WCRP Workshop on Seasonal Prediction

Barcelona Spain June 4-7, 2007

 

Abstracts

Talks

Day 1 - Monday 4th June

Key-Note Talks

Dynamical Seasonal Prediction: Current Status, Prospects and Challenges
Jagadish Shukla
IGES, George Mason Uni.

This paper will briefly review the current status of dynamical seasonal prediction. Results of hindcasts by research groups, and forecasts by operational centers will be summarized. The paper will present the following conjecture:

"The dominant obstacle in realizing the potential predictability of intraseasonal and seasonal variations is inaccurate coupled ocean - atmosphere models."

The paper will establish the validity of this conjecture by presenting evidence from a variety of model experiments. In particular the paper will show that the forecast skill is related to model fidelity, and that the model's ability to simulate tropical heating is particularly important.

Finally, the paper will suggest that the traditional strategy of parameterizing the (unresolved) deep convection should evolve towards developing the next generation of ultra high resolution models which can resolve cloud systems. A major challenge is to develop data assimilation systems for cloud resolving models. An unknown scientific question is whether the rapid growth of the unpredictable cloud systems will overwhelm the predictable large scale flow.

Improved methods of initializing coupled models, even at the current resolution, are expected to improve the skill in seasonal prediction.

Seasonal Forecasting and Seamless Climate Prediction
Tim Palmer
ECMWF
Paco Doblas Reyes (ECMWF); Antje Weisheimer (ECMWF); Mark Rodwell (ECMWF); Thomas Jung (ECMWF); Judith Berner (ECMWF)

Seasonal Climate Forecasting is of enormous value in its own right. However, as stressed in many of the IPCC assessment reports, seasonal forecasting can also be important as a means of assessing the reliability of climate change predictions. The reliability of the DEMETER multi-model seasonal forecasts will be reviewed from both of these perspectives. New results will be shown, illustrating the use of DEMETER-based reliability diagrams to calibrate the IPCC AR4 probabilistic climate-change projections. Finally, initial results from the more recent ENSEMBLES integrations will be shown. These will allow comparison of the effectiveness of different methods for representing model uncertainty in seasonal forecast mode: multi-model methods, perturbed parameter methods and stochastic physics. Our results argue for the development of seamless forecast tools, as stressed in WCRP's strategic framework.

ENSEMBLES: Seamless seasonal-to-decadal forecasting
Francisco Doblas Reyes
ECMWF

The EU-funded ENSEMBLES project uses the seamless framework to perform global ensemble dynamical predictions across different time scales, from a season up to a decade. The issue of model uncertainty is addressed using three different approaches: multi-model ensemble, stochastic physics and perturbed parameters, while the initial condition uncertainty is taken into account using the last generation of ocean analyses as well as ocean and atmosphere perturbations to create the ensembles. A thorough comparison of the merits of the different approaches to model uncertainty is being carried out and should lead to the development of an improved forecast system. Realistic natural and anthropogenic forcings have been used for the first time in this sort of exercise. The link with a range of applications, such as crop yield forecasting and health management, that use climate information as an extrinsic factor is undertaken through the development of a web-based public statistical downscaling tool and with a large set of multi-model dynamical downscaling experiments. The global model simulations are made publicly available from a series of servers at ECMWF to cater for the different needs of the users of climate information. The project also investigates projections of anthropogenic climate change and the evaluation of the processes most relevant for climate sensitivity, so the talk will mention the link between verifiable predictions of climate variability at the seasonal-to-decadal time scale and of climate change at longer time scales.

Why should we care about the stratosphere?
Mark Baldwin
NWRA

In this talk I will discuss how the circulation of the stratosphere affects the troposphere, on time scales relevant for seasonal forecasting. In many ways the stratosphere acts as a boundary condition for the troposphere. The stratospheric circulation can be highly variable, with a time scale much longer than that of the troposphere. The variability of the stratospheric circulation can be characterized mainly by the strength of the polar vortex, or equivalently the high-latitude westerly winds. During Northern winter and southern late spring, stratospheric variability peaks. When the flow just above the tropopause is anomalous, the tropospheric flow tends to be disturbed in the same manner, with the anomalous tropospheric flow lasting up to ~two months. The surface pressure signature looks very much like the North Atlantic Oscillation or Northern Annular Mode.

The stratospheric aspects of seasonal prediction can only be captured by models that properly simulate stratospheric variability. Thus far, the stratosphere's potential to improve seasonal forecasts is largely untapped. It is essential that seasonal forecast models simulate the intense, rapid shifts in the stratospheric circulation, as well as the downward propagation of circulation anomalies through the stratosphere. In addition, models must be able to simulate the poorly understood connections between the lower stratosphere and the tropospheric circulation.

Validation of Seasonal Forecasts: Statistical Methods and Downscaling
Jose Manuel Gutierrez
University of Cantabria

In this talk we present a simple statistical method to validate seasonal forecasts, comparing them with random predictions. This method provides an estimation of the statistical significance of the skill and, hence, allows us to find out predictable situations where the seasonal system significantly outperforms a random forecast. We also analyze the advantages of post-processing the predictions with some appropriate statistical downscaling method. The method is applied to precipitation and temperature forecasts considering two regions with different seasonal behavior: Peru (in the tropics) and Spain (mid-latitudes). Results show high predictability over Peru during El Nino periods. Here, the use of a downscaling method clearly improves the forecast skill. Over Spain the forecast signal is much weaker, but some predictability related to El Nino and La Nina events is found.

Finally, some sensitivity studies are presented. On the one hand, we compare raw station data with high resolution gridded interpolated data. On the other hand, different temporal aggregation patterns are used for the analog downscaling method (daily, weekly and monthly), comparing the obtained results.

Towards integrating applications within end-to-end seamless ensemble prediction systems - A case for Africa first?
Andy Morse
University of Liverpool

The integration of application models within seasonal ensemble prediction systems (EPS) in an end-to-end manner has been achieved within the DEMETER project and is continued to be refined in the ENSEMBLES project. At the same time there has been the growth in user integration within medium range EPS supported in programs such as WWRP THORPEX. Both communities have made strong connections with user, application and forecasting needs in Africa, as well as in Europe and worldwide, and have recently forged links with the AMMA project investigating the West African Monsoon and its impacts. A number of initiative to link EPS to Africa's needs have been initiated e.g. AMMA-THORPEX, THORPEX-Africa and more recently the start towards a more formal connection between ENSEMBLES and AMMA. Integrating application models and user methodologies with EPS has lead to a number of novel research questions addressing amongst other things biases, ensemble dressing, downscaling, required forecast skill, forecast verification, communication of forecasts and the need to overcome the deterministic inertial of many quasi-operational application users! What is clear is that only application users can determine the real world value of any forecasting system and have an integral role to play in the development and refinement of EPS in a true end-to-end approach. Linkage with decision makers and the socio-economic community is still underdeveloped and the lack of capacity to use EPS in many parts of the world, including Africa, points to the need for more training provision to use readily disseminated products. It is likely that the THORPEX mission to improve the forecasting of high impact weather will add to its current two week time horizon the newly developed 30 day EPS. It is paramount that the seasonal forecasting community meets halfway and that the two communities strive to produce seamless EPS with integrated user applications. The seamless prediction of the climate system and their practical application for the benefit of and value to society is a key mission of the WCRP. This paper will make the case that Africa is the prime candidate region to build the capacity to utilize the first operational seamless system with integrated applications, with a forecast window of days to years. Africa is a continent that suffers more than anywhere else from humanitarian crises that are directly connected to climate variability, and in the future climate change, and these impacts are often triggered by climatic variability that acts across a range of time scales rather than for a single forecast product window. Naturally the social, economic, nutritional status and health wellbeing of the society needs study and consideration and these are often linked to climate too. One key issue is to ensure that the forecast predictions are integrated into these programmes too. This paper will address how a future seamless EPS could be used to produce timely responses to future climate induced disasters and how this could aid the economic development of the continent.

Session I - Synthesis of Seasonal Prediction Skill

More Certain Uncertainty in Seasonal Climate Forecasts
Lisa Goddard
International Research Institute for Climate & Society/Columbia University
S. Mason (IRI/Columbia University)

The better quantified the uncertainty in climate forecasts, the more reliable the forecasts will be, and hence the more potential value the information has. Uncertainty in seasonal forecasts can arise both from model errors, which one wishes to minimize, and from the uncertainty inherent in the climate system, which one wishes to capture faithfully. Proper identification and treatment of errors, together with the use of multiple models, can greatly reduce the impact of model error on forecast uncertainty. Common types of errors in models used for seasonal climate prediction and current approaches to reducing their impact on forecast uncertainty are briefly discussed. The primary focus will be placed on a new approach being developed at the IRI to minimize error in the boundary-forced component of the climate variability (i.e. in the ensemble mean or model 'signal'). Our approach involves conditional exceedance probabilities (CEPs), which define the probability of the observed climate exceeding the forecast given the value of the forecast. The CEPs can be used to re-center an ensemble distribution, and thus redefine the probabilities of predefined events. This allows us to isolate better the contribution of biases in model signal (i.e. ensemble mean) from biases in model noise (i.e. ensemble spread), providing diagnostic information for model evaluation and recalibration information for forecasting. Minimizing model biases allows for much more reliable and flexible probabilistic forecasts, as desired by decision makers in climate risk management, from malaria early warning to agricultural insurance markets.

