European Commission logo
English English
CORDIS - EU research results
CORDIS
Content archived on 2024-05-23

Origin and fate of biogenic particle fluxes in the ocean and their interaction with the atmospheric co2 concentration as well as the marine sediment

Deliverables

A model of the marine carbon cycle (HAMOCC3, Hamburg Model of the Ocean Carbon Cycle) was driven off-line by glacial oceanic circulation and dust input fields. As a novelty developed within ORFOIS, the model incorporates both the open and the closed system components that constitute the marine carbon cycle. The newly added open system component consists of Aeolian dust and fluvial CaCO3 input (the latter kept constant in both experiments) and subsequent burial and dissolution of CaCO3 in the sediment. The closed system component consists of the biological pump, the counter pump as well as the internal redistribution by the oceanic circulation. Thus the model used here can be considered an improvement over the more often used and easier to handle but less realistic closed system models, in which river input of CaCO3 is used to balance the loss from burial to the sediment. The closed system models proved to be fairly insensitive to changes in the potential driving mechanisms for glacial pCO2 reduction. The oceanic response to the glacial forcing is decreased total CO2, increased total alkalinity, and decreased PO4. The atmospheric pCO2 drops to some 230 ppm after 8,000 years of integration. The drawback is that the lysocline is below the sea floor everywhere (not shown). We still think that we have identified and installed a potentially crucial and efficient mechanism to lower pCO2. This is one step forward to identify the control mechanisms for glacial/ interglacial pCO2 changes, which have been tried to identify and simulate now for more than 20 years.
We estimated the marginal damage costs of carbon dioxide emissions at 7-33/tC, with a maximum of 50/tC. We estimated the marginal abatement costs of carbon dioxide emissions at 10-20/tC, with a maximum of 100/tC. We predict that Denmark, Iceland, New Zealand and Norway will advocate the inclusion of ocean sinks in the successor of the Kyoto Protocol, and will meet little opposition.
The physical ocean model fields driving the biogeochemical module HAMOOC5 were computed under anomaly forcing using the ECHAM3/OPIC results from the GHG experiment (Roeckner et al., 1999). Changes in the ocean carbon uptake due to an increased atmospheric CO2-concentration were determined from being integrated with the HAMOCC5 model. The atmospheric CO2-concentration was prescribed in the period 1860-2050 according to the IS92a-scenario. This IS92a implies a "business-as-usual" scenario resulting in a rapid increase in the atmospheric CO2-level such that the pCO2 level has reached 519ppm by 2050. The carbon uptake experiments were started from a 1200-year spin-up of the physical model. The nutrients, oxygen and biomasses were started from a 650-year integration. Initial conditions for total CO2 and alkalinity were obtained by starting integration from present day total CO2 and alkalinity fields (gridded results from the GLODAP project, R. Key et al., 2004). The gridded data of total CO2 and alkalinity was interpolated onto the HAMOCC5 grid for each vertical level, in the same way as the two-dimensional ECHAM4/OPYC3 forcing fields were interpolated. Finally, the fields were linearly interpolated onto the depth levels in the HAMOCC5 grid. The GLODAP data only covers the ocean up to 64 N. North of this latitude the total CO2 and alkalinity was estimated from linear relationships to salinity. These relations were determined from total CO2, alkalinity and salinity measured in the Nordic seas. Initially the model was integrated with an atmospheric pCO2 = 280ppm to obtain a DIC-field in equilibrium with pre-industrial conditions. The HAMOCC5 IS92a-experiment presented in this report is from a 35 year spin-up with a pCO2 = 280ppm. This implies that the upper ocean has not equilibrated to the pCO2 = 280ppm. Therefore the first century of the integration is not representative for the ocean uptake in this period, and in the final part of the integration only the time dependent trends are representative for the oceanic carbon uptake. The surface alkalinity field is relaxed toward the observed field in a way similar to the relaxation performed on the surface salinity. The CO2 air-sea gas exchange depends on the fractional ice-cover. The initial conditions for the IS92a were not equilibrated with the pre-industrial conditions. However, a longer 100-year integration was made with an atmospheric pCO2 level of 280. This solution was closer to equilibrium than the solution used for the initial conditions.
