WCRP Workshop on Seasonal Prediction

Barcelona Spain June 4-7, 2007

Abstracts

Poster Session III

Day 3 - Wednesday 6th June

POSTER SESSION III - Seasonal Prediction Regional Skill and Applications

International Cooperation in Climate for Development between Africa and Europe (INCLUDE) A competitive proposal for the EU FP7
Alberto Troccoli
ECMWF

Introducing Climate and Crop-Growth Modelling Tools to Provide Climate-Risk Mitigation Options to European Agriculture: The Role of Local Researchers, Seasonal Forecasts and Demonstration Proposals
Angel Utset
Agrarian Technological Institute of Castilla y Leon (ITACyL)

Benchmarks for Probabilistic Seasonal Forecasts
Leonard Smith
LSE / Oxford
R. Binter, J. Broecher, L. Clarke and H. Du, LSE

Attribution of changes in extreme weather risk: a study of the European Summer 2003 Heatwave
Helen Hanlon
AOPP Department of Physics, University of Oxford
Daithi Stone (1,2), Myles Allen (1), Peter Stott (3) and Alberto Troccoli (4)
(1) AOPP Department of Physics, University of Oxford, (2) Tyndall Centre, University of Oxford, (3) Hadley Centre, UK Met Office, (4) European Centre for Medium-Range Weather Forecasts

An Expert Prediction Market for the North Atlantic Oscillation Index
Alberto Arribas
Met Office
Mark Roulston (1), Anthony Kwasnica (2), Gary Bolton (2) and Andrew Kleit (3)
(1) Met Office, (2) Smeal College of Business, Penn. State Univ, (3) Dpt. Meteorology, Penn. State Univ.

Seasonal forecast verification using the Climate Explorer
Geert Jan van Oldenborgh
KNMI
C.A.S. Coelho (CPTEC), C.A.T. Ferro (U. Reading)

A modified method for detecting incipient bifurcations in a dynamical system
Valerie N. Livina
University of East Anglia
Tim M. Lenton (UEA)

Empirical and Numerical Approaches to Estimating the Potential Predictability of Tropical Rainfall
Vincent Moron
CEREGE UMR 6635, University of Aix-Marseille and IRI
Andrew W. Robertson (IRI, Columbia University); M. Neil Ward (IRI, Columbia University)

Use of Linear Discriminant Methods for Calibration of Seasonal Probability Forecasts
Andrew Colman
Hadley Centre, Met Office, UK
Richard Graham, Hadley Centre, Met Office, UK

Comparison of the potential skill of GCM and downscaled GCM output for river flow forecasting: a UK case study
David Lavers
Centre of Ecology and Hydrology
Christel Prudhomme, Centre of Ecology and Hydrology, David Hannah, School of Geography, Earth and Environmental Sciences, University of Birmingham

Probabilistic Crop Yield Forecasts using the Upgraded FSU/COAPS Regional Climate Model
Dong-Wook Shin
COAPS/FSU
J. Bellow, E. Chassignet, S. Cocke, T. E. LaRow, Y.-K. Lim and J. J. O'Brien, COAPS

Local 2m T seasonal hindcasts over France using statistical downscaling
Laurent Dubus
EDF R&D
Marta Nogaj (EDF R&D), Bastien Chapuis (ENSTA)

Prediction of moisture availability in agricultural soils using probabilistic monthly forecasts
David Bolius
(1) Agroscope Reckenholz-Tnikon ART; Reckenholzstrasse 191, CH-8046 Zurich
D. Bolius (1), P. Calanca (1), A. Weigel (2), M. A. Liniger (2)
(2) Federal Office of Meteorology and Climatology (MeteoSwiss), Switzerland

Repercussion of seasonal prediction models in water resources management during drought periods in the Mediterranean climate of Catalunya
J. Helmbrecht
Agència Catalana de l'Aigua
A. Manzano-Rojas, J.J. Pastor-Justo and E. Velasco-Cabre (Agència Catalana de l'Aigua)

Precipitation Forecast Comparison over Spain from the System 2 and System 3 ECMWF Operational Seasonal Forecasts.
Bartolome Orfila
I.N.M. (Spain)
E. Diez (INM) and F. Franco (INM)

Seasonal predictability over the Iberian Peninsula associated with ENSO events
Maria Dolore Frias
Applied Meteorology Group. Department of Applied Mathematics and Computer Science, University of Cantabria. Santander (SPAIN).
S. Herrera(1), A. S. Cofiño(1), J. M. Gutierrez(1)
(1)Applied Meteorology Group. Department of Applied Mathematics and Computer Science, University of Cantabria. Santander (SPAIN)

Potential sources of seasonal climate predictability in the Mediterranean Basin
Joan Ballester
Climate Research Laboratory, Spain
Xavier Rodo (1), Ben Cash (1,2)
(1) Climate Research Laboratory, Spain (2) Center for Ocean-Land-Atmosphere Studies, USA.

An adaptive multi-regressive method for summer seasonal forecast in the Mediterranean area
Massimiliano Pasqui
Institute of Biometeorology - National Research Council (IBIMET - CNR)
L. Genesio, A. Crisci, J. Primicerio, R. Benedetti and G. Maracchi, IBIMET - CNR

Downscaling Techniques Applied to Seasonal Forecast Outputs of the EU ENSEMBLES (FP6) Project (RT2B)
Elia Diez
INM, Spain
B. Luaces, E. Criado and B. Orfila, INM

Downscaling the seasonal forecasting information : the example of New-Caledonia
Jean-Pierre Céron
Météo-France - Direction de la Climatologie
LEROY Anne, Météo-France, Direction Interrégionale de Nouvelle-Calédonie

Developing prediction models for the primary Caribbean Dry Season
Tannecia Stephenson
Climatic Research Unit
A. Anthony Chen and Michael A. Taylor, The University of the West Indies

Seasonal forecasts of Canadian winter precipitation by post-processing GCM integrations
Hai Lin
Environment Canada
H. Lin (1), G. Brunet (1) and J. Derome (2)
(1) Environment Canada, (2) McGill

VAMOS

Prediction-ability of Seasonal Climate over North America
Predictability Prediction and Applications Interface Panel (PPAI) of US CLIVAR
Lisa Goddard*, Simon Mason, Ben Kirtman, Kelly Redmond, Randy Koster, Wayne Higgins, Marty Hoerling, Alex Hall, Jerry Meehl, Tom Delworth, Nate Mantua and Gavin Schmidt
*Presenting Author: Lisa Goddard, The International Research Institute for Climate & Society, The Earth Institute of Columbia University

Prospects for seasonal prediction of the North American monsoon
David Gutzler
University of New Mexico, USA

Impacts of Ocean-Atmosphere Interactions and Climate Change on the North American Monsoon
Ruth Cerezo Mota
University of Oxford

EUROBRISA: A EURO-BRazilian Initiative for improving South American seasonal forecasts
Caio Coelho
Centro de Previsão de Tempo e Estudos Climáticos (CPTEC)

Climatological features of SAMS based on SMIP simulations
Iracema F.A. Cavalcanti
CPTEC/INPE
Fernanda Cerqueira (CPTEC/INPE) Alice M. Grimm (Universidade Paraná)

Seasonal Trend of Maximum and Minimum Temperature in Argentina During the Period 1961-2000
José Luis Stella
Servicio Meteorologico Nacional de Argentina

AAMP

Multidecadal variability of the Indo-Pacific oceanic teleconnection
Ge Shi
Department of Biological and Physical Sciences, University of Southern Queensland
Joachim Ribbe (1), Wenju Cai (2), Tim Cowan (2)
(1) Department of Biological and Physical Sciences, University of Southern Queensland, (2) CSIRO Marine and Atmospheric Research

The stochastic simulation of boreal summer time ISO for the post-processing of dynamical seasonal prediction
Jin Ho Yoo
ICTP
Andrew W. Robertson (IRI), In-Sik Kang (CES/SNU)

Performance of climate prediction models on annual cycles and its relation with seasonal prediction
June-Yi Lee
University of Hawaii/ IPRC
Bin Wang, University of Hawaii/ IPRC, USA In-Sik Kang, Seoul National University, Korea J. Shukla, George Mason University, USA C.-K. Park, APCC, Korea

Role of combined effect of NAO, SO and MJO in seasonal prediction of Indian summer monsoon
Surendra Dugam
Indian Institute of Tropical Meteorology
S. B.Kakade, Indian Institute of Tropical Meteorology

Seasonal Forecasts of Indian Summer Monsoon Rainfall using Statistical Models: Problems and Prospects
Madhavan Nair Rajeevan
National Climate Centre, India Meteorological Department, India
D.S.Pai and R Anil Kumar, National Climate Centre, India Meteorological Department, India

On the Relationship between the Indian Summer Monsoon and River Flow in the Aral Sea Basin
Reinhard Schiemann
Institute for Atmospheric and Climate Science, ETH Zurich
Mariya G. Glazirina, Christoph Schor

Seasonal Predictions of the Indian Summer Monsoon using the NCMRWF Global Modeling System
Sarat Kar
National Centre for Medium Range Weather Forecasting (NCMRWF), India
A. K. Bohra, NCMRWF, India

Improved hindcasts of Indian monsoon rainfall using a Tier 1.5 approach
Juergen Kroeger
Abdus Salam ICTP, Trieste, Italy
F. Kucharski (ICTP), A. Bracco (Georgia Tech), J. H. Yoo (ICTP) and F. Molteni (ECMWF)

Probabilistic forecast of All-India monsoon rainfall on the basis of monthly rainfall
Dilip Kothawale
Indian Institute of Tropical Meteorology
A. A. Munot, Indian Institute of Tropical Meteorology

Asian Monsoon Variability and Predictability: WCRP CMIP3 multi-model Simulations
Ramesh Kripalani
Indian Insititute of Tropical Meteorology
A Kulkarni and S.S. Sabade, Indian Institute of Tropical Meteorology

VACS

North African climate variability prediction
Abebe Yeshanew
National Meteorological Agency, Ethiopia

Changes in the Frequency and Intensity of Extremes over NorthEast Africa
Gebru J. Endalew
Wageningen University
B. van den Hurk and G. J van Oldenborgh, KNMI

Coupled SST-West African Monsoon precipitation patterns in AGCM simulations
Elsa Mohino
UCM
Teresa Losada (UCM), Belen Rodriguez-Fonseca (UCM), Carlos Roberto Mechoso (UCLA)

Verification of seasonal forecasts from a multi-model ensemble system over West Africa
Alexandre Gagnon
Environmental Research Institute, UHI
Anne E. Jones and Andrew P. Morse Department of Geography, Roxby Building University of Liverpool

Climate Extremes over the Equatorial East Africa and their Predictability with the Indian-Ocean Dipole Mode
Zablone Owiti
Department of Meteorology, University of Nairobi

Regional rainfall anomalies in Central Equatorial Africa and their links to the Ocean-Atmosphere dynamic: Explicative variables identification using cross wavelet analysis.
Elie Zihindula Kagayo
Department of Statistics and Operation Research. University Polytechnic of Cataluña, Barcelona-Spain.
M. Martí-Recober (University Polytechnic of Cataluña)