Predictions of temperature and rainfall anomalies with the ECMWF Seasonal Forecast System-3.
Franco Molteni
ECMWF, Reading, U.K.
T. Stockdale, L. Ferranti, M. Balmaseda, F. Doblas-Reyes (ECMWF)

In March 2007, ECMWF has implemented a new version of its seasonal forecast system, referred to as System-3. The system is based on ensemble integrations of a global coupled model, composed by a recent cycle of the IFS atmospheric model, the HOPE ocean model and the OASIS-2 coupler. A new ocean data assimilation system was designed to initialise the operational seasonal forecasts, as well as a set of hindcast ensembles started at the beginning of every month from January 1981 to December 2005. The set of forecast products include maps of ensemble mean anomalies and tercile probabilities, calibrated using the 25-year hindcast set, and time series of 2-metre temperature and rainfall anomalies averaged over 25 regions in the world. After a brief review of the skill of the system in predicting tropical SST anomalies, which are the main source of seasonal predictability, an overview of the forecast skill for 2m temperature and rainfall anomalies is presented for the northern winter and summer season, as derived from the hindcast dataset. Skill scores are compared with potential predictability estimates derived under perfect-model assumption. A more detailed analysis of results will be presented for some specific regions and seasons, both in the tropics and in the extratropics. It will be shown that, in some cases, simple spatial or temporal filtering methods can improve the forecast skill obtained from the direct model output. Cases of successful and unsuccesful predictions of major, near-continental scale anomalies will be highlighted.

Overview of the APEC Climate Center for Climate Information Services
Chung-Kyu Park
APEC

Recently, many parts of the world have experienced serious natural disasters associated with unusual climate. It is anticipated that the occurrence of extreme weather and climate events may become more frequent under the present global warming conditions. There is growing recognition that the international exchange of climate information is essential for minimizing natural disasters and their negative economic impacts. Scientists and policy makers are unanimous in concluding that concerted actions need to be taken to develop climate early warning systems and climate information networks at the regional scale to improve the monitoring and prediction of climate variations.

The APEC Climate Center (APCC) aims at realizing the APEC vision of regional prosperity through the enhancement of economic opportunities, the reduction of economic loss, and the protection of life and property by responding effectively to natural disasters and mitigating economic losses in the case of extreme climate events.

As a first step towards this goal, APCC aims to provide reliable and timely climate forecasts to a wide range of users in the APEC region through its operational multi-model ensemble-based system for climate prediction. This system comprises of 15 global climate models, each running from an ensemble of initial conditions. Based on these optimized forecasts are made 4 times a years in the form of deterministic and probabilistic multi-model ensembles. Planned enhancement to the climate prediction capability includes regional downscaling, increased frequency of forecasts, operational 12 month forecasts and extreme event forecasts.

APCC will enhance the socio-economic well-being of the APEC member economies by utilizing up-to-date scientific knowledge and applying innovative climate prediction techniques.

ENSO-related variability in Northeast Asia - An advantageous location for seasonal predictability
Tomoaki Ose
Meteorological Research Institute/Japan Meteorological Agency
Tamaki Yasuda (MRI); Yuhei Takaya (JMA); Shuhei Maeda (JMA); Chiaki Kobayashi (JMA); Hirotaka Kamahori(JMA)

Statistical analysis using the JRA-25 reanalysis data (Onogi,2007) indicate that Northeast Asia seems to be one of the advantageous locations in the extra-tropics for high seasonal predictability because ENSO-related variability over this region is relatively significant both for the northern summer and winter seasons as compared with other extra-tropical regions. This is attributed to the fact that Northeast Asia climate near the Pacific is directly influenced by the tropical western Pacific precipitation anomalies over the ENSO-created Indo-Pacific SST anomalies.  For the northern winter, the situation is relatively simple. Precipitation anomalies over the western Pacific occur closely related to local SST anomalies during ENSO events. These precipitation anomalies are successfully reproduced in an AGCM (mj98; Shibata et al.,1999) experiment. Note that there are the exceptional ENSOs showing almost no precipitation anomalies in the western Pacific (Ose,2000). These ENSOs tend to continue toward the next winter and have little impact on Northeast Asia climate. For the northern summer, local SST anomalies are not related to precipitation anomalies in the western Pacific. These precipitation anomalies are statistically related to Nino3.4 SST anomalies in the previous winter rather than in the summer. But, the AGCM succeeded again in reproducing the summer precipitation anomalies. The result indicates that the influences of the previous winter Nino3.4 SST anomalies are left in the summer SST anomalies over the Indo-Pacific domain (Ose, 2004). Questions rise about how robustly the above ENSO-related variability in precipitation and other fields is simulated with AGCMs and CGCMs. The operational JMA AGCM succeeds in predicting precipitation anomalies even in the summer western Pacific and give good scores for summertime three-month forecast over Japan (Kobayashi, 2005). The 20- and 21-century experiments by MRI-CGCM2 (Yukimoto,2001) show the above inter-annual relationship between model-ENSOs and the western Pacific precipitation for the northern winter but not for the northern summer. Further analysis on the differences from the real will be also reported.

Seasonal forecasting in Canada
George J Boer
CCCma, Environment Canada
Team SIP (EC)

The current Canadian operational seasonal forecasting system has grown out of a modest collaborative effort involving climate and weather models and researchers in government and the universities. The result is purely objective, 2-tier, multi-model, 1-season deterministic and probabilistic 3-category forecasts for Canada with accompanying skill measures. The forecast system is predicated on the Canadian CLIVAR HFP2, the second Historical Forecasting Project, consisting of a suite of 10-member retrospective forecasts for the 35-year period 1969-2003 produced with each of four different models. Skill assessment follows WMO recommendations. Studies are underway into several aspects of the predictability exhibited by the results and into the possibility of enhancing skill through various post-processing approaches including adding information from long-timescale atmospheric processes, fitting probability distributions, and Bayesian adjustment. Future activities include the Global Ocean-Atmosphere Analysis and Prediction (GOAAP) project involving coupled regional and global ocean and atmosphere analysis and prediction for weather and climate. Part of this effort is a Coupled Historical Forecasting Project (CHFP) analogous to HFP2 and congruent to the TFSP seasonal forecasting experiment.

IAP Dynamical Seasonal Prediction System and its applications
Zhaohui Lin
Institute of Atmospheric Physics, Chinese Academy of Sciences
Hong CHEN(Institute of Atmospheric Physics, Chinese Academy of Sciences); Guangqing ZHOU(Institute of Atmospheric Physics, Chinese Academy of Sciences); Zhengkun QIN(Institute of Atmospheric Physics, Chinese Academy of Sciences); Qingcun ZENG(Institute of Atmospheric Physics, Chinese Academy of Sciences)

The IAP dynamical seasonal-to-interannual prediction system (IAP DCP) will be briefly described, and its application to the real-time prediction of climate anomalies over China will also be summarized. Generally, IAP DCP is comprised of five components, i.e., IAP ENSO Prediction system, Prediction Integrations and Anomaly Coupling Technique, Ensemble Prediction Technique, Correction System, Prediction Products and Analyses. The skill of IAP DCP for the seasonal prediction of drought/flood situations over China has been assessed by several sets of hindcast experiments during the period 1980-2000. Hindcast experiments show that IAP DCP can well predict the large summer rainfall anomalies around China, especially over Yangtze River and Huaihe River Valley, and the predictability for the East Asian winter monsoon and the climate anomalies during springtime have also been investigated. The IAP DCP-II has been applied to the real-time prediction of summer rainfall anomalies over China, and the occurence freqency of spring dust storm acitivites over Northern part of China. During the real-time extraseasonal prediction, the ensemble technique have been used, with sea surface temperature anomalies (SSTA) taken from the IAP ENSO prediction system, and different atmospheric conditions from NCEP reanalysis, and the final prediction product is obtained by averaging the total ensemble after correction. Verification with the observed summer rainfall anomalies shows that, IAP DCP can quite well predict the large-scale patterns of summer flood and drought conditions over China. For example, the severe flood over Yangtze River Valley and Northeast China for year 1998, the positive summer rainfall anomalies over Southern part of China for year 1999, and the rainfall maximum over lower reach of Yangtze River valley for year 2001, all these are quite well predicted by IAP DCP. Meanwhile, the persisted drought conditions over North China from 1999 to 2002 have also been pretty well predicted by IAP DCP-II. The occurrence frequency of spring dust-storm activities have also been quite well predicted since 2003. Even the application of IAP DCP to seasonal prediction of summer precipitation in China is encouraging. However, considerable efforts are also needed, which include the improvement of climate system model, the improvement of oceanic and land data assimilation system, and exploitation of generalized ensemble technique, etc.

