Upload
jeffry-carter
View
221
Download
0
Tags:
Embed Size (px)
Citation preview
© Crown copyright Met Office
The EN4 dataset of quality controlled ocean temperature and salinity profiles and monthly objective analysesSimon Good
© Crown copyright Met Office
What is the EN series of datasets?
• Climate datasets of ocean temperature and salinity profiles.• Include:
1. The observed profiles with quality flags.
2. Monthly objective analyses formed from the data.
• Data are in NetCDF files.
• Freely available for scientific research and private study.
• www.metoffice.gov.uk/hadobs
© Crown copyright Met Office
Contents
• History of the dataset
• The EN4 dataset
• Objective analysis uncertainty estimates
• Ensemble of opportunity
• Conclusions
© Crown copyright Met Office
History of the dataset
© Crown copyright Met Office
How did they originate?
• The first version of the dataset was produced for the ENACT project (ENhanced ocean data Assimilation and Climate predicTion) for global ocean data assimilation.
• Data were collected from a number of sources.• To get consistent quality control, all flags were
discarded from the source.• Then, all data were run through the Met Office
quality control system.• The system was shared with real time systems –
necessary to have good automated checks.
© Crown copyright Met Office
Updates to EN2 and EN3
• A new version was produced for the ENSEMBLES project (EN2; Ingleby and Huddleston, 2007, Journal of Marine Systems, 65).• Changes were updated data from sources; tweaks to
quality control checks; updated background error covariances.
• EN3 (with various subversions) followed.• Updated data again.• Revised climatology.• Tweaks to quality control checks again.• Added some manual rejects.
© Crown copyright Met Office
Update to EN4
• Targeted development of a duplicate check and new quality control checks to solve specific issues with the dataset.• Work done for the ERA-CLIM project (European
Reanalysis of Global Climate Observations; http://www.era-clim.eu/).
• Uncertainty estimates produced to accompany the objective analyses.
• Good et al. (2013), JGR-Oceans, 118.
© Crown copyright Met Office
The EN4 dataset
© Crown copyright Met Office
Input data
• Profiles obtained from • WOD• ASBO (Arctic Synoptic Basinwide Observations)• Argo• GTSPP
• Argo and GTSPP are used for monthly updates.
• Duplicate check applied to remove the multiple versions of profiles.• Based on Gronell and Wijffels (2008), JAOT, 25.
© Crown copyright Met Office
Number of profiles
© Crown copyright Met Office
Thinning of levels
CTD profile
50.9°W, 58.2°N
July 1995
Original profile: 3596 levels
EN3 version: 127 levels(~5m spacing near surface)
EN4 version: 390 levels(~1m spacing near surface)
Thinned profiles have (if available)
EN3: 5m spacing near surface
EN4: 1m spacing near surface
© Crown copyright Met Office
Monthly processing cycle
Quality control
Analysis
Persistence forecast
Output as NetCDF file
One month of observations
processed per cycle
Output as NetCDF file
Available from Met Office
website
Observations
© Crown copyright Met Office
Quality control tests
Manual exclusions
Track check
Profile check (spikes etc.)
Thinning (informational)
Stability check
Background checks
Buddy checkMulti level check
Argo delayed mode flags
Argo grey list
Argo altimetry quality control
External quality information
Automatic quality checks
Bathymetry check
Measurement depths check
Waterfall check
Near surface and deep BTs
Range check
© Crown copyright Met Office
Background check
Quality control examples
Measurement depths check
© Crown copyright Met Office
Quality control examples
Track check
Waterfall check
Spike check
© Crown copyright Met Office
Objective analysis system
• Uses a iterative scheme equivalent to optimal interpolation.• But no need to subselect a limited number of
observations close to a grid box.
• Monthly.• Potential temperature and salinity.• 1 degree grid, 42 levels in the vertical.• Uses a persistence forecast as background.• Relaxes to a climatology in the long term
absence of observations.
© Crown copyright Met Office
Example fields – March 2014 at 5 m
© Crown copyright Met Office
Objective analysis uncertainty estimates
© Crown copyright Met Office
Method
• The scheme used to produce the objective analyses makes it impossible to calculate uncertainty using the optimal interpolation equations.• These are only valid if background error covariances are well
known.
• Use an analysis quality method instead where observation values are set to 1 and background values to 0 (based on Donlon et al. 2012, Remote Sensing of Environment, 116).
• All other aspects of the analysis are the same except length scales are shortened.
• Result is linearly related to analysis error variance.
© Crown copyright Met Office
Example of observation influence
© Crown copyright Met Office
Example of observation influence
Dots are the locations of observations
© Crown copyright Met Office
Example of observation influence
© Crown copyright Met Office
Example of observation influence
© Crown copyright Met Office
Example of observation influence
© Crown copyright Met Office
Linear relationship between ‘observation influence’ and analysis error variance
• Using simulated data points it is possible to show this relationship.
• Length scales have to be shortened by a factor 1.75 to get a linear relationship.
Simulations
Line fit
© Crown copyright Met Office
Coefficients of the linear relationship• For EN4 I developed an empirical scheme to do
this.• The observations are split into two groups and
analyses made from each.• The differences between analyses allow the
uncertainty to be estimated.• The method also allows some checks to be
done:• Intercept should match the background error
variance – method can even be used to improve this• Gradients should be negative – some issues with the
EN4 analyses were identified
© Crown copyright Met Office
Gradients of the linear relationship
© Crown copyright Met Office
Example uncertainty estimates
© Crown copyright Met Office
Ensembles of opportunity
© Crown copyright Met Office
Uncertainty in construction of an ocean dataset• There are a number of global ocean profile
databases:• WOD• GTSPP• CORA• EN4• Etc.
• These are constructed differently, e.g.• Different quality control.• Different choices about which profiles to include.• Bias adjustments.
• This tells us about the uncertainty in the data.
© Crown copyright Met Office
Example
• From Lyman et al. (2010), Nature, 465.
© Crown copyright Met Office
Conclusions
© Crown copyright Met Office
Conclusions
• The EN series of datasets were originally started for global ocean data assimilation.
• This has been incrementally updated to the latest version EN4.
• It is available freely for scientific research and private study from www.metoffice.gov.uk/hadobs.
• The new version includes objective analysis uncertainty estimates produced using an ‘observation influence’ method.
© Crown copyright Met Office
Questions and answers