Multi-model simulations of streamflow into the river basins shared by South Africa and Mozambique
Willem Landman
South African Weather Service
Francois Engelbrecht - University of Pretoria, South Africa, Trevor Lumsden - University of KwaZulu-Natal, South Africa

Multi-Model ensemble inter-seasonal to seasonal forecasting and probability outcomes in prediction system
Nana Ama Kum Browne
Climate System Analysis Group, ENGEO Department, University of Cape Town

Regional Outlook Forums in Africa
Leonard Njogu Njau
ACMAD

Combating Flood Crisis with Geographic Information System (GIS): An Example from Akure, Southwest Nigeria
Okuku Ediang
Nigerian Meteorological Agency
A.O Eludoyin, Department Of Geography, Obafemi Awolowo University, Ile-Ife, Nigeria O. M. Akinbode, Department of Geography and Planning Sciences, Adekunle Ajasin University, Akungba, Akoko, Ondo state, Nigeria

Seasonal forecasting of Ethiopian rainfall using SST
Gulilat Tefera Diro
Department of Meteorology, University of Reading
David Grimes and Emily Black, Department of Meteorology, University of Reading

Optimisation of the Manantali dam water management using a seasonal forecasting information
Jean-Pierre Céron
J.C. Bader (1), J.P. Lamagat (2), J.P. Piedelièvre (3) and J.P. Céron (4)
(1) Institut de Recherche pour le Développement, Montpellier, France (2) Institut de Recherche pour le Développement, Dakar, Sénégal (3) Météo-France, CNRM, Toulouse, France (4) Météo-France, Direction de la Climatologie, Toulouse, France

Simulation of the sub-Saharan Africa climate with the Hadley Centre Regional Climate Modeling System: Validation for the period 1961-1990
André Kamga Foamouhoue
African Centre for Meteorological Applications to Development
Etienne Buscarlet University of Toulouse, Environmental sciences department.

Verification of climate outlooks in Africa: proposed methodology for West Africa and prospects for all Africa
André Kamga Foamouhoue
African Centre for Meteorological Applications to Development

West African Weather System In The Development Of Tropical Cyclones
Tairu Salami
Nigerian Meteorological Agency
O.s Idowu E,E Balogun

International Cooperation in Climate for Development between Africa and Europe (INCLUDE) A competitive proposal for the EU FP7
Alberto Troccoli
ECMWF

Economic, Social and Environmental Development are top priorities for most nations as promulgated for Europe within the 'Lisbon Strategy' and for Africa with the Climate for Development agenda recently endorsed by the African Union - with Sustainable Development high on the agenda for most, not least those African countries pursuing the United Nations Millennium Development Goals (MDGs). Given the vulnerabilities of European and African societies, amongst all others, to the impacts of current Climate Variability, and recognising the challenges of establishing sustainable future Economic, Social and Environmental Development within the constraints of a changing Climate, there is an urgent need to examine and restructure the integration of Climate into strategy, policy and investment decision making at all levels throughout Europe and Africa. With the medium-term objective to achieve improved Climate informed Policies and Investment Strategies for Development, this project will initially concentrate on two first steps: i) the setting up a network for International Cooperation between Africa and Europe tasked to strengthen or, where required, initiate dialogue regarding the nexus of Climate and Development between all appropriate groupings in Europe and in Africa which pertain to Climate, Development and Policy sectors and ii) the appraisal of the statuses in Africa and in Europe of Climate information and knowledge in relation to Investment Strategies, in Planning for Development, and in existing Policy and in Policy Creation. The two main expected results would be: 1. Recommendations to the EU Framework Programme on priorities for Science and Technology research aimed at the improvement of Climate information to be used in a Development context both in Africa and Europe. 2. Strengthened Africa-Europe bi-regional collaboration on the integration of Climate information for Investment and Policies Strategies.

The climate-change due weather variability significantly affects European agriculture. Negative climate impacts could be reduced by adopting mitigation options obtained from crop-model simulations combined with climate scenarios. The usefulness of such simulation tools has been proved in manifold papers that appeared in the last years, usually produced in Universities or similar centres. However, despite of the considerable public concern about climate variability, European stakeholders and farmers are not yet using these scientific results for agricultural decision-making. Researchers from local agricultural-services can effectively realize which practical decisions should be taken for mitigating the possible climate risks on their local conditions. Nevertheless, those local institutions are not usually connected to high-level researches neither to EU funding procedures and they need support before being able to use the current climate and crop-growth modelling tools. According to this, the EU-funded proposal AGRIDEMA provided initial contacts and feedback mechanisms between high level research centres and Universities; where modelling tools have been developed and tested ('developers'); with their potential users of these tools, located mainly in regional Agricultural Research Services ('users'). Connections between 'developers' and 'users' were made through short courses and about 20 pilot assessments conducted in several European countries. Local researchers knew in the AGRIDEMA courses how to access to GCM data and seasonal forecasts. They receive also basic knowledge on weather generators, statistical and dynamical downscaling; as well as several crop models. An illustrative example of an AGRIDEMA pilot assessment shows how weather variability could be an important shortcoming for Sugarbeet irrigation management, as well as the potential usefulness of a seasonal forecast for such management. Stakeholders will adopt climate-risk mitigation options only if they realize the reliability of such options on their specific cases. To achieve this, the 'users' of the modelling tools must develop local demonstration proposals, aimed to model calibration and validation, etc. Such demonstration proposals should be initially included in the FP7 Cooperation calls, since stakeholders are still unfamiliar with this technology and the 'users' need the 'developers' support. Particularly, the outputs of the DEMETER and ENSEMBLE 'developers' can be used as input in several 'users' applications. The FP7 Capacities calls, involving SMEs, could be the precise way to introduce such tools in a second step, although large validations are needed before practical introductions. All this effort could enable European agricultural research services, as ITACyL, to play an important role in the framework of Lisbon strategy and in the European rural development.

Benchmarks for Probabilistic Seasonal Forecasts
Leonard Smith
LSE / Oxford
R. Binter, J. Broecher, L. Clarke and H. Du, LSE

Evaluating the skill of probabilistic seasonal forecasts requires benchmark distributions with which dynamic model based forecasts can be compared. Methods for the construction of an unconditioned distribution (climatology) of monthly average quantities from historical data are discussed and illustrated. Forecasts from the DEMETER forecast-observation archive are evaluated against this unconditioned distribution which, of course, varies with the target month, noting that outperforming this monthly climatology is not sufficient to establish the value of ensemble forecasts from a simulation model(s). Dynamic climatologies, specifically simple statistical models conditioned on the current observations, are also discussed and shown to yield distributions with significantly more skill than the unconditioned climatological distribution. The development of dynamic climatologies for the evaluation of forecast applications (forecasts in user relevant variables) is discussed.

In 2003 the average summer temperature in continental Europe exceeded the 1961-1990 European summer mean by 2.3K. Many regions experienced a large number of deaths due to the elevated temperatures, and attribution studies have determined that human activity have at least doubled the risk of such a heatwave compared to pre-industrial times. However, other, non-linear processes could also have amplified summer 2003 temperatures: feedbacks between reduced cloud cover and precipitation and reduced soil-moisture may have prevented the usual convective disruption of the high pressure system. This study will build on previous attribution work by attempting to further isolate the change in risk of the heatwave due to anthropogenic influences. A large ensemble of the ECMWF IFS model will be performed at higher resolution than previous studies, with improved simulation of land surface processes.

In the same way that the FTSE100 index summarises the behaviour of the London stock market in a single number, the North Atlantic Oscillation (NAO) index summarises the state of the North Atlantic winter weather. In this experiment we will analize the ability of markets to aggregate and combine information into a single, probabilisitic prediction of the NAO index. Participants will be able to buy and sell contracts corresponding to conditions where the NAO index falls into one of four categories during each of the winter months. Contracts will be traded online, with participants wishing to buy/sell contracts posting bids/offers online.

Seasonal forecast verification using the Climate Explorer
Geert Jan van Oldenborgh
KNMI
C.A.S. Coelho (CPTEC), C.A.T. Ferro (U. Reading)

Seasonal climate forecasts are made using multi-model ensembles. Contrary to climate change projections, the skill of the forecasts can be verified against observations using old forecasts and hindcasts. In practice the small number of forecasts (15-45) is a severe limitation, as the skill depends strongly on the region and season. We present a web-based system to produce charts and maps of the skill of operational seasonal forecast systems using a variety of measures. It is part of the KNMI Climate Explorer (climexp.knmi.nl), and presently contains data from the ECMWF S3, UKMO GloSea and NCEP CFS operational forecast systems, as well as the Demeter research experiment. The verification measures have been developed in the RCLIM project, and include deterministic measures such as the ensemble mean correlation, RMSE and MAE, as well as probabilistic measures such as the Brier Score, its decomposition into resolution, reliability and uncertainty, and the ROC curve. These are available both for time series (area-averaged or all grid points in a region) and as spatial maps. More measures, and estimates of the uncertainties of the skill scores, are planned. The verification system allows seasonal forecasters and climate researchers to quickly explore the predictability of the short-term climate with current state-of-the-art models.

A modified method for detecting incipient bifurcations in a dynamical system
Valerie N. Livina
University of East Anglia
Tim M. Lenton (UEA)

We assess the proximity of a system to a bifurcation point using a degenerate fingerprinting method that estimates the declining decay rate of fluctuations in a time series as an indicator of approaching a critical state. The method is modified by employing Detrended Fluctuation Analysis (DFA) which improves the estimation of short-term decay, especially in climate records which generally possess power-law correlations. When the modified method is applied to GENIE-1 model output that simulates collapse of the Atlantic thermohaline circulation, the bifurcation point is correctly anticipated. The technique could be used to anticipate future bifurcations in systems with high-resolution time series.

Empirical and Numerical Approaches to Estimating the Potential Predictability of Tropical Rainfall
Vincent Moron
CEREGE UMR 6635, University of Aix-Marseille and IRI
Andrew W. Robertson (IRI, Columbia University); M. Neil Ward (IRI, Columbia University)

This study examines space-time characteristics of seasonal rainfall predictability in tropical regions (Senegal, Brazilian Nordeste, Kenya, Northwestern India, Queensland). Potential predictability is analyzed in terms of the spatial coherence of observed interannual variability at the station scale, and within-ensemble coherence of general circulation model (GCM) simulations with observed sea surface temperatures (SSTs) prescribed. Seasonal-mean rainfall anomalies are decomposed in terms of daily rainfall frequency (= number of wet days), and daily mean intensity (= mean rainfall on wet days). The observed spatial coherence is computed from the network of daily rainfall including 9-81 stations, in terms of (a) interannual variability of a standardized anomaly index (i.e. the average of the normalized anomalies of each station), (b) the external variance (i.e. the fraction of common variance amongst stations) and, (c) the number of spatio-temporal degrees of freedom. Spatial coherence of interannual anomalies at regional-scale is found to be much stronger for seasonal rainfall amount and daily occurrence frequency, compared to daily mean intensity of rainfall. Combinatorial analysis of the station observations suggests that, for occurrence and seasonal amount, the empirical number of spatial degrees of freedom is largely insensitive to the number of stations considered. The spatial coherence of frequency of occurrence is sometimes far larger than the one of seasonal amounts when large daily amounts > 50 mm are especially frequent and intermittent (ex : Northwestern India). For daily mean intensity, by contrast, each station is found to convey almost independent information, and the number of degrees of freedom would be expected to increase for a denser network of stations. The GCM estimates of potential predictability and skill associated with the SST forcing are found to be remarkably consistent with those inferred from the observed spatial coherence for Senegal: there is a moderate-to-strong skill at reproducing the interannual variations of seasonal amounts and rainfall occurrence whereas the skill is weak for the mean intensity of rainfall.

