Summary of Project

--

Overview over this Repository

Even though this repo contains only thesis/code/ it works with the whole thesis/-directory. E.g. thesis/latex/ is initialized during the initialization and figures are stored in thesis/latex/figures. The working directory for python/R code is thesis/code/ unless specified otherwise.

Directories - description

An description of the most important directories in thesis/...:

code/data/ -- here all the data is stored
code/data/yieldmapping_data/ -- the original untouched data (compressed with pickle for faster loading)
code/data/computation_results/ -- results of computations (to avoid unnecessary computations each time)
code/data/data_description/ -- description of yieldmapping data
code/interpol/ -- all python scripts regarding the interpolation-chapter
code/my_utils/ -- help functions to be used in other scripts (by import my_utils.<fname.py>)
code/ndvi_corr/ -- scripts demonstrating and computing the NDVI-correction discussed in the chapter NDVI-correction
code/plots_witzwil/ -- plots related to satellite image view of witzwil
code/shell_scripts/ -- bash-scripts for efficient reproducibility

latex/ -- the latex-directory for writing the thesis
latex/figures -- all figures used during the thesis (and the beamer-presentation)
latex/sty -- style-files to give the ETH-look
latex/tex -- all the written text visible in the thesis

beamer/ -- a latex (beamer) presentation

Pixel Class

Since we heavily focus on individual pixels and their time series we treat each pixel as an object (my_utils.pixel.Pixel)

Functionalities of such a pixel object are:

• Providing the (corrected) NDVI-time-series
• The application of various interpolation-methods (e.g. smoothing splines; c.f. my_utils.itpl) with various interpolation-strategies (e.g. identity, cross-validation; c.f. my_utils.strategies), while applying various weighting/filtering methods
• Plot ndvi/interpolation results
• Parallelization: computations on a list of pixels can be parallelized by my_utils.pixel_multiprocess.pixel_multiprocess(<list-of-pixels>)

How to Reproduce

Requirements: A linux machine with recent versions of R and python Also the python librarys pandas, seaborn and Jinja2 should be installed for the user (for R to use it)

Directory setup

thesis/code/data/yieldmapping_data/<data here>

now make sure that you are in the thesis directory, i.e. cd .../thesis

Initialize

run: ./code/shell_scripts/init_environment.sh this initializes the python environment and get the newest version from GitHub

Run python jobs

run: ./code/shell_scripts/reproduce.sh