Abstract

In this study, the Florida State University Coupled Global Spectral Model (FSUCGSM) is utilized to examine the possible regulation of the warm pool SST and its contributors. The model is run for 1 yr to obtain the residue-free time evolution of the warm pool SST. The results are verified against NCEP SST analysis for the period of the model integration. The best agreement was seen over the western equatorial Pacific.

The initial analysis of the model output has suggested that the warm pool SST is derived mainly by three important types of oscillations, namely, semiannual, 10–25-, and 30–60-day oscillations. Further examination using Butterworth bandpass filter and EOF analysis has revealed that the tendency of solar radiation is the primary cause of the high-frequency oscillations (20–25 and 30–60 day) and secondary cause for the low-frequency oscillations. Moreover, the evaporative cooling is found to be the primary cause of the low-frequency oscillations and secondary cause for the high-frequency oscillations. The variations of these two forcings were found to be strongly related to convective activities. At high frequencies, convective activities are associated with equatorial waves, whereas at low frequencies such conditions are derived by the migration of the ITCZ.

In relation to the atmospheric moisture content, it was found that the cloud shortwave forcing plays the most crucial role in the solar radiation. The connection between convective activities and the changes in the evaporative cooling is found to be through the humidity deficit at low-frequency oscillations and surface wind speed at high-frequency oscillations. A careful examination of the SST–convection interaction has revealed that the warm pool SST may have an upper limit as suggested by earlier authors.

Abstract

The superensemble technique has been proven to be successful in producing a deterministic forecast superior not only to any of the individual models going into it, but also to the multimodel ensemble forecast. Research so far has been done on the superensemble as a deterministic forecast, and it has been shown that using the superensemble method leads to a significant reduction in rms errors. This paper investigates the skill of the superensemble as a probabilistic forecast, and it compares it with that of the multimodel ensemble. Using the Atmospheric Model Intercomparison Project (AMIP I) seasonal multimodel precipitation forecasts, probability forecasts are defined for the multimodel ensemble and for the multimodel superensemble. The Brier skill score of these forecasts is calculated for different thresholds of precipitation anomaly. It is shown that both the multimodel ensemble and the superensemble probability forecasts are much better than climatological forecast and that the superensemble probability forecast outperforms the multimodel bias-removed ensemble at any threshold level.

Abstract

The goal of this study is to utilize several recent developments on rainfall data collection, downscaling of available climate models, training and forecasts from such models within the framework of a multimodel superensemble, and first a detailed examination of the seasonal climatology. The unique aspect of this study is that it became possible to use the forecast results from as many as 16 state-of-the-art coupled climate models. A downscaling component, with respect to observed rainfall estimates, uses a very dense Asian rain gauge network. This feature enables the forecasts of each model to be bias corrected to a common 25-km resolution. The downscaling statistics for each model, at each grid location, are developed during a training phase of the model forecasts. This is done wherever the observed rainfall estimates are available. In the “forecast phase,” the forecasts from all of the member models use the downscaling coefficients of the “training phase.” The downscaling and the extraction of the superensemble weights are done during the training phase. This makes use of the cross-validation principle. This means that the season to be forecasted is left out of the entire forecast dataset. Thus all of the statistics for downscaling and the superensemble construction are done separately for the forecasts of each season for all the years. The forecast phase is the season that is being forecast, where the aforementioned statistics are deployed for constructing the final downscaled superensemble.

These forecasts are next used for the construction of a multimodel superensemble. The geographical distributions of the downscaling coefficients provide a first look at the systematic errors of the member model forecasts. This combination of multimodels, the vast rain gauge dataset, the downscaling, and the superensemble provides a major improvement for the rainfall climatology and anomalies for the forecast phase. One of the main results of this paper is on the improvement of rainfall climatology of the member models. The downscaled multimodel superensemble shows a correlation of nearly 1.0 with respect to the observed climatology. This high skill is important for addressing the rainfall anomaly forecasts, which are defined in terms of departures from the observed (rather than a model based) climatology. This first part of the paper provides a description of the member models, the length of the training and forecast phases, the sensitivity of results as the numbers of forecast models are increased, and the skills of the downscaled climatology forecasts.