Use of Linear Discriminant Methods for Calibration of Seasonal Probability Forecasts
Andrew Colman
Hadley Centre, Met Office, UK
Richard Graham, Hadley Centre, Met Office, UK

The linear discriminant method is presented as a tool for calibrating probability bias in seasonal forecasts, such that the probabilities provide an improved estimate of the likelihood of the predicted event occurring. Results of calibration of the Met Office GloSea model and Multi-model (ECMWF, Meteo-France, GloSea ) hindcasts produced as part of the DEMETER project will be presented. Results of calibrated GloSea model hindcasts combined with SST based statistical predictions will also be presented. When tested using quintile and tercile hindcasts from the GloSea model, replacing raw ensemble probabilities with discriminant calibrated probabilities improves reliability but reduces forecast resolution - as measured by the Relative Operating Characteristic (ROC) diagnostic. A pragmatic approach that recovers ROC skill whilst retaining the reliability improvements involves generating weighted averages of calibrated and uncalibrated probabilities and results will be presented Multivariate (MV) discrimanant analysis has also been used to produce calibrated probabilities. The predictors are combinations of GCMs or of GCM and SST based statistical predictions. This method is used in prediction of North and East African rainfall, NE Brazil rainfall and European winter temperature. The MV probabilities are not quite as reliable as single variable discriminant forecasts and can be disproportionate to predictor skill in certain cases. Procedures are being tested to correct these anomalies by for example significance tests on predictors prior to their selection or adapting the discriminant equation to take more account of uncertainty in observations. Examples will be presented and discussed.

Forecasting of river flows at seasonal time scales, defined as lead times of 1 to 6 months, is an area of growing research interest in the UK (and elsewhere). This is because seasonal forecasts of river flows may be used practically to inform water management decisions and to prepare better for hydrological extremes (floods or droughts), which may have major societal and economic impacts. The drought conditions experienced in parts of the UK from 2004-06 brought the need for seasonal forecasting of water resources into sharp focus. The aim of this paper is to compare the potential skill of river flow forecasting using (1) Global Climate Model (GCM) output and (2) downscaled precipitation data from GCMs. The study focuses upon the River Dyfi in central West Wales, UK. This river basin was chosen as a test case because it has limited anthropogenic influence, hence the climate-flow signal should be stronger than a basin with greater human impact. The input meteorological data used are taken from ERA-40, which is a re-analysis of meteorological observations produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). ERA-40 precipitation and observed potential evaporation data drive a rainfall-runoff model (Probability Distributed Model, PDM) to simulate river flow series. PDM is a lumped rainfall-runoff model that transforms rainfall and potential evaporation data to river flow at the catchment outlet. PDM was calibrated with observations over a 10 year (1980 - 1990) period, and then validated over an independent 10 year (1991 - 2001) period. Owing to the coarser resolution of the ERA-40 data (2.5° x 2.5°, reduced to 1.5° x 1.5°) in comparison with the spatial scale of the river basin, a statistical downscaling tool is utilised to produce rainfall data at the catchment scale, therefore more appropriate for hydrological applications. The ERA-40 and downscaled ERA-40 precipitation data are inputted to PDM to assess their relative river flow modelling skill. The findings of this research have important implications for assessing (1) the potential of coarse GCM output for the seasonal forecasting of river flows and (2) the relative merits of downscaled vs non-downscaled ERA-40 data for river flow forecasting. Additionally, the results obtained using ERA-40 will represent a benchmark for future seasonal forecasts based on GCM output.

Probabilistic Crop Yield Forecasts using the Upgraded FSU/COAPS Regional Climate Model
Dong-Wook Shin
COAPS/FSU
J. Bellow, E. Chassignet, S. Cocke, T. E. LaRow, Y.-K. Lim and J. J. O'Brien, COAPS

An advanced land model (NCAR CLM2) is coupled to the Florida State University/Center for Ocean-Atmospheric Prediction Studies (FSU/COAPS) regional spectral model to improve seasonal surface climate outlooks at very high spatial and temporal resolution and to examine its potential for crop yield estimation. The regional model domain is over the southeast United States and run at 20 km resolution, roughly resolving the county level. Warm season (March-September, 7-month simulation) and cold season (October-March, 6-month simulation) ensemble simulations are performed for the period of 19 years (1987-2005) to characterize uncertainty in the forecast. Twenty member ensembles of the regional model are generated using different initial conditions and model configurations (i.e., the ensemble methods based on different convective schemes). These ensembles are used to make probabilistic crop yield forecasts. Outputs from the model such as max/min surface temperatures, precipitation, and shortwave radiation at the surface are analyzed and used as inputs into the crop models (e.g., CROPGRO-Peanut and CERES-Maize) to determine crop yields. Results will be presented in the workshop.

Local 2m T seasonal hindcasts over France using statistical downscaling
Laurent Dubus
EDF R&D
Marta Nogaj (EDF R&D), Bastien Chapuis (ENSTA)

The University of Cantabria and Instituto Nacional de Meteorologia of Spain web portal for statistical downscaling has been used to downscale the DEMETER/system 2 and ENSEMBLES/stream 1 seasonal hindcasts. The aim is to predict local 2mT on 25 local stations in France, using daily historical observations and multi model seasonal hindcasts. Compared to the raw hindcasts at the stations nearest gridpoints, the downscaled temperatures reduce the error and improve the ACC with regard to observed data, from 10% to more than 50%, depending on the area, the lead team and the season. Work is still in progress, but such an improvement in 2mT predictions may be enough to improve substantially applications such as energy demand predictions.

Despite technological advances in breeding and agricultural practice, crop yield remains subject to considerable inter-annual variability related to short-term (seasonal) climate fluctuations. Of utmost importance in this context are variations in soil water availability. Extreme conditions such as the heat wave observed in Europe during the summer of 2003 can lead to anomalous soil moisture depletion and induce considerable losses in crop production. The prediction of soil water levels using monthly forecasts could provide valuable means for risk assessment and mitigation. We present first results of a prediction system for soil moisture forecasts with a lead time of up to one month. The system uses dynamical, probabilistic forecasts of daily temperature, precipitation and global radiation from the European Centre for Medium-Range Weather Forecasts. The forecasts drive a bucket model of the water balance in the root zone. We show that the seasonal evolution of the soil water available to crops is well reproduced by the system. The prediction system was tested for a grid point in Switzerland (47°N, 8°E) using monthly hindcasts covering the time period 1994-2005. Simple downscaling of raw model data was performed by applying model anomalies from the model climatology to the observed climatology. Resulting soil moisture forecasts showed to be skilful over climatology (0<skill<0.6) up to a lead time of three weeks. The system allows for the probabilistic estimation of reaching a critical level of soil moisture within such a forecast period. This can serve as a valuable information for the farmers' decision-making process.

Like floods, droughts are one of the climatic manifestations of greater socioeconomic repercussion in the Catalan territory. This happens especially due to the strong spatial and temporary irregularity that characterizes the pluviometric regime, one should also take into account for a high interannual variability. In this scenario, the results of the numerical models of seasonal weather forecasts usually have a wide range of variability, that is translated in a greater uncertainty of the prediction of the evolution of the stored reserves and the satisfaction of the dependent uses. The lack of precision of these forecasts of hydric reserves implies the assumption of a risk unknown at the time of taking (or not taking) preventive measures of saving or applying water use restrictions before a possible episode of drought. Although these measures can avoid an emergency situation and guarantee the supply of potable water to the population, they also have an important socioeconomic repercussion for the nonhigh-priority users. Therefore, the minimization of the uncertainty of numerical models weather forecasts in the Mediterranean climate would cause a more efficient use of the hydric resources and a decrease in the negative effects of drought. The Catalan Water Agency has analyzed several seasonal models throughout 2 years having concluded a great dispersion in predictions and also in the degree of coincidence with the registered real situations.

Precipitation Forecast Comparison over Spain from the System 2 and System 3 ECMWF Operational Seasonal Forecasts.
Bartolome Orfila
I.N.M. (Spain)
E. Diez (INM) and F. Franco (INM)

Since March 2007, the operational seasonal forecasts from the ECMWF are produced by the System 3, which replaced the System 2. Both are produced by the ECMWF Global Coupled Ocean Atmospheric Model. However, an EPS of 41 members is also available from October 2006. The performance of both Systems for precipitation over Spain has been studied by comparing direct model outputs and outputs from applying ANALO-ONE, the downscaling statistical tool already used in the DEMETER Project and operationally for seasonal forecast in INM The period covered for this comparison is Autumn 2006 and Winter 2006. The results are calibrated using a 25-year hindcasting (1981-2005) of 11 members. Results of two cases of dynamical downscaling using the Rossby Centre Atmospheric model (RCA) nested to the System 3 outputs are also presented. For it a 25-year hindcasting of 5 members has been prepared. In future operational seasonal forecast are aimed to use 11 members.

