Traditionally, weather-sensitive businesses have managed their weather-related risk through operational decision-making and hedging strategies using commodities, futures and options. Recent advances in climate research and technologies, and the emergence of the weather derivatives market, present various businesses with a significant exposure to variations in year-to-year climate, with new opportunities to quantify and manage their climate risk.

There is convincing evidence that a growing number of weather-sensitive companies are taking a more active role in managing their weather and climate-related risk, a risk that was previously considered a cost of doing business.

Technical expertise
Many organisations are now building in-house technical climate expertise and recognise the fundamental need to remain up-to-date with the latest developments in the field. The increasing interest in weather and climate risk management is leading to a growing demand for high quality historical weather data, customised climate analysis and climate forecasts in the US and other regions in the world.

Over the past two decades, the scientific community has discovered a number of large-scale ocean-atmospheric phenomena which have been shown to be the leading drivers of year-to-year climate variability in many regions around the world. These include the El Niño-Southern Oscillation (ENSO)1,2 the Arctic Oscillation (AO) and its regional manifestation the North Atlantic Oscillation (NAO)3, and the Northern Pacific Decadal Oscillation (PDO)4 also known as NPO. The ability to forecast these phenomena with useful lead-times will lead to significant improvements in the skill of seasonal forecasts over the US and many other regions worldwide.

With the growing market for weather risk management in many weather-sensitive industries, understanding the impacts of ENSO, NAO and PDO on regional climate variability in regions of relevance to a company would result in significant improvements in assessment and management of weather-related risk. Furthermore, the ability to forecast these phenomena would lead to capabilities in forecasting regional climate variability, providing ‘at risk' companies with useful information on how climate varies over time in their regions of interest.

The scientific community has focused on building state-of-the-art systems capable of forecasting year-to-year climate variability with lead times of one to several seasons. In the past few years, these systems have been improved to a level where climate forecasts in certain regions and for certain periods during the year are reliable enough to be used in business decision-making. However, effective use of these forecasts depends on the specific application. In addition, it requires an understanding of the underlying drivers of climate variability in regions of interest; selection and evaluation of forecasting methodologies for specific applications; and correct interpretation and utilisation of the information provided by the forecasts.

Climate predictability
Climate research published in the late 1970s and 1980s has indicated that slow variations in the lower boundary conditions of the atmosphere such as sea surface temperature (SST), soil moisture, etc., give rise to significant predictability of the variations in the climate from year to year. It has also been shown that several large-scale ocean-atmospheric phenomena, including ENSO, PDO and NAO and their interactions, play an important role in the variations of climate from year-to-year variability in different regions of the world. This leads to the premise for predictability of year-to-year climate variability in regions where the link between climate variability and these phenomena has been demonstrated.

As a result, it is critical to understand the extent to which each of these phenomena plays a role in the variation of various atmospheric variables in different regions and for periods of interest to a weather-sensitive company. This information alone could have significant use in a number of different applications.

Traditionally, climate and its variability have been quantified using statistical analysis of historical data. Over the past decade, development of forecasting schemes based on the physics of the Earth's climate system (generally referred to as dynamical forecasting systems) has led to the capability of forecasting climate variability with lead times of up to a year. These systems are producing probabilistic short-term climate forecasts, such as forecasts of climate with one to several seasons' lead-time. For example, a short-term climate forecast could provide information on the probabilities of average temperature, aggregate Heating Degree Days (HDD) or Cooling Degree Days (CDD) to be higher or lower that normal in the season ahead, or on the probability of extreme temperatures being higher or lower than normal for the upcoming summer, etc.

Note that the leading climate forecasting systems are currently ENSO-based and are unable to forecast phenomena such PDO and NAO. However, scientific efforts are directed towards understanding the dynamics of these phenomena and their impacts on climate variability on different time-scales.5 A significant portion of the skill of climate forecast models will arise from the ability to directly forecast PDO and NAO and their interactions with ENSO within the forecast model. Currently, leading climate forecasting groups combine the results of the ENSO-based dynamical models with outputs of statistical tools to develop the best possible forecast. For example, the 1998-1999 El Niño and its climatic impacts were forecast six months in advance with relatively high accuracy.

