Archive Meta Data

Author(s)

Kenneth Thorø Martinsen - Theis Kragh - Kaj Sand-Jensen

Title

Carbon dioxide partial pressure and emission throughout the Scandinavian stream network

Description

# Carbon dioxide partial pressure and emission throughout the Scandinavian stream network

This repository contains data and scripts used and presented in the manuscript titled:

*Carbon dioxide partial pressure and emission throughout the Scandinavian stream network*

Rawdata is not included in this repository but all data is cited in references, openly available and should be downloaded prior to processing. 

The repository contains:

* The "data_products" folder contain grid products covering the stream network (masked by inland waters) in Denmark, Sweden and Finland (25 m resolution .tif raster files):

  - pco2_preds_DNK_SWE_FIN contain predicted carbon dioxide partial pressure (pCO~2~, unit micro-atmosphere)
  
  - kgas_co2_DNK_SWE_FIN contain gas exchange velocity (k, unit m d-1)
  
  - flux_intensity_co2_DNK_SWE_FIN contain carbon dioxide flux intensity (F~CO2~, unit g c m-2 d-1)

* The "processed_data" folder contains other objects part of the analysis:

  - test_preproc.rds and train_preproc.rds the preprocessed training and test data sets used for predictive modeing
  - recipe_fit.rds is the preprocessing instructions from the recipes package
  - all_data_carb_site.rds contains aggregated (annual averages) values of ph, alk, pco2 and water temperature from stream sites
  - site_attr.rds is site catchment data extracted from processed rasters
  - all_data_carb_site_attr.rds is raw (prior to preprocessing) carb and attribute data joined with some observations excluded
  - rf_predlist.rds is an R list object with the final trained Random Forest model (as an object from the mlr package) and its predictions and performance on the test set
  - algorithms_benchmark.rds is the benchmark of the candidate predictive models evaluated on the training set
  - leave_country_out_result.rds is the result of the "leave-one-country-out cross-validation" analysis

* The scripts in the folder can be processed (R version 3.5 and Python 3.7 used) in the following order:
  - pco2_rawdata.R
  - pco2_calculation.R
  - gis_rawdata_1.R
  - gis_processing_1.py
  - gis_rawdata_2.R
  - modeling_preprocessing.R
  - modeling_training.R
  - modeling_predictions.R
  - modeling_temperature.R
  - gis_processing_2.py
  - gis_dataproduct_extraction.R
  - Other: libs_and_funcs.R and gis_processing_imports.py contain libraries and functions used in analysis
  - Other: load_processed_data.R loads data objects resulting from analysis to use for figures etc.
  - Presenting results: figures.R, tables.R