Recent Development in Ocean-Atmosphere Modelling for Seasonal Prediction in Australia
Guomin Wang
Bureau of Meteorology Research Centre, Australia
O. Alves (BMRC); H. Hendon (BMRC)

The Predictive Ocean Atmosphere Model for Australia (POAMA) is a state-of-the-art seasonal to interannual seasonal forecast system based on a coupled ocean/atmosphere model. The first version of POAMA (POAMA-1) has been running in operational mode for over four years. Recently a new version of the POAMA system (POAMA-1.5) has been developed and is now in transition to operations. This talk will a) describe new features in POAMA-1.5, b) give preliminary assessment on the system performance and forecast skill, c) present several application examples of seasonal predictions using the POAMA outputs, and d) present plans for future versions of the system.

Seasonal Prediction Activities at the South African Weather Service
Willem A. Landman
South African Weather Service
Asmerom Beraki; Mary-Jane Kgatuke; Maluta Mbedzi

Prior to 1994, seasonal rainfall and temperature outlooks issued by the South African Weather Service (SAWS) were primarily based on the state of the equatorial Pacific Ocean: El Nino meant dry and hot summer conditions, and La Nina a wet and cool summer season for South Africa. Since 1994 the SAWS started to issue seasonal forecasts based on statistical models. These forecasts incorporated the evolutionary and steady state features of global sea-surface temperature (SST) anomalies as predictors of seasonal rainfall and temperature. However, changes in association between South African seasonal rainfall and Indian Ocean SSTs were observed over the most recent decades. These changes in rainfall-SST associations can be successfully simulated using general circulation models (GCMs). Moreover, statistical post-processing of GCM output produces skill levels that outscore raw forecasts from GCMs and also model forecasts based on SST-rainfall relationships. The GCMs run operationally by the SAWS are the ECHAM4.5 and C-CAM. These models produce an ensemble of forecasts and are forced with predicted and persisted SST anomalies. In addition to the output data from these GCMs, model forecast fields from local universities (Pretoria and Cape Town) and international centres (International Research Institute for Climate and Society and the United Kingdom Meteorological Office) are combined into a multi-model seasonal forecast system. Forecast skill levels of such a multi-model system are further enhanced through statistical post-processing and combining them with statistical forecasts based on rainfall-SST relationships. Recently, the SAWS issued its first ever regional climate model (ECHAM4.5-RegCM3) forecasts for the mid-summer months of December 2006 to February 2007. This paper will discuss these seasonal forecast systems of the SAWS in some detail, and present forecast verification statistics for South Africa.

Day 2 - Tuesday 5th June

Coupled Model Performance - Chair M. Davey

Impact of ocean initial conditions on seasonal forecasts skill
Magdalena Alonso Balmaseda
ECMWF

It is widely accepted that skilful forecasts of ENSO at seasonal time scales rely on the knowledge of the ocean subsurface. However, there is open debate about the relative merits of the different ocean initialization strategies, usually in the context of initialization shock versus 'realistic' initial conditions. In this work we present the the impact of 3 different initialization strategies on the ECMWF System 3 seasonal forecasting system. In addition, we present the impact of the different ocean observing systems on the seasonal forecast skill of SST is quantified.

Coupled Model Data Assimilation and ENSO Forecasting
Anthony Rosati
GFDL

What are the limiting factors governing ENSO forecast skill in prediction and predictability in coupled models? There are several possible sources of error. One major source of error is the uncertainty in initial conditions. This issue will be examined by comparing the SST ENSO forecasts, in the GFDL CM2.1 coupled model, from two different ocean data assimilation systems and initialization techniques. In addition the role of stochastic noise and systematic bias will also be discussed. The role of stochastic noise on the system will be estimated by a modification in the atmospheric convective parameterization that leads to a damped ENSO.

A fully coupled model Ensemble Kalman Filter (EKF) data assimilation system has been developed in the GFDL CM2.1 framework. First results from perfect model forecasts will be shown. The impacts of the multivariate EKF on retrospective forecasts for the 1980-2005 period will be compared to the forecasts from the univariate three dimensional variational (3Dvar) system, along with perfect model potential predictability results.

Understanding El Nino in Ocean-Atmosphere General Circulation Models
Eric Guilyardi
LOCEAN/IPSL & Walker Institute

El Nino events represent major disruptions of the annual cycle of the tropical Pacific Ocean-atmosphere system on interannual timescales, with severe societal impacts on the whole planet. Predicting the characteristics of El Nino occurrence, amplitude and remote impacts for the next decades and centuries, is still a scientific challenge. State-of-the-art Ocean-Atmosphere General Circulation Models (OAGCMs) are advanced tools to both analyze El Nino mechanisms and predict its behavior on a range of timescales. Recent improvements show that El Nino is now an emergent model of variability in complex models. However, the diversity of their simulations of El Nino contributes to large uncertainty in projections. The recent multi-model approach, derived from the IPCC AR4, allows, to an unprecedented scale, latest generation OAGCMs to be analyzed together and compared. This requires new community efforts to develop appropriate metrics to assess the performance of individual models in reproducing El Nino events.

Session II - Seasonal Prediction Skill and the Coupled System

Stratospheric Processes and Seasonal Prediction (SPARC) - Chair: M. Baldwin

The Role of the Stratosphere in Seasonal Prediction: Background and Session Overview
Andrew Charlton
Dept. Meteorology, University of Reading

This talk will examine the current literature on the use of stratospheric information in seasonal forecasting. The talk will describe the reasons for the current interest in this topic and the small number of studies which have demonstrated an impact of stratospheric information on predictive skill. The talk will conclude by posing the main current research questions in this area which will be addressed by the other speakers in the session.

Downward propagation from the stratosphere: Physical mechanism and potential for seasonal prediction
Bo Christiansen
Danish Meteorological Institute

In the cold seasons the intraannual variability in the stratosphere is dominated by downward propagation. Zonal mean zonal wind anomalies are born in the mesosphere and propagate down through the stratosphere and into the troposphere on a time scale of weeks to months. We explain the mechanism as a consequence of nonlinear interactions between the zonal mean and large scale waves and show that even simple models can represent the basic features. We show the potential for using the downward propagation of anomalies from the stratosphere to the troposphere in extended range statistical forecasts. Considering the near surface zonal mean zonal wind at 60 N as predictand we find that the inclusion of stratospheric information improves the daily forecast on lead times larger than 5 days. The best forecasts are obtained for predictors in the lower stratosphere. Similar predictions can not be obtained if the statistical forecast only includes tropospheric information. The simple statistical forecast based on stratospheric winds compares favorably to the forecasts of a state-of-the-art dynamical ensemble prediction system.

Occurrence patterns of stratospheric sudden warming events in view of the stratosphere-troposphere coupled system and their predictability
Toshihiko Hirooka
Department of Earth and Planetary Sciences, Kyushu University
T. Ichimaru (Kyushu Univ.); H. Mukougawa (Kyoto Univ.)

Stratospheric sudden warmings are a typical example of stratosphere-troposphere coupled system in which planetary wave driving gives rise to its rapid time evolution. In some sudden warmings, only planetary waves with zonal wavenumber 1 essentially contribute to their occurrence, while wavenumber 2 and/or 3 components play an important role in the development of other sudden warmings, which are often accompanied with split polar vortices. Such a difference in the course of time evolution would make a difference in predictable periods of the sudden warmings. In this study, we examine five warming events occurring in recent five Northern Hemisphere winters using the Japan Meteorological Agency (JMA) ensemble one-month forecast data. Each predictable period is carefully estimated on the basis of the prediction of zonal mean temperatures in the polar stratosphere. It is found that predictable periods crucially depend on the time evolution of the warmings; the lead time for the prediction of the wavenumber-1 warmings is relatively long, say, 2-3 weeks in advance, compared with that of the warmings contributed to by wavenumber-2 and/or 3, say, 7-10 days. The short predictability of the latter might be connected to the difficulty in the prediction of wavenumber-2 and 3 evolution compared with that of wavenumber-1 evolution. Even though the time change of zonal mean temperatures is successfully predicted, that of zonal-mean zonal winds is often difficult to predict; this may be due partly to the failure in precise prediction of the wavenumber-2 and 3 evolution. Furthermore plausible factors giving rise to different time evolution of each warming event will be discussed.

