On the representation error in data assimilation Janjić, T.; Bormann, N.; Bocquet, M. ...
Quarterly journal of the Royal Meteorological Society,
April 2018 Part B, Letnik:
144, Številka:
713
Journal Article
Recenzirano
Odprti dostop
Representation, representativity, representativeness error, forward interpolation error, forward model error, observation‐operator error, aggregation error and sampling error are all terms used to ...refer to components of observation error in the context of data assimilation. This article is an attempt to consolidate the terminology that has been used in the earth sciences literature and was suggested at a European Space Agency workshop held in Reading in April 2014. We review the state of the art and, through examples, motivate the terminology. In addition to a theoretical framework, examples from application areas of satellite data assimilation, ocean reanalysis and atmospheric chemistry data assimilation are provided. Diagnosing representation‐error statistics as well as their use in state‐of‐the‐art data assimilation systems is discussed within a consistent framework.
Representation, representativity, representativeness error, forward interpolation error, forward model error, observation‐operator error, aggregation error and sampling error are all terms used to refer to components of observation error in the context of data assimilation. This article is an attempt to consolidate the terminology that has been used in the earth sciences literature and was suggested at a European Space Agency workshop held in Reading in April 2014. We review the state of the art and, through examples, motivate the terminology.
Phytoplankton functional‐type (PFT) data are assimilated into the global coupled ocean‐ecosystem model MITgcm‐REcoM2 for two years using a local ensemble Kalman filter. The ecosystem model has two ...PFTs: small phytoplankton (SP) and diatoms. Three different sets of satellite PFT data are assimilated: Ocean‐Color‐Phytoplankton Functional Type (OC‐PFT), Phytoplankton Differential Optical Absorption Spectroscopy (PhytoDOAS), and SYNergistic exploitation of hyper‐ and multi‐spectral precursor SENtinel measurements to determine Phytoplankton Functional Types (SynSenPFT), which is a synergistic product combining the independent PFT products OC‐PFT and PhytoDOAS. The effect of assimilating PFT data is compared with the assimilation of total chlorophyll data (TChla), which constrains both PFTs through multivariate assimilation. While the assimilation of TChla already improves both PFTs, the assimilation of PFT data further improves the representation of the phytoplankton community. The effect is particularly large for diatoms where, compared to the assimilation of TChla, the SynSenPFT assimilation results in 57% and 67% reduction of root‐mean‐square error and bias, respectively, while the correlation is increased from 0.45 to 0.54. For SP the assimilation of SynSenPFT data reduces the root‐mean‐square error and bias by 14% each and increases the correlation by 30%. The separate assimilation of the PFT data products OC‐PFT, SynSenPFT, and joint assimilation of OC‐PFT and PhytoDOAS data leads to similar results while the assimilation of PhytoDOAS data alone leads to deteriorated SP but improved diatoms. When both OC‐PFT and PhytoDOAS data are jointly assimilated, the representation of diatoms is improved compared to the assimilation of only OC‐PFT. The results show slightly lower errors than when the synergistic SynSenPFT data are assimilated, which shows that the assimilation successfully combines the separate data sources.
Key Points
Phytoplankton‐type assimilation corrects the phytoplankton fields stronger than total chlorophyll assimilation in particular for diatoms
The phytoplankton‐type assimilation reduces the exclusion of the different groups so that both types coexist
The joint assimilation of two phytoplankton‐type satellite products leads to further improvement of the assimilation results
The coupled ocean circulation‐ecosystem model MITgcm‐REcoM2 is used to simulate biogeochemical variables in a global configuration. The ecosystem model REcoM2 simulates two phytoplankton groups, ...diatoms and small phytoplankton, using a quota formulation with variable carbon, nitrogen, and chlorophyll contents of the cells. To improve the simulation of the phytoplankton variables, chlorophyll‐a data from the European Space Agency Ocean‐Color Climate Change Initiative (OC‐CCI) for 2008 and 2009 are assimilated with an ensemble Kalman filter. Utilizing the multivariate cross covariances estimated by the model ensemble, the assimilation constrains all model variables describing the two phytoplankton groups. Evaluating the assimilation results against the satellite data product SynSenPFT shows an improvement of total chlorophyll and more importantly of individual phytoplankton groups. The assimilation improves both phytoplankton groups in the tropical and midlatitude regions, whereas the assimilation has a mixed response in the high‐latitude regions. Diatoms are most improved in the major ocean basins, whereas small phytoplankton show small deteriorations in the Southern Ocean. The improvement of diatoms is larger when the multivariate assimilation is computed using the ensemble‐estimated cross covariances between total chlorophyll and the phytoplankton groups than when the groups are updated so that their ratio to total chlorophyll is preserved. The comparison with in situ observations shows that the correlation of the simulated chlorophyll of both phytoplankton groups with these data is increased whereas the bias and error are decreased. Overall, the multivariate assimilation of total chlorophyll modifies the two phytoplankton groups separately, even though the sum of their individual chlorophyll concentrations represents the total chlorophyll.