Seasonal predictability over the Iberian Peninsula associated with ENSO events
Maria Dolore Frias
Applied Meteorology Group. Department of Applied Mathematics and Computer Science, University of Cantabria. Santander (SPAIN).
S. Herrera(1), A. S. Cofino(1), J. M. Gutierrez(1)
(1)Applied Meteorology Group. Department of Applied Mathematics and Computer Science, University of Cantabria. Santander (SPAIN)

There has been an increasing interest in seasonal forecast since it was proved that a certain predictability exists over the tropical Pacific at interannual time scale. Nowadays, the most important meteorological institutions around the world have developed coupled atmosphere-ocean general circulation models which allow to predict climate anomalies at interannual time scales, such as El Nino phenomenon, several months in advance. Nevertheless, the predictability is limited at higher latitudes far from the influence of tropical Pacific, so the models should be validated to provide end-users with accurate information about the value of the seasonal predictions at extra-tropical latitudes. Several studies (Rodo et al 1997; Pozo-Vazquez et al 2001) have shown the connection between El Nino phenomenon and the precipitation and temperature variability over Spain. In this study, particular attention is given to the predictability during strong El Nino and La Nina events. The connection between precipitation and maximum and minimum temperature variability and the ENSO events is first analysed by seasons. To this end a 0.2°x0.2° latitude-longitude gridded dataset developed over Spain using data from 3000 stations of the Spanish Institute of Meteorology is considered. The connection with the ENSO events is analysed in terms of order statistics (terciles and quintiles) taking into account the strongest La Nina and El Nino events for the period 1950-2000. Thus, the frequency of, e.g., dry and wet terciles during El Nino (and La Nina) events can be compared with the uniform climatological frequency, obtaining an estimation of the statistical significance. Besides the well known teleconnection in Spring, some other weaker signals have been found. According to these results, seasonal predictability is studied in Spain using a combination of a multimodel ensemble system plus a statistical downscaling method The goal of the study is twofold: The validation of seasonal forecasts from the DEMETER project at middle latitudes, and the need to provide local information in order to improve seasonal forecasts. Regional information is required in many impact assessment studies in which the coarse resolution of the general circulation models is not enough. Statistical donwscaling methods translate large-scale information to local or regional scale by means of empirical relationships between large-scale variables and the target local variables. The statistical downscaling method considered in this study is based on the search of analogs (Gutierrez et al 2004). The method differs from the standard analog approach by applying a clustering technique to the predictors. It allows the definition of meaningful subgroups (weather types) each associated with a reference pattern which represents a specific atmospheric scenario. This approach is applied to assess the DEMETER skill to forecast seasonal precipitation and maximum and minimum temperature over Spain taking into account the observed influence of ENSO events. Seasonal ensemble forecasts from the direct output from DEMETER and the downscaled values are verified against observations from a probabilistic point of view taking into account order statistics (terciles and quintiles). Bibliography Gutierrez, J. M., A. S. Cofino, R. Cano, and M. A. Rodriguez (2004), Clustering methods for statistical downscaling in short-range weather forecasts. Mon. Wea. Rev., 132: 2169-2183. Pozo-Vazquez D., M. J. Esteban-Parra, F. Rodrigo and Y. Castro-Diez (2001). The association between ENSO and winer atmospheric circulation and tempeartue in the North Atlantic region. Journal of Climate, 14: 3408-3420.

Current seasonal climate predictions for the Mediterranean basin derived from statistical and dynamical models show a very low skill and limited application. The absence of an accurate and exhaustive diagnosis describing all possible sources of predictability seems to also add to that limitation. As an example, although the ENSO influence over the Mediterranean region has been already described, the impact that other oceanic regions of the planet might also have are far to be fully determined. The present work aims to explore tropical ocean regions in the search of useful sources of seasonal predictability for the Mediterranean Basin. The Extended Reconstructed global Sea Surface Temperature dataset (NOAA NCDC ERRSST v2) was used for the 1951-2000 time interval in order to describe the ocean structures yielding the required ocean memory. The study was complemented by the use of the NCEP/NCAR reanalysis searching for dynamical atmospheric patterns associated to the former teleconnections. According to the results, not only the Tropical Pacific leads to predictable Mediterranean anomalies, but the same applies for specific regions in the Tropical Atlantic. Coherent atmospheric patterns in response can be found, although characteristic time of the atmosphere is smaller. We compare the resulting projections with those arising from simulations using a high-resolution AGCM-only model with prescribed-only conditions in the eastern tropical Pacific region. Results show how at large the prescribed AGCM successfully reproduces main observed patterns in the Tropics, though extratropical responses are largely underestimated and the large dipole anomaly forming at high latitudes is not reproduced.

Seasonal forecasting represent the attempt to predict, in a statistical framework, the spatial and temporal distribution of weather anomalies a few months into the future. Even though the detailed dynamical evolution of atmospheric systems is not predictable on those time scales, some of their statistical features and behaviours can be predicted. In particular it is possible to infer on the average behaviour over a month or season, and how much the probability distribution of such averages, or anomalies, differs from the "climatology". Since late 90's seasonal forecasts experienced a growing role, despite the large uncertainties still present. Precipitation and temperature anomalies knowledge, available a few months early, can be useful for technical services and organizations on managing water resources, crop and energy. At the same time, methods and results of this recent branch of atmospheric sciences must be as simple and accessible as possible for any potential users. For this reasons, the Institute of Biometeorology developed a simple, physically-based, statistical approach to obtain monthly outlooks, regarding rainfall and temperature anomalies over the Mediterranean basin, tuned for the summer period and based on the NCEP-NCAR Reanalysis dataset. The forecasting strategy is a multi-regressive method based on physical atmospheric indices and sea surface anomalies. Using the state of the art of the atmospheric behaviour knowledge, at monthly time scale, for the Mediterranean basin, we select potential predictors among a list of monthly large scale circulation indexes (SV-NAM, Modified Zi, NAO), sea surface temperature anomalies (Atlantic Tripole, Guinea Gulf). The selected indexes and their coefficients, in the multi-regressive model, have been chosen according to a maximization of the regression values between observed and forecasted rainfall and temperature anomalies. Therefore the 'adaptation' is performed through the best choice of predictors which provides a maximum probability of detection for the selected anomalies. A validation strategy was developed with respect to the 1979-2005 period based on the "information entropy" theory. Skill scores analysis show encouraging results of the method useful for a monthly outlook evaluation.

Downscaling Techniques Applied to Seasonal Forecast Outputs of the EU ENSEMBLES (FP6) Project (RT2B)
Elia Diez
INM, Spain
B. Luaces, E. Criado and B. Orfila, INM

One of the objectives of ENSEMBLES (ENSEMBLE-based Predictions of Climate Changes and their Impacts) is the production of probabilistic predictions from seasonal to decadal timescales through the use of ensembles. Dynamical and statistical methods have been used as a downscaling tool for seasonal forecast over Europe. Dynamical downscaling is performed using the Rossby Centre Climate Atmospheric Model (RCA3), which has been nested to outputs of two ENSEMBLES models: the ECMWF and UKMO UM. and run in climate mode for six months periods during 1991-2001.

New Caledonia is a priori an ideal framework to downscale seasonal forecasts issued by dynamical models. First, we establish the forcing by the large scale circulation of different local meteorological parameters observed in New Caledonia. A zoning is proposed for the New Caledonia and the 3 studied parameters (rainfall, Maximum temperature and Minimum temperature). Then 3 models of the Demeter experience; namely the models from CEPMMT, Met Office and Météo-France are used and merged in different ways. The use of a linear downscaling method shows the reduced interest of various attempt to correct systematic biases of dynamical models in a region where the models forecasts are of good quality and the spatial limits of downscaling are identified. Finally, the performances of different statistical downscaling methods (linear, neural and by analogues) are compared to each other as well as to a method of non-adaptation.

Developing prediction models for the primary Caribbean Dry Season
Tannecia Stephenson
Climatic Research Unit
A. Anthony Chen and Michael A. Taylor, The University of the West Indies

Two statistical models are created for the Caribbean during its dry season. Canonical correlation analysis (CCA) confirms that the El Nino Southern Oscillation (ENSO) is the dominant mode of variability for the dry season. The mode manifests itself as oppositely signed precipitation anomalies over the north and south Caribbean. The south-eastern Caribbean becomes dry in response to a warm event. Consequently the predictability of mid-dry season rainfall over the south-eastern Caribbean is investigated. A model which retains an ENSO proxy as one of two predictors shows reasonable skill with hindcast predictions for the region. A second model is created using a Jamaican rainfall index as predictand. Jamaica falls in the transition zone i.e. between the oppositely signed north-south precipitation anomalies characteristic of the ENSO dry season mode. This model retains no ENSO related predictor. Composite analysis of select atmospheric variables for anomalously high and low rainfall years (for the dry season) provide insight into the dynamics of the Caribbean dry season during phases of the ENSO, particularly those which lead to the creation of the transition zone.

Seasonal forecasts of Canadian winter precipitation by post-processing GCM integrations
Hai Lin
Environment Canada
H. Lin (1), G. Brunet (1) and J. Derome (2)
(1) Environment Canada, (2) McGill

In the second phase of the Canadian Historical Forecasting Project (HFP2), four global atmospheric models (GCMs) were used to perform seasonal forecasts over the period of 1969-2003. Little predictive skill was found from the GCM ensemble seasonal predictions for the Canadian winter precipitation. This study is an effort to improve the precipitation forecasts through a post-processing approach. Canadian winter precipitation is significantly influenced by two of the most important atmospheric large scales patterns, namely, the Pacific-North American pattern (PNA) and the North Atlantic Oscillation (NAO). The time variations of these two patterns were found to be significantly correlated with those of the leading singular value decomposition (SVD) modes that relate the ensemble mean forecast 500 mb geopotential height over the Northern Hemisphere and the tropical Pacific SST in the previous month (November). A statistical approach to correct the ensemble forecasts was formulated based on the regression of the model's leading forced SVD patterns and the observed seasonal mean precipitation. The performance of the corrected forecasts was assessed by comparing its cross-validated skill with that of the original GCM ensemble mean forecasts. The results show that the corrected forecasts predict the Canadian winter precipitation with statistically significant skill over the southern prairies and a large area of Quebec-Ontario. Most of the skill comes from the NAO.

A necessary step towards understanding and improving our ability to predict the climate is first documenting the current levels of predictability in real-time forecast systems. This task was deemed an important activity for the US CLIVAR PPAI panel to undertake. Our panel sought to:

-Review current state of predictability for SI forecasting systems, in coordination with WGSIP, motivated in part by the questions: What is the "limit of predictability"? How do we know it is the "limit", and why is there a limit?
-Recommend and promote appropriate metrics for use in validating forecast systems, guided by WMO recommendations, including associated pros and cons of different metrics.
-Identify outstanding research questions potentially limiting the skill and/or reliability of current forecast systems.

Using retrospective forecast data provided by WCRP SMIP-2, DEMETER, CFS, and IRI, we have produced a skill assessment for seasonal mean 2m air temperature and total precipitation for June-July-August and December-January-February for the period 1981-2000. The model skills are assessed individually as well as for a simple multi-model combination. We examine both deterministic skill of the ensemble means, as judged by mean squared error (MSE), and its decomposition (correlation, bias, and variance ratio), and probabilistic skill, as measured by reliability and relative operating characteristic areas. Pros and cons are suggested for the metrics, and in some cases recommendations are given for improving the usability of the metric's information for those who use the models and their predictions. In addition, we examine the ability of these dynamical prediction tools to capture aspects of the climate variability of particular interest to society, such as regional temperature trends and multi-year regional drought.

We find that although models vary in their performance metrics, there is no clear preference between coupled models and atmosphere-only models. Further, as found in many previous studies, a simple combination of many reasonable quality models improves most performance metrics relative to any individual model. The one aspect of the forecast degraded by the multi-model combination is sharpness of the probabilities, most likely owing to our simple and un-calibrated combination of the models. A surprising finding is that although the skill metrics would suggest temperature is better predicted than precipitation, the models predicted the persistent drought conditions of the southwest North America much better than the upward temperature trend, suggesting the potential importance of missing sources of climate variability in these seasonal prediction models.

Skill assessment exercises such as this can be applied to other regions, and updated when new or improved prediction tools become available. In addition to providing information on expected prediction skill to those who might use the predictions for climate risk management, the assessments provide a valuable baseline against which improvements to prediction tools and methodologies can be tested.

