Curated model results of control simulation for FESOM2.1-REcoM3This dataset is linked to the GMD Model Assessment publication of Zeising et al. (2026), https://doi.org/10.5194/gmd-19-2077-2026.

Curated model results of control simulation for FESOM2.1-REcoM3This dataset is linked to the GMD Model Assessment publication of Zeising et al. (2026), https://doi.org/10.5194/gmd-19-2077-2026.

The dataset contains the analysis and sea ice forecast results from the Alfred Wegener Institute Coupled Prediction System (AWI-CPS). Note that due to the large ensemble used (30) in the AWI-CPS, the output data is generated with only monthly mean data. One who has further interest in the daily output could reach the authors through emails. AWI-CPS.DA.tar.gz has both ocean and sea ice variables, where the variable has a suffix of 'f' means forecast, 'a' stands for analysis, and 'ini' stands for initialized states. AWI-CPS.FCST.tar.gz contains the sea ice forecasts for different lead times initialized in January, April, July and October. The axis 'fcst' for each variable actually represents forecasts with 4 different lead times of L0-2, L3-5, L6-8, L9-11. For example, L0-2 indicates forecasts with lead times of 0, 1, and 2 months, and so forth for other lead times. Both sea ice probability and sea ice concentration before and after the calibration are given in the data file. Specifically, SIP_FCST_CORR is for sea ice probability after calibration, while SIP_FCST_RAW is the raw forecast; SIC_Cal is after calibration, while SIC_RAW is the raw forecast. To obtain the mesh file for FESOM2 and visualize the data, one is referred to https://doi.org/10.5281/zenodo.6335530.

The dataset contains the analysis and sea ice forecast results from the Alfred Wegener Institute Coupled Prediction System (AWI-CPS). Note that due to the large ensemble used (30) in the AWI-CPS, the output data is generated with only monthly mean data. One who has further interest in the daily output could reach the authors through emails. AWI-CPS.DA.tar.gz has both ocean and sea ice variables, where the variable has a suffix of 'f' means forecast, 'a' stands for analysis, and 'ini' stands for initialized states. AWI-CPS.FCST.tar.gz contains the sea ice forecasts for different lead times initialized in January, April, July and October. The axis 'fcst' for each variable actually represents forecasts with 4 different lead times of L0-2, L3-5, L6-8, L9-11. For example, L0-2 indicates forecasts with lead times of 0, 1, and 2 months, and so forth for other lead times. Both sea ice probability and sea ice concentration before and after the calibration are given in the data file. Specifically, SIP_FCST_CORR is for sea ice probability after calibration, while SIP_FCST_RAW is the raw forecast; SIC_Cal is after calibration, while SIC_RAW is the raw forecast. To obtain the mesh file for FESOM2 and visualize the data, one is referred to https://doi.org/10.5281/zenodo.6335530.

This data set contains the mean diffuse attenuation coefficient of the downwelling plane irradiance over the first optical depth and over three different wavelength regions: 312.5 - 338 nm (Kd-UVAB), 356.5 - 390 nm (Kd-UVA), and 390 - 423 nm (KD-blue) as retrieved from the Sentinel-5P TROPOMI sensor from 11 May to 9 June 2018 in the Atlantic Ocean. The retrieval for the products is based on Differential Optical Absorption Spectroscopy (DOAS) extended to the ocean domain (PhytoDOAS). The spectral integrated Kd are derived from the Vibrational Raman Scattering (VRS) signal of the ocean which is retrieved by DOAS fits in three different fit windows. Kd-UVAB corresponds to DOAS VRS fits in the wavelength regions of 349.5 - 382 nm, Kd-UVA to 405 - 450 nm, and Kd-blue to 450 - 493 nm. VRS fit factors in the blue fit window (450 - 493 nm) were offset-corrected (an offset of 0.186 was added to the VRS fit factor of all processed S5P ground pixels). Derived Kd-blue are otherwise unrealistically high. The offset was determined with the help of Kd data at 490 nm from the Ocean and Land Color Instrument (OLCI) onboard Sentinel-3A. Fit results from the DOAS retrieval are converted into physical quantities using look-up-tables which were established with coupled atmosphere-ocean radiative transfer modeling using the software SCIATRAN version 4.0.8 (Rozanov et al. 2017, https://www.iup.uni-bremen.de/sciatran/). Only TROPOMI data with a cloud fraction smaller 0.01 were processed by the algorithm. Output data within the Atlantic Ocean (55°N-55°S, 70°W-10°E) were gridded daily into 0.083° latitudinal/longitudinal bins. Details on the algorithm can be found in the related publication by Oelker et al. (2022).