Plain Language Summary
Different types of plankton are simulated globally with ocean ecosystem models. To further increase their prediction quality, we combine the model with satellite observations of chlorophyll using modern methods called data assimilation. This method allows us not only to improve the modeled total chlorophyll but also the simulation of the different plankton types. Further, we can fill gaps in the satellite data that results, for example, from clouds. Thus, we are able to better predict the ocean ecosystem, which in turn helps to understand climate change patterns and carbon cycle processes.
Key Points
Multivariate assimilation of satellite chlorophyll data improves simulation of phytoplankton functional groups and influences them differently
Small phytoplankton is weakly deteriorated in the Southern Ocean, while diatoms are improved globally
Regional variability of assimilation leads to stronger improvement at midlatitudes and equator than at high latitudes
Weak constraint parameter estimation is applied to tune a 0-D 4-component (phytoplankton, zooplankton, detritus, dissolved inorganic nitrogen) model for the ecosystem of the upper mixed layer in the ...North Atlantic. The model is constrained by the monthly mean Nimbus-7 chlorophyll data. The basin under consideration is separated into 5×5° boxes where the model is optimized independently. Adjusted model parameters exhibit significant spatial variations. We also show how to evaluate a posteriori the relative credibility of the data and the model. A method for estimating the accuracy of model parameterizations is suggested. For Bermuda Station “S”, the major contribution to model errors comes from uncertainties in parameterizations for the phytoplankton mortality, zooplankton losses and breakdown of detritus.
A Sequential Importance Resampling filter (SIR) is applied to assimilate data of the Bermuda Atlantic Time-Series Study for the period December 1988 to January 1994 into a nine-compartment ecosystem ...model. The filter provides an opportunity to combine state and parameter estimations. We detected notable seasonality of some model parameters. A filtered solution is in close agreement with the data and is superior to that obtained with fixed model parameters. The seasonal dependence of the initial slope of the
P–
I curve is similar to other known estimates. The seasonality of the phytoplankton specific mortality rate obtained can point out that either the phytoplankton mortality parameterization has to be improved or the Chl:C ratio varies in time. Being of the same computational cost as the Ensemble Kalman filter, the data assimilation approach used can be implemented for on-line tuning and operational prediction the ecosystem dynamics with a coupled hydrodynamical–ecosystem model.
Data assimilation experiments that aim at improving summer ice concentration and thickness forecasts in the Arctic are carried out. The data assimilation system used is based on the MIT general ...circulation model (MITgcm) and a local singular evolutive interpolated Kalman (LSEIK) filter. The effect of using sea ice concentration satellite data products with appropriate uncertainty estimates is assessed by three different experiments using sea ice concentration data of the European Space Agency Sea Ice Climate Change Initiative (ESA SICCI) which are provided with a per-grid-cell physically based sea ice concentration uncertainty estimate. The first experiment uses the constant uncertainty, the second one imposes the provided SICCI uncertainty estimate, while the third experiment employs an elevated minimum uncertainty to account for a representation error. Using the observation uncertainties that are provided with the data improves the ensemble mean forecast of ice concentration compared to using constant data errors, but the thickness forecast, based on the sparsely available data, appears to be degraded. Further investigating this lack of positive impact on the sea ice thicknesses leads us to a fundamental mismatch between the satellite-based radiometric concentration and the modeled physical ice concentration in summer: the passive microwave sensors used for deriving the vast majority of the sea ice concentration satellite-based observations cannot distinguish ocean water (in leads) from melt water (in ponds). New data assimilation methodologies that fully account or mitigate this mismatch must be designed for successful assimilation of sea ice concentration satellite data in summer melt conditions. In our study, thickness forecasts can be slightly improved by adopting the pragmatic solution of raising the minimum observation uncertainty to inflate the data error and ensemble spread.
Satellite sensor configurations enabling high spatial and high spectral resolution could be far more suited for monitoring inland and coastal water ecosystems' water quality than common ocean color ...sensors. It is expected that even the composition of phytoplankton, the primary producer in these ecosystems, could be determined. Up to now, atmospheric correction is limiting the exploitation of these sensors' data. Here, we evaluate the atmospheric correction method Polymer applied to hyper- and multispectral satellite data (HICO, DESIS and OLCI) over coastal and inland waters. We assess the quality of retrieved water reflectance and biomass for all and specific phytoplankton groups by comparison to in-situ matchup data and results from other retrieval methods.
