ISOTOPES IN PRECIPITATION - STATISTICS

The python script (ISO.py) contained in this repository is a program that can be used for calculating basic statistical properties of (monthly collected) isotopic composition (δ²H, δ¹⁸O, ³H) of precipitation. This script was created as a subproduct when initiating a dynamical website called SLONIP (SLOvenian Network of Isotopes in Precipitation) - https://slonip.ijs.si/. Major part of the code presented here (ISO.py) was firstly implemented in the backend of the mentioned website, and is here reused for the purpose of creating a program that can be run on any machine locally.

INPUT

• Excel files containing data on isotope composition of precipitation (default file in map 'data')

WHAT IS CALCULATED

• means: annual, mothly and seasonal (precipitation weighted and unweighted)
• regression coefficients describing local meteoric water line (precipitation weighted and unweighted):
  • MA (Major axis regression)
  • RMA (Reduced major axis regression)
• in addition to what is calculated in the backend of the SLONIP webpage, the present program also calculates regression coefficients for every year represented in data and plots the results on graphs

The complete description of used methods can be found on https://slonip.ijs.si/data/ under the segment 'Evaluation of data'.

INSTRUCTIONS

Excel files:

• excel files must be in folder 'data'
• excel files must be constructed in a specific manner in order for program to work - an example data sheet can be found in the 'data' folder.
• there should be only one 'sheet' in excel file
• the program supports the following type of excel files: xls
• the names of excel files must be without dots (e.g. Zg. Radovna -> Zg Radovna). Spaces are allowed.
• in excel file there should be dots (not commas) used for decimal numbers
• if no errors, the results will be saved in /results/<name_of_input_excel_file>

Start the program:

• make sure you have installed python3 on your computer, with the following libraries:

  • xlrd (version >= 2.1.0)
  • numpy
  • matplotlib.pyplot

• open the terminal at the location of README.txt and type in: python ISO.py (to start the program)

  • after executing the program the command prompt will ask you to:
    • specify the name of the station (excel file), located in folder 'data/' (e.g. data/Murska_Sobota_16-18)
    • specify if precipitation amount data exists for all months ( * )
    • specify which types of statistics you want to calculate (e.g. annual means, weighted monthly means,...)
    • results will be saved under 'results/Ljubljana16-18/'
    • note: if presented data describes only one year, then the calculation 'Regression by years' will throw an error (at minimum two years are required to compute this type of regression)

Filters:

• if, for a given year, there is less than 8 (out of 12) existing values for some type of data, the annual statistics won't be calculated for this type of data (e.g. δ¹⁸O‰ has missing data in year 2020 for months Jan-May -> only 7/12 values for the year 2020 -> the annual (2020) δ¹⁸O‰ statistics won't be calculated)
• if, for a given year, the existing data (some type of data) represent less than 70% of total annual precipitation, the annual statistics won't be made for this type of data (e.g. only 10/12 data for δ¹⁸O‰ are available for the year 2020, but this represents less than 70% of all precipitation collected that year -> the annual (2020) δ¹⁸O‰ statistics won't be calculated)
• when calculating regression coefficients if one of the pairs (δ¹⁸O‰, δ²H‰) is missing, then this pair of values won't be taken into account when calculating regression coefficients
• for given data (δ¹⁸O‰, δ²H‰), at least two years must be valid according to the above stated filters, for annual regression calculation to be executed
• in the case when all precipitation data is not available ( * ), the '70% filter' will not be taken into account and the weighted statistics will not be calculated

Rounding:

• by default the results are rounded on:
  • 2 decimals (δ¹⁸O)
  • 1 decimal (δ²H)
  • 1 decimal (deuterium)
  • 1 decimal (³H)
  • 2 decimals (regression coefficients)
• note: rounding is done with python's function round, which rounds *.*n5 down to *.*n (and not to *.*(n+1) e.g. 2.675 is rounded to 2.67 and not to 2.68)

Prepared by Aljaž Pavšek & Polona Vreča, IJS, 2022