This data set provides the collocated data of remote sensing reflectance (Rrs) at 9 bands extracted from the merged ocean color products from GlobColour archive (https://www.globcolour.info/), satellite sea surface temperature from CMEMS (https://marine.copernicus.eu/), and chlorophyll a concentrations (Chl-a) derived from a global database of in situ HPLC pigment data collected from 2002 to 2012. The total Chl-a, Chl-a of six phytoplankton functional types (PFTs) that are diatoms, dinoflagellates, haptophytes, green algae, prokaryotes and Prochlorococcus, and two fractions of prokaryotes and Prochlorococcus are included in this data set. PFT Chl-a and fractions are derived using an updated diagnostic pigment analysis (DPA) method (Soppa et al., 2014; Losa et al., 2017), that was originally developed by Vidussi et al. (2001), adapted in Uitz et al. (2006) and further refined by Hirata et al. (2011) and Brewin et al. (2015). Matchups of satellite Rrs to in situ PFT data (which were also matchups to SST) were extracted from global 4-km daily merged products. Extraction and averaging protocol including quality control were described in detail in Xi et al. (2020).

This data set is composed of model output from Darwin-MITgcm, of chlorophyll-a concentration, coloured dissolved organic matter absorption, sea-ice concentration, sea surface and subsurface temperature, surface heat flux, ice-covered days, mixed layer depth, meridional advection of temperature. It covers a time period from January 2007 to January 2017, while some fields cover only parts of the summer of 2012.

The simulated sea ice drift data is a by-product from a sea ice thickness assimilation system that generates the Arctic 'Combined Model and Satellite sea ice Thickness (CMST; doi:10.1594/PANGAEA.891475) ' dataset.The data also provide the ocean current velocity where ice free. To obtain the sea ice drift on the geographic coordinate, a transformation must be done as following:uE = AngleCS * SIuice - AngleSN * SIvice;vN = AngleSN * SIuice + AngleCS * SIvice;where uE and vN are two velocity components on the geographic coordinate; AngleCS and AngleSN can be found in 'grid.cdf'; SIuice and SIvice are sea ice velocity on model mesh.

An Arctic sea ice thickness record covering from 2010 to 2016 is generated by assimilating satellite thickness from CryoSat-2 and Soil Moisture and Ocean Salinity (SMOS). The model is based on the Massachusetts Institute of Technology general circulation model (MITgcm) and the assimilation is performed by a local Error Subspace TransformKalman filter (LESTKF) coded in the Parallel Data Assimilation Framework (PDAF).

This phytoplankton group (PFT) concentration a (Chl a) data are output from the algorithm PhytoDOAS version 3.3 applied to SCIAMACHY data from 2 Aug 2002 to 8 Apr 2012. Data have been gridded monthly on 0.5° latitude to 0.5°. For cyanobacteria (includes all prokaryotic phytoplankton) and diatoms the PhytoDOAS PFT retrieval algorithm by Bracher et al. (2009) and for coccolithophores the algorithm by Sadeghi et al. (2012) have been used. However, these methods have slightly been improved which includes:- Data during SCIAMACHY instrument decontamination are excluded in the analysis.- SCIAMACHY level-1b input data for PhytoDOAS are now version 7.04 data (instead of version 6.0).- The wavelength window for all three phytoplankton groups (PFTs) fit factor starts at 427.5 nm (instead of 429 nm).- Coccolithophores fit factors are retrieved in a retrieval fitting simultaneously diatoms and coccolithophores (instead of a triple fit with also fitting dinoflagellates as in Sadeghi et al. 2012).- Vibrational Raman Scattering (VRS) is now fitted directly in the blue spectrum (450 to 495 nm), following Dinter et al. (2015), (instead of in the UV—A region as in Vountas et al. 2007) except that here the daily solar background spectrum measured by SCIAMACHY and the VRS pseudo absorption spectrum calculated based on a SCIAMACHY solar spectrum following Vountas et al. (2007) was used in order to correct for the variation of instrumental effects over time (this is not achieved when using the RTM simulated background spectrum as done in Dinter et al. 2015).- The PFT Chl a are derived from the ratio of the PFT fit factor to the VRS fit factor multiplied by a LUT (Look Up Table). The LUT is based on radiative transfer model (RTM) SCIATRAN simulations (see Rozanov et al. 2014) accounting also for changing solar zenith angle (SZA).