The importance of the coastal ocean and river-dominated depositional environments as sinks of nutrients such as silica is uncertain. Studies in tropical deltaic systems (i.e. Amazon delta) have shown that biogenic silica particles upon deposition to the sediments undergo rapid conversion to cation-rich alumoninosilicate phases. The reactions that lead to the formation of new cation-rich aluminosilicate phases are considered as potentially important for the global elemental budgets of major cations (K, Mg) , minor elements (F) and nutrients (Si). During the cation uptake reactions, there is a return of CO2 from the oceans to the atmosphere, a feedback mechanism that returns CO2 consumed during silicate weathering on land. Thus these reactions are known as "reverse weathering reactions." As part of ORFOIS we have investigated the role of biogenic silica as a substrate for the formation of newly formed aluminosilicate phases in the Ganges-Brahmaputra and Mississippi deltaic systems. Operational analytical methods, both standard and modified, were used for the determination of reactive silica content in these sediments. Fresh, unaltered biogenic silica (BSi) is measured with the most generally accepted operational method that utilises the time-dependent release pattern of dissolved Si from sediment exposed to 1% Na2CO3 (~0.1 M) at 85°C over ~ 2 - 10 hrs. For the measurement of altered biogenic silica particles and other authigenic aluminosilicate phases in addition to fresh biogenic silica, a two-step operational leach method is used. First, a mild acid digestion (0.1N HCl for 18 hours, room T) is used in order to remove metal coatings, poorly crystalline authigenic minerals, and to activate biogenic silica surfaces for subsequent alkaline dissolution, prior to leaching in alkaline solution. Subsequently the sample is analysed with the time-dependent release alkaline method described before. On the basis of area-weighted sedimentation rate patterns and the corresponding reactive silica concentration in the two deltaic systems we were able to determine the unaltered biogenic silica and the total reactive silica (i.e. biogenic silica + converted biogenic silica+silica in newly formed aluminosilicates) stored in these systems. Extrapolation of the existing results for the Amazon and the newly acquired results for the Mississippi and Ganges-Brachmaputra to the global coastal ocean show that deltaic systems are significantly more important sinks of riverine dissolved silica than previously thought. In addition observations and chemical analyses on newly formed cation rich aluminosilicate phases on siliceous substrates were obtained with electron microscopy in an attempt to estimate on the importance of these reactions as a CO2 source to the atmosphere on geologic time scales.
Partner 1 completed the work on the forcing data sets. The complete dataset of the GSDIO experiment (ECHAM3/OPYC coupled climate model warming run including aerosols as used for input to anthropogenic future runs with BOGCM I) are now stored on 14 DVD's. Also the data set for the GHG experiment (same as GSDIO but without aerosols) was prepared based on monthly averaged dataset available for the period 1860-2100 was compiled. The GSDIO and GHG-forcing fields have been compared with the OMIP-forcing fields and verified.
During OROIS, a sediment module was implemented to the standard PISCES model. The development of the sediment module closely followed Heinze et al. (1999). The description of POC remineralisation, limited to oxide degradation in Heinze et al. (1999), was extended to sub-oxide nitrate reduction. Off-line simulations were forced with bottom water conditions of dissolved Si, oxygen, phosphate, alkalinity, nitrate, and DIC, as well as yearly average fluxes of POC, clay, opal and carbonate provided by the PISCES standard version. Stable distributions of dissolved and solid sedimentary tracers were obtained after 50kg of integration. The computed sediment composition was then used as initial conditions for the fully coupled simulations. Due to the long equilibration time of the coupled water-column/surface sediment system, the simulation is still running. Results of the coupled run are expected to be exploitable early in 2005. In parallel and with the objective to validate the particle dynamics, we used yearly averaged fluxes computed by the PISCES standard version STD and the modified version STD2 to force the sediment module in an off-line mode. This approach takes advantage of the benthic compartment as the ultimate sediment trap to validate deep-water fluxes. Based solely on the bulk sediment composition, it is not possible to discriminate between PISCES STD and PISCES STD2. The spatial pattern of SiO2 and CaCO3 distributions are reproduced by both versions. Computed values are however to low for areas rich in CaCO3 like for instance the mid-Atlantic ridge. The underestimation of maximum CaCO3 levels is stronger for PISCES STD2. On the contrary, this model version overestimates the POC concentration of surface sediments. Since results are represented on a weight percent basis, the increase in one component will result in a corresponding decrease of the other. The straightforward comparison between model output and observations of bulk sediment variables has thus a limited value. We extended our evaluation of model output to the benthic oxygen fluxes. The latter are linked to the mineralization activity. They integrate POC supply, oxygenation of bottom waters and oxygen demand due to aerobic metabolism. Figure 4 compares estimates of benthic O2 fluxes by Jahnke (1996) to model output. Oxygen fluxes computed by the model forced with fluxes from PISCES STD are largely underestimated. The spatial pattern present in Jahnke's estimates is not reproduced. In comparison, the model version STD2 yields a satisfying spatial distribution of fluxes. They are however overestimated. These results are consistent with the conclusions drawn from the comparison of modelled and observed particle fluxes in the water column. They emphasize the correct representation of the fate of sinking fluxes in order to represent the coupled pelagic-benthic system.
