An interpretable machine learning model for seasonal precipitation forecasting

Seasonal climate forecasting is important for societal welfare, as it supports decision-makers in taking proactive steps to mitigate risks from adverse climate conditions or to take advantage of favorable ones. Here, we introduce TelNet, a sequence-to-sequence machine learning model for short-to-medium lead seasonal precipitation forecasting. The model takes past seasonal precipitation values and climate indices to predict an empirical precipitation distribution for every grid point of the target region for the next six overlapping seasons. TelNet has a simple encoder-decoder-head architecture, allowing the model to be trained with a limited amount of data, as is often the case in climate forecasting. Its deterministic and probabilistic performance is thoroughly evaluated and compared with state-of-the-art dynamical and deep learning models in a prominent region for seasonal forecasting studies due to its high climate predictability. The training, validation, and test sets are resampled multiple times to estimate the uncertainty associated with a small dataset. The results show that TelNet ranks among the most accurate and calibrated models across multiple initialization months and lead times, especially during the rainy season when the predictable signal is strongest. Moreover, the model allows instance- and lead-wise forecast interpretation through its variable selection weights.

Evaluating constraints on offshore wind farm installation across the Taiwan Strait by exploring the influence of El Niño-Southern Oscillation on weather window assessment

The transition to renewable energy sources, such as offshore wind farms, is essential in mitigating climate change. Taiwan has set ambitious targets to harness wind energy from the Taiwan Strait, but offshore wind farm installations are highly dependent on weather conditions, particularly wind speeds. This study examines the relationship between the El Niño-Southern Oscillation (ENSO) and offshore wind farm installation by assessing weather windows-periods with wind speeds below 12 m per second at a height of 100 m for at least 12 h. Our analysis shows that during La Niña years, the number of feasible weather windows decreases by up to 40 %, particularly between October and June, compared to neutral and El Niño years. This decrease can be as high as fourfold in December, significantly impacting installation schedules. Seasonal variations are also notable, with wind speeds exceeding 12 m s-1 in winter 66.4 % of the time, compared to 29.4 % in spring, making spring and summer the most favorable periods for installation. However, even during these favorable seasons, La Niña years can bring higher wind speeds, necessitating careful planning. These results underscore the importance of integrating ENSO forecasts into project planning to avoid installation delays and optimize installation timelines. By leveraging seasonal and interannual climate variability predictions, decision-makers can improve the resilience of offshore wind farm projects and ensure efficient energy transition strategies.

A self-attention-based neural network for three-dimensional multivariate modeling and its skillful ENSO predictions

Large biases and uncertainties remain in real-time predictions of El Niño-Southern Oscillation (ENSO) using process-based dynamical models; recent advances in data-driven deep learning algorithms provide a promising mean to achieve superior skill in the tropical Pacific sea surface temperature (SST) modeling. Here, a specific self-attention-based neural network model is developed for ENSO predictions based on the much sought-after Transformer model, named 3D-Geoformer, which is used to predict three-dimensional (3D) upper-ocean temperature anomalies and wind stress anomalies. This purely data-driven and time-space attention-enhanced model achieves surprisingly high correlation skills for Niño 3.4 SST anomaly predictions made 18 months in advance and initiated beginning in boreal spring. Further, sensitivity experiments demonstrate that the 3D-Geoformer model can depict the evolution of upper-ocean temperature and the coupled ocean-atmosphere dynamics following the Bjerknes feedback mechanism during ENSO cycles. Such successful realizations of the self-attention-based model in ENSO predictions indicate its great potential for multidimensional spatiotemporal modeling in geoscience.

Advances and challenges of operational seasonal prediction in Pacific Island Countries

Seasonal climate forecasts play a critical role in building a climate-resilient society in the Pacific Island Countries (PICs) that are highly exposed to high-impact climate events. To assist the PICs National Meteorological and Hydrological Services in generating reliable national climate outlooks, we developed a hybrid seasonal prediction system, the Pacific Island Countries Advanced Seasonal Outlook (PICASO), which has the strengths of both statistical and dynamical systems. PICASO is based on the APEC Climate Center Multi-Model Ensemble (APCC-MME), tailored to generate station-level rainfall forecasts for 49 stations in 13 countries by applying predictor optimization and the large-scale relationship-based Bayesian regression approaches. Overall, performance is improved and further stabilized temporally and spatially relative to not only APCC-MME but also other existing operational prediction systems in the Pacific. Gaps and challenges in operationalization of the PICASO system and its incorporation into operational climate services in the PICs are discussed.

