Data & Tools

Data assimilation tools

Data assimilation tools for objectively combining information in observations with models

Data assimilation (DA) is a powerful tool for combining the information contained in observations with models. The output can be a more accurate estimate of the earth system state than can be provided by either the observations or model alone.

Differences between DA algorithms

Most data assimilation methods can be derived from Bayes’ theorem but crucial differences exist in the algorithms:

  • Variational (Var) techniques aim to find the maximum a posteriori estimate of the state assuming Gaussian error statistics (i.e. they can be fully quantified by their mean and covariances) and near-linear models. Var often makes use of a climatological estimate of the prior error covariances and considers a time window of observations.
  • Ensemble data assimilation techniques, such as EnKF, assimilate observations sequentially in time and use an ensemble of model states to evolve the prior uncertainty. The EnKF is similar to Var in that it still assumes that the error statistics are Gaussian.
  • Lastly there are techniques such as the particle filter which avoid making any assumptions about the error distributions and try to provide a complete representation of the posterior, whether that be Gaussian, skewed or multi-modal.

The best method to use depends on the intended application, as the validity of assumptions and the data available differ substantially between e.g. monitoring changes to the land surface, numerical weather prediction and forecasting the ocean ecosystem.

Our collection of tools for DA

The tools provided here include EMPIRE which allows for any model to be used with ensemble data assimilation techniques such as the EnKF and the particle filter. Also included are application-specific DA tools such as those for the ocean (NEMOVar) and for the land surface (EOLDAS and CARDAMON).

Use the links below to auto scroll to content
DA applied to land surface
CARDAMOM merges various versions of the DALEC ecosystem model with data-assimilation and inverse procedures. It also includes a variety of wrappers / functions written in Python and R to create the input data and process the output of the DA procedure.

Contact: Jeff Exbrayat (University of Edinburgh).

View Here
EOLDAS (Earth Observation Land Data Assimilation Scheme)
A data assimilation framework developed initially via ESA funding as part of the Support to Science Element (STSE) of the Earth observation Envelope Programme. The aim of EOLDAS is to allow (inter alia) merging of data with different observation characteristics (particularly bands, and angular sampling, but also spatial and temporal), and models of the surface state (observation operators). Development of EOLDAS is continuing under various streams of funding, particularly NCEO, to include spatial constraints on model inversion as well as extension of the operator aspect into the thermal domain.

Contact: Philip Lewis (UCL) or Jose Gomez-Dans (UCL)
View Here
DA applied to the ocean
This is a data assimilation code suite that has been developed through collaboration between the Met Office, ECMWF and CNRS Toulouse. Operationally it is currently run as a 3DVar FGAT scheme but the project team has ongoing development of many new aspects, including 4DVar, ensemble management methods, and the use of EOFs for background error modelling. The code is currently being ported to Archer by NCAS-CMS staff for wider use in the academic community.

Contact: Keith Haines (University of Reading)

View Here
Ensemble DA techniques
EMPIRE (Employing MPI for Researching Ensembles)
This tool allows an easy connection of models to a number of different data assimilation methods. The ocean models NEMO and GOTM-ERSEM are currently embedded in this tool. In particular, biogeochemical ocean data are assimilated into ERSEM by using the Equivalent Weights Particle filter (EWPF) and the Localised Ensemble Transform Kalman Filter (LETKF).

Contact: Peter Jan (University of Reading)

View Here
EnKF (Ensemble Kalman filter)
This is a sequential data assimilation method that applies an ensemble of model states to represent the error statistics of the model estimates. The ensemble is integrated to predict the error statistics forward in time, and it uses an analysis scheme which operates directly on the ensemble of model states when observations are assimilated.

Contact: Stefano Ciavatta (Plymouth Marine Laboratory)

View Here

Discover More