Abstract

This is the second part of a paper on the improved seasonal precipitation forecasts for the Asian monsoon using 16 atmosphere–ocean coupled models. This study utilizes a large suite of coupled atmosphere–ocean models; this second part largely addresses the skill of rainfall anomaly forecasts. These include both deterministic and probabilistic skill measures such as the RMS errors, anomaly correlations, equitable threat scores, and the Brier skill score. It was possible to improve the skills of rainfall climatology from the use of a downscaled multimodel superensemble to very high levels, and it is of interest to ask how far this methodology would go toward improving the skills of seasonal rainfall anomaly forecasts. It is possible to go through a sequence of multimodel post processing to improve upon these skills by using a dense rain gauge network over Asia, downscaling forecasts for each member model, and constructing a multimodel superensemble that benefits from the persistence of errors of the member models. This paper addresses the spinup issues of the downscaling and the superensemble results where the number of years of model data needed for training phase, for the downscaling, and for the construction of the superensemble, is addressed. In the context of cross validation, the training phase includes 14 seasons of monsoon data. The forecast phase is only one season; it is this season that was not included in the training phase each time.

The relationship between data length and the number of models needed for enhanced skills is another issue that is addressed. Seasonal climate forecasts over the larger monsoon Asia domain and over the regional belts are evaluated. The superensemble forecasts invariably have the highest skill compared to the member models globally and regionally. This is largely due to the presence of large systematic errors in models that carry low seasonal prediction skills. Such models carry persistent signatures of systematic errors, and their errors are recognized by the multimodel superensemble. The probabilistic skills show that the superensemble-based forecasts carry a much higher reliability score compared to the member models. This implies that the superensemble-based forecasts are the most reliable among all the member models. It is possible to examine the performance of models and of the superensemble during periods of heavy monsoon rainfall versus those for deficient monsoon rainfall seasons. One of the conclusions of this study is that given the uncertainties in current modeling for seasonal rainfall forecasts, post processing of multimodel forecasts, using the superensemble methodology, seems to provide the most promising results for the rainfall anomaly forecasts. These results are confirmed by an additional skill metric where the RMS errors and the correlations of forecast skills are evaluated using a normalized precipitation anomaly for the forecasts and the observed estimates.

Abstract

The diurnal mode of the Asian summer monsoon during active and break periods is studied using four versions of the Florida State University (FSU) global spectral model (GSM). These versions differ in the formulation of cloud parameterization schemes in the model. Observational-based estimates show that there exists a divergent circulation at 200 hPa over the Asian monsoon region in the diurnal time scale that peaks at 1200 local solar time (LST) during break monsoon and at 1800 LST during active monsoon. A circulation in the opposite direction is seen near the surface. This circulation loop is completed by vertical ascending/descending motion over the monsoon domain and its surroundings. This study shows that global models have large phase and amplitude errors for the 200-hPa velocity potential and vertical pressure velocity over the monsoon region and its surroundings. Construction of a multimodel superensemble could reduce these errors substantially out to five days in advance. This was on account of assigning differential weights to the member models based on their past performance. This study also uses a unified cloud parameterization scheme that inherits the idea of a multimodel superensemble for combining member model forecasts. The advantage of this model is that it is an integrated part of the GSM and thus can improve the forecasts of other parameters as well through improved cloud cover. It was seen that this scheme had a larger impact on forecasting the diurnal cycle of cloud cover and precipitation of the Asian summer monsoon compared to circulation. The authors show that the diurnal circulation contributes to about 10% of the rate of change of total kinetic energy of the monsoon. Therefore, forecasting this pronounced diurnal mode has important implications for the energetics of the Asian summer monsoon.

Abstract

The superensemble technique has previously been demonstrated to provide an improved seasonal forecast compared to the bias-removed ensemble of equally weighted models. This paper offers a further improvement to the superensemble method by modifying the regression coefficients used in the weighting of the models for the construction of the superensemble. The improvement is achieved by use of singular value decomposition of the covariance matrix, and selecting only the largest singular value, corresponding to maximal explained variance, for the calculation of the regression coefficients. The results shown here are based on calculations done with 10 yr worth of monthly forecasts from the Atmospheric Model Intercomparison Project (AMIP) dataset, using cross validation.