Compilation of multi-tracer data sets: At the ORFOIS Modelling Meeting (March 21-22, 2002 in Yerseke, The Netherlands) a decision was taken concerning "What variables (and in what units) do the modellers want to be seen in the database?" During the ORFOIS First annual workshop (January 20-21, 2003 in Hamburg, Germany) this decision was even more specified that this compilation of multi-tracer data sets for the model validation would be desirable for the following sites (in alphabetical order); AESOPS (Southern Ocean, Pacific sector); BATS: Bermuda Atlantic Time-series Study & Hydrostation "S" Research; BENGAL: Benthic Biology and Geochemistry of a Northeastern Atlantic Abyssal Locality; BIGSET: Biogeochemische Stoff- und Energietransporte in der Tiefsee / Biogeochemical Transport of Matter and Energy in the Deep Sea; EqPac: Equatorial Pacific Process Study; ESTOC: European Station for Time-series in the Ocean, Canary Islands; EUMELI: - no transcription found, yet; HOT: Hawaii Ocean Time-Series; KERFIX (Southern Ocean, Indian sector): A fixed time-series station in the Southern Ocean near the Kerguelen Islands; NABE: North Atlantic Bloom Experiment; OMEX: Ocean Margin exchange project; PAP: The Porcupine Abyssal Plain Observatory - time series observations in the open ocean; PAPA: Ocean Station PAPA; WS1: Weddell Sea Station 1. Compilation of global sediment trap flux data sets: The work on the global sediment trap flux data compilation is still under work and should be published soon. A preliminary version, however, is available upon request from O. Ragueneau, UBO Brest, France. (cf. Ragueneau et al. (2000): A review of the Si cycle in the modern ocean: recent progress and missing gaps in the application of biogenic opal as a paleoproductivity proxy, Global and Planetary Change, 26(4), 317-365) Compilation of global solid surface sediment data sets. Through the model development, ORFOIS provide tools for the study of global and regional environmental issues (marine CO2 source/sink distribution and its spatial as well as temporal change, carbon nutrient redistribution in the water column and the sediment, atmospheric CO2 partial pressure, see p. 16 of the Work Programme EESD, Part A). There are many thoughts what depth would be the best to retrieve a maximum of valuable data. We decided to [wrongly] define the sediment surface as the depths interval between 0 cm and 35 cm sediment depth. This is scientifically not correct and corresponds rather to a technical working solution. Thus: If you refer to other sediment depths as given here in order to model the surface sediment of the global ocean, please, cut the data that you need out of the data files. Compilation of global water column data sets: Through the model development, ORFOIS provide tools for the study of global and regional environmental issues (marine CO2 source/sink distribution and its spatial as well as temporal change, carbon nutrient redistribution in the water column and the sediment, atmospheric CO2 partial pressure, see p. 16 of the Work Programme EESD, Part A). Around the globe, there are many data collections available that contain water column concentrations of Oxygen, Phosphate, Silicate, Alkalinity, Dissolved Inorganic Carbon, Dissolved Organic Carbon, and Nitrate, as it is requested by the modelling group. However, it is up to the single scientist to decide whether she/he wants to include other data than the present data compilations. WDC-MARE data collections yet explicitly exclude WOD 2001 and WOA 2001. Since we offer global data compilations as ODV collections, you may add data as necessary. In the aftermath of ORFOIS the several MB comprising high quality data collections eWOCE Electronic Atlas of WOCE Data (Schlitzer, 2000) and Hydrographic Atlas of the Arctic Ocean (Nikiforov, 2001) will be included in the ORFOIS compilation at WDC-MARE and then multiply the actual data compilation.