Understanding the origin of extreme events in El Niño southern oscillation

We investigate a low-dimensional slow-fast model to understand the dynamical origin of El Niño southern oscillation. A close inspection of the system dynamics using several bifurcation plots reveals that a sudden large expansion of the attractor occurs at a critical system parameter via a type of interior crisis. This interior crisis evolves through merging of a cascade of period-doubling and period-adding bifurcations that leads to the origin of occasional amplitude-modulated extremely large events. More categorically, a situation similar to homoclinic chaos arises near the critical point; however, atypical global instability evolves as a channellike structure in phase space of the system that modulates variability of amplitude and return time of the occasional large events and makes a difference from the homoclinic chaos. The slow-fast timescale of the low-dimensional model plays an important role on the onset of occasional extremely large events. Such extreme events are characterized by their heights when they exceed a threshold level measured by a mean-excess function. The probability density of events' height displays multimodal distribution with an upper-bounded tail. We identify the dependence structure of interevent intervals to understand the predictability of return time of such extreme events using autoregressive integrated moving average model and box-plot analysis.

Enhancing the ENSO Predictability beyond the Spring Barrier

El Niño-Southern Oscillation (ENSO) is the dominant interseasonal–interannual variability in the tropical Pacific and substantial efforts have been dedicated to predicting its occurrence and variability because of its extensive global impacts. However, ENSO predictability has been reduced in the 21^st century, and the impact of extratropical atmosphere on the tropics has intensified during the past 2 decades, making the ENSO more complicated and harder to predict. Here, by combining tropical preconditions/ocean–atmosphere interaction with extratropical precursors, we provide a novel approach to noticeably increase the ENSO prediction skill beyond the spring predictability barrier. The success of increasing the prediction skill results mainly from the longer lead-time of the extratropical–tropical ocean-to-atmosphere interaction process, especially for the first 2 decades of the 21^st century.

A Modal Rendition of ENSO Diversity

The El Nino and Southern Oscillation (ENSO) ‘diversity’ has been considered as a major factor limiting its predictability, a critical need for disaster mitigation associated with the trademark climatic swings of the ENSO. Improving climate models for ENSO forecasts relies on deeper understanding of the ENSO diversity but currently at a nascent stage. Here, we show that the ENSO diversity thought previously as ‘complex,’ arises largely as varied contributions from three leading modes of the ENSO to a given event. The ENSO ‘slow manifold’ can be fully described by three leading predictable modes, a quasi-quadrennial mode (QQD), a quasi-biennial (QB) mode and a decadal modulation of the quasi-biennial (DQB). The modal description of ENSO provides a framework for understanding the predictability of and global teleconnections with the ENSO. We further demonstrate it to be a useful framework for understanding biases of climate models in simulating and predicting the ENSO. Therefore, skillful prediction of all shades of ENSO depends critically on the coupled models’ ability to simulate the three modes with fidelity, providing basis for optimism for future of ENSO forecasts.

Progress in ENSO prediction and predictability study

ENSO is the strongest interannual signal in the global climate system with worldwide climatic, ecological and societal impacts. Over the past decades, the research about ENSO prediction and predictability has attracted broad attention. With the development of coupled models, the improvement in initialization schemes and the progress in theoretical studies, ENSO has become the most predictable climate mode at the time scales from months to seasons. This paper reviews in detail the progress in ENSO predictions and predictability studies achieved in recent years. An emphasis is placed on two fundamental issues: the improvement in practical prediction skills and progress in the theoretical study of the intrinsic predictability limit. The former includes progress in the couple models, data assimilations, ensemble predictions and so on, and the latter focuses on efforts in the study of the optimal error growth and in the estimate of the intrinsic predictability limit.