Prospects for seasonal prediction of the North American monsoon
David Gutzler
University of New Mexico, USA

Like its more intense counterparts on other continents, the North American monsoon exhibits a high degree of seasonal regularity and extremely irregular interannual variability. The recent CLIVAR/GEWEX NAME program (North American Monsoon Experiment) was designed to improve the observed description of the monsoon, providing enhanced monitoring in and near complex terrain, and improved numerical simulation capabilities. The ultimate goal of NAME is improved seasonal prediction skill, which currently is very low. Some empirical predictors of regional monsoon variability are discussed, and two coordinated atmospheric model assessments (called NAMAP and NAMAP2) are reviewed. We are eager to share results with ongoing CLIVAR efforts to improve monsoon predictions on other continents.

Impacts of Ocean-Atmosphere Interactions and Climate Change on the North American Monsoon
Ruth Cerezo Mota
University of Oxford

The North American Monsoon (NAM) is a regional scale convective phenomenon that affects not only the northwest portion of Mexico but the southeast of the United States (Stensrud et al, 1997). It contributes to 60-80% of the annual precipitation of the NAM core region (Douglas, et al, 1993). Agricultural practices are traditionally linked to the annual cycle while the regular warm, moist, cool and dry phases of the monsoon seem to be ideal for the agricultural societies. Small variations in the rainfall timing and quantity could mean low crop yield and too much rainfall may produce devastating floods (Webster et al, 1998). Therefore an appropriate forecast of the beginning and intensity of the NAM is a priority for a country such as Mexico since the two main producers of crops of Mexico (source: INEGI, Statistics, Geographic and Computer National Institute) are located within the NAM region.

In this work we implemented the PRECIS (Providing Regional Climate for Impact Studies) model within the NAM region. PRECIS is a regional atmospheric and land surface model. After one year of simulation we found that PRECIS captures reasonably well the mesoscale circulation of the region in comparison with other high resolution models. We also compare the model precipitation with observed data. Observed data was obtained from the COOP meteorological net over USA.

After PRECIS validation, in order to determine the impact of sea surface temperature (SST) in the NAM, we included a linear time-relaxation of the heat fluxes. The SSTs would no longer be prescribed, thus there would be exchange via the heat fluxes. Preliminary results of this code modification will be presented and we expect that this modification will create differences when compared with the standard version of the model.

EUROBRISA: A EURO-BRazilian Initiative for improving South American seasonal forecasts
Caio Coelho
Centro de Previsão de Tempo e Estudos Climáticos (CPTEC)

The South American continent is located in a region particularly benefited by El Nino-Southern Oscillation (ENSO) atmospheric teleconnections in terms of seasonal predictability. Mainly because of ENSO teleconnections there is potential skill in seasonal forecasts for some regions of South America. Good quality seasonal forecasts are fundamental for local governments to plan their actions in order to minimize human and economical losses that may be caused by anomalous climate events such as ENSO. In South America these forecasts are useful for civil defence, agriculture, fishery, health and water resources (reservoir management) planning. Brazil, the largest and most populated country of South America, produces more than 90% of its electricity with hydropower stations, which are administered by the government. This emphasizes the need of good quality rainfall seasonal forecasts. EUROBRISA is a multi-national initiative, involving European and Brazilian climate researchers, that aims to provide South American local governments with improved and well-calibrated probabilistic seasonal forecasts. In EUROBRISA, empirical and dynamical coupled model rainfall seasonal forecasts for South America are currently being combined and calibrated using a Bayesian approach know as forecast assimilation. Empirical forecasts are produced using a simple multivariate regression model that uses the previous month Pacific and Atlantic sea surface temperature as predictors for South American rainfall of the following three month season. Dynamical forecasts are from the European Seasonal to Inter-annual Prediction (EURO-SIP) coupled multi-model system composed by ECMWF, UK Met Office and MÈtÈo-France, in addition to the recently developed CPTEC coupled model. EUROBRISA envisages to produce improved quality seasonal forecasts for South America, by objectively combining and calibrating empirical and multi-model dynamical forecasts into a single probabilistic forecast that gather all available information at the time the forecasts are issued. This presentation will illustrate preliminary results of rainfall seasonal forecasts for South America produced by EUROBRISA together will probabilistic skill assessment measures that indicate regions where forecasts have best quality. EUROBRISA real-time forecasts are intended to become operational in the second half of 2007, producing realiable probabilistic rainfall estimates for South America. The societal value of these forecasts is planned to be demonstrated by integrated activities involving end-users of seasonal forecasts.

Climatological features of SAMS based on SMIP simulations
Iracema F.A. Cavalcanti
CPTEC/INPE
Fernanda Cerqueira (CPTEC/INPE) Alice M. Grimm (Universidade Paraná)

The systematic errors of a climatological simulation using the CPTEC/COLA AGCM with resolution of T62L28 were previously identified, showing some deficiencies over South America. Typical features of the South America Monsoon System (SAMS) were also identified in that simulation, as well as interannual variability of precipitation in several areas of the region. That simulation consisted on a long-range run of 10 years using the observed SST as boundary condition. Other long-range run of 50 years with this model presents similar errors. The purpose of this study is to evaluate the model response considering the seasonal simulation mode of SMIP, analyzing mainly the SAMS region. The analyzed season is DJF considering the 1st and 2nd season potential predictability. The climatological features and systematic errors are evaluated considering observational data from NCEP/NCAR reanalysis data for atmospheric circulation and CMAP/CAMS ground observed precipitation. A multi-model analysis is performed taking other SMIP models with the same resolution of CPTEC model to assess the improvements over a single model results.

Seasonal Trend of Maximum and Minimum Temperature in Argentina During the Period 1961-2000
José Luis Stella
Servicio Meteorologico Nacional de Argentina

The object of this work is to analyze the deportment of the seasonal trend (autum, winter, spring and summer) of the maximum and minimum temperature in Argentina. Nowadays is very common to hear about the global warming, but this trend is really observed all over the world? As a general result of this study it could be observed a different deportment between both temperatures in a regional way, while taking each parameter alone, they show a similar trend in all the seasons. To complement this work, monthly trends have been analysed too, where it could be found some singularitys to emphasize.

Since 1980, transmission of El Nino-Southern Oscillation (ENSO) signals into the Indian Ocean invloves an equatorial, and a subtropical North Pacific (NP) Rossby wave pathway. We examine the robustness of the amount of energy that leaves the Pacific via each of the pathway using the Simple Ocean Data Assimilation with the Parallel Ocean Program (SODA-POP) reanalysis and a multi-century coupled model control experiment. We find that in the pre-1980 period, little ENSO signal is transmitted to the Indian Ocean and does not involve the subtropical NP pathway. Such multidecadal variability is periodically produced by the climate model. Examinations reveal that when ENSO is weak as determined by Nino3.4, their meridional extent is narrow, the associated discharge-recharge does not involve the subtropical NP pathway; further, weak ENSO events have a low signal-to-noise ratio, making the transmission hard to detect. The dynamics of multidecadal variability.

The stochastic simulation of boreal summer time ISO for the post-processing of dynamical seasonal prediction
Jin Ho Yoo
ICTP
Andrew W. Robertson (IRI), In-Sik Kang (CES/SNU)

The Asian summer monsoon exhibits pronounced variability on various time scales. Among them, an intraseasonal oscillation (ISO) with a period of 30-60 days is a key phenomenon determining seasonal mean weather statistics, which should be a next target of seasonal prediction system beyond seasonal mean value. Although remarkable improvement of seasonal prediction system has been made, the dynamical seasonal forecasting of summer monsoon and ISO is a still difficult task due to the uncertainties in physical parameterizations. To overcome the limitation of dynamical models, the statistical methods are often adopted and combined with dynamical seasonal prediction. Also statistical models are used for spatial and/or temporal downscaling of seasonal prediction by GCMs. In this study, a simple but flexible probabilistic model for intraseasonal variability of summer monsoon rainfall is developed. The modeling framework is provided by a Hidden Markov Model (HMM), in which a latent state variable (hidden state) is introduced to enable a simplified factorization of the probability distribution function (PDF) of pentad rainfall. The HMM selects consecutive phases of ISO as distinctive states and the transition probabilities among the states clearly support a cyclic transition of ISO phase. This stochastic model is combined with dynamical model forecast by allowing changes of transition probabilities with respect to the input variable derived from seasonal mean forecast of GCM. In the stochastic simulation, the canonical ISO propagation and the level of irregularity are reasonably represented by HMM up to 20 days time lag. And the simple model also shows comparable skill with GCMs for the seasonal mean precipitation field. The sensitivity of selecting input variables and predictability of ISO will be also discussed.

The performances of ten coupled atmosphere-ocean models in their retrospective predicting annual modes of precipitation and rainy season characteristics are assessed by using one-month-lead hindcast products for the period 1981-2001. [To compare with the coupled model predictions, five uncoupled two-tier modelsí predictions were also evaluated. The current coupled models predict the annual mean and the first annual cycle (AC) mode reasonably well, while they have difficulty in capturing the 2nd AC mode faithfully, especially over the Indian Ocean and western North Pacific (WNP). The uncoupled models have large biases in the leading AC mode over the WNP where the atmosphere has significant negative feedback to the ocean. Over the WNP region, most coupled models capture the mean and AC more realistically than the uncoupled models do; but the coupled models tend to underestimate the precipitation amount and the interannual variability, which degrades seasonal prediction skills. In the Asian-Australian sub-monsoon domains, even though the coupled models have difficulty in capturing climatological intraseasonal variation, they can predict climatological onset and withdrawal dates realistically, especially over the Indian monsoon region. Using quantitative measures of the performance in the annual modes, we show that the skills in predicting seasonal anomalies are positively correlated with their performances on both the annual mean and annual cycle in the coupled climate models. Improvement of the model climatology has a significant positive impact on the models' skill in seasonal forecast.

The Indian summer monsoon exhibits prominent 30-40 day fluctuations with “active periods of heavy rain interrupted by dry periods i.e. “Breaks”. The circulation anomalies associated with active/break monsoon cover the entire Indian Ocean influences remote tropics and North Pacific Ocean. A prolonged dry/wet period will result in severe drought/flooding, which have profound influences on the south Asia water cycle, agriculture and societal activity of over more than one billion people. The atmospheric general circulation models have great difficulty in simulating the Intra-seasonal oscillation (ISO). Therefore, it is necessary to study the empirical relationship between various atmospheric processes, which are responsible for the ISO. In this paper, the analysis of North Atlantic Oscillation Index (NAOI) and Madden Julian Oscillation Index (MJOI) on daily scale is carried out in relation to daily Indian summer monsoon rainfall (June-September). The analysis is carried out for period 1979-2001. Since the potential predictability limit for monsoon break is about 20 days, the 20 days running lag/lead correlation analysis between the NAOI and MJOI is found out for each year. It is observed that 20-day lag relationship between NAO and MJO is inverse and significant (0.1 level) and this relationship remains negative throughout the break monsoon period and in active phase it reverses. This twenty days lag relationship between NAO and MJO is potential predictor for break/ active monsoon condition over Indian region. The analysis is verified for major drought year 2002 and it is observed that prolonged break in July is fairly well predicted. In this paper, the simultaneous effect of North Atlantic Oscillation (NAO) and Southern Oscillation (SO) on monsoon circulation over Indian subcontinent is also studied. The simultaneous effect of NAO and SO on Indian summer monsoon rainfall (ISMR) is more important rather than their individual impact. To represent the simultaneous impact of NAO and SO, an index called effective strength index (ESI) has been defined based on monthly NAO and SO indices. The variation in the tendency of ESI from January through April has been analyzed and its use in the prediction of ISMR is examined. The analysis shows an inverse association of the tendency of ESI from January through April with summer monsoon rainfall over India as a whole. The simple linear regression model is developed, using this ESI tendency, for predicting summer monsoon rainfall over India. The performance of this equation is tested and it reveals that this tendency can be used as a pre-cursor for the prediction of monsoon rainfall over India as a whole. During 2002, this parameter has indicated reduced rainfall activity over India.