Abstract

In this paper we have examined the evolution of a number of parameters we believe were important for our understanding of the drought over India during the summer of 1987. The list of parameters includes monthly means or anomalies of the following fields: sea surface temperatures, divergent circulations, outgoing longwave radiation, streamfunction of the lower and upper troposphere, and monthly precipitation (expressed as a percentage departure from a long-term mean). The El Niño related warm sea surface temperature anomaly and a weaker warm sea surface temperature anomaly over the equatorial Indian Ocean provide sustained convection, as reflected by the negative values of the outgoing longwave radiation. With the seasonal heating, a pronounced planetary-scale divergent circulation evolved with a center along the western Pacific Ocean. The monsoonal divergent circulation merged with that related to the El Niño, maintaining most of the heavy rainfall activity between the equatorial Pacific Ocean and east Asia. Persistent convective activity continued south of India during the entire monsoon season. Strong Hadley type overturnings with rising motions over these warm SST anomaly regions and descent roughly near 20° to 25°S was evident as early as April 1987. The subtropical high pressure areas near 20° to 25°S showed stronger than normal circulations. This was revealed by the presence of a counterclockwise streamfunction anomaly at 850 mb during April 1987. With the seasonal heating, this anomaly moved northwards and was located over the Arabian Sea and India. This countermonsoon circulation anomaly at the low levels was associated with a weaker than normal Somali jet and Arabian Sea circulation throughout this summer. The monsoon remained active along northeast India, Bangladesh, northern lndochina, and central China during the summer monsoon season. This was related to the eastward shift of the divergent circulation. An eastward shift of the upper tropospheric anticyclone bell near 25° to 30°N resulted in the continued presence of a westerly wind anomaly north of India. The westerly winds brought in very dry air over the tropical upper troposphere. The dry air penetrated eastwards to central Uttar Pradesh and this seemed to have a major role in inhibiting organized deep convection over most of central, northern and western parts of the Indian subcontinent. The westward extension of the planetary-scale divergent circulation over North and South Africa and the continued drought over the regions are also briefly addressed.

Abstract

Improved seasonal prediction of sea surface temperature (SST) anomalies over the global oceans is the theme of this paper. Using 13 state-of-the-art coupled global atmosphere–ocean models and 13 yr of seasonal forecasts, the performance of individual models, the ensemble mean, the bias-removed ensemble mean, and the Florida State University (FSU) superensemble are compared. A total of 23 400 seasonal forecasts based on 1-month lead times were available for this study. Evaluation metrics include both deterministic and probabilistic skill measures, such as verification of anomalies based on model and observed climatology, time series of specific climate indices, standard deterministic ensemble mean scores including anomaly correlations, root-mean-square (RMS) errors, and probabilistic skill measures such as equitable threat scores for seasonal SST forecasts. This study also illustrates the Niño-3.4 SST forecast skill for the equatorial Pacific Ocean and for the dipole index for the Indian Ocean. The relative skills of total SST fields and of the SST anomalies from the 13 coupled atmosphere–ocean models are presented. Comparisons of superensemble-based seasonal forecasts with recent studies on SST anomaly forecasts are also shown.

Overall it is found that the multimodel superensemble forecasts are characterized by considerable RMS error reductions and increased accuracy in the spatial distribution of SST. Superensemble SST skill also persists for El Niño and La Niña forecasts since the large comparative skill of the superensemble is retained across such years. Real-time forecasts of seasonal sea surface temperature anomalies appear to be possible.