Predictability of Stratosphere-Troposphere Coupling during Stratospheric Sudden Warming Events in the Northern Hemisphere
Hitoshi Mukougawa
Disaster Prevention Research Institute, Kyoto University
Y. Kuroda(MRI); T. Hirooka(Kyushu Univ.)

Recently downward influence of the stratospheric circulation on the troposphere has been a central issue in the stratosphere-troposphere dynamical interaction inspired by the work of Baldwin and Dunkerton (1999). In the current study, the predictability of stratosphere-troposphere dynamical coupling during stratospheric sudden warming (SSW) events is examined using an atmospheric general circulation model (GCM) and operational 1-month ensemble forecast data set provided by Japan Meteorological Agency (JMA). First, we discuss the predictability of a wavenumber-1 SSW event in December 2001 by conducting a series of hindcast experiments using MRI/JMA-GCM. The SSW is predictable from at least 2 weeks in advance, and high sensitivity to the initial condition for the SSW prediction is observed during the onset period of the SSW. Characteristic zonal wind anomalies in the troposphere are detected as the precursory event for the occurrence of the SSW through a regression analysis on the GCM experiment. The anomalies are closely associated with a persistent blocking over the Atlantic sector during this period. Furthermore, it is revealed from a series of GCM hindcast experiments that the response of stratospheric circulation to the magnitude of the precursory anomaly is nonlinear, which suggests the existence of a threshold magnitude of the precursory anomaly for the occurrence of the SSW. Detailed investigation on the precursory event will enable us to reveal the dynamical relationship between tropospheric circulation anomaly and the subsequent SSW. Second, we are going to discuss the relationship between the variation of the prediction skill of tropospheric Northern Hemisphere Annular Mode (NAM) and the lower stratospheric NAM index using JMA 1-month ensemble forecasts from 2001/2002 to 2005/2006 winter seasons. Our preliminary results suggest that the forecast skill of the tropospheric NAM index tends to improve when the lower stratospheric circulation is characterized by persistent negative NAM anomalies.

Stratosphere-troposphere interaction and plans for the GloSea4 Seasonal Forecast system
Adam Scaife
Hadley Centre, UK Met Office
Alberto Arribas and Sarah Ineson, UK Met Office

The influence of stratospheric variability on surface climate in winter is estimated using modelling experiments and observational datasets. Stratospheric changes appear to be important for the very rapid warming of European winters between the 1960s and 1990s and associated changes in the frequency of climate extremes. The stratosphere is also shown to play a key role in transmitting ENSO signals to Europe in winter. The winter of 2005/6 is used as a case study to illustrate how stratospheric influence occurs in an individual year. Changes to the research and development plans for the GloSea4 seasonal forecast system in the light of these results will also be described.

Stratosphere-troposphere interaction and the GloSea4 Seasonal Forecast system
Mike Keil
UK Met Office

A commonly held view is that the stratosphere is primarily important on seasonal and climate timescales, but not so important at shorter time ranges. However, recent upgrades to the Met Office operational global NWP model have shown a strong relationship between an improved stratospheric analysis and more accurate tropospheric weather forecasts on the 0-5 days timescales.

This talk discusses the main reasons behind this result, which came as a surprise to many, and considers its implications in the context of a seamless prediction system over multiple timescales. In addition, other aspects from improved assimilation of data which potentially impact on seasonal timescales, such as stratospheric ozone assimilation, will be highlighted.

Prospects for wintertime European seasonal prediction
Warwick Norton
The Walker Institute; PCE Investors

Statistical analysis shows that for non-ENSO years Atlantic SSTs can provide a useful predictor of wintertime anomalies over Europe (an example being the winter of 2005/6). However dynamical seasonal forecast models have so far failed to reproduce this predictive capability. A new high resolution coupled model HiGEM is presented and examined to see if it produces the connection between the Atlantic SSTs and wintertime anomalies over Europe. It is also examined to see if the NAO has added persistence in the 10-25 day range which potentially could arise from stratospheric anomalies.

Land-Atmosphere Interactions and Seasonal Prediction (GEWEX) - Chair: R. Koster

GLACE-2: A Study of the Impacts of Land Initialization on Forecast Skill
Randal Koster
Global Modeling and Assimilation Office, NASA/GSFC

Much of today's research into land-atmosphere interaction has the underlying, eventual goal of determining the extent to which land surface state initialization can improve a forecast. The first GLACE (Global Land Atmosphere Coupling Experiment) project a few years back addressed one critical element of this goal, namely, quantifying land-atmosphere coupling across a broad range of AGCM systems. While important, the first GLACE project did not address the full question of initialization impacts. This suggests the need for GLACE-2, an international model intercomparison experiment that will provide a multi-model, comprehensive view of the impact of land state initialization on forecast skill. Participants in GLACE-2 will learn, probably for the first time ever, the quantitative benefits that can stem from the incorporation of realistic land surface state initialization into their forecast algorithms.

Land surface contribution to seasonal climate variability and predictability
Hervé Douville
Météo-France/CNRM

Besides sea surface temperatures, the land surface hydrology is also likely to contribute to atmospheric predictability at the monthly to seasonal timescale. Such a contribution has been suggested by case studies and idealized experiments, but is difficult to assess in a more comprehensive and/or realistic framework given the lack of global land surface climatologies. Recently, the second phase of the Global Soil Wetness Project (GSWP-2) has provided a unique opportunity to produce 10-year global soil moisture and snow mass climatologies by driving land surface models with a combination of meteorological analyses and observations. Such a climatology has been produced at CNRM and has been used to relax the Arpege-Climat atmospheric GCM towards a realistic land surface hydrology. Ensembles of AMIP-type simulations have been designed to assess the influence of soil moisture / snow mass boundary and/or initial conditions on climate variability over the GSWP-2 period (1986-1995). Though too short to draw quantitative conclusions about the land surface contribution to seasonal climate predictability, these experiments suggest its relevance in the summer hemisphere mid-latitudes, while the sea surface temperature signal is generally dominant in the tropics and in the winter extratropics. A particular attention has been paid to the West African monsoon, where the soil moisture influence highlighted by former studies could be a stochastic artefact rather than a genuine land surface memory effect.

Seasonal Hydrological Prediction: Activities and progress under GEWEX's Hydrologic Application Project
Eric F Wood
Princeton University

The goal of the Global Energy and Water Cycle Experiment (GEWEX) is to reproduce and predict, by means of observations and suitable models, the variations of the global hydrological regime.  As part of this goal, GEWEX has a research element that focuses on 'demonstrating skill in predicting variability in water resources and soil moisture on seasonal to annual as an element of WCRP's prediction goals for the climate system'. As part of this element, GEWEX has developed the Hydrologic Application Project (HAP) with the emerging goals being:

1. Developing procedures for assessing current hydrologic conditions (nowcasting) through application of GEWEX supported data products, including remotely sensing;

2. Developing and testing of reliable, skillful hydrologic ensemble forecast procedures based on seasonal climate model forecasts;

3. Demonstrating that the procedures can be applied at scales useful for water resources through test-bed sites and demonstration projects;

Thus HAP will help foster and develop the science behind skillful ensemble hydrologic seasonal forecasts, and demonstrating their usefulness.

This presentation will present the current seasonal hydrologic prediction activities that include: (i) the strategies being developed and tested for bias correcting and downscaling ensemble seasonal climate model predictions, including multi-model ensemble predictions; (ii) multi-model reforecast (hindcast) results for the U.S., Africa and China; (iii) planned activities for the GEWEX HAP testbeds and the relationship to Hydrological Ensemble Prediction Experiment (HEPEX testbeds and activities.

Cryospheric Processes and Seasonal Prediction (CLIC) - Chair: J. Christensen

Seasonal Forecast of Antarctic Sea Ice
X. Yuan
LDEO, Columbia University

Long-range forecasts of Antarctic sea ice are very much in demand, not only because of the potential importance of sea ice in global climate, but also for the practical purpose of exploring the Antarctic continent. Unfortunately, such forecasts are not yet feasible with any state-of-the-art general circulation models, because the complex air-sea-ice interaction processes on long timescales are still not well understood and are by no means well simulated by these models. An alternative is to apply statistical methods to Antarctic sea ice prediction. The linear Markov model used in this study represents one of the first attempts in this direction.

