Representing uncertainty in ocean observations and the ocean model, for coupled ensemble data assimilation and ensemble extended-range prediction

Lead Research Organisation: University of Oxford
Department Name: Oxford Physics


There is currently a large effort in the development of general circulation model (GCM)-based seasonal to decadal prediction systems to provide climate forecasts. Such techniques are rather complex, technically challenging and still in their infancy. Any weather or climate forecast will be subject to three sources of uncertainty, namely observation uncertainty, the model-component of initial uncertainty, and model uncertainty over the forecast period.

The aim of this proposal is to improve the reliability of extended range forecast of weather and climate, mainly focusing on the ocean component of the coupled system. We propose to develop and incorporate various tools based on stochastic physics to improve the reliability of forecasts focusing on a more accurate representation of ocean observations and model uncertainties. The individual impacts of the different developments on the reliability of the forecasts will be quantified to provide estimates of the different sources of uncertainties in the forecasts. The development of reliable extended range forecasts can be extremely beneficial with major economical and societal consequences.

Planned Impact

A key goal for any weather and climate forecast centre is production of reliable predictions. Since the basis of this proposal is to improve the reliability of extended range forecasts of weather and climate, the proposed developments will be directly beneficial to an enormous number of scientists and users alike - the latter ultimately encompassing anyone that consults, makes economic decisions based on, or otherwise uses weather and climate forecasts on monthly, seasonal and decadal timescales.

This is because the proposal has the potential to improve significantly the reliability of forecasts that make use of coupled ocean atmosphere models. The developments proposed will have such wide benefits because so many international and national operational forecasting bodies (UK Met Office, MPI, Meteo France, IPSL, EC-Earth etc) use the NEMO coupled ocean atmosphere model as part of their forecasting methodology.

Naturally the results will be of particular interest to climate scientists at these research institutions and meteorological agencies, who are charged with developing projections of future climate, as well as producing the next generation of coupled ocean-atmosphere climate models.

The general public, particularly in the UK, has always shown a keen interest in weather forecasting. Climate change on all timescales, and the role of the ocean and atmosphere in it, has captured the attention of the public and media, and there is a clear need for the relevant complex science to be communicated in a form that can be understood by a non-specialist. While the technical aspects of the project are not suitable for wider dissemination, the project will produce some highly relevant information regarding forecast reliability on monthly, seasonal and decadal timescales.


10 25 50
Description Through comprehensive testing of the state of the art ocean-atmosphere model, used by the European Centre for Medium Range Forecasts (ECMWF) for seasonal prediction, we have gained an understanding of the impact of current stochastic representation, or parameterisation, of sub grid ocean processes. In addition to the application of two methods (Brankart 2013 & Williams 2012), involving perturbation of the equation of state and ocean surface fluxes, two additional methods have been developed that perturb the representation of sub-grid scale eddies. We now have an improved understanding of the relative importance of these processes.

We also investigated an alternative method: Wilks (2005) and Arnold et. al (2013), applied regression to estimate the sub-grid scale deterministic and stochastic forcing in an idealised chaotic system. We found that extension of this method to the case of a dynamical fluid model is guaranteed to lead to unstable solutions and is therefore not a suitable approach.

We have developed an alternative algorithm, guaranteed to be stable, that for the first time is able to calculate the sub-grid scale forcing required for a running fluid model to have the correct mean state (climatological mean) and variability (climatological variance). See Andrejczuk et al, 2016. To develop this practical algorithm, a key problem in the optimisation of fluids was solved, enabling the accuracy of current fluid nudging methods to be arbitrarily increased. In addition we developed a new and accurate method for the calculation of the necessary sub-grid variability and have demonstrated its accuracy in a chaotic dynamical fluid model.

There is a theory, often referred to as the fluctuation-dissipation theorem, that suggests that correction of the mean state and variability of a fluid model, leads to a more accurate model response to forcing. For example, in a climate change experiment, where heating is expected to have an impact upon the atmospheric and oceanic dynamics, a model with empirically corrected climatological timescales will exhibit a much more accurate response to forcing in comparison with a model without such a correction. We have tested this theory and have found it to be true in a simplified model of the North Atlantic Ocean.

Empirical models of the ocean have been developed (Penland, 1989, and Newman, Alexander & Scott 2011) that are currently the most accurate method of forecasting tropical sea surface temperatures ( However severe simplifications have been necessary which have limited model complexity and have prevented their application in a wider context. We have discovered the physical assumptions required to eliminate these simplifications and have developed an algorithm for accurate empirical model development on any scale, Juricke et al (2017), submitted to the Journal of Climate. Further work to combine these empirical methods with the calculation of model sub-grid forcing is well advanced. We hope to submit our results for publication within the next two to three months.
Exploitation Route A large potential impact is through taking the algorithms we have developed and implementing them in state of the art atmosphere-ocean general circulation models. Our results can also be applied to fluid dynamical optimisation problems in the field of engineering.

Our algorithms involve an improvement of the integration of measurement data with state of the art dynamical models. This can lead to an increase in accuracy: The accuracy and uncertainty in seasonal weather forecasts can be improved. Forecasts of global climate change under various forcing scenarios can be made more accurate through an improvement in the models response to forcing. With an improved model, attribution of climatological changes can be more accurately estimated.

Our algorithms can be used as guidance for future model development: Our methods give the location and magnitude of forcing and variability missing from a model. For example where correction of current and introduction of new physical processes or an increase in resolution in the model is required. In a realistic ocean context, our results can be used to further understanding of the statistical properties of turbulence.
Sectors Aerospace, Defence and Marine,Agriculture, Food and Drink,Chemicals,Communities and Social Services/Policy,Construction,Creative Economy,Digital/Communication/Information Technologies (including Software),Education,Energy,Environment,Financial Services, and Management Consultancy,Healthcare,Leisure Activities, including Sports, Recreation and Tourism,Manufacturing, including Industrial Biotechology,Retail,Transport

Description • Our results have been used to inform scientists and researchers working worldwide in both blue sky and applied research. This has been achieved through meetings and presentations at national and international scientific conferences and at both the UK Meteorological Office and the European Centre for Medium Range Forecasts (ECMWF). • Two scientific papers have been submitted for publication: Andrejczuk et al (2016) which appeared in Monthly Weather Review and Juricke et al (2017), submitted to the Journal of Climate. • In a research context, we have used our results to inform the implementation of new modelling techniques in a realistic ocean model. This effort is on going and required that significant problems were first solved using a simpler prototype algorithm before integration with a state of the art ocean model could be attempted. These major obstacles have now been overcome and work on implementing our results in the European Nucleus for an Ocean Model ( and the MITgcm ( is underway. • Outside of the current project, through close collaboration with other researchers, the insight gained has been used to help steer the direction of other research projects. In addition two undergraduate project students have gained valuable experience working on these problems.
Sector Digital/Communication/Information Technologies (including Software),Education
Impact Types Societal