Optimising the model by increasing the Fe-scavenging rate: The comparison of the vertical distribution of dissolved Fe in the standard model HAMOCC5 with observed Fe data revealed, that although the model reflects the deep concentrations of dissolved iron, it strongly overestimates the dissolved iron in the surface ocean layers. Given the model's quite high iron input via dust deposition mainly in the North Atlantic, and the long integration time which distributes the iron within the whole ocean, this of course leads to a weak (if any) iron limitation in the model. The parameterisation of dissolved iron scavenging in the standard configuration, was based on the paper by Johnson et al. (1997), with a time constant of 0.005 yr-1 for relaxation of excess dissolved iron to values of 0.6nmol L-1. However, Johnson et al. estimated this rate constant from a model that applied an upper (50m) boundary condition of 0.05 nmol/L for dissolved iron. There is no such upper boundary in the model presented here, but surface iron concentrations can become whatever the model suggests, and thus the low scavenging rate constant may be inappropriate in the context of HAMOCC5.1. Further, this may influence the small effect of Fe-fertilization in the standard scenario. In order to more thoroughly investigate the effect of Fe-fertilization on CO2 gas exchange in an experiment we have increased the scavenging rate by a factor of 100 in year 870 of the standard BGC model simulation, and ran it for 400 years, i.e. both the standard run and the "high Fe-complex" run have the same age. The resulting dissolved iron concentrations in the upper layers much better fit the observations that the standard run does. The increased iron limitation especially in the Southern Ocean (and also equatorial and northern Pacific) leads to lower Chl concentrations, which also better fits the observations. Reduced production leads to reduce export; this is evident e.g., from the plots of fluxes vs. latitude. Differences between the two scenarios appear mainly in the deep sedimentation in the southern latitudes, where the model with increased Fe-scavenging rate shows a better agreement with observations. Summarizing, it seems that the parameterisation of high Fe-scavenging leads to a better agreement of model surface and deep concentrations and fluxes especially in the Southern Ocean. The effect of Fe-fertilization: At first, an experiment with the standard model was carried out. Simply simply switching the Fe-limitation for the whole ocean tested the effect of Fe-fertilization. This has been done for ten years of model simulation; after that, Fe-limitation was switched on again, and the model simulated for another 20 years. Omitting Fe-limitation basically has an effect on the fluxes in the Southern Ocean; here, all fluxes are increased, and the ocean takes up more CO2. There is but little effect in the other oceanic regions. After an initial strong increase of global fluxes (the ocean even turns from source to sink), the model even within the Fe-fertilization period starts to relax back to the initial values (e.g., for air sea gas exchange, the mode in the end only shows a very weak net uptake any more). After the standard model run, an experiment with the high Fe-scavenging rate was performed. Enhancing the effect of Fe-limitation by increasing the Fe-scavenging rate over the 400 years prior to the fertilization, the initial response of the ocean to fertilization is much stronger; but the effect vanished quickly, and in the end (after a relaxation of 20 years), the air sea gas exchange is more or less the same as at the start of the fertilization period.
Development/validation of a biogeochemical ocean general circulation model –PISCES is a biogeochemical model which simulates the marine biological productivity and describes the biogeochemical cycles of carbon and the main nutrients (P, N, Si, Fe) in the ocean. In 2004, a stable release of the model became available to the community on the OPA website (hhtp://www.lodyc.jussieu.fr/opa). Model developments within ORFOIS essentially consisted in improving the description of biogeochemical fluxes below the euphotic zone and in adding a sediment model (Heinze et al. 1999). Thanks to the large database of sediment traps and sediment composition put together, the ORFOIS project enabled us to test / validate the new developments. Model studies focused on: - Particle fluxes in the mesopelgaic: sensitivity experiments and validation. We have tested the parameterisations / parameters of some of the processes represented in the PISCES standard version by carrying out sensitivity experiments. We have found that many of the processes (sinking speed, aggregation, flux feeding) introduced in the model have the ability to substantially modify deep fluxes. In particular, the way we parameterise flux feeding on particles has been found to be very sensitive. While the flux feeding intensity strongly modulates particle fluxes in the intermediate and deep water column, it has only little impact on surface ocean productivity. - Iron fertilization experiments. Example of experiment: patchy Fe additions. Patchy Fe additions were performed at various locations of the three main Fe limited regions to mimic the small-scale in-situ fertilization experiments. Except in the sub-arctic Pacific, the model broadly reproduces the main observed features of the biological response to the supply of iron (increased in total chlorophyll, diatoms relative abundance, export production and decreased pCO2). In the Southern Ocean, the magnitude and the timing of the response display large variations between the different sites. The major controlling factors are the background iron and silicate concentrations, the physical environnement and the timing relative to the growing season. We show in particular a strong correlation between the maximum of the chlorophyll response and the depth of the mixed layer at the fertilized point. - Global warming experiment: PISCES was coupled to the IPSL climate coupled model (IPSL-CM4). A global warming experiment was carried out from 1xCO2 (pre-industrial state) to 4xCO2, with a 1% per year increase in atmospheric pCO2. The objective was to investigate how global warming may affect the surface ecosystem, uptake of carbon from the atmosphere, recycling and export of carbon from the euphotic zone and re-mineralisation of carbon in the water column.