Statistical Evidence for the Role of Southwestern Indian Ocean Heat Content in the Indian Summer Monsoon Rainfall

This study examines the benefit of using Ocean Mean Temperature (OMT) to aid in the prediction of the sign of Indian Summer Monsoon Rainfall (ISMR) anomalies. This is a statistical examination, rather than a process study. The thermal energy needed for maintaining and intensifying hurricanes and monsoons comes from the upper ocean, not just from the thin layer represented by sea surface temperature (SST) alone. Here, we show that the southwestern Indian OMT down to the depth of the 26 °C isotherm during January–March is a better qualitative predictor of the ISMR than SST. The success rate in predicting above- or below-average ISMR is 80% for OMT compared to 60% for SST. Other January–March mean climate indices (e.g., NINO3.4, Indian Ocean Dipole Mode Index, El Niño Southern Oscillation Modoki Index) have less predictability (52%, 48%, and 56%, respectively) than OMT percentage deviation (PD) (80%). Thus, OMT PD in the southwestern Indian Ocean provides a better qualitative prediction of ISMR by the end of March and indicates whether the ISMR will be above or below the climatological mean value.

Conditional Gaussian Systems for Multiscale Nonlinear Stochastic Systems: Prediction, State Estimation and Uncertainty Quantification

A conditional Gaussian framework for understanding and predicting complex multiscale nonlinear stochastic systems is developed. Despite the conditional Gaussianity, such systems are nevertheless highly nonlinear and are able to capture the non-Gaussian features of nature. The special structure of the system allows closed analytical formulae for solving the conditional statistics and is thus computationally efficient. A rich gallery of examples of conditional Gaussian systems are illustrated here, which includes data-driven physics-constrained nonlinear stochastic models, stochastically coupled reaction–diffusion models in neuroscience and ecology, and large-scale dynamical models in turbulence, fluids and geophysical flows. Making use of the conditional Gaussian structure, efficient statistically accurate algorithms involving a novel hybrid strategy for different subspaces, a judicious block decomposition and statistical symmetry are developed for solving the Fokker–Planck equation in large dimensions. The conditional Gaussian framework is also applied to develop extremely cheap multiscale data assimilation schemes, such as the stochastic superparameterization, which use particle filters to capture the non-Gaussian statistics on the large-scale part whose dimension is small whereas the statistics of the small-scale part are conditional Gaussian given the large-scale part. Other topics of the conditional Gaussian systems studied here include designing new parameter estimation schemes and understanding model errors.

Percolation Phase Transition of Surface Air Temperature Networks: A new test bed for El Niño/La Niña simulations

In this work, we studied the air-sea interaction over the tropical central eastern Pacific from a new perspective, climate network. The surface air temperatures over the tropical Pacific were constructed as a network, and the nodes within this network were linked if they have a similar temporal varying pattern. Using three different reanalysis datasets, we verified the percolation phase transition. That is, when the influences of El Niño/La Niña are strong enough to isolate more than 48% of the nodes, the network may abruptly be divided into many small pieces, indicating a change of the network state. This phenomenon was reproduced successfully by a coupled general circulation model, Flexible Global Ocean-Atmosphere-Land System Model Spectral Version 2, but another model, Flexible Global Ocean-Atmosphere-Land System Model Grid-point Version 2, failed. As both models have the same oceanic component, but are with different atmospheric components, the improperly used atmospheric component should be responsible for the missing of the percolation phase transition. Considering that this new phenomenon is only recently noticed, current state-of-the-art models may ignore this process and induce unrealistic simulations. Accordingly, percolation phase transition is proposed as a new test bed, which deserves more attention in the future.

Percolation Phase Transition of Surface Air Temperature Networks under Attacks of El Niño/La Niña

In this study, sea surface air temperature over the Pacific is constructed as a network, and the influences of sea surface temperature anomaly in the tropical central eastern Pacific (El Niño/La Niña) are regarded as a kind of natural attack on the network. The results show that El Niño/La Niña leads an abrupt percolation phase transition on the climate networks from stable to unstable or metastable phase state, corresponding to the fact that the climate condition changes from normal to abnormal significantly during El Niño/La Niña. By simulating three different forms of attacks on an idealized network, including Most connected Attack (MA), Localized Attack (LA) and Random Attack (RA), we found that both MA and LA lead to stepwise phase transitions, while RA leads to a second-order phase transition. It is found that most attacks due to El Niño/La Niña are close to the combination of MA and LA, and a percolation critical threshold Pc can be estimated to determine whether the percolation phase transition happens. Therefore, the findings in this study may renew our understandings of the influence of El Niño/La Niña on climate, and further help us in better predicting the subsequent events triggered by El Niño/La Niña.

