Compendium of R Scripts and Data for Lidar Feature Classification Accuracy Assessment

Archaeo-Lidar-Error-Quant.zip Data and R code for assessing the relationship between underlying lidar data quality and archaeological feature classification accuracy. Logistic regression models (created in R) test whether data quality, as measured by the Digital Feature Model Confidence Map presented by Štular et al. (2021, correlates with points sampled along correctly or incorrectly identified archaeological features. Archaeological features were detected through lidar-based prospection and validated by pedestrian survey. The Confidence Map is a classification of each pixel in a DFM into six categories based on topographic and point-cloud characteristics, with 1 representing poor quality and 6 representing high quality data. There are three folders: R-scripts: this folder houses the R scripts used to run the multinomial logistic regression models that tested statistical relationships. The naming convention is MultinomialLogit_[buffer distance on feature line]_[% of buffer area sampled with points] (e.g., MultinomialLogit_1m_33p.R) results: this folder houses the analysis results as .jpg and .html files. Models name, buffer distance, and sample percentage are indicated by the file name. data: this folder houses the point data used for analysis as .csv files. Each sheet is one buffer-sample pairing, for example SamplePoints_1m_33percent.csv houses all points from the 1m buffer with 33% pixel sample. Note: original archaeological feature vectors are not available to the public at the moment, as it is sensitive cultural heritage information. Please follow the Small Cycladic Islands Project (https://smallcycladicislandsproject.org/) for more information and updates on data availability. The primary dataset used for analysis in published material is the 1m buffer & 33% pixel sample dataset. DFM-Creation-and-Visualization.zip The "DFM" and "Viz" folders store the R and Python Notebook (.ipynb) files that, respectively, were used to create original Digital Feature Models and their visualization suites. DFM creation relies on the lidR R package and the Visualizations requires ArcPy and the Relief Visualization Toolkit. File paths stored in the scripts are not universal and will require modification to function.