ERror Growth in Operational Data Initialised Coupled Systems (ERGODICS)

Lead Research Organisation: University of Reading
Department Name: Meteorology

Abstract

Coupled initialisation of climate models is a challenging new area requiring the application of sophisticated data assimilation methods, while at the same time practically managing complex interacting physics and dynamics of the coupled system, particularly at the air-sea interface and boundary layers. This proposal focuses on understanding and then reducing coupled hindcast bias errors at a range of lead times. It will exploit hindcast sets already performed or planned at ECMWF and the Met Office, making modifications to the assimilation increments or bias schemes. Full-field hindcasts will be used where the model is forced as close to the observations as possible. The lead-time dependent climate drift contains potentially important information about the model process errors. This proposal will develop and test several methods to analyse and reduce this lead-time dependent drift, including matching initial condition drift to perturbed physics drifts, matching atmosphere-ocean initialization increments to coupled model covariances, modifying covariances to focus on longer lead time predictions, and modifying already used bias schemes within operational systems. This diversity of approaches, focused on reducing coupled bias, is a strength of the proposal allowing promising results to be developed more fully. This approach, the experience of the team, and the focus on modifications to already existing operational coupled systems will also allow this project to operate effectively as an integrating project for the whole NGWCP program, allowing promising results from other funded projects to be brought through and tested within operational model environments.

Planned Impact

The main impact of this work will be in the testing of new methods to reduce bias in coupled models used in operational forecasting at the Met Office and at ECMWF. This project will also act as a coordinating project for the NGWCP program and as such will seek to facilitate and arrange for the testing of useful new methods emerging from the whole research program within an operational environment.
It will do this by organising program meetings inviting ECMWF and MetOffice coupled model development representation. It will also offer to represent the work of the NERC program at coupled modelling meetings both at the met office and internaitonally where appropriate.
The interests of this program is strongly aligned with the interests of both the NCAS and NCEO programs

Publications

10 25 50
 
Description (1) We have shown how shocks (sudden changes) arise in coupled atmosphere-ocean forecast models that are used for longer term weather predictions, due to the methods chosen to start the forecasts. We provide recommendations about how to detect these shocks and avoid them, which could lead to improved forecastss on longer timescales.
A paper has been published on this topic.
(2) Further work has studied the role of equatorial correction of biases in the upper oceans. We have found that allowing for these biased during the first month of a seasonal forecast allows the forecasts to have improved skill compared with current operational practice .
A paper is published. This method will now be rigorously tested at ECMWF and if confirmed will lead to changes in operational forecast practices.
(3) A third piece of work studies the initial drifts fo coupled ocean-atmosphere models and compares these with drifts induced by modifying the physics parametrisations. this comparisons allows us to identify which physical processes may be in error in model regions, potentially giving insight that may help future model developments.
A paper is now published
Exploitation Route This will be of value to the operational agencies particularly Met Office and ECMWF in improving weather prediction systems and improving the underlying coupled model processes.
Sectors Environment

 
Description Changes to seasonal forecasting are now being tested at ECMWF. Strictly the impacts have not yet been achieved
First Year Of Impact 2015
Sector Environment
Impact Types Societal

 
Description Implications of the opening of Arctic sea routes: A government commissioned Foresight evidence report on the future accessibility of the Arctic for shipping and the implications for national policy. https://www.gov.uk/government/publications/future-of-the-sea-implications-from-opening-arctic-sea-routes
Geographic Reach National 
Policy Influence Type Implementation circular/rapid advice/letter to e.g. Ministry of Health
 
Description Climate model with spatially varying parameters 
Organisation Meteorological Office UK
Country United Kingdom 
Sector Public 
PI Contribution Provided modified climate model parameters based on work in the Ergodics project
Collaborator Contribution Dr Simon Wilson of Met Office and NCAS made modifications to the HadCM3 climate model code to allow the Convective entrainment parameter to be spatilly varying according to a field we have provided. It took several rounds of testing to ensure the code was working ~ 2 weeks work
Impact One or two paper will be written on the results from the new code
Start Year 2016
 
Description ClimatePrediction.net 
Organisation University of Oxford
Department Atmospheric, Oceanic and Planetary Physics
Country United Kingdom 
Sector Academic/University 
PI Contribution Analysing a set of CP.net experiments set up by the Oxford team and run by distributed computing partners. Exoperiments establish a new method for identifying errors in model processes that lead to forecast drifts
Collaborator Contribution Setting up CP-net infrastructure. to allow hindacst sets wiht HadCM3 model to be performed
Impact Paper in draft should be submitted within 1 month
Start Year 2013
 
Description Testing and improving ECMWF seasonal forecasting system 
Organisation European Centre for Medium Range Weather Forecasting ECMWF
Country European Union (EU) 
Sector Public 
PI Contribution Using special project resources we carried out a series of hindacst studies of ECMWF coupled prediction system and compared skill. 1 paper is published and 1 other is being submitted now.
Collaborator Contribution Data access. Providing code, project meeitngs. ECMWF will test out new code improvements using operational suite and a large number of test experiments (using much more resources than we have had available till now). If results are successful this could lead to updates to the operaitonal forecast suite.
Impact 1 paper published, another submitted
Start Year 2013
 
Title New climate model code 
Description New climate model version allowing spatially varying process parametrisations 
Type Of Technology Software 
Year Produced 2017 
Impact Primarily research code at the moment but with potential to be further developed for operational use 
 
Title Test of Coupled model forecasting methodology 
Description We have tested how new coupled data assimilation system being implemented at ECMWF for reanalyses and forecasts, responds under different methods of initialisation. We show where the origin of the largest biases emerge from when the model configuraitons are changed for either atmosphere of ocean components. Joint conference rpesentaiotns of these results have already been made with ECMWF staff. A journal paper has been prepared and will be submitted shortly. 
Type Of Technology New/Improved Technique/Technology 
Year Produced 2014 
Impact A better understanding of the origin of initial condition shocks in coupled model forecasts has been demonstrated. There is particular danger from using reanalysis results prepared from a different model when initialising a coupled forecast as this leads to the largerst shocks to the system.