On the progress of the 2015–2016 El Niño event

Abstract. It has been recently reported that the current 2015–2016 El Niño could become "one of the strongest on record". To further explore this claim, we performed the new analysis described in detail in Varotsos et al. (2015) that allows the detection of precursory signals of the strong El Niño events by using a recently developed non-linear dynamics tool. In this context, the analysis of the Southern Oscillation Index time series for the period 1876–2015 shows that the running 2015–2016 El Niño would be rather a "moderate to strong" or even a "strong" event and not “one of the strongest on record", as that of 1997–1998.

Seasonal forecasting of lightning and thunderstorm activity in tropical and temperate regions of the world

Thunderstorms are convective systems characterised by the occurrence of lightning. Lightning and thunderstorm activity has been increasingly studied in recent years in relation to the El Niño/Southern Oscillation (ENSO) and various other large-scale modes of atmospheric and oceanic variability. Large-scale modes of variability can sometimes be predictable several months in advance, suggesting potential for seasonal forecasting of lightning and thunderstorm activity in various regions throughout the world. To investigate this possibility, seasonal lightning activity in the world’s tropical and temperate regions is examined here in relation to numerous different large-scale modes of variability. Of the seven modes of variability examined, ENSO has the strongest relationship with lightning activity during each individual season, with relatively little relationship for the other modes of variability. A measure of ENSO variability (the NINO3.4 index) is significantly correlated to local lightning activity at 53% of locations for one or more seasons throughout the year. Variations in atmospheric parameters commonly associated with thunderstorm activity are found to provide a plausible physical explanation for the variations in lightning activity associated with ENSO. It is demonstrated that there is potential for accurately predicting lightning and thunderstorm activity several months in advance in various regions throughout the world.

Technical note: On the progress of the 2015–2016 El Niño event.. [DOI: 10.5194/acpd-15-35787-2015]

Abstract. It has been recently reported that the current 2015–2016 El Niño could become "one of the strongest on record". To further explore this claim, we performed the new analysis described in detail in Varotsos et al. (2015) that allows the detection of precursory signals of the strong El Niño events by using a recently developed non-linear dynamics tool. In this context, the analysis of the Southern Oscillation Index time series for the period 1876–2015 shows that the running 2015–2016 El Niño would be rather a "moderate to strong" or even a "strong" event and not "one of the strongest on record", as that of 1997–1998.

Salinity anomaly as a trigger for ENSO events

According to the classical theories of ENSO, subsurface anomalies in ocean thermal structure are precursors for ENSO events and their initial specification is essential for skillful ENSO forecast. Although ocean salinity in the tropical Pacific (particularly in the western Pacific warm pool) can vary in response to El Niño events, its effect on ENSO evolution and forecasts of ENSO has been less explored. Here we present evidence that, in addition to the passive response, salinity variability may also play an active role in ENSO evolution, and thus important in forecasting El Niño events. By comparing two forecast experiments in which the interannually variability of salinity in the ocean initial states is either included or excluded, the salinity variability is shown to be essential to correctly forecast the 2007/08 La Niña starting from April 2007. With realistic salinity initial states, the tendency to decay of the subsurface cold condition during the spring and early summer 2007 was interrupted by positive salinity anomalies in the upper central Pacific, which working together with the Bjerknes positive feedback, contributed to the development of the La Niña event. Our study suggests that ENSO forecasts will benefit from more accurate salinity observations with large-scale spatial coverage.

Replicated throughfall exclusion experiment in an Indonesian perhumid rainforest: wood production, litter fall and fine root growth under simulated drought

