generally reproduces these observations but overestimates those over the North Pacific. Lightning was the largest source of these species in the equatorial and South Pacific tropospheric column and in the tropical North Pacific upper troposphere. The oceanic source of acetone implied by high observations of acetone concentrations (mean 431 pptv) allows an improved simulation of PAN/NO x chemistry. However, the high acetaldehyde concentrations (mean 78 pptv) measured throughout the troposphere are inconsistent with our understanding of acetaldehyde and PAN chemistry. Simulated concentrations of HNO 3 and HNO 3 /NO x are highly sensitive to the model representation of deep convection and associated HNO 3 scavenging. Chemical losses of NO x during PEMTropics B exceed chemical sources by a factor of 2 in the South Pacific upper troposphere. The chemical imbalance, also apparent in the low observed HNO 3 /NO x ratio, is explained by NO x injection from lightning and by frequent convective overturning which depletes HNO 3 . The observed imbalance was less during the PEM-Tropics A campaign in September 1996, when aged biomass burning effluents over the South Pacific pushed the NO x budget toward chemical steady state.

In this paper model-generated data sets are examined to address the question of seasonal precipitation forecast skill of the Asian and the North American monsoon systems. In this context the seasonal climate forecast data from a set of coupled atmosphere-ocean models were used. The main question we ask is if there is any useful skill in predicting seasonal anomalies beyond those of climatology. The methodology for prediction is the 'FSU Superensemble', which is applied here to the anomalies of the predicted multimodel data sets and the observed (analysis) fields. The skills of seasonal forecasts are evaluated using two different types of parameters: anomaly correlations and root mean square errors. Comparison of skill of the coupled model forecasts and the AMIP hindcasts yields encouraging results. It is noted that the superensemble based anomaly forecasts have somewhat higher skill compared to the bias-removed ensemble mean of member models, individually bias removed ensemble mean of the member models and the climatology. This skill comes partly from the forecast performance of multimodels and partly from the training component built into this system that is based on past collective performance of these multimodels. These components are separated to assess the improvements of the superensemble. Though skill of the forecasts from the superensemble is found to be higher than that of the bias-removed ensemble mean and has shown some usefulness over the climatology, the issue of forecasting a season in advance in quantitative terms still remains a challenge and demands further advancement in climate modeling studies.