System issues
A typical state-of-the-art climate forecasting system has several components. These include: historical and real-time observations of ocean, atmosphere and land; sophisticated initialisation and data assimilation schemes; coupled ocean-atmosphere dynamical (physical)6 and statistical prediction models; and validation techniques.

Over the past decade, a hierarchy of ENSO-based prediction schemes has been developed.7 These are now being extended to include other key phenomena. The currently operational forecasting systems can be divided into three classes:
1) purely statistical;
2) physical ocean model/statistical atmosphere model (also known as hybrid systems); and
3) physical coupled ocean-atmospheric models (also known as coupled dynamical models). The statistical models can be classified further into models which can use either low-frequency variations in the ocean or the atmosphere as predictors. The physical models consist of models of varying degrees of complexity for the Earth's climate system.

Consensus
There is consensus among the scientific community that the most promising climate prediction schemes are based on the physical models. Currently, leading forecasting groups use a variety of dynamical and statistical techniques to develop their official forecasts. Statistical tools are used to include the impacts of phenomena such as the observed warming trends on to the results of the ENSO-based dynamical models. Efforts are underway to incorporate phenomena such as PDO and NAO in the dynamical forecast systems.

The decision to adopt forecast information in business decision-making should depend on:
1) understanding the drivers of climate variability in the regions and for the periods of interest to a company as it is critical to identify sources of predictability for factors relevant to the company;
2) understanding different forecasting methodologies, their underlying assumptions and capabilities. A key component is the quantification of the skill of the selected forecasting system(s) for the key variables in the region(s), for the period(s) of interest, and for different climate regimes. Forecast skill depends on a number of factors including lead-time, season, decade, geographic region, the state of the climate (phases of ENSO, PDO, NAO), strength of the sea's surface temperature anomaly (e.g., the amplitude of the ENSO event), and the variable forecasted; and
3) quantifying the skill of different model(s) for key parameters most relevant to a company's applications, and develop a decision tree for use of various models for different scenarios and applications.

References
1 ENSO is a coupled ocean-atmosphere phenomenon centred in and over the tropical Pacific and refers to large-scale fluctuations in a number of oceanic and atmospheric variables (e.g. sea surface temperature, sea level pressure, rainfall, etc.). El Niño and La Niña episodes are the opposite extremes of the ENSO phenomenon. During an El Niño, above normal sea surface temperatures (SSTs) extend across the central and eastern equatorial Pacific Ocean. During a La Niña, below normal SSTs extend through much of central and eastern tropical Pacific.

2 Both El Niño and La Niña events impact climate, though their impacts are not exactly opposite since their effects are in different regions of the tropical Pacific Ocean.

3 NAO, which is known to be the dominant mode of atmospheric variability in the North Atlantic sector, refers to the fluctuations in the sea level pressure at two stations across the North Atlantic (Azores and Iceland). NAO is centred on the North Atlantic Ocean basin, where atmospheric circulation displays a strong north-south pressure contrast, with low pressure centred on Iceland and high pressure centred near the Azores in southern edge of the basin.

4 PDO refers to an ENSO-like pattern involving the decade-long changes in the North Pacific atmosphere and ocean with 20 to 30 years of persistence.

5 Seasonal climate variability over North America exhibits rather large fluctuations between years characterised by the same ENSO phase. A number of recent studies have indicated that both NAO and PDO exert a modulating effect on the impacts of ENSO on the climate in the US and other regions around the world.

6 The terms ‘physical' and ‘dynamical' both refer to models which integrate some version of equations of motion, which have been modified for ocean and atmosphere dynamics, in time. The term ‘coupled' refers to incorporation of the interactions between ocean and the atmosphere, and atmosphere and land in these models. A dynamical forecast is a real-time estimate of the future state of the ocean or atmosphere obtained by running a dynamical model forward in time.

7 The development of these forecasting systems has become possible because of a better understanding of the climate system and factors influencing inter-annual climate variability together with progress in computing systems, earth-ocean-atmosphere observing systems, and various data processing techniques.