Climate change scenarios predict increases in the frequency and duration of ENSO-related droughts for parts of South-East Asia until the end of this century exposing the remaining rainforests to increasing drought risk. A pan-tropical review of recorded drought-related tree mortalities in more than 100 monitoring plots before, during and after drought events suggested a higher drought-vulnerability of trees in South-East Asian than in Amazonian forests. Here, we present the results of a replicated (n = 3 plots) throughfall exclusion experiment in a perhumid tropical rainforest in Sulawesi, Indonesia. In this first large-scale roof experiment outside semihumid eastern Amazonia, 60% of the throughfall was displaced during the first 8 months and 80% during the subsequent 17 months, exposing the forest to severe soil desiccation for about 17 months. In the experiment's second year, wood production decreased on average by 40% with largely different responses of the tree families (ranging from -100 to +100% change). Most sensitive were trees with high radial growth rates under moist conditions. In contrast, tree height was only a secondary factor and wood specific gravity had no influence on growth sensitivity. Fine root biomass was reduced by 35% after 25 months of soil desiccation while fine root necromass increased by 250% indicating elevated fine root mortality. Cumulative aboveground litter production was not significantly reduced in this period. The trees from this Indonesian perhumid rainforest revealed similar responses of wood and litter production and root dynamics as those in two semihumid Amazonian forests subjected to experimental drought. We conclude that trees from paleo- or neotropical forests growing in semihumid or perhumid climates may not differ systematically in their growth sensitivity and vitality under sublethal drought stress. Drought vulnerability may depend more on stem cambial activity in moist periods than on tree height or wood specific gravity.

Status and challenges for conservation of small mammal assemblages in South America

South America spans about 44° latitude, covers almost 18 million km(2) , and is second only to Africa in continental mammal species richness. In spite of this richness, research on the status of this fauna and on the nature and magnitude of contemporary threats remains limited. Distilling threats to this diverse fauna at a continental scale is challenging, in part because of the limited availability of rigorous studies. Recognizing this constraint, we summarize key threats to small mammals in South America, emphasizing the roles of habitat loss and degradation, direct persecution, and the increasing threat of climate change. We focus on three regional 'case studies': the tropical Andes, Amazonia and adjacent lowland regions, and the southern temperate region. We close with a brief summary of recent findings at our long-term research site in north-central Chile as they pertain to projected threats to this fauna. Habitat alteration is a pervasive threat that has been magnified by market forces and globalization (e.g. extensive agricultural development in Amazonia), and threatens increasing numbers of populations and species. Climate change poses even greater threats, from changes in rainfall and runoff regimes and resulting changes in vegetative structure and composition to secondary influences on fire dynamics. It is likely that many changes have yet to be recognized, but existing threats suggest that the future may bring dramatic changes in the distribution of many mammal taxa, although it is not clear if key habitat elements (vegetation) will respond as rapidly as climatic factors, leading to substantial uncertainty. Climate change is likely to result in 'winners' and 'losers' but available information precludes detailed assessment of which species are likely to fall into which category. In the absence of long-term monitoring and applied research to characterize these threats more accurately, and to develop strategies to reduce their impacts, managers already are being faced with daunting challenges. As the line between 'pure' and 'applied' research blurs in the face of converging interests of scientists and society we hope that solutions to these critical issues will be incorporated in addressing anticipated conservation crises.

Relationship between increase rate of human plague in China and global climate index as revealed by cross-spectral and cross-wavelet analyses

Plague has caused the death of hundreds of millions of people throughout the human history. Today this disease is again re-emerging and hence is again becoming an increasing threat to human health in several parts of the world. However, impacts of global climate variation (e.g. El Nino and Southern Oscillation [ENSO]) and global warming on plagues are largely unknown. Using cross-spectral analysis and cross-wavelet analysis, we have analyzed the relationship between increase rate of human plague in China during 1871-2003 and the following climate factors (as measured by the Southern Oscillation Index [SOI], Sea Surface Temperature of east Pacific equator [SST] and air Temperature of the Northern Hemisphere [NHT]). We found in the frequency domain that increase rate of human plague was closely associated with SOI and SST. Cross-spectral analysis reveals that significant coherencies between increase rate of human plague and ENSO were found over short periods (2-3 years), medium periods (6-7 years) and long periods (11-12 years, 30-40 years). Cross-wavelet analysis reveals that increase rate of human plague oscillates in phase with SOI, but in anti-phase with SST over periods of 2-4 years and approximately 8 years (6-10 years). These results indicate that ENSO-driven climate variation may be important for occurrences of human plague in China. However, there is a need for a further analysis of the underlying mechanism between human plague in China and ENSO.

