I do have a pretty skimpy experience in programming. As many pupils in Russia, I had been learning Pascal for a couple of years in school. Then it was Fortran in university. During my PhD in Russia, I had written a couple of scripts also in Fortran. After that, I have stick to R for a couple of projects and finished my small journey with Python, which I use as a primary programming language since 2015.

Python does have a pretty powerful standard library, but Python’s real power is in the utilization of additional libraries that can provide you an endless functionality.

Below I categorize Python libraries I use in my research.

General

  • numpy – the fundamental library for scientific computing.
  • pandas – the fundamental library for data analysis.
  • scipy – provides efficient numerical routines for, e.g., interpolation and optimization.

Geographical analysis

  • geopandas – pandas for geographical data. If you have .shp or .geojson files to analyze, you definitely should give geopandas a try!
  • xarray – the best library for manipulation of meteorological datasets provided in netcdf or grib formats.
  • cdo – I use it exclusively for remapping meteorological data from one grid to another.
  • wradlib – the core library for weather radar data processing.
  • opencv – the fundamental library for computer vision problems (e.g., computation of optical flow).
  • scikit-image – image processing without a pain.
  • trackpy – tracking particles made easy.

Machine learning

  • scikit-learn – the core package for machine learning applications.
  • tensorflow – deep neural networks development in a few of lines of code with built-in keras API.
  • fbprophet – time-series forecasting library developed by Facebook. I use it for water levels forecasting in OpenLevels.

Plotting

Code development

  • jupyter – interactive IDE running in your web-browser.
  • pylint – library developed for code analysis.

Misc

  • h5py – make use of HDF files.
  • numba – accelerates your code for fast compute.
  • requests – API requests for humans.

Dependencies (non-direct usage)

  • pytables – HDF files processing with pandas.
  • xlrd – reading xls with pandas.
  • dask – recommended xarray dependency.
  • netCDF4 – recommended xarray dependency.
  • bottleneck – recommended xarray dependency.
  • cartopy – recommended xarray dependency.
  • pynio – recommended xarray dependency.
  • pseudonetcdf – recommended xarray dependency.
  • nc-time-axis – recommended xarray dependency.
  • cfgrib – reading grib files.
  • eccodes – cfgrib dependency.
  • descartes – plotting polygons with geopandas.
  • pims – recommended trackpy dependency.

The environment

You can use conda to install all the libraries mentioned above in one isolated programming environment. Everything you need is just to create a file environment.yml copy-pasting the text below and then run conda env create -f environment.yml in your terminal.

name: megaenv
channels:
  - conda-forge
dependencies:
  - numpy
  - pandas
  - scipy
  - geopandas
  - xarray
  - cdo
  - wradlib
  - opencv
  - scikit-image
  - trackpy
  - scikit-learn
  - tensorflow
  - fbprophet
  - matplotlib
  - seaborn
  - bokeh
  - folium
  - branca
  - jupyter
  - pylint
  - h5py
  - numba
  - requests
  - pytables 
  - xlrd 
  - dask 
  - netCDF4 
  - bottleneck 
  - cartopy 
  - pynio 
  - pseudonetcdf 
  - nc-time-axis 
  - cfgrib 
  - eccodes 
  - descartes 
  - pims

After the installation (can take really long), you need to activate the environment you created by running conda activate megaenv in your terminal. Have fun!