Seasonal Predictability over the Arctic Region & exploring the role of boundary conditions (SPAR)
Rasmus Benestad
The Norwegian Meteorological Institute
Yvan Orsolini (Norwegian Institute for Air Research); Ina T Kindem (Bjerknes Centre for Climate Research); Arne Melsom (The Norwegian Meteorological Institute)

Thus far, seasonal prediction efforts have not made an optimal use of the information that is inherent in the initial states and memories of the cryosphere and the ocean. The impact of stratospheric events also needs to be considered in this context. The objective of the proposed project is to exploit how the low frequency forcing agents sea-ice, SST, snow cover, stratospheric conditions and oceanic heat anomalies affect the predictability of weather pattern statistics on seasonal time scales. Our focus will be to examine this predictability for the Arctic, but our results will also be used to illuminate predictability associated with Arctic conditions for northern Europe. The working hypothesis is that several factors act simultaneously and that the response is non-linear over a combination of conditions. In the SPAR project (http://spar.met.no), the sensitivity to the above-mentioned boundary conditions will be explored both individually and in combination. The proposed work will provide complementing results that together with the EU projects DEMETER/ENSEMBLES advance the science of seasonal predictability within Europe.

Verification of hemispheric-wide winter temperature forecasts based on fall snow and atmospheric anomalies
Judah Cohen
AER, Inc., Lexington, MA, USA
Christopher Fletcher Atmospheric Physics Group, Department of Physics, University of Toronto

One outstanding issue in seasonal climate prediction is whether there is any robust predictability beyond ENSO dynamics. We have operationally produced real-time winter forecasts for the US based on fall Eurasian snow cover and atmospheric anomalies for the past eight years. Operational forecasts have been expanded to include Europe for the past four years and East Asia for the past three. In addition, hindcasts have been produced for the winters 1972/73-2004/05. We assess the skill of these forecasts, up through the most recent winter season. These snow-based forecasts appear to provide considerable additional information beyond the standard-ENSO based forecasts and even the most sophisticated dynamical models.

Day 3 - Wednesday 6th June

Session III - Seasonal Prediction: Applications and Regional Skill

Methods and Applications - Chair: J. M. Gutierrez

Verification of Seasonal to Secular Climate Forecasts
David Stephenson
University of Exeter, UK

Verification is a key aspect in developing good scientific forecast systems. Without reliable procedures for forecast evaluation, we are unable to judge whether a forecasting system is better than just chance or whether a new forecasting system is really an improvement compared to previous schemes.

Climate forecasts pose particular problems for forecast verification. Small sample sizes and non-stationarity due to long-term trends can make verification scores extremely unreliable and uninformative and put serious limits on how well we can assess such types of forecast.

This talk will discuss various issues that affect the verification of climate forecasts and will present simple statistical models for understanding verification of climate forecasts on lead-times from seasons to centuries.

On downscaling methodologies for seasonal forecast applications
V. Moron
CEREGE UMR 6635, University of Aix-Marseille and IRI
A. W. Robertson and J. Qian, International Research Institute for Climate as Society, Palisades, NY

Emerging strategies for climate risk management lend a "demand-driven" approach to the seasonal forecasts produced with GCMs. Calibration and downscaling of such forecasts are central to their potential usability in sectoral societal settings that are often local or regional. The most effective strategy may often be to interface GCM output with sectoral models, such as crop or reservoir management models. In this talk, we will outline some of the downscaling methodological issues (both statistical and dynamical) that arise in the context applications with examples from on-going work at IRI.

Ensembles of crop yield at seasonal and multi-decadal timescales
Andrew Challinor
Institute for Atmospheric Science, School of Earth and Environment, University of Leeds

The increasing availability of climate model ensembles is enabling probabilistic simulations of crop growth and yield. Crop simulations can themselves include perturbed-parameter ensembles, which allow quantification of uncertainty in both crop and climate. To do this, an understanding of crop-climate processes is needed. Processes, methods and results relevant to seasonal and multi-decadal prediction are reviewed, and future directions identified.

New simulations of crop growth under climate change are presented. These highlight the interaction between water stress and CO2 concentration. From a physiological perspective, water-stressed crops are expected to show greater CO2 stimulation than well-watered crops. However, this result is not seen consistently in observations or in the crop models. An analysis of the evidence from the models and from the data suggests that scale (canopy versus leaf) and model complexity are factors in determining the sign and magnitude of the interaction between CO2 and water stress. This interaction has clear implications for the productivity of irrigated and rainfed systems.

Exploring potential sources of seasonal predictability for climate-driven diseases
Xavier Rodó
University of Barcelona

Seasonality is a major driving force of the spatiotemporal variation of certain relevant infectious diseases. For those diseases with a clear link to climate, regional seasonal forecasting appears as a promising tool for the near future. However, several problems may prevent its wide range of applications for prediction of disease incidence, if those are not properly taken into account. Among the main ones, there is the current scarcity of long and accurate time series of epidemics, as there are only a few number of exceptional datasets for cholera, malaria and a couple other relevant vector-borne diseases, and after this it is only slim-pickings. Therefore, there is an urgent need for recovering records yet hidden in old books in libraries. The immediate consequence is the current inability to derive useful and testable disease models for some of the major threats to human population, both coming from long-known endemic diseases as well as from new emerging ones which distribution may be altered by ongoing climate change. That is, there is a lack of knowledge of disease dynamics (and not only those linked to climate), and in some dramatic cases, ther is not even an appropriate knowledge of those factors modulating seasonal changes.

Other problems are derived from the typology of disease data (e.g. long disease records usually come from hospital facilities that centralize disease surveillance and control in a large area, therefore, climate drivers relevant for the disease dynamics should be integrated among similar spatial and temporal scales the disease data covers, which appears far from trivial). Difficulties and relevance of approaching timings of epidemics in climate-linked diseases will be discussed (as the phase of an epidemics in an outbreak proves to be a useful parameter carrying important epidemiological information). Similarly, the importance of matching amplitude changes is crucial, as the role of climate on postepidemic dynamics is determined by the susceptibility of the population after the last outbreak, and thus knowledge of previous disease evolution and the state of the population are extremely relevant to assert whether or not and when the regional climate will be a useful parameter for prediction.

Issues related to coupling disease prediction with climate predictions, that is adapting climate outputs to disease needs should also be accurately taken into account. Topics such as the control in error propagation when downscaling climate simulations from global models to relevant disease scales appear important, as well as the ability of seasonal forecast models to properly simulate climate extremes (namely if tails of distribution sseverely differ between rainfall simulations and observations).

State-of-the-art in disease models as well as a summary of the global distribution of some major climate-driven diseases will be shown, to stir discussion on why and when seasonal forecasting can be of use.

Importance of knowledge of some of the relevant climate-local weather relationships for diseases will be highlighted as well as the utility of the recent approaches on the windows-of-opportunity concept and the use of pacemaker models

I will finally present insights gained on the nonlinear dynamics of endemic and emerging diseases through the analysis of the classical forced SIR (susceptible, infectious, recovered) epidemic model. Endemic vs epidemic behavior and the role of climate in modulating those two will be briefly introduced.

ENSEMBLES Web Portal for Seasonal Statistical Downscaling. Description and Demo
Antonio S. Cofiño
University of Cantabria
Daniel San-Martìn (University of Cantabria); Jose M. Gutiérrez (University of Cantabria)

The demand for high-resolution seasonal predictions is continuously increasing due to the multiple end-user applications in a variety of sectors (hydrology, agronomy, energy, etc.) which require regional meteorological inputs. To fill the gap between the coarse-resolution lattices used by global weather models and the regional needs of applications, a number of statistical downscaling techniques have been proposed. Statistical downscaling is a complex multi-disciplinary problem which requires a cascade of different scientific tools to access and process different sources of data, from GCM outputs (see 1 and 2) to local observations (3 and 4). One of the ENSEMBLES project's aims is maximizing the exploitation of the results by linking the outputs of the ensemble prediction system to a range of applications. In order to accomplish this task in an interactive and user-friendly form, we have developed a Web portal which integrates the necessary prediction and validation tools. In this form, users can obtain their downscaled data testing and validating different statistical methods (from the categories 'weather typing', 'regression' or 'weather generators') in a transparent form, not worrying about the technical details of the downscaling methods and data formats and access. One of the new features of the portal is the automatic selection of predictors and the validation of the resulting downscaling methods using re-analysis data (5 and 6). In this talk we will describe the features of the portal and will illustrate its practical application for end-users using an online demo. Datasets (1) DEMETER dataset: http://www.ecmwf.int/research/demeter/ (2) ENSEMBLES s2d integrations for stream 1 dataset: http://www.ecmwf.int/research/EU_projects/ENSEMBLES/ (3) MARS-STAT Data Base: http://mars.jrc.it/marsstat/datadistribution/ (4) European Climate Assessment & Dataset project: http://eca.knmi.nl/ (5) ECMWF Re-Analysis ERA-40: http://www.ecmwf.int/research/era/ (6) NCEP/NCAR Reanalysis Project: http://www.cpc.ncep.noaa.gov/products/wesley/reanalysis.html

WMO CLIPS PROJECT : A GLOBAL FRAMEWORK FOR SEASONAL FORECAST APPLICATIONS
R. Kolli, B. Nyenzi and L. Malone
World Climate Programme Department
World Meteorological Organization

Over the past few decades, several important developments have taken place, leading to a characteristic change in the traditional perception of climate services. These include:

Major advances in the three main areas of monitoring, prediction and communications, leading to enhanced climate applications and services, as well as the dissemination of climate information;

Increased awareness that, in addition to mean weather conditions, monitoring of the current state of climate to determine the evolution of climate anomalies and prediction of the near future climate were essential parameters for the decision-making process; and

Recognition that regional/global collaboration was crucial for the development and sharing of knowledge and the prediction of global and regional climate.