Probable physical linkage for using this relationship for predicting the dry/wet spell in monsoons period could be like this, previous studies have established that the active/break monsoon are triggered by organized northward propagation of heavy precipitating or cloud free zones from the equatorial regions towards the continental land mass. This northward propagating mode (NPM) over the Indian monsoon region was weaker (stronger) during drought (flood). It is also known that the MJO interacts with the NPM in the Indian monsoon region; it is plausible that the different phases of NAO modulate the temperature gradients cold/warm in upper troposphere. Hence, it might be responsible for the increase in the period of the NPM. A longer period NPM could possibly lead to longer monsoon break periods causing the major drought condition and vice versa

Inter-annual variability of Indian summer monsoon rainfall (ISMR) has a profound influence on agriculture and national economy. In spite of an increase in share from the service sector to India's growth, the performance of the agricultural sector is a decisive factor to the growth of Gross Domestic Product (GDP) of India. Therefore, seasonal forecast of ISMR becomes very crucial. India Meteorological Department has been issuing seasonal forecasts of ISMR using statistical models for many years. Recently, attempts also have been made to prepare experimental forecasts using dynamical models (AGCM). However, our experience with dynamical models is not very inspiring and satisfactory (Gadgil and Sajini 1998, Sperber and Palmer, 1996, Kang et al. 2002 and Wang et al. 2005). Therefore, till the dynamical models improve to meet our expectations, statistical methods are to be used for preparing operational forecasts of ISMR. At the India Meteorological Department, continuous efforts are made to develop and improve state of art statistical models for operational forecasts. However, statistical methods have inherent problems and limitations. One of the serious limitations is the secular variations of the predictor-rainfall relationships. Exact treatment of non-linearity of the relationships also poses a serious problem. Recently, new statistical models have been developed using a new training methodology (Rajeevan et al. 2006). An innovative method for the identification of the best predictor data set was also used. New approaches like Ensemble Multiple Regression (EMR) and Projection Pursuit Regression (PPR) have been also adopted. The new statistical models showed better performance, compared to the model based on climatology. These models were able to provide correct forecasts of the recent two deficient monsoon rainfall events (2002 & 2004). The experimental forecasts for the 2005 and 2006 seasons based on these models were also found to be accurate. The statistical models are based on different global predictors, which are ultimately interlinked with the ENSO phenomenon. However, the relationship between ENSO and ISMR is not perfect. Only less than half of El Nino events are associated with deficient rainfall over India. In other El Nino years, ISMR was either normal or excess. In a recent study, Krishna Kumar et al. (2006] analyzed the observations as well as atmospheric model simulations and suggested that the El Nino events with the warmest SST anomalies in the central Pacific (such as 2002) are more effective in focusing drought producing subsidence over India than events with the warmest SST anomalies in the eastern equatorial Pacific (such as 1997). It turns out that there are major exceptions to this association. For example a severe drought occurred in 1972, although the SST anomaly pattern of 1972 was similar to that of 1997. Krishna Kumar et al. [2006] have further suggested that the incorporation of additional information on the spatial distribution of SST anomalies over the equatorial Pacific Ocean in the statistical models should improve monsoon forecast skill. This hypothesis on the skill of monsoon forecasts was examined by analyzing SST data for the period 1880-2004 (Rajeevan and Pai 2007). The study reveals that incorporation of information on the spatial pattern of SST anomalies (by incorporating the Trans Nino Index) does not improve the association between El Nino and Indian monsoon rainfall. Simply using the SST index over the central Pacific (Nino- 3.4) may be a better indicator for the association than the Combined Nino Index derived from Nino-3 and Trans Nino Index (TNI). The study concluded that there are still some unresolved issues on the El Nino-ISMR relationship. References: Gadgil, S. and S.Sajini, 1998, Climate Dyn, 14, 659-689. Kang, I-S., et al., 2002, J.Climate, 15, 2791-2805. Krishna Kumar, K. et al. 2006, Science, 314, 115-119. Rajeevan,M., D.S.Pai, R.Anil Kumar and B.Lal, Climate Dynamics, 2006, DOI 10.1007/s00382-006-019706. Rajeevan, M., and D.S.Pai, 2007, Geophys.Res.Letters, 34, L04704, doi 10.1029/2006GL028916 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.

We study the contemporaneous relationship between the intensity of the Indian Summer Monsoon (ISM) and runoff in the major rivers of the Aral Sea basin (Amudarya, Syrdarya) and some of their subcatchments. To this end, we use All-India rainfall (AIR) data, CRU surface observations of precipitation and temperature, ERA40 atmospheric data, and natural discharge data corrected for human interference. We show that there is a highly significant positive correlation between ISM intensity and Amudarya runoff. This finding cannot be explained by the spill-over of ISM precipitation over the Hindu Kush into the Amudarya basin. Instead, we suggest that the observed co-variability is mediated by tropospheric temperature variations due to fluctuations in the ISM intensity. These variations are known to be due to Rossby-wave propagation in response to condensational heating during monsoon precipitation. We hypothesise that the corresponding anomalies in surface temperatures imply anomalies in meltwater formation. While the focus of the study is a better understanding of the natural climate variability in Central Asia, it is conceivable that the link we describe can be applied in order to add further skill to seasonal forecasts of summer runoff in the Aral Sea basin.

Seasonal Predictions of the Indian Summer Monsoon using the NCMRWF Global Modeling System
Sarat Kar
National Centre for Medium Range Weather Forecasting (NCMRWF), India
A. K. Bohra, NCMRWF, India

The NCMRWF global modeling system has been used to carry out multi-ensemble simulations of the Indian summer monsoon. Hindcast simulations have been carried out for each monsoon season from 1982 to 2004 using observed SST (SMIP type), persistent SST anomalies (SMIP-HFP type), and using climatological SST. For each case, 12 member ensemble runs have been made with initial conditions in April and May of each year. The model simulations have been carried out till end of September for each year. Model simualtions show that the model used has a reasonably good summer rainfall climatology. The interannual standard deviation of rainfall over the Indian region is comparable to that of observed rainfall. However, with climatological SST also, the model simulates interannual variability comparable to the run forced with observed SST. The model does reasonably well in simulating anomalous rainfall over India during 2002, 1997, 1994, 1988 and 1987, when the model is forced with observed SST. However, temporal correaltion of model rainfall with observation over the Indian region is low (0.3-0.4). With persisted SST anomalies, the skill goes down further. With climatological SST, the correlation over India becomes negative. This suggests that eventhough the model has large variability due to inernal dynamics, ensemble members of the model integrations tend to produce rainfall which are more realistic when forced with observed SST than in the climatological SST. Systematic errors in the model runs have been identified using EOF analysis of the model and observed rainfall. Several interesting features come out when the leading modes are examined. A probabilstic seasonal prediction system has been developed assuming that the ensemble mean has the signal component and the ensemble spread is the stochastic noise having normal distribution. It is seen that reliability of seaosnal predicions can be improved, if the probability values are properly calibrated.

In this work it is shown that using what we call a 1.5-tier approach leads to considerably improved Indian Summer Monsoon rainfalls (IMR) hindcasts as compared to the original coupled model hindcasts from the DEMETER project. The expression 1.5-tier indicates that the DEMETER sea surface temperature (SST) anomalies are used everywhere outside the Indian Ocean region to force an Atmospheric General Circulation Model (AGCM), whereas inside the Indian Ocean, a fully coupled OAGCM is used to simulate the effect of air-sea coupling. Further experiments are devised to perform a comparison with a pure 2-tier approach, where DEMETER SST anomalies are prescribed everywhere.

Probabilistic forecast of All-India monsoon rainfall on the basis of monthly rainfall
Dilip Kothawale
Indian Institute of Tropical Meteorology
A. A. Munot, Indian Institute of Tropical Meteorology

The distribution of southwest monsoon rainfall in space and time is very important for country's agriculture production, industries and generation of hydroelectric power. Timely and spatially well distributed rainfall gives good amount of food grain production, otherwise the food grain production is adversely affected. It is very much essential to know how the performance of monsoon rainfall will be well in advance to take precautionary measure in case of abnormal monsoon. Generally, the abnormal performance of monsoon is seen from June, the starting month of monsoon itself. In view of this, an attempt has been made to understand what may be seasonal (June+July+August+September) rainfall of All-India and subdivisions when monthly rainfall is reported as excess or deficient. For all-India and 30 meteorological subdivisions, the probabilities of occurrence of excess, deficient and normal seasonal rainfall are worked out when each of monsoon months viz. June, July, August and September rainfall reported to be excess or deficient, using long homogeneous monthly and seasonal rainfall series of 135 years (1871-2005). The seasonal rainfall is also quantified when monthly rainfall is excess or deficient for All-India and 30 sub-divisions for entire period 1871-2005. When the rainfall of any monsoon month is excess for any subdivision or All-India as whole, the probability of seasonal rainfall to be deficient is almost nil and seasonal rainfall is either excess or normal., If deficient rainfall occurs in any monsoon month for any sub-division then the season's rainfall may be either normal or deficient but not excess. All-India June, July and August rainfall is reported as deficient, the probabilities of occurrence of seasonal rainfall to be deficient are 34%, 66%, 33% respectively. It is also seen that during the period 1871-2005, in any sub-division, most of the years' seasonal rainfall was above/below it's long term mean rainfall when rainfall of any of the monsoon month was reported as excess/deficient.

The Asian Monsoon variability and predictability are examined from the outputs of the WCRP CMIP3 (World Climate Research Program Coupled Model Inter-comparison Project-3) multi-model dataset. Over South Asia only 6 of the 23 models are able to generate the most realistic 20th century monsoon climate. The South Asian monsoon variability is examined with respect to the model-simulated heat low, pressure gradient, monsoon circulation, Eurasian Snow Cover and the El Nino Southern Oscillation phenomenon (Kripalani et al. 2007a). Most of the models are able to simulate inter-annual and spatial monsoon variability over East Asia reasonably well. The variability over this region is examined with respect to the monsoon circulation, North Pacific Subtropical High and the Meiyu-Changma-Baiu frontal zone (Kripalani et al. 2007b). Over South East Asia the variability is examined with respect to the Indian Ocean Dipole Mode and the El Nino Southern Oscillation phenomenon (Kripalani et al. 2007c) Analyses reveal that the predictability is highest over the Southeast Asian domain and the least over the South Asian domain. References: Kripalani RH, JH Oh, A Kulkarni, SS Sabade and HS Chaudhari 2007a: South Asian summer monsoon precipitation variability: Coupled climate model simulations and projections under IPCC AR4. Theoretical and Applied Climatology, Published Online DOI 10.1007/s00704-006-0282-0 Kripalani RH, JH Oh and HS Chaudhari 2007b: Response of the East Asian summer monsoon to doubled atmospheric CO2: Coupled climate model simulations and projections under IPCC AR4. Theoretical and Applied Climatology 87,1-28. DOI 10.1007/s00704-006-0238-4 Kripalani RH, A Kulkarni and SS Sabade 2007c: South-east Asian monsoons and its connections with IODM and ENSO in the WCRP CMIP3 multi-model dataset: Simulations and Projections (under preparation).