The processes controlling preservation and recycling of particulate biogenic silica in superficial sediments must be understood in order to calculate oceanic mass balances for this element. Three models representing early digenesis of silica were used to calculate the vertical distributions of pore water and solid phase silica in the sediment. The first model contains one type of biogenic silica and dissolution rate is constant, the second model contains one type of biogenic silica and dissolution rate is variable with sediment depth (term of aging), the third contains two types of biogenic silica which differ by their reactivity (fast dissolving and slow dissolving are constants). The porosity is variable exponentially with sediment depth. An explicit term of re-precipitation was incorporated into these models. The distributions of pore water and solid silica predicted by the steady state models were compared to experimental data reported from the Southern Ocean, the Equatorial Pacific and the North Atlantic (Porcupine Abyssal Plain). Models cannot reflect the observed data for dissolved and solid silica using the measured biogenic silica flux and dissolution rates. After parameter estimations, the first model cannot reflect the observed data for dissolved and solid silica. For the second and third models, good agreements were found between predicted and measured dissolved and solid silica, over a wide range of oceanic areas and sediment composition, from opal rich to opal poor sediments. The apparent silica dissolution rate constants, the biogenic silica flux deposited at the sediment-water interface, the dissolved silica re-precipitation rate constant, and the saturation concentration for silica re-precipitation were the critical parameters. The non steady state of the second and third models was tested on the Southern Ocean, the Equatorial Pacific and the North Atlantic.
The HAMOCC5 model in its basic version includes the three reservoirs atmosphere, water column, and bioturbated sediment. The full carbon, nitrogen and silicon cycles are simulated. The sediment part includes the dissolved and solid components for organic carbon (dissolved form TCO2), silicon, inorganic carbon (CaCO3) and clay. The particle aggregation mechanism according Kriest and Evans (1999, 2000) was implemented. As water column processes and sediment accumulation occur on different time scales, an acceleration method was developed which allows the build-up of the model sediment within the normal spin-up times of the model (within a few months turnaround time on the computer corresponding to 1000-1500 model years).