The Climate Information and Prediction Services (CLIPS) project was established in 1995 by the Twelfth World Meteorological Congress as an implementing project of the World Climate Applications and Services Programme (WCASP), a constituent of the World Climate Programme (WCP). The principal objective of CLIPS is to develop the capacity of the National Meteorological and Hydrological Services (NMHSs) to take advantage of the rapid advances in the science of climate and in the processing and dissemination of climate information, and to pass along the benefits of the improved climate services to the user community. CLIPS provides an essential link between climate prediction/information and their applications - bridging the gap between the science and the applications to promote development activities in a manner beneficial to both producers and users of climate information and prediction products. CLIPS pioneered the concept of regional climate outlook forums (RCOFs), which are actively pursuing consensus-based climate outlook preparation and end-user liaison in different parts of the world notably in Africa, South America and Asia, with support from WMO and other partners. CLIPS has also been active in capacity building, with a number of training workshops held in different parts of the world.

Employing climate information in leveraging the decision-making process at different levels in the society, both in exploiting the opportunities and managing the risks, is increasingly being recognized as a crucial factor in sustainable development. Concepts like climate-related risk management are rapidly developing, in which the NMHSs have a critical role to play. The Espoo Statement of 2006 as well as the Madrid Action Plan of 2007, resulting out of two key WMO Conferences focusing on climate applications, called for an approach driven by the needs and requirements of the users, for which effective partnerships with the decision sectors and development of local capacities are crucial. CLIPS, being an integral component of WCASP pursuing applications and partnerships, and with a global network of Focal Points and sustained RCOFs, provides an excellent global framework to promote seasonal forecasting applications. In this context, it is important to make concerted and sustained efforts to bridge the gap between research and operational activities in seasonal forecasting as well as between climate providers and users.

African Climate System and Seasonal Prediction (VACS) - Chair: A. Morse

Taking the Shorter Route: Climate Prediction and Africa
Richard Washington
Oxford University Centre for the Environment

Seasonal prediction in Africa takes on a significance which is not seen in many regions of the world owing to the extent of the reliance of subsistence farmers on rain fed agriculture. Recently a range of institutions both within and beyond the continent have embraced SIP as the most pragmatic way to prepare for climate change - a step which further endorses the importance of these shorter timescales of prediction for Africa. In this paper we investigate the background for this recent step and identify the major constraints to skilful prediction. In part these constraints link to systematic errors in numerical models and some leading errors are discussed. The implications of the longevity and duplication of these errors in models are discussed in terms of the best strategy for improving SIP on the continent.

Development and Applications of Climate Diagnostic and Prediction Tools in Africa
Leonard Njogu Njau
African Centre of Meteorological Application for Development

The climate varies naturally on all time-scales. The variations may occur due to external forces such as changes in the sun's energy output. They may also be generated by interactions among the different components of the global climate systems: the atmosphere, ocean, biosphere and cryosphere. Natural variability of weather and climate on time-scales of days, months, and years, can produce extreme events such as droughts, floods, severe storms, heat waves and frosts among other disasters. Accurate climate prediction and early warning depends on monitoring, collection and analysis of meteorological data. The monitoring which involves assessment of the state of climate in recent past and present, is an important component in the provision of climate information for decision making as it provides useful insight regarding the current state and evolution of future climate. However, due to complexity of the climate generating systems and the different climate regimes in the Africa, various indicators as well as climate parameters have been developed to monitor and assess the current state to predict the future climate. For example the provision of accurate seasonal climate prediction and early warning requires timely availability of reliable diagnostic indicators with sufficient lead time as precursors of impending events. However, the recognition of which elements of the climate system that significantly influence future seasonal climate has grown considerably through the last century as diagnostic, modelling and emperical studies continue to add our knowledge and understanding on the climate predictability. The current diagnostic tools for seasonal climate prediction and early warning include determining the linkages between the seasonal climate and evolution of monsoons, El Nino/Southern Oscillation (ENSO), medium and upper level tropospheric variables (winds, temperature and geopotentials), Madden-Julian Oscillation (MJO), North Atlantic Oscillation (NAO), Quasi biennial oscillation (QBO), tropical cyclones activity, Ocean basins sea surface temperatures (SSTs) and associated SSTs gradients and Dioples, interactions between large and local scales circulation systems among others.

The development of climate diagnostic and prediction tools for seasonal forecasts and the factoring of these climate information into decision making processes will improve the climate related socio-economic activities and enhance development in various sub-regions and countries in Africa. However, the applications of climate information as management tool is being practised at various levels and with varying effectiveness across various parts of the African region. This paper presents a success story on development and application of climate prediction tools (CPTs) based on the recent research findings by Njau (2006).

Progress in the Seasonal Rainfall Prediction for the Greater Horn of Africa Region
Philip Omondi
Intergovernmental Authority on Development - Climate Prediction and Applications Centre
L. A. Ogall

The climate of the Greater Horn of Africa (GHA) may be classified as Arid and semi arid (ASALs) with a very high degree of rainfall variability both in space and time. Subsistence rain-fed agriculture is the mainstay of most economies of the GHA countries. Climate related climate events such as floods, droughts, etc are very common in Greater Horn of Africa (GHA) with devastating effects on all major sustainable development sectors and often retard national economic growth. The IGAD Climate Prediction and Applications Centre (ICPAC) is a specialized Institution of the Inter Governmental Authority on Development (IGAD) charged with the responsibility of climate monitoring, prediction and applications in the Greater Horn of Africa (GHA) region. GHA comprises the following 10 countries namely: Burundi, Djibouti, Eritrea, Ethiopia, Kenya, Rwanda, Somalia, Sudan, Tanzania and Uganda. Nine of the ten countries are classified within the Least Developed Countries (LDCs) that are characterized by extreme poverty as well as inadequate human and institutional capacity.

Some of the key products and services being provided by ICPAC and the NMHSs include the following among others: Updating and expanding regional data bank for risk mapping and various applications; Generation and dissemination of ten day, monthly and seasonal climate products and services; Dissemination of regional climate products / services and potential associated impacts ; Development of new climate prediction products and tools; Outreach, Awareness, Education, and Interactions with users of climate products and services;  Enhance applications of climate products and services in all climate sensitive sectors for disaster risk reduction and sustainable development; Addressing regional climate change challenges including climate change science (monitoring, detection, and attributions, regional climate modeling and scenarios development, etc), impacts/vulnerability and adaptation studies; Research related to regional climate and specific sector applications; and Enhancing capacity building needs of climate scientists and users

This paper presents recent progress, experiences and lessons that have been learnt by ICPAC regarding various aspects of seasonal climate prediction over the Greater Horn of Africa region using statistical and dynamical methods. The issues addressed in the paper include improvement of regional model skills; verification of the forecast skills by both climate scientists and users; applications of the seasonal climate outlooks in disaster risk management, together with challenges associated with the use of the probability based seasonal climate prediction products; capacity building needs of climate scientists and users; assessment of the benefits of the generated products, among many others.

Seasonal climate prediction activities at the Drought Monitoring Centre
Bradwell Garanganga
Southern African Development Community Drought Monitoring Centre

The Drought Monitoring Centre (DMC) was established in 1990 as part of the initiative of African governments and the cooperating partners in order to combat perennial calamities arising out of the recurrent extremes of climate variations. Its principal goal is, therefore, to contribute to the reduction of negative impacts of adverse weather and climate conditions such as drought, floods and other extreme events on sustainable socio- economic development, and to the rational use, conservation and protection of national resources in the Southern African Development Community (SADC). The purpose of the DMC, therefore, is to ensure the operation of a SADC mechanism for monitoring and predicting in seasonal climate conditions, taking advantage of advances in climate science and technology, which underpins extended range predictions that now span multi-timescales. The DMC uses seasonal climate prediction approaches that are possible within the limits of its resources.

