A novel spectral index designed for drone-based mapping of fire-damage levels: demonstration and relationship with biophysical variables in a peatland

Introduction This study introduces novel spectral indices specifically designed for drone-based data to identify and differentiate between varying levels of fire damage. Their application was demonstrated in an Estonian peatland, where its effectiveness was compared with that of traditional vegetation indices. Methods Four drone surveys were conducted at different post-fire intervals, and biophysical variables, including surface and soil temperatures, soil moisture, and aboveground biomass, were measured. The proposed triangular-area indices (TAI) were derived from reflectance maps obtained using a multispectral sensor. Damage classes were defined using binary and multi-level classification approaches, and decision trees were trained and evaluated for accuracy. Results The TAI1 index achieved classification accuracies between 80.6% and 90.9%, comparable to those of more complex machine learning techniques. TAI1 exhibited strong correlations with biophysical variables, highlighting its potential for post-fire assessment. Although TAI1 showed some limitations in distinguishing moderate damage levels, it demonstrated improved capability in detecting severely damaged areas. Discussion TAI1 can provide ecologically relevant insights, enhance the interpretation of fire damage, and support rapid, high-resolution assessments of vegetation health. Further research is required to validate its interchangeability with other indices across different scales, sensors, and environmental contexts.