The quantification of in-situ carbonate dissolution rates relies on an accurate parameterisation of carbonate dissolution kinetics, which implies an improved knowledge of the rate constant and the solubility product. Dissolution, both it in the water column and surface sediments, is driven by the departure from solubility with respect to the particular carbonate phase. The solubility product evolves with depth under the combined influence of pressure and temperature. Because biogenic marine carbonates are complex solid-solution mixtures, their solubility product and dissolution kinetics are poorly known. Experimental Work: - Solubility: [Abstract reproduced from "Reassessing the dissolution of marine carbonates, part 1" by Gehlen et al., submitted to Deep Sea Research I]. We studied the solubility of the [63 - 150 microm] and the greater than 150 microm size fractions of sediments from two bathymetric transects in the eastern tropical Atlantic (Sierra Leone rise and Cape Verde Plateau). Both fractions are mainly made of foraminiferal shells and fragments. We determined the calcite crystallinity (width at half weight of XRD (104) peak) of the >150µm size fraction. Equilibration experiments were carried out in artificial seawater (20 degrees Celsius, pCO2 = 3100ppm) for up to 57 days starting from under-saturation with respect to calcite and super-saturation with respect to aragonite. Experiments starting from super-saturation indicate the presence of trace amounts of aragonite in samples from the shallowest stations of both transects. Concentration products computed for the deeper stations are intermediate between aragonite and calcite solubility. Our results are compatible with the precipitation of a high-Mg coating. The equilibration period was too short to allow the complete re-crystallization of these Mg-rich overgrowths. Experiments initiated from under-saturation yield concentration products similar to previously reported estimates for biogenic calcite (Keir, 1976). They are between 4 to 24% higher than the stoichiometric concentration product of synthetic calcite (Mucci, 1983). These differences between estimates of calcite stoichiometric solubility products are explained in terms of variations in experimental conditions (artificial versus natural seawater) and related choices of carbonic acid dissociation constants. They do not reflect a true difference in solubility between biogenic and synthetic calcite. The thinning of foraminiferal calcite (104) XRD peak from 0.168°(2theta) to 0.148°(2theta) along the depth transects is interpreted as reflecting an improvement in calcite crystallinity. This and the change in specific surface area are consistent with the progressive change of the carbonate assemblage. The evolution of the bulk composition of the carbonate fraction is not paralleled by a significant change in its stoichiometric concentration product. It reflects ongoing differential dissolution due to kinetic effects. - Reaction Kinetics: [Abstract reproduced from "Reassessing the dissolution of marine carbonates, part 2" by Gehlen et al., submitted to Deep Sea Research I] We studied dissolution kinetics of the carbonate fraction > 150 microm of sediments sampled along two bathymetric transects in the eastern tropical Atlantic: the Sierra Leone Rise (SLR) and the Cape Verde Plateau (CVP). The reaction was followed by monitoring solution of pH during free-drift experiments lasting between 46 and 50 hours (20 degrees Celsius, pCO2; 3100ppm and 1atm pressure). The alkalinity reached at the end of the dissolution experiments ranged between 2.444 and 2.798meq/kgsw. The dissolution time series was extrapolated to equilibrium by fitting an empirical relation to the data. The estimated asymptotic concentration products range from 4.27 X 10-7 to 6.77 X 10-7 mol2kgsw-2. These asymptotic concentration products are compatible with the stoichiometric concentration product of aragonite (6.56 X 10-7 mol2kgsw-2) and calcite (4.37 (+/- 0.22) X 10-7 mol2kgsw-2) derived for the same sediment material during long term equilibration experiments (Gehlen et al., submitted). They are indicative of the presence of trace amounts of aragonite in sediments of the shallow stations (SLR station A, 2637m; CVP station M, 3104m). In sediments from deeper stations calcite is the main dissolving carbonate mineral. The order of reaction is always greater than unity. It varies between 1.4 (SLR station C) and 2.8 (CVP station M2), with an average n = 2.3 +/- 0.4. The higher order reaction is explained in terms of a multiphase system. Specific rate constants range from 0.09 to 0.53meq/m2/d. They are comparable to earlier results.
Quantifying mass exchanges within the marine environment is essential for understanding its role in the global carbon budget and how it will respond to perturbations in climate. Mathematical models are most frequently used for this quantification and prediction. However, as modelling of aquatic systems is not an exact science, ecologists have been struggling for decades to improve their models and to find the most appropriate balance between model complexity, ease of calibration, and precision. On the one hand, complex models may better represent the data, whilst on the other hand, they are more difficult to understand, and prediction uncertainty may increase with model complexity. To facilitate the selection of the complexity of models, a modelling framework has been implemented during the course of the ORFOIS project. This framework allows creating 1-D water-column and sediment models with varying complexity, from simple NPZD models to more complex models that combine unbalanced formulations for (several groups of) algae, DOM and POM with balanced formulations for (several groups of) zooplankton. Two manuals were written and disseminated with the source code, model executables and model output, which were put on the ORFOIS CD. The first manual (Soetaert and Greenwood, a) describes how to create 1-D models within the framework, the second manual (Soetaert and Greenwood, b) describes the various components that are incorporated in the framework. Several applications of the 1-D model framework were run, demonstrating the impact of model complexity on model outcome. The output of these runs was also put on the ORFOIS CD, together with software that can be used to visualize this output. The visualization software and the model executables run on a pc.

Searching for OpenAIRE data...

There was an error trying to search data from OpenAIRE

No results available