Abstract

Recently the National Aeronautics and Space Administration (NASA) Tropical Rainfall Measuring Mission (TRMM) project office made available a new product called the convective–stratiform heating (CSH). These are the datasets for vertical profiles of diabatic heating rates (the apparent heat source). These observed estimates of heating are obtained from the TRMM satellite’s microwave radiances and the precipitation radar. The importance of such datasets for defining the vertical distribution of heating was largely the initiative of Dr. W.-K. Tao from NASA’s Goddard Laboratory. The need to examine how well some of the current cumulus parameterization schemes perform toward describing the amplitude and the three-dimensional distributions of heating is addressed in this paper. Three versions of the Florida State University (FSU) global atmospheric model are run that utilize different versions of cumulus parameterization schemes; namely, modified Kuo parameterization, simple Arakawa–Schubert parameterization, and Zhang–McFarlane parameterization. The Kuo-type scheme used here relies on moisture convergence and tends to overestimate the rainfall generally compared to the TRMM estimates. The other schemes used here show only a slight overestimate of rain rates compared to TRMM; those invoke mass fluxes that are less stringent in this regard in defining cloud volumes. The mass flux schemes do carry out a total moisture budget for a vertical column model and include all components of the moisture budget and are not limited to the horizontal convergence of moisture. The authors carry out a numerical experimentation that includes over a hundred experiments from each of these models; these experiments differ only in their use of the cumulus parameterization. The rest of the model physics, resolution, and initial states are kept the same for each set of 117 forecasts. The strategy for this experimentation follows the authors’ previous studies with the FSU multimodel superensemble. This includes a 100-day training and a 17-day forecast phase, both of which include a large number of forecast experiments. The training phase provides a useful statistical database for tagging the systematic errors of the respective models. The forecast phase is designed to minimize the collective bias errors of these member models. In these forecasts the authors also include the ensemble mean and the multimodel superensemble. In this paper the authors examine model errors in their representations of the heating (amplitude, vertical level of maximum, and the geographical distributions). The main message of this study is that some cumulus parameterization schemes overestimate the amplitude of heating, whereas others carry lower values. The models also exhibit large errors in the placement of the vertical level of maximum heating. Some significant errors were also found in the geographical distributions of heating. The ensemble mean largely mimics the model features and also carries some large errors. The superensemble is more selective in reducing the three-dimensional collective bias errors of the models and provides the best short range forecasts, through hour 60, for the heating. This study shows that it is possible to diagnose some of the modeling errors in the heating for individual member models and that information can be important for correcting such features.

Abstract

The Tropical Rainfall Measuring Mission (TRMM) satellite supplemented with the Defense Meteorological Satellites Program (DMSP) microwave dataset provides accurate rain-rate estimates. Furthermore, infrared radiances from the geostationary satellites provide the possibility for mapping the diurnal change of tropical rainfall. Modeling of the phase and amplitude of the tropical rainfall is the theme of this paper. The present study utilizes a suite of global multimodels that are identical in all respects except for their cumulus parameterization algorithms. Six different cumulus parameterizations are tested in this study. These include the Florida State University (FSU) Modified Kuo Scheme (KUO), Goddard Space Flight Center (GSFC) Relaxed Arakawa–Schubert Scheme (RAS1), Naval Research Laboratory–Navy Operational Global Atmospheric Prediction System (NRL–NOGAPS) Relaxed Arakawa–Schubert Scheme (RAS2), NCEP Simplified Arakawa–Schubert Scheme (SAS), NCAR Zhang–McFarlane Scheme (ZM), and NRL–NOGAPS Emanuel Scheme (ECS). The authors carried out nearly 600 experiments with these six versions of the T170 Florida State University global spectral model. These are 5-day NWP experiments where the diurnal change datasets were archived at 3-hourly intervals. This study includes the estimation of skills of the phase and amplitudes of the diurnal rain using these member models, their ensemble mean, a multimodel superensemble, and those from a single unified model. Test results are presented for the global tropics and for some specific regions where the member models show difficulty in predicting the diurnal change of rainfall. The main contribution is the considerable improvement of the modeling of diurnal rain by deploying a multimodel superensemble and by constructing a single unified model. The authors also present a comparison of these findings on the modeling of diurnal rain from another suite of multimodels that utilized different versions of cloud radiation algorithms (instead of different cumulus parameterization schemes) toward defining the suite of multimodels. The principal result is that the superensemble does provide a future forecast for the total daily rain and for the diurnal change of rain through day 5 that is superior to forecasts provided by the best model. The training of the superensemble with good observed estimates of rain, such as those from TRMM, is necessary for such forecasts.