A successful real-time forecast of the 2010–11 La Niña event

During 2010–11, a La Niña condition prevailed in the tropical Pacific. An intermediate coupled model (ICM) is used to demonstrate a real-time forecast of sea surface temperature (SST) evolution during the event. One of the ICM's unique features is an empirical parameterization of the temperature of subsurface water entrained into the mixed layer (T_e). This model provided a good prediction, particularly of the "double dip" evolution of SST in 2011 that followed the La Niña event peak in October 2010. Thermocline feedback, explicitly represented by the relationship between T_e and sea level in the ICM, is a crucial factor affecting the second cooling in 2011. Large negative T_e anomalies were observed to persist in the central equatorial domain during 2010–11, inducing a cold SST anomaly to the east during July–August 2011 and leading to the development of a La Niña condition thereafter.

Effect of ENSO-driven precipitation on population irruptions of the Yangtze vole Microtus fortis calamorum in the Dongting Lake region of China

The Yangtze vole (Microtus fortis Buechner, 1889) is a small herbivore species that inhabits lake beaches in the Dongting Lake region along the Yangtze River in Southern China. Its population shows strong oscillations during the wet season due to summer precipitation-induced immigration away from the lake into adjacent rice fields. The effect of El Niño-Southern Oscillation-driven precipitation on population abundance and growth of the vole species is not fully understood. We undertook an analysis of the combined data of 4 time series covering 1981-2006 from 4 different sites and a separate analysis on a single time series (1981-2006) from one site. Our results demonstrate that a dual effect of El Niño-Southern Oscillation-driven precipitation on the population abundance of voles is time-dependent: precipitation in the current year has a positive effect, whereas precipitation in the previous year has a negative effect. The dual effect of precipitation on vole population is well explained by the unique interactions among vole population, precipitation water level and the lake beach habitat around Dongting Lake. We found that drier than average weather of the previous year benefited voles because their breeding habitats, lake beaches, were exposed for long stretches of time. Wet weather was found to increase the number of voles inhabiting rice fields because as the water level of the lake rose they were forced from beaches into surrounding rice fields. Summer precipitation in the Dongting Lake region was found to be positively associated with the sea surface temperature (SST) of the eastern tropical Pacific Ocean of the previous year and winter SST and spring SST of the current year. Annual rates of increase in the vole population of the reconstructed time series are negatively associated with the vole abundance and autumn precipitation of the previous year and winter precipitation of the current years. These results suggest that both extrinsic and density-dependent intrinsic factors may affect population dynamics of the Yangtze voles.

Population dynamics of two sympatric rodents in a variable environment: rainfall, resource availability, and predation

Precipitation plays an important role in the dynamics of species found in arid and semiarid environments. However, population fluctuations generally are driven by a combination of multiple factors whose relative contribution may vary through time and among species. We monitored fluctuations of species in three trophic levels for >17 years at a semiarid community in north-central Chile. The region is strongly affected by the El Niño Southern Oscillation, resulting in high variation in rainfall that triggers dramatic changes in food resource availability, with strong effects on upper trophic levels. We focused our analyses on the role played by endogenous and exogenous (climatic) factors on the dynamics of two important rodent species in the community, Octodon degus and Phyllotis darwini. We documented population fluctuations of several orders of magnitude in response to wet and dry episodes of different strength and duration. P. darwini reached similar maximum densities, regardless of the duration of high-rainfall events, whereas O. degus showed additive effects of multiple wet years. Time series diagnostic tools revealed oscillations with a 5-year periodicity in rainfall, which may be the cause of the same periodicity and a weak second-order signal observed in the rodent dynamics. However, the dynamics of both rodent species were dominated by strong first-order processes, suggesting an important role of direct density dependence. Intraspecific competition, expressed as the ratio of rodent density/rainfall (or food resources) explained more than two-thirds of the variation in the population rate of change, whereas less than one-third was explained by lagged rainfall (or food resources). We detected no significant effects of predation. Our results contribute to a growing number of examples of dynamics governed by the combined effect of density dependence and climatic forcing. They also reveal strong bottom-up regulation that may be common in other arid environments.

