Skip to content

geco-bern/flx_wb2

Repository files navigation

Actual Evapotranspiration in Arid and Semi-arid Regions

This project uses Eddy Covariance FluxDataKit, Condensation, Soil Thickness Data from Pelletier and Compound Topographic Index (CTI) Data to explain deviations of high aridity in combination with high actual evapotranspiration from the theoretical Budyko framework.

Scenarios:

in order to compare different impacts of condensation (with or without) and letant energy (corrected or not), 4 scenarios are distinguished:

  • without condensation + gap filled latent energy
  • wihtout condensation + corrected latent energy
  • with condensation + gap filled latent energy
  • with condensation + corrected latent energy

Data

FluxDataKit

The FLuxDataKit (version 3.4) provides flux tower records from Zenodo and delivers half hourly and daily data from PLUMBER2 (Ukkola, 2021), Ameriflux (Ameriflux, 2023; Pastorello, 2020), ICOS Drought2018 (ICOS, 2018), ICOS WarmWinter2020 (ICOS, 2020). The data is available for each site.

path to files:

- ~/data_2/FluxDataKit/v3.4/zenodo_upload/fdk_site_info.csv
- ~/data_2/FluxDataKit/v3.4/zenodo_upload/fdk_site_fullyearsequence.csv

Topography from Marthews (2015)

This NetCDF dataset is at grid layer which was upscaled to a resolution of 15 arcsec. Each grid cell has a geographical location holding a specific value for CTI.

path to file:

- ~/data/archive/gti_marthews_2015/data/ga2.nc

Soil Thickness from Pelletier (2016)

This additional variable is an estimated thickness from bedrock to land surface. It comes with a spatial resolution of 1km grid, also provided in the NetCDF format. With exact coordinates, thickness is extracted and assigned to the sitenames of the FluxDataKit.

path to file:

- ~/data/archive/soil_pelletier_2016/data/Global_Soil_Regolith_Sediment_1304/data/average_soil_and_sedimentary-deposit_thickness.tif

Condensation

The P-Model calculates condensation by converting night time total net radiation into the equivalent mass of water. In order to check robustness of condensation which is added to precipitation, this additional factor helps to distinguish different scenarios.

path to file:

- ~/data_2/FluxDataKit/v3.4/zenodo_upload/rsofun_driver_data_v3.4.2.rds

Structure

This project is structured as followed and shall be run in sequenced order:

data-raw/
   ├─ 00_Download_Data.R
   ├─ 01_P_Model_RSOFUN.R
data/
   ├─ 01_Mean_Table.R
   ├─ 02_Budyko.R

analysis/
   ├─ 00_Budyko_Plots.R
   ├─ 01_Aridity_Plots.R
   ├─ 02_Site-Characteristics.R
   ├─ 03_Statistics.R
   ├─ 04_RandomForest_Model.R
   ├─ pics/     #folder for outputs
   ├─ tables/   #folder for outputs

/data-raw

  • 00_Downolad_Data.R:

This script gatters site infos and fullyearsequences from FluxDataKit. In a next step, needed Value from CTI is extracted and finally merged and written into the df_sites.csv file which will be stored in the /data folder.

  • 01_P_Model_RSOFUN.R:

Values of condensation are extracted and written into df_cond_mean_ann_2.csv. finally, these values are joined into the df_sites.csv file.

/data

  • 01_Mean-Table.R:

The daily FluxDataKit data is loaded. With the function fun_read_onesite()entries matching FLX_<sitename>_FLUXDATAKIT_FULLSET_DD* are read from the CSV and tagged with the sitename. looping over all sitenames in df_sites$sitename, purr::map then combines them into the nex dataframe daily_data.

functions for converting latent energy to evapotranspiration are execuded.

having mass flux in mm day-1, final values of precipitation prec, actual evapotranspiration aet, corrected aet aet_corr, each with and without condensation, and potential evapotranspration pet are joined to the main dataframe df_sites.csv

  • 02_Budyko.R:

Here, the Budyko by Fu (1981) equations are introduced. For each scenario, the new dataframe df_budyko transmutes the variables of precipitation and actual aswell as potential evapotranspiration.

finally, the residuals of the actual evapotranspiration from the theoretical (potential) evapotraspiration in the Budyko framework can be calculated, which delivers a value for each scenario.

/analysis

  • 00_Budyko-Plots.R: The aim of this script is the visualization of the scenarios. In this sort, the plotting of the residuals in the in the Budyko framework and according histogram are the output of this script.

see:

analysis/
   ├─ pics/  
      ├─ Budyko_Comparison_ALL.png
      ├─ Budyko_Residuals_ALL.png
  • 01_Aridity-Plots.R:

In order to filter out arid and semi-arid as well as high cti, the data is plit into quintiles.

Based on the PET/P ratio, following splitting is defined for aridity: H (Humid): PET/P < 1 SH (Semi-humid): 1 ≤ PET/P < 2 SA (Semi-arid): 2 ≤ PET/P < 4 A (Arid): PET/P ≥ 4

CTI is plit into quintiles as followed: Very Low Low Medium High Very High

In this sort, aridity_class_nocond, aridity_class, and cti_class are the objects for the use of the extraction of the Fluxnet sites with high aridity and high CTI.

The outputs are:

analysis/
   ├─ pics/  
      ├─ AI_Boxplot_ALL.png
      ├─ Heatmap_ALL.png
  • 02_Site-Characteristics.R:

This script includes high CTI and aridity in a diagram showing the degree of deviations from Budyko. Also, site information (coordinates) with soil thickness data from Pelletier et al. (2016) is linked and integrated to result in two excel tables. Here, in addition to high CTI and aridity, soil thickness expands the dataset.

The data is merged into df_budyko

analysis/
   ├─ pics/  
      ├─ AI_CTI_Diagram_ALL.png
      ├─ Pelletier_LE_F_MDS.png
      ├─ Pelletier_LE_CORR.png
   ├─ tables/       
      ├─ table_topsites.xlsx
      ├─ table_topsites_corr.xlsx
  • 03_Statistics.R:

Here, the linear model statistical analysis is executed for the estimation of coefficients with gap-filled and corrected latent energy.

analysis/
   ├─ pics/  
      ├─ stat_lm_coefficients.png
      ├─ stat_lm_coefficients_corr.png
  • 04_RandomForest_Model.R:

To visualize the distribution of R² and RMSE, bowplots show the behavior of the different predictors (CTI, thickness and landcovertype). With the help of random forest outputs,

analysis/
   ├─ pics/  
      ├─ stat_rf_vip.png
      ├─ stat_rf_vip_corr.png
      ├─ stat_rf_comparison_ALL.png

Bibliography

About

Flux DataKit Waterbalance Bachelor Thesis

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages