Climate Forecasting Unit

The Climate Forecasting Unit (CFU) undertakes advanced research to forecast global climate variations from one month to several years into the future (also known as seasonal-to-decadal predictions). The unit members also investigate the risk of future climate variability in socio-economic sectors, and the management of such risk via the development of climate services for renewable energy, water resources etc.
Emphasis is made on forecasting changes in high-impact climate events such as the extreme frequency of rare or persistent winds, floods, droughts and temperature extremes.  
Using state-of-the-art dynamical climate forecast systems, climate is seamlessly predicted in a probabilistic way, based on the forecast quality.  The qualilty is assessed and compared to other forecast systems available worldwide, while the physical processes at the origin of the prediction skill are studied using sensitivity and idealised experiments carried out by CFU climate scientists. The unit has a particular interest at advancing the frontier of climate prediction by exploring the impact of substantial increases in model resolution, the use of stochastic methods to address model inadequacy and the development of postprocessing methods that combine climate information from multiple sources.
The CFU is involved in several national and European FP7-funded projects.  The substantial computing resources needed to achieve the ambitious objectives of the unit are obtained from the IC3 cluster, and the supercomputers of the Barcelona Supercomputing Center (BSC, Spain) and the European Centre for Medium-Range Weather Forecasts (ECMWF, United Kingdom) .
CFU General Objectives:
  • Developing climate forecast systems and methodologies: develop climate models, and assess dynamic and statistical climate forecasting methods over seasonal to decadal timescales.
  • Understanding and quantifying error in climate models and limitations of the methodologies: identify predictability and processes at the source of climate model error, which plays a major role in the climate forecast assessment.
  • Formulate reliable climate forecast to meet specific user needs: including the development and implementation of techniques to statistically downscale, calibrate and combine dynamic ensemble and simple statistical forecasts. 
  • Develop a competitive climate service based on seasonal to decadal climate forecasts: by demonstrating the benefits and limitations of climate predictions and its role within the assessment and management of risk for different sectors.
CFU Research groups
  • Seasonal prediction research group. Seasonal forecasts provide information about how average weather conditions on a monthly or seasonal basis are likely to be, from a few months up to one year into the future. These forecasts have a great number of applications and help decision-making in agriculture, energy, health and hydrology, amongst other sectors of society.
    Interactions between ocean, sea ice, atmosphere and land, influence the weather conditions at the seasonal time-scales. Therefore, a suitable way to represent the Earth Climate system at these time-scales, is to use a dynamical coupled ocean-seaice-atmosphere-land model.
    Due to the complexity of the climate and the model processes involved, as well as the impact of small-scale processes, that cannot be explicitly represented, on the large-scale processes of interest, it is not possible to describe the state of the climate system with complete accuracy. A way of accounting for the uncertainties associated to seasonal prediction is to use an “ensemble” of several climate models to represent the most likely states of the climate system over the coming months, given the past and current climate conditions from which they evolve. To account for the uncertainty in observational estimates of the past and current climate conditions the climate models are started from, several predictions are run which each model starting from slightly different initial conditions.
    An alternative or complement to dynamical coupled models are statistical models. These use statistical methods that combine the different sources of forecast information, including information provided by dynamical prediction systems, to obtain a better estimate of future climates.
    The CFU is following these three approaches: dynamical, statistical and combined prediction, with an ongoing participation in a number of projects and coordinated efforts.
  • Decadal prediction research group. Decadal prediction lies between seasonal/interannual forecasting (few months-2 years), and longer-term climate change projections (50 years +), by focusing on time-evolving regional climate conditions over the next 2–30 years. Quantities of interest are multi-year averages of climate variables such as temperature and precipitation which determine the local climate conditions. Decadal predictions therefore fill the temporal gap between the seasonal forecasts and climate change projections.
    Decadal predictions are of increasing scientific interest as an additional validation tools for our climate models, but are also potentially a benefit to society. Many investments in large scale infrastructure are made for over decadal time scales. Knowing how climate could change regionally during this period may help improve cost-benefit analyses of new investments. The probabilistic character of the predictions is also potentially useful in risk analyses.
    The premise of decadal predictions is two-fold:
    The decadal internal fluctuations in the Earth System are partly predictable (such as the Pacific Decadal Variability and Atlantic Multidecadal Variability), and
    That by initialising the climate system with the best possible observed state we obtain better estimates of the anthropogenically forced climate change.
    Technically, decadal predictions are a challenge, although the methodology to produce them is in its infancy and considerable advancements are expected as climate science develops further.
  • Forecast System Development research group. Forecast system development focuses on the set of techniques or tools required for analysis of historical data, selection of the most appropriate modeling structure, model validation, development of forecasts, and monitoring and adjustment of forecasts.
    The main goal is to develop and assess dynamical and statistical methods for the prediction of global and regional climate on time scales ranging from a few weeks to several years.
    Coupled climate models are sophisticated tools designed to simulate the Earth climate system and the complex interactions between its components. CFU is currently working with EC-Earth, NEMO and IFS.
    A typical climate forecast experiment executes tens or even hundreds of jobs, a task that can obviously not be done manually. Usually the jobs have multiple dependencies between them. Therefore CFU is working with the models described above by using High Performance Computing (HPC) machines, and a tool to create, manage and monitor experiments.
    The computational resources to carry out those experiments consists of a sizeable allocation in the IC3 linux cluster, as well as competitive allocations on the ECMWF, the Barcelona Supercomputing Centre (BSC), and other PRACE super-computers.
  • Climate Services research group. Climate services facilitates the use of relevant climate information across different sectors, such as energy and water, to manage and adapt accordingly the socio-economic risks and opportunities caused by future climate variability.
    Changes in regional climatic patterns over time are already recognised, and climate variability is expected to increase with the near-term effects of global warming, which could result in unexpected changes to climate patterns or extremes.
    With successful climate risk management, the economic and social sectors that can understand and adapt to the impacts of increasing climate variability will be best positioned for continued growth.
    Recent advances in science and technology has significantly contributed to improve the ability to forecast climate over the coming seasons, years or decades. These timescales can now be modelled with increasing probability, or skill, to evaluate the vulnerability and manage the risks and opportunities from future climate variability.
    By providing access to useful and clear climate information, climate services can facilitate timely, actionable and decision relevant outcomes to guide operational and planning decisions, enhance productivity and increase the safety and security across many different sectors at both local and global scales.

Research Themes