The DMC takes the basic approaches to climate prediction, which by no means mutually exclusive, such as: statistical, analogical and dynamical modelling techniques. At the end of these methods lies the human who interprets the results from the various schemes taking into account experiences accumulated by the science community. An understanding of the correlation of statistics is essential in deriving statistical models for climate prediction. Invariably, multi-variate regression models are constructed. Principal Component Analysis (PCA) technique is used at DMC to identify homogenous rainfall zones in seasonal climate forecasting. Model Output Statistics method from the Climate Prediction Tool is also used. Analogue methods, which allow practitioners to select as many as possible past events that were behaving in a similar fashion, are also used.

One of the activities of the DMC is organizing Southern Africa Regional Climate Outlook Forum (SARCOF) where new climate prediction techniques are reviewed and applied to generate seasonal forecasts for the user-community. The different techniques used to generate climate forecast requires that some consensus be developed. This leads to seasonal forecasts that are presented in a probabilistic manner. Blending the outputs from different models achieves the consensus that SARCOF demands. Verification of these probabilistic seasonal climate forecasts is routinely carried out to assess their skill, reliability and usability. Resource constraints are one of some of the limiting factors in the case of routinely applying dynamical models.

American Monsoon System and Seasonal Prediction (VAMOS) - Chain: I. Cavalcanti

An overview of Seasonal Forecast-related issues over the Pan-VAMOS domain
Celeste Saulo
CIMA/FCEN (CONICET-UBA)
Jose Marengo (CPTEC/INPE); Hugo Berbery (U. of Maryland); Wayne Higgins (CPC/NOAA)

The CLIVAR/VAMOS panel has recently celebrated its 10th Anniversary and concomitantly an overview of the main accomplishments has been done. Particularly, the VAMOS Modeling component has been recognized as a valuable effort that the panel seeks to promote provided its unique capability to integrate the different science programs inside VAMOS. Modeling within VAMOS has been continuously evolving, and is characterized by the maintenance of a multi-scale approach in which local processes are embedded in, and are fully coupled with, larger-scale dynamics. Most of the activities have taken advantage of the enhanced observations provided by field experiments, including the availability of enhanced re-analyses. On the other hand, particular model intercomparison activities have been carried out and a super-model ensemble over South America is experimentally operational since 2005. VAMOS community has been moving forward from the model assessment at different time/spatial scales to the model development and the hypothesis testing with a strong emphasis in the improvement of the warm season precipitation representation as well as the understanding of the main mechanisms underlying its variability. This work will present some of the results that highlight what has been done and is actually being done in relation with Seasonal Prediction over the Pan-VAMOS region. Special importance will be given to the discussion of the topics that require more effort and also of the issues that the VAMOS community is more interested on, so as to provide guidance for future activities.

Do seasonal forecasts reproduce the link between early and peak monsoon rainfall in South America?
Alice Grimm
Department of Physics - Federal University of Parana (UFPR)
Marcia T. Zilli (PGERHA-UFPR); Iracema F. Cavalcanti (CPTEC-INPE)

Previous observational studies have disclosed a link between peak summer monsoon rainfall in central-east Brazil, comprising part of the South American monsoon core region, and antecedent conditions in spring. Rainfall in this region during part of spring shows significant inverse correlation with rainfall in peak summer, especially during ENSO years. The corresponding precipitation anomalies appear in the first modes of spring and summer variability. A surface-atmosphere feedback hypothesis involving soil moisture in spring has been proposed to explain this relationship, and a crucial role of the mountains in Southeast Brazil is suggested by modeling experiments. There has not been any assessment of the climate models' ability in reproducing that relationship between early and peak summer monsoon rainfall in South America. This study analyzes the austral spring/summer seasonal forecasts with focus on the interannual variability and on the relationship between the spring conditions and the summer forecast during extreme events of that inverse relationship. Outputs of CPTEC/COLA AGCM seasonal simulations of the SMIP2 project are used in this study. This spectral atmospheric model was integrated with T62L28 resolution for the SMIP2 period (1979 to 2001), applying as boundary conditions the observed SST. The model is run at each year for the four seasons considering simulations of 6 months. In this study, the simulations of SONDJF will be considered to analyze the relationship between spring and summer.

SUMMER RAINFALL FORECAST SKILL OVER SOUTH AMERICA DERIVED FROM WCRP SMIP-2 MODELS RESULTS
Lincoln Muniz Alves
Center for Weather Forecasting and Climate Studies (Centro de Previsão de Tempo e Estudos Climáticos (CPTEC)
Paulo Nobre (CPTEC); Ana Claudia Prestes (CPTEC); José Fernando Pesquero (CPTEC)

This work analyses the forecast skill of summer (DJF) rainfall for several SMIP-2 AGCMs over South America, relative to CPTEC's both atmospheric and coupled ocean-atmosphere GCMs. The main climatological features of South America rainfall are evaluated, as well as the model's ability to predict seasonal rainfall interannual variability over the region. Particular attention is paid to the South Atlantic Convergence Zone, which modulates summer rainfall over most of the continent, and which AGCMs have difficulty both simulating and predicting. The results suggest that part of the seasonal rainfall predictability over the region emerges from coupled ocean-atmosphere interactions. Results of multi-model ensemble techniques used to reduce systematic errors are also reported.

Performance of Seasonal Forecast along South America during 2006-2007 period: progress, challenges and the impacts on Society
Rodney. Martínez Güingla
Centro International para la Investigacion del Fenomeno de El Nino

From the middle of 2006, outputs of relevant Seasonal Forecast models, from ECMWF, NOAA-CFS, TCC, BCC, METEOFRANCE, UKMO, CPTEC, IRI and CIIFEN were analyzed in order to get the regional skill and identify critical areas.

The analysis shows good results during the last half of 2006, when the last ENSO event started its evolution, however, several difficulties were evidenced between December 2006 and March 2007, in most of Western South America region. The unusual evolution of the mature and final ENSO phase, especially over the Eastern Pacific was followed by a continuous discrepancy with most of the seasonal forecast.

Some comparison between statistical and dynamical models in the region suggest a stronger than normal influence of the physical processes over the Atlantic in South America including the west coast. In the same way a non typical upper atmosphere circulation pattern was persistent between the Caribbean and the northeast of South America, and block most of the time the convection due to the ENSO warming in NINO 1+2 region. For a new consecutive year, SST over NINO 1+2, shows a different evolution of the Tropical Pacific, however, some improvements in the prediction of the SST were evidenced after the abrupt switch  from warm to cold temperatures in January 2007, in spite of this switch could not be solved by most of the forecast in advance.

The documented experience of this year suggests the importance of the processes next to the Eastern Pacific coast, and the influence of the Atlantic and the Caribbean circulation over South America. The statistical model used in the region with data from more than 126 stations along the western Coast of South America allowed to identify the relevant role of the atmosphere during the last ENSO event and remark the need  of data assimilation in our region to improve the model skill.

The application of the Seasonal Forecast, was considerably positive in South America region, in spite of the limitation of models. The available information allowed to make opportune decisions that in some cases were followed by adequate prevention actions, however, extreme events and existent vulnerability caused several damages and social and economical impact in several countries.

It is concluded that, combining the global seasonal prediction power with the regional statistical and dynamical model could be improved if the regional data could be available for assimilation by global models. CIIFEN is working in close cooperation with National Institutions of the region to work on this direction and strength the regional capacities of statistical and dynamical Seasonal forecast.

Asian-Australian Monsoon System and Seasonal Prediction (AAMP) - Chair: E. Jin

How accurately do coupled climate models predict the Asian-Australian Monsoon interannual variability?
Bin Wang
University of Hawaii/ IPRC, USA
June-Yi Lee, University of Hawaii, USA; In-Sik Kang, Seoul National University, Korea; J. Shukla, George Mason University, USA; C.-K. Park, APCC, Korea

Accurate prediction of the Asian-Australian monsoon (A-AM) seasonal variation is one of the most important and challenging tasks in climate prediction. In order to understand the causes of the low accuracy in the current prediction of the A-AM precipitation, this study strives to determine to what extent the 10 state-of-the-art coupled atmosphere-ocean-land climate models and their multi-model ensemble (MME) can capture the two observed major modes of A-AM rainfall variability which account for 43% of the total interannual variances during the retrospective prediction period of 1981-2001. The first mode is associated with the turnabout of warming to cooling in the El Nino-Southern Oscillation (ENSO), whereas the second mode leads the warming/cooling by about one year, signaling precursory conditions for ENSO. We show that the MME one-month lead prediction of the seasonal precipitation anomalies captures the first two leading modes of variability with high fidelity in terms of seasonally evolving spatial patterns and year-to-year temporal variations, as well as their relationships with ENSO. The MME shows a potential to capture the precursors of ENSO in the A-AM domain about five seasons prior to the maturation of a strong El Nino. However, the MME underestimates the total variances of the two modes and the biennial tendency of the first mode. The models have difficulties in capturing precipitation over the maritime continent and the Walker-type teleconnection in the decaying phase of ENSO, which contributes in part to a monsoon 'spring prediction barrier'. The NCEP/CFS model hindcast results show that, as the lead time increases, the fractional variance of the first mode increases, suggesting that the long-lead predictability of A-AM rainfall comes primarily from ENSO predictability. The correlation skill for the first principal component remains about 0.9 up to six months before it drops rapidly, but for the spatial pattern it exhibits a drop across the boreal spring. This study uncovered two surprising findings. First, the coupled models' MME predictions capture the first two leading modes of variability better than those captured by the ERA-40 and NCEP-2 reanalysis datasets, suggesting that treating the atmosphere as a slave may be inherently unable to simulate summer monsoon rainfall variations in the heavily precipitating regions (Wang et al. 2004). It is recommended that future reanalysis should be carried out with coupled atmosphere and ocean models. Second, the NCEP/CFS ensemble hindcast overperforms the MME in terms of the biennial tendency and the amplitude of the anomalies, suggesting that the improved skill of MME prediction is at the expense of overestimating the fractional variance of the leading mode. Other outstanding issues are also discussed.