North African climate variability prediction
Abebe Yeshanew
National Meteorological Agency, Ethiopia

Tropical North Africa depends on rain-fed agriculture as the main economic driver. The variability of climate-sensitive resources is investigated with a goal to develop statistical long-lead prediction models with reasonable skill. Climate data from NCEP is analysed in conjunction with agricultural and economic production in various sectors, in addition to the traditional climatic indices: temperature and rainfall. Key predictors for statistical models include the lower-level zonal wind over the Atlantic and Pacific Oceans. These exhibit a 'memory' that is consistent with sea surface temperatures (SST) through equatorial upwelling dynamics. Kinematic predictors outperform SST in hindcast fit by an average 33% with respect to various tropical north African resource indices. A multi-decadal oscillation induces long-term trends in rainfall that contribute to apparently skilful forecasts based on the interaction of Pacific ENSO and the Atlantic zonal overturning circulation. The skill of statistical forecasts is lower when the drying trend is removed.

Changes in the Frequency and Intensity of Extremes over NorthEast Africa
Gebru J. Endalew
Wageningen University
B. van den Hurk and G. J van Oldenborgh, KNMI

Northeastern Africa (with lat.3^o-15^oand long. 33^o-48^o) climate varies from humid to semi-arid with abundant and scarce moisture. Hence, flooding and drought are frequent phenomena which have a direct impact on the agriculture, health, water and other socio-economic sectors of the region. Analysis for the expected changes in the frequency and intensity of extreme rainfall induced by climate change is the aim of this study.

The area has been divided into three homogeneous rainfall regimes (Zone A-C).The seasonal classification of the region, especially over Ethiopia, is from February to May, June to September and October to January, called Belg, Kiremt and Bega, respectively. Here, more emphasis is given to the Kiremt (JJAS) seasons.

Anomalous wet and dry decades were identified using ERA-40 data. The condition of the main rain bearing systems during these decades has been assessed and from this result systems which favor more for the extremes have been detected. Among the systems that control the weather activity in the region are: ITCZ, TEJ, EALLJ, ENSO as well as Mascarena, St. Helena, Azores and Arabian High pressure systems. Then using an ensemble of ECHAM5/OM, simulations were analyzed.

During JJAS season, 90-96% wet anomalous were mostly situated in the 1958-1979 time range. The dry anomalous decades were mostly found in 1980-2001 for both zones.

Further results relating to the dynamical conditions affecting the extreme precipitation occurrences were:

-For the wettest decades, the Mascarena and St. Helena high pressure systems have positive anomaly during the onset and cessation periods (June and September) and the means of transport ( the jet ) has a positive anomaly most of the time.

-For the driest decades, the Mascarena and St. Helena high pressure systems show positive anomaly most of the time. However, the jet has negative anomaly.

-For most of the wet decades the centre of the Mascarena high pressure system is centered over the ocean 30^o-50^oE. However, for most of the dry decades, it was either stretched east-west or centered over the land.

-During July and August, the ITCZ reaches an extremely northern position and has more convergence accompanied by a formation of a heat low over the Arabian and Sudan area. This contributes to the incursion of moisture from the Congo Basin. The opposite occurs during dry spells. As the heat low formed over the Sudan area shifts to the west and is over Chad. Besides, there is a formation of cyclone over the Gulf of Guinea, which affects the incursion of moisture towards the area.

-Also the TEJ is shifted to the equator (south of 10.5^oN) compared to dry decades.

The 3-month lag Nino 3.4 condition during most of the season show negative to neutral and positive to neutral condition for most of the wettest and driest decades, respectively.

Coupled SST-West African Monsoon precipitation patterns in AGCM simulations
Elsa Mohino
UCM
Teresa Losada (UCM), Belen Rodriguez-Fonseca (UCM), Carlos Roberto Mechoso (UCLA)

Our project aims to study the seasonal predictability of precipitation in the West African Monsoon (WAM) region using atmospheric general circulation models (AGCMs). At this stage we are investigating the potential contribution to seasonal predictability of WAM precipitation by sea surface temperature (SST) anomalies in different ocean basins. We use the results of WAM simulations by several AGCMs that have been run with observed SSTs for the period 1950-2000 as part of an intercomparison exercise in the framework of EU AMMA project. The presentation will focus on the results of Extended Maximum Covariance Analysis (EMCA) between the WAM precipitation simulated by the models and lagged monthly mean SSTs from different basins (Atlantic, Pacific and Indian) for the enhanced observational period 1979-1998. The results are compared with the same analysis done with the observational dataset given by CMAP.

Sub-Saharan Africa is a continent where human activity is closely linked to inter-annual rainfall variability. There is growing evidence of the benefits that seasonal forecasts would provide for the management of the consequences of climate fluctuations in the region, notably for the control of vector-borne diseases such as malaria. This study undertakes a meteorological verification of rainfall forecasts over West Africa using retrospective forecasts available from the EU DEMETER project for the period 1980-2001. Probabilistic forecasts of rainfall totals for three precipitation thresholds are produced for the summer rainy season (June-July-August) using model outputs from May, corresponding to forecasts with a 2-4 month lead-time. The skill of those forecasts is then computed on the forecastsí anomalies using the Brier and ROC skill scores with both CMAP and ERA-40 used as reference climatology. This analysis reveals positive skill scores in some areas of West Africa, particularly for years with high precipitation, signifying that over many realizations the DEMETER forecasts are better than a forecast based on climatology. This assessment has also showed how sensitive skill scores are to the reference climatology. Given the large discrepancies obtained between the gridded and reanalysis datasets, further analysis was performed on ground-based weather stations at a number of Sahelian grid points. This paper concludes with a number of comments on the potential applications of those forecasts by the public health community in view of the current forecasting capabilities.

Climate Extremes over the Equatorial East Africa and their Predictability with the Indian-Ocean Dipole Mode
Zablone Owiti
Department of Meteorology, University of Nairobi

This study was devoted to the improvement of prediction skill of east African seasonal rainfall through the use of improved knowledge of the teleconnections between Indian Ocean Dipole (IOD) evolution phases and the East African rainfall variability. The specific objectives of the study included determination of teleconnections between east African seasonal rainfall extremes and space-time evolutions of the various IOD phases; investigating the physical mechanisms of the teleconnections; and determination of the predictability potentials of IOD indices in order to enhance our understanding of interannual variability and advancement of climate prediction skills over sub-region.

Statistical and dynamical methods were adopted in the study. The statistical methods used in the study include standardization of anomaly indices; together with correlation, regression and composite analyses. Space-time characteristics of a series of dynamical parameters were investigated using products from ECHAM4.5 general circulation model (GCM) and the NCEP/NCAR reanalysis to ascertain the physical reality of the IOD-seasonal rainfall linkages derived from the statistical approaches.

We show that the pattern in SST anomalies reflected by the IOD indices affects the climate system during the OND season. It may be concluded from this study that the linkages that were observed from the statistical analyses had some physical reality. Such information will help to improve monitoring, prediction and early warning of extreme rainfall events over east Africa, reduce the vulnerability, and improve the resilience of the society of the region to negative impacts of extreme rainfall events that are common in the region.

Some critical references for the study included: Reason, CJC et al 1996, Saji, NH et al., 1999, Saji, NH and Yamagata, T 2003a and 2003b, Vinayachandran, PN S. et al, 2002, Webster, PJ 1981, Webster, PJ., et al 1999, Yamagata, TSK. et al, 2002, 2003

Future climate knowledge, on seasonal and interannual scales is required for economic and social activities. The identification of the predictors of the climate for a given land area is one important field of investigation in meteorology. Currently, adapted statistical techniques are provided for the analysis of the Oceanic and atmospheric links to land climate. The objective of this paper is to identify the Oceanic and atmospheric variables that are more significantly governing the anomalies of the precipitation on the Congo basin in the Central Equatorial Africa.We use a combined transfer function and wavelet transform approach for detecting the time dependence coherence between oceanic and atmospheric variables and the land rainfall during the last 55 years (1951-2006) for this region. Interannual cycles with a period of approximately 3 years are identified in the northern Congo basin rainfall. The dominant frequencies in the data are detected using the methods of periodogram and maximum entropy spectral density with smoothing splines and singular spectrum. Using cross-wavelet phases analysis, the interannual cycles of 3 years in the rainfall coincide with those present in NINO3 SST3 (2.7 years) for the same time period, suggesting a possible association between the sea climate and the northern Congo rainfall. The cross-wavelet coherency analysis reveals that the rainfall are significantly coherent with NINO3 Sea Surface Temperature and with NAO at periods around 3 years for NINO3 and 0-2 years with NAO; NINO3 and NAO leading the rainfall. The rainfall is also significantly coherent at the seasonal scale, with the oceanic (NIÑO3) and atmospheric (NAO) variables, these variables leading the rainfall. The results of cross-wavelet coherence and phase are consistent with those of the cross-correlation approach. The seasonal dynamics of the rainfall are led by NINO3 with a 3-4 months lag with significant negative correlation (R2=73), and by NAO with a 1 month lag with significant negative correlation (R2=65). From 1950 to 2005, the monthly average variance trend in the NINO3 SST and NAO are both similar with the monthly average variance trend in the northern Congo basin rainfall. Both variance was increasing for the periods 1950-70, 1975-85, 1995-2005, and decreasing for the periods 1970-75 and 1985-95. Finally, combining these findings, we modeled a transfer function for the Rainfall forecast for this region, with NINO3 as exogenous Oceanic variable and NAO as exogenous atmospheric variable. From the identified transfer function, an equivalent Dynamic Ocean-Atmosphere-Land climate model is formulated in state space form and is described. From large scale atmospheric circulation, NINO3 and NAO govern the local and regional wind anomalies, and the rainfall is dependent with the local wind. Our planned future works need to well understand the mechanism from which the wind variable seem to be the main intermediate variable which regulate the transfer of the water vapor between the Ocean and the Congo basin region, using combined lags correlations, cross wavelets phases and cross

South Africa shares a number of river basins with its neighbours. Owing to increased water demands in the region the water in these rivers is increasingly under pressure. A multi-model forecasting system is presented that downscales simulations from two GCMs, the ECHAM4.5 and C-CAM, to streamflow of the Limpopo, Incomati and Maputo river basins. Simulation data from these GCMs are used to test the potential of a model output statistics (MOS) approach to forecast seasonal streamflow into the catchments of these basins. Preliminary results from the ECHAM4.5-MOS system show that useful skill is found for DJF for the South African catchments adjacent to Mozambique: Retro-active area-averaged ensemble mean simulations over a 40-year period produce a correlation of 0.38. In this paper, simulation skill, when combining simulations of the two GCM-MOS systems and the output from a statistical model that uses SST as the only predictor, is presented.