This study presents an algorithm for globally retrieving chlorophyll a (Chl-a) concentrations of phytoplankton functional types (PFTs) from multi-sensor merged ocean color (OC) products or ...Sentinel-3A (S3) Ocean and Land Color Instrument (OLCI) data from the GlobColour archive in the frame of the Copernicus Marine Environmental Monitoring Service (CMEMS). The retrieved PFTs include diatoms, haptophytes, dinoflagellates, green algae and prokaryotic phytoplankton. A previously proposed method to retrieve various phytoplankton pigments, based on empirical orthogonal functions (EOF), is investigated and adapted to retrieve Chl-a concentrations of multiple PFTs using extensive global data sets of in situ pigment measurements and matchups with satellite OC products. The performance of the EOF-based approach is assessed and cross-validated statistically. The retrieved PFTs are compared with those derived from diagnostic pigment analysis (DPA) based on in situ pigment measurements. Results show that the approach predicts well Chl-a concentrations of most of the mentioned PFTs. The performance of the approach is, however, less accurate for prokaryotes, possibly due to their general low variability and small concentration range resulting in a weak signal which is extracted from the reflectance data and corresponding EOF modes. As a demonstration of the approach utilization, the EOF-based fitted models based on satellite reflectance products at nine bands are applied to the monthly GlobColour merged products. Climatological characteristics of the PFTs are also evaluated based on ten years of merged products (2002−2012) through inter-comparisons with other existing satellite derived products on phytoplankton composition including phytoplankton size class (PSC), SynSenPFT, OC-PFT and PHYSAT. Inter-comparisons indicate that most PFTs retrieved by our study agree well with previous corresponding PFT/PSC products, except that prokaryotes show higher Chl-a concentration in low latitudes. PFT dominance derived from our products is in general well consistent with the PHYSAT product. A preliminary experiment of the retrieval algorithm using eleven OLCI bands is applied to monthly OLCI products, showing comparable PFT distributions with those from the merged products, though the matchup data for OLCI are limited both in number and coverage. This study is to ultimately deliver satellite global PFT products for long-term continuous observation, which will be updated timely with upcoming OC data, for a comprehensive understanding of the variability of phytoplankton composition structure at a global or regional scale.
•Multi-sensor ocean color remote sensing data are used to estimate six PFTs.•Chlorophyll a concentrations of multiple PFTs are well retrieved globally.•The retrieved PFTs are very comparable to other equivalent products.•Plausible PFT dominance information can be derived on a global scale.•First application to OLCI shows high potential in a continuous long-term observation.
The impact of assimilating weekly CryoSat‐2 sea ice thickness data together with daily SMOS sea ice thickness and daily SSMIS sea ice concentration data on the sea ice fields of a coupled sea ...ice–ocean model of the Arctic Ocean is investigated. The sea‐ice model is based on the Massachusetts Institute of Technology general circulation model (MITgcm) and the assimilation is performed by a localized Singular Evolutive Interpolated Kalman (LSEIK) filter coded in the Parallel Data Assimilation Framework (PDAF). A period of three months from 1 November 2011 to 30 January 2012 is selected to assess the skill of the assimilation system in the cold season. Compared to the unassimilated solution and a solution where only sea ice concentration is assimilated, the model–data misfits are substantially reduced in areas of both thick and thin ice. The sea ice thickness estimates agree significantly better with in situ observations in the central Arctic Ocean than the sea ice thickness obtained from assimilating SMOS data alone, while the sea ice concentration shows very small improvements. The sea ice fields obtained by the joint assimilation of SMOS and CryoSat‐2 data also have lower errors in thickness and concentration than those obtained from directly assimilating a statistically merged SMOS and CryoSat‐2 sea ice thickness product. These lower errors suggest that model dynamics play a significant role in data blending.
Temporal evolution of RMSE between Exp_Ctrl (grey solid), Exp_SSMIS (blue solid), Exp_SM&CS2 (black solid), Exp_SM (red dashed) and (a) SMOS sea ice thickness (0–1.0 m), (b) CryoSat‐2 sea ice thickness from 1 November 2011 to 30 January 2012. For thickness over valid CryoSat‐2 area (b), the RMSE are computed relative to weekly CryoSat‐2 data.
A search for cosmic neutrino sources using the data collected with the ANTARES neutrino telescope between early 2007 and the end of 2015 is performed. For the first time, all neutrino ...interactions-charged- and neutral-current interactions of all flavors-are considered in a search for point-like sources with the ANTARES detector. In previous analyses, only muon neutrino charged-current interactions were used. This is achieved by using a novel reconstruction algorithm for shower-like events in addition to the standard muon track reconstruction. The shower channel contributes about 23% of all signal events for an E−2 energy spectrum. No significant excess over background is found. The most signal-like cluster of events is located at (α,δ)=(343.8°,23.5°) with a significance of 1.9σ. The neutrino flux sensitivity of the search is about E2dΦ/dE=6×10−9 GeV cm−2 s−1 for declinations from −90° up to −42°, and below 10−8 GeV cm−2 s−1 for declinations up to 5°. The directions of 106 source candidates and 13 muon track events from the IceCube high-energy sample events are investigated for a possible neutrino signal and upper limits on the signal flux are determined.