Data and R scripts for "Disentangling season-specific contributions of drone-derived spectral and structural data to mire plant community discrimination"

This repository contains the minimum dataset and R scripts used to reproduce the analyses for the study “Disentangling season-specific contributions of drone-derived spectral and structural data to mire plant community discrimination”. The dataset includes 20-cm resampled UAV-derived raster data, polygon data for study areas and masks, and plant community training point data from two mires in Hokkaido, northern Japan: Kanekinto Mire and Ukishima Mire. The raster data consist of RGB orthomosaics, multispectral bands, and digital surface models acquired in spring and summer. Derived predictor variables, model input tables, random forest models, figures, tables, and appendices can be generated by running the included R scripts. The workflow evaluates how season-specific optical and surface-structural data contribute to plant community classification. The scripts derive optical indices and DSM-based variables, extract raster values at training locations, construct random forest models using different combinations of predictor groups, evaluate classification performance, and generate the figures and tables used in the manuscript.