UAS Soil Moisture Sensing

This repository contains R codes and data to accompany Araya et al. 2021. HESS.

Brief descrions of the file strucutre and important files.

The *.html and *.md files are reports of corresponding *.R scripts with the same file name.

• uav_data_preprocess.R: compiles and prepared the master table (i.e. uav_data_v4.rds)
• uav_data_split.R: takes uav_data_v4.rds and creates testing and training data splits.
• Data_Processed/: has the tabular data including used for training and testing of models
  • ML_Training_Metada.xlsx: Description of column headers in dataset.
  • VWC_Met_GIS_table4.rds: Merged data table of soil moisture, hydrologic variables, and terrain variables
  • uav_data4.rds: Cleaned data for machine learning training. Produced from uav_data_preprocess.Rscript.
• uavtune/: This contains the main tunning codes that were run on merced cluster.
  • uavout/: has the built machine learning models. Can be downloded from here.
  • universal_tune_doParallel.R is the main file that runs the training. It is called by individual training scripts. For example: to tune a BRT62 model, run the BRT62.R file on cluster with all it's dependencies (the other R files, the parameter .csv file, etc.).
• model_analysis/: This folder has files that produce model performance reports for each model. Uses drake R package. Understanding of drake package is requred to use the scripts.
  • Reports/: where produced reports are saved in *.html file and performance summaries in *.csv filetype.