ENSO and IOD predictions in the SINTEX-F coupled GCM
Jing-Jia Luo
FRCGC/JAMSTEC, Japan
S. Masson (IPSL); S. Behera (FRCGC); T. Yamagata (FRCGC)

The prediction of ENSO (and its related climate impacts) sufficiently prior to its onset is vital for effective reduction of climate disasters. Retrospective forecasts for the period 1982-2004 show a high predictability of ENSO. All past El Nino and La Nina events, including the strongest 1997/98 warm episode, are predicted successfully with anomaly skill scores above 0.7 at 12-month lead time. Spatial sea surface temperature (SST) anomalies, teleconnections, and global drought/flood associated with ENSO at its different evolution stages are realistically predicted at 9-12 months lead (Luo et al. J. Clim, 2005). Our extended retrospective forecasts show that several ENSO events over the past two decades can be predicted even at lead times of up to two years. The El Nino condition in 1997/98 winter can be predicted to some extent up to about 1.5-year lead but with a weak intensity and large phase delay in the prediction of the onset of this exceptionally strong event. This is attributed to the influence of active and intensive stochastic westerly wind bursts during late 1996 to mid-1997 which are generally unpredictable at seasonal timescales. The cold signals in 1984/85 and 1999/2000 winter during the peak phases of the past two long-lasting La Nina events are predicted well up to 2-year lead. Amazingly, the mild El Nino-like event of 2002/03 is also predicted well up to 2-year lead, suggesting a link between the prolonged El Nino and the tropical Pacific decadal variability. Seasonal climate anomalies over vast parts of the globe during specific ENSO years are also realistically predicted up to 2-year lead for the first time (Luo et al. J. Clim, 2007, under review). The Indian Ocean Dipole (IOD), an air-sea coupled mode similar to ENSO, has also profound socio-economic impacts on various parts of the world as well as countries surrounding the Indian Ocean. Our model is able to predict the extreme positive IOD event in 1994 at 2-3 seasons lead. However, predictability of IOD is limited by the active and chaotic intraseasonal disturbances in the Indian Ocean. Retrospective ensemble forecasts of IOD index for the past two decades showed skillful scores only up to 3-4 months lead and a winter prediction barrier associated with its intrinsic strong seasonal phase-locking. Prediction skills of SST anomalies in both the eastern and western Indian Ocean are higher than those of the IOD index; this is due to influences of the highly predictable ENSO. Encouragingly, increasing ensembles may improve IOD predictions. Our experimental real time forecasts with 18 members successfully predicted the weak negative IOD in 2005 fall and La Nina in 2005/06 winter at two or three seasons lead. The strong positive IOD event in 2006 has been successfully predicted up to one year ahead (http://www.jamstec.go.jp/frcgc/research/d1/iod/). It is found that strong cold subsurface signals in the Southwest Indian Ocean can provide an important preconditioning (and hence long-lead predictability) for the IOD development (Luo et al. J. Clim, 2006, in press).

EU-Ensembles Project: Prediction skill of Indian monsoon rainfall in two anomalous years, 1994 and 1997
Madhavn Nair Rajeevan
National Climate Centre, IMD, Pune, INDIA
O.P.Sreejith, National CLimate Centre, IMD Pune; Jyoti Bhate, National Climate Centre, IMD Pune

Seasonal forecasts of Indian summer monsoon rainfall is very important in view of its importance to Indian economy. At present, operational forecasts are generated using advanced statistical methods with a reasonable accuracy. However, to meet the demands of users for forecasts in higher spatial resolution, dynamical models are needed. Unfortunately, the experience with all the dynamical models (atmospheric as well as coupled models) for the Indian monsoon region is not inspiring and (Gadgil and Sajani 1998, Sperber and Palmer 1996, Kang et al. 2002 and Wang et al. 2005). Analysis of numerous model results (AMIP I and II, PROVOST, DEMETER) showed that all the models failed to capture the excessive monsoon rainfall in two anomalous years of 1994 and 1997. In both the years, positive Indian Ocean dipole event (Saji et al. 1999) prevailed over the Indian Ocean. Some studies (for e.g., Gadgil et al. 2005) have suggested that above normal rainfall over the Indian sub-continent in these two years, in spite of positive SST anomalies, was likely due to favourable conditions over the equatorial Indian Ocean.

In this study, we have analyzed the hindcast results of EU-Ensembles Project for these two years, 1994 and 1997. We have considered four models, IFS/HOPE (ECMWF), Glosea (UK MET office), ECHAM5/OM1 (ECHAM) and ARPEGE/OPA (Meteo France). We have analyzed the results of runs with May 1 initial dates for 9 ensemble members. We have considered seasonal rainfall (June to September) runs for the period 1991-2001.

Both in 1994 and 1997, monsoon season rainfall was excess over the Indian sub-continent, especially over the central parts. In 1994, all-India rainfall was 10% above normal and in 1997, it was 2% above normal. An interesting aspect of rainfall distribution over the equatorial Indian Ocean was a dipole structure with negative (positive) rainfall anomaly over the east (west) equatorial Indian Ocean associated with below (above) normal SSTs. In 1994, over the equatorial central and east Pacific normal rainfall was observed and in 1997, above normal rainfall was observed due to positive SST anomalies.

The model analysis reveals that both in 1994 and 1997, UK Met office and MeteoFrance models were able to capture the observed above normal rainfall over the Indian sub-continent, which is quite interesting and encouraging. However, the ECMWF and ECHAM models predicted below normal rainfall in both these years. The ECMWF model was able to capture the below normal rainfall over east equatorial Indian Ocean in 1997 but not in 1994. The UK Met office model was able to capture the below normal rainfall over east equatorial Indian Ocean in both 1994 and 1997. However, both these models were not able to capture the above normal rainfall over the west equatorial Indian Ocean and thus the dipole structure in the rainfall pattern. The MeteoFrance model was able to simulate well the dipole structure of rainfall anomalies over the equatorial Indian Ocean, which is very encouraging. The ECHAM model was able to capture negative SST anomalies over the equatorial Indian Ocean in 1997. But the model was not capable of simulating above normal rainfall over the Indian sub-continent in 1994 and 1997.

More results on the model errors of SSTs over the Indian Ocean and the relationship with rainfall will be analyzed and presented.

References: Gadgil, S. and S.Sajini, 1998, Climate Dyn, 14, 659-689. Gadgil, S., Rajeevan, M. and Nanjundiah, R., 2005. Kang, I-S., et al., 2002, J.Climate, 15, 2791-2805. Sperber and Palmer, 1996, J.Climate, 9, 2727-2750. Wang, B., Q.Ding, et al., 2005, Geophys.Res.Lett. 32, L15711, doi 10.1029/ 2005GL022734.

Seasonal Prediction of the Indian summer monsoon rainfall
Sulochana Gadgil
Centre for Atmospheric and Oceanic Sciences, Indian Institute of Science

In this lecture I shall present my perception of the present understanding of the interannual variation of the Indian summer monsoon rainfall (ISMR) and the role of two critical modes viz. ENSO and the Equatorial Indian Ocean Oscillation (EQUINOO). Possible reasons for the rather poor skill of simulation of the interannual variation of ISMR by AGCMs, run with the observed SST as boundary condition will be discussed on the basis of analysis of AMIP 2 and the results of a project (SPIM) for assessment of the skill of the AGCMs used in India for generating seasonal prediction. It appears that the potential of two tier prediction is yet to be achieved. This leads to a suggestion about the problems that need to be addressed for achieving an improvement in the skill of simulation and hence prediction of ISMR.