Multi-Model ensemble inter-seasonal to seasonal forecasting and probability outcomes in prediction system
Nana Ama Kum Browne
Climate System Analysis Group, ENGEO Department, University of Cape Town

Multi-model ensemble forecasting aims to improve individual forecast systems and provide estimation of the reliability of the forecast. Seasonal forecasting as well as decadal and multi-decadal climate prediction are of economic and public interest; value in such forecasts may be found in planning the future in many sectors that depend on the climate and may include the agricultural management, infrastructure plans, energy sectors, etc. It has therefore become necessary for an increased understanding of inter-seasonal to seasonal climate variability and the need for reliable forecasts systems.

Three atmospheric models CAM2, CCAM and HadAM3 will be considered. 30-year model runs will be made for all models with 2.5 degree resolution over Southern Africa. Observed Sea Surface Temperature forcing will be used in generating the ensemble members for each model. However, constructing ensembles by using perturbed initial conditions as input to a single model as commonly employed carries the implicit assumption that all of the uncertainty resides in the initial conditions. This cannot be true because the nature of the climate at these time scales is only quasi-deterministic, while also being a boundary condition problem, the quality of the projected sea surface temperature used in forcing the models is equally important. Consequently, multi-model ensemble members are normally generated to capture the envelope of possible time-evolutions of the climate. These are generated from all the global models leading to the extended multi-model ensemble forecasting.

The prediction skill of each model will be measured in hindcast mode by comparing the model predictions to observed data. The quality of the hindcast ensembles estimated using skill measures such as the brier skill score. Building on this, the thesis will seek to develop an improved method of using the multiplicity of the ensemble data to enhance the value of the derived seasonal forecast.

Regional Outlook Forums in Africa
Leonard Njogu Njau
ACMAD

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 climate related events. The current African Centre of Meteorological Applications for Development (ACMAD) diagnostic tools for climate prediction and early warning include determining the linkages between the seasonal climate and evolution of monsoons, tropospheric energy indices, medium and upper level winds, North Atlantic Oscillation (NAO), Madden-Julian Oscillation (MJO), Quasi biennial oscillation (QBO), El Nino Southern Oscillation (ENSO), Indian Ocean Dipole (IOD), Ocean basins Sea Surface Temperatures (SSTs) anomalies and interactions between large and local scales circulations. The recent development of new diagnostic prediction tools for seasonal climate prediction and early warning has to some extent improved the climate related socio-economic activities in several parts of the subregions in the Continent. The West Africa subregion has in recent years become highly vulnerable to extremes climate events compared to other parts of the globe. The extreme rainfall events such as droughts and floods have occurred from time to time with devastating impacts on various socio-economic activities and livelihoods in the subregion. Such impacts have been associated with food shortages, famine, lack of energy and water among other major basic needs resulting in loss of life and property and many other socio-economic disruptions. The economies of West African countries largely depend on agriculture, which is highly vulnerable to the rainfall onset, duration, amounts and distribution. The studies on the interannual variability patterns on rainfall in the subregion indicate that the coherence in rainfall variations is greater at some periods of the year. For example the July-September rains are associated with large-scale climate anomalies, which are thus more predictive than those of the April-June, but are not homogeneous in both time and space over several parts of the subregion. Accurate monitoring, prediction and early warning of seasonal rainfall performance can be used to improve planning and management of various rainfall dependent socio-economic activities, and enhance the resilience and livelihood of the communities in the subregion. The occurrence of extreme climate events such as drought is a slow process that may takes months before causing acute food insecurity and famine. The impact is felt when there is no consensus among governments, NGOs, UN and other humanitarian agencies as to when and how to respond to an early warning on a developing climate event such as drought. This has often resulted in delayed interventions affecting the effectiveness of humanitarian aid. The seasonal climate forecasts for example on possible drought and food shortages should help governments and humanitarian agencies to plan ahead so that a timely humanitarian assistance could be provided. In an attempt to use climate information and forecast for food security, Climate Outlook Forums (COF) have been regularly organized for linking the Climate Outlook with food security outlook. The purpose of the Forums is to provide a consensus forecast for the region before each major rainy season. For West Africa countries the Climate Outlook Forum is the PRESAO while for Greater Horn of Africa (GHA) countries is GHACOF and for Southern Africa countries, the SACOF.

Combating Flood Crisis with Geographic Information System (GIS): An Example from Akure, Southwest Nigeria
Okuku Ediang
Nigerian Meteorological Agency
A.O Eludoyin, Department Of Geography, Obafemi Awolowo University, Ile-Ife, Nigeria O. M. Akinbode, Department of Geography and Planning Sciences, Adekunle Ajasin university, Akungba, Akoko, Ondo state, Nigeria

Flood is a natural environmental disaster which could be aggravated by man's unguided development. It may subsequently cause destruction of properties and loss of life. Therefore it needs to be controlled and human influences controlled. This study attempts to describe an application of GIS as decision support to flooding problems in an urban area in Nigeria. The objective of the study is to describe the efficacy of GIS in monitoring of development on floodplains in an urban area in Nigeria. Topographic features were digitised from an existing 1:5,000 topographic map of Akure, with some position data collected and map updated using a handheld GPS. A database was created using both cartographic and attributes data collected from these and other sources. Spatial analyses were carried out using a PC based Integrated Land and Water Information System (ILWIS), version 3.2. The results obtained implicated dumpsites within the river channel as well as structural development within the River Ala floodplain as the major causes of inundation in this section of the city, especially, in the wet season. The study concluded that GIS could provide adequate decision support information to policy makers.

Rainfall is the most important climate parameter in many part of Africa since the economy is based mainly on rain fed agriculture. Predicting the likelihood of rainfall with a reliable accuracy with a lead time of several months is therefore of great significance. In this study, a seasonal forecasting Method using Multiple Linear Regression (MLR) and Linear Discriminant Analysis (LDA) has been developed for Ethiopia using Sea Surface temperature (SST) as a predictor. Considering the spatial variability of rainfall over the region, homogeneous rainfall zones have been identified for a given season and forecasting has been carried out for each zone separately. This kind of localized forecast approach could be used for operational purpose in the National Meteorological Services Agency (NMSA) of Ethiopia. The forecasts made by the above techniques were compared with persistence and climatology. Two sensitivity experiments were done by including and excluding SSTs from the synchronous season to the Models. Results show that the performance of both methods [MLR and LDA] is generally better than Climatology or persistence. Including contemporaneous SST predictors was also observed to have a significant impact in some zones. To understand the link between the predictors (Remote SSTs) and Ethiopian Rainfalls two sets of composites have been analysed. The first set was composite of global atmospheric fields based on excess and deficit rainfall years to identify the large scale atmospheric parameters responsible for rainfall variability. The second set was also composite of global atmospheric fields but based on warm and cold SST of the regions used as predictors in the forecasting models. This analysis aimed to identify the link between the remote SST and the Ethiopian rainfall via large scale atmospheric features.

Optimizing the satisfaction of requirements generated by a multi-purpose dam, while minimizing its impact on the environment and on traditional human activities, in a context of low water resources, are the goals laid down for the Manantali dam on the Senegal river. Because of the reduced length of the rainy season, the water managers must decide by mid-August on the flood support level required over the September-October period in order to preserve flood recession crops, as well as hydro-electric energy production till the next rainy season. In such a framework, the seasonal forecasting of rainfall for the coming September and October is a crucial information for allowing the best decision. Accordingly, the IRD, Météo-France and the OMVS have developped a water management model which makes use of the seasonal forecast issued beginning of August from the ARPEGE climate model, adapted both in space and time for optimisation purposes. This application works since 2005 in an operational mode. The benefits of such an application are important in terms of economical, environmental and social points of view.

A series of Climate Outlook Forums (COFs) seasonal rainfall forecasts for sub-regions of Africa have been prepared since 1997. Global dynamical models and statistical tools are used in the seasonal forecasting process. COFs recommendations have stressed the needs for development of more local seasonal forecasts. Regional climate models can be tested for local seasonal forecasting over the region. This paper compares regional model simulations of recent climate of sub-Saharan with observations and analysis of observations. It is shown that the multi-decadal dry period over the Sahel, the mean temperature patterns and annual cycle are well reproduced by the regional model. More interestingly, the significant droughts of 1972,1983 and 1984 over the Sahel are well simulated by the regional model. The regional model does reproduce realistic inter annual variability of temperature over the Sahel and inter annual rainfall variability in Gulf of Guinea. The peak of the rainy season in August is quite realistic in the simulations. However, a warm bias is observed over the region during the hottest months of the year. Rainfall season begins one to two months earlier than normal in the model atmosphere. It is concluded that interpretation and use of the Hadley centre regional climate model for seasonal forecasting and climate projections should consider the strengths and weaknesses diagnosed over sub-Saharan Africa.

Verification of climate outlooks in Africa: proposed methodology for West Africa and prospects for all Africa
André Kamga Foamouhoue
African Centre for Meteorological Applications to Development

Since the beginning of Climate Outlook Forums (COFs) in Africa in1997, a comprehensive verification experiment has not yet been implemented. A growing number of potential users and donors are requesting evaluation of COFs products. Given the substantial investment provided to organize the past nine COFs in Africa, it is proposed to assess these COFs.

More importantly, a better interpretation and use of seasonal forecasting systems significantly rely upon regular verification. A set of metrics have been proposed by the European Commission funded project AMMA-EU for seasonal forecast verification over the monsoon region of West Africa. Implementation of these metrics requires data sets that are being or have been generated by other programs and institutions ( DEMETER, ENSEMBLE, IRI,..). This paper discusses the metrics and datasets to pave the way for exchanges on relevant collaborations needed to build an operational seasonal forecast verification network for Africa.

West African Weather System In The Development Of Tropical Cyclones
Tairu Salami
Nigerian Meteorological Agency
O.s Idowu E,E Balogun

Tropical Cyclones have their origins from areas of low atmospheric pressure over warm waters in the tropics or subtropics. We have have carefully studied the interconnection between the West African Weather Systems (WAWS) and their subsequent development into Tropical Cyclones. Between 2004 and 2005, we studied the interconnection and the teleconnection between the WAWS and the various occurrences of Tropical Cyclones and their eventual development into Hurricanes. We noted that critical synoptic characteristics and the environmental properties of the Systems;the thermodynamic conditions of the storms trajectory and the conditions of the ocean are all closely linked. It is therefore believed that proper understanding and monitoring of these systems will play a very vital role in early detection of potential WAWS that may develop into Tropical Cyclones and even Hurricanes. More practical issues will be presented. It was recorded that over the period 1992-2001, weather and climate-related disasters especially those of Tropical Cyclones origin killed about 622000 people, affected more than two billion, left millions more homeless, devastated arable land and spread diseases.