ENSO as an integrating concept in earth science

The El Niño-Southern Oscillation (ENSO) cycle of alternating warm El Niño and cold La Niña events is the dominant year-to-year climate signal on Earth. ENSO originates in the tropical Pacific through interactions between the ocean and the atmosphere, but its environmental and socioeconomic impacts are felt worldwide. Spurred on by the powerful 1997-1998 El Niño, efforts to understand the causes and consequences of ENSO have greatly expanded in the past few years. These efforts reveal the breadth of ENSO's influence on the Earth system and the potential to exploit its predictability for societal benefit. However, many intertwined issues regarding ENSO dynamics, impacts, forecasting, and applications remain unresolved. Research to address these issues will not only lead to progress across a broad range of scientific disciplines but also provide an opportunity to educate the public and policy makers about the importance of climate variability and change in the modern world.

Neural network forecasts of the tropical Pacific sea surface temperatures

A nonlinear forecast system for the sea surface temperature (SST) anomalies over the whole tropical Pacific has been developed using a multi-layer perceptron neural network approach, where sea level pressure and SST anomalies were used as predictors to predict the five leading SST principal components at lead times from 3 to 15 months. Relative to the linear regression (LR) models, the nonlinear (NL) models showed higher correlation skills and lower root mean square errors over most areas of the domain, especially over the far western Pacific (west of 155 degrees E) and the eastern equatorial Pacific off Peru at lead times longer than 3 months, with correlation skills enhanced by 0.10-0.14. Seasonal and decadal changes in the prediction skills in the NL and LR models were also studied.

Predictability of El Niño over the past 148 years

Forecasts of El Niño climate events are routinely provided and distributed, but the limits of El Niño predictability are still the subject of debate. Some recent studies suggest that the predictability is largely limited by the effects of high-frequency atmospheric 'noise', whereas others emphasize limitations arising from the growth of initial errors in model simulations. Here we present retrospective forecasts of the interannual climate fluctuations in the tropical Pacific Ocean for the period 1857 to 2003, using a coupled ocean-atmosphere model. The model successfully predicts all prominent El Niño events within this period at lead times of up to two years. Our analysis suggests that the evolution of El Niño is controlled to a larger degree by self-sustaining internal dynamics than by stochastic forcing. Model-based prediction of El Niño therefore depends more on the initial conditions than on unpredictable atmospheric noise. We conclude that throughout the past century, El Niño has been more predictable than previously envisaged.

Regional climate impacts of the Northern Hemisphere annular mode

The Northern Hemisphere annular mode (NAM) (also known as the North Atlantic Oscillation) is shown to exert a strong influence on wintertime climate, not only over the Euro-Atlantic half of the hemisphere as documented in previous studies, but over the Pacific half as well. It affects not only the mean conditions, but also the day-to-day variability, modulating the intensity of mid-latitude storms and the frequency of occurrence of high-latitude blocking and cold air outbreaks throughout the hemisphere. The recent trend in the NAM toward its high-index polarity with stronger subpolar westerlies has tended to reduce the severity of winter weather over most middle- and high-latitude Northern Hemisphere continental regions.

Variability in the El Niño-Southern Oscillation through a glacial-interglacial cycle

The El Niño-Southern Oscillation (ENSO) is the most potent source of interannual climate variability. Uncertainty surrounding the impact of greenhouse warming on ENSO strength and frequency has stimulated efforts to develop a better understanding of the sensitivity of ENSO to climate change. Here we use annually banded corals from Papua New Guinea to show that ENSO has existed for the past 130,000 years, operating even during "glacial" times of substantially reduced regional and global temperature and changed solar forcing. However, we also find that during the 20th century ENSO has been strong compared with ENSO of previous cool (glacial) and warm (interglacial) times. The observed pattern of change in amplitude may be due to the combined effects of ENSO dampening during cool glacial conditions and ENSO forcing by precessional orbital variations.

Predictability in the midst of chaos: A scientific basis for climate forecasting

The Earth's atmosphere is generally considered to be an example of a chaotic system that is sensitively dependent on initial conditions. It is shown here that certain regions of the atmosphere are an exception. Wind patterns and rainfall in certain regions of the tropics are so strongly determined by the temperature of the underlying sea surface that they do not show sensitive dependence on the initial conditions of the atmosphere. Therefore, it should be possible to predict the large-scale tropical circulation and rainfall for as long as the ocean temperature can be predicted. If changes in tropical Pacific sea-surface temperature are quite large, even the extratropical circulation over some regions, especially over the Pacific-North American sector, is predictable.