https://doi.org/10.5061/dryad.44j0zpcn5

Description of the data and file structure

The data folder contains all the necessary data to run the models and analyses in R.

All the files, variables and units in the folder ./data are described in the metadata.xlsx file

• climate_models.parquet: Set climate models on a target spatio-temporal grid that involves interpolating on the chosen grid and averaging weekly. Physical product (sea temperature, salinity, and the zonal (u) and meridional) are from the CNRM-RCSM4 (Centre National de Recherches Météorologiques - Centre Européen de Recherche et de Formation Avancée en Calcul Scientifique 4) coupled regional climate system model itself driven by the CMIP5 CNRM-CM5 global climate mode following a high greenhouse gas emission scenario (RCP8.5). Biogeochemical variables (nitrate, phosphate, dissolved oxygen, and chlorophyll-a concentrations) were simulated by the Eco3M-S model

  Simulated Period: current (1999-2017), mid-century (2041-2060), and end-century (2081-2100).

• climate_models-corrected-slope.parquet : Corrected climate models (climate_models.parquet) with CDF-t approach. We associated derived variables such as currents relative to coast and recent change of each variable (slope fitted to the last few values)

• coastal_env.parquet : in situ measurements of temperature, salinity, and concentrations of nitrate, phosphate, dissolved oxygen, and chlorophyll-a, were gathered from the database of the SOMLIT networkfor three stations (Point B: 43°41N, 7°19E, Frioul: 43°15N, 5°18E, and SOLA 42°29N, 03°09E) in the period 1999-2017. NA = Missing values which correspond to no measurements or a a bad quality check (See breton et al, 2023)

• coastline.parquet : coastline data for mapping

• GSHHS_shp/h/GSHHS_i_L1.shp : shapefile of the continent

• habitat_model.RData : best xgboost model

• habitat_model-data.parquet : data ready to be modelled intersect between the sampling point from monitoring and environmental data of corrected reanalysis

• habitat_model-pred_for_climate_models.parquet : prediction of the best xgboost model on corrected climate models

• habitat_model-pred_for_reanalyses.parquet : prediction of the best xgboost model on corrected reanalysis

• local_temperature.parquet : temperature measurements were taken with a microprocessor conductimeter WTW (Model LF197)on the day and at the site where biological samples were collected, in Monaco and Villefranche-sur-Mer (see ostreopsis.parquet file).

• ostreopsis.parquet : 10 year monitoring programs from which O. cf. ovata abundances carried out in France (Villefranche-sur-Mer), Monaco (Larvotto beach), and Spain (Llavaneres) between 2007 and 2017

• pixels_suitable.parquet : grid along the coast on which the niche model will be fitted

• pred_boost_m.parquet : model initially fitted on bootstraps of the data

• reanalyses.parquet : Set reanalyses on target spatio-temporal grid that involves interpolating on the chosen grid and averaging on a weekly basis. Reanalysis are distributed by the Copernicus Marine Environment Monitoring Service (CMEMS, MEDSEA_MULTIYEAR_BGC_006_008 http://marine.copernicus.eu/services-portfolio/access-to-products, accessed 2020-03-20). Biogeochemical data (nitrate, phosphate, dissolved oxygen, and chlorophyll-a concentrations) were provided as weekly averages (G. Cossarini, pers. comm.).

• reanalyses-corrected-anomalous_values_cropped.parquet : Corrected reanalysis with CDF-t approach by putting NA on unrealistic values (such as negative values for oxygen), only used to perform the CDF't on climate models

• reanalyses-corrected-anomalous_values_shifted.parquet : Corrected reanalysis with CDF-t approach

• reanalyses-corrected-slope.parquet : Corrected reanalysis with CDF-t approach. We associated derived variables such as currents relative to coast and recent change of each variable (slope fitted to the last few values), NA= missing values  for the date "1999-01-01" "2041-01-04" "2081-01-03" because recent changes (fitted slope) cannot be calculated for the first date because there are no values before. 

• habitat_model-data.parquet : data used to fit cell abundance to the reanalysed data through habitat modelling. NA= missing values  for the date "1999-01-01" "2041-01-04" "2081-01-03" because recent change (fitted slope) cannot be calculated for the first date because there are no values before. 

• habitat_model-pred_for_climate_models.parquet : projection with just the best model on corrected climate models. NA= missing values  for the date "1999-01-01" "2041-01-04" "2081-01-03" because recent change (fitted slope) cannot be calculated for the first date because there are no values before. 

• habitat_model-pred_for_reanalyses.parquet : projection with just the best model on the corrected reanalysis. NA= missing values  for the date "1999-01-01" "2041-01-04" "2081-01-03" because recent change (fitted slope) cannot be calculated for the first date because there are no values before. 

SCRIPT AND ANALYSIS

_project.Rproject : R project to run the analysis

• 0.setup.R : common variables, loading packages, function
• 1.1_process_data-transform_data_and_reanalyses.R : correction of coastal and reanalysis data
• 1.2_.process_data-transform_climate_models.R : correction of climate models using the CDF-t approach
• 2.1.model-prepare_data : data preparation for habitat model

2.2model-fit_model_abundance.R : model fitting

3.Explore_results.Rmd : Analysis and figures