High Resolution Soil Moisture Prediction Using Machine Learning and Unmanned Aircraft Systems-Based Remote Sensing
Pending the publication of this work in a peer-reviewed journal, this repository will host the machine learning models, training data, and R scripts used to build the models that predict topsoil moisture from surface properties remotely sensed by unmanned aircraft system (drone). Detailed description of this work is available on the fourth chapter of my dissertation.
Highlights
Drone remote sensing
In addition to reflectance in four separable bands, a high-resolution digital surface model was developed from the drone-based multispectral imaging and used to develop tens of topographic parameters.
Reproducible and efficient coding
The model testing and tuning scripts were done using drake R package which ensures reproducibility and efficiency. Click the image and download the html file to explore a sample dependency graph of a script that produces a model report.
Model analysis and interpretation
Model comparisons and analyses were done to understand the relationships and interactions between the variables. Click the image to explore a sample report for one set of model tuning iteration.