# Study on physiological indicators and spectral response characteristics of alfalfa under simulated spontaneous combustion of coal gangue dumps

## TL;DR

This study examines how alfalfa responds to heat stress from coal gangue spontaneous combustion, identifying key growth stages and spectral indicators for early detection.

## Contribution

The study introduces growth-stage-dependent spectral responses and identifies optimal monitoring methods for spontaneous combustion in reclaimed coal gangue dumps.

## Key findings

- SPAD values and photosynthetic parameters showed consistent trends during branching, budding, and flowering stages.
- Vegetation indices based on original spectra achieved the highest prediction accuracy for heat stress monitoring.
- The budding stage was identified as the optimal observation period, with the SVR model showing the best performance (R² = 0.77).

## Abstract

Coal gangue dumps may still pose a risk of spontaneous combustion even after reclamation, threatening both ecological restoration and surrounding environments. As a dynamic and complex process, vegetation responses to spontaneous combustion heat stress vary across growth stages. Therefore, identifying growth-stage-dependent spectral responses of physiological indicators is critical for the timely and accurate detection of spontaneous combustion in coal gangue dumps.

In contrast to field-based investigations, this study selected alfalfa (Medicago sativa L.), a typical reclaimed herbaceous species on coal gangue dumps, as the study object. Constant high-temperature conditions were established indoors to simulate heat stress associated with coal gangue spontaneous combustion. The spectral response differences of chlorophyll relative content (SPAD) and photosynthetic parameters at different growth stages were analyzed. Correlation analysis combined with the SPA algorithm was applied to identify sensitive physiological indicators, spectral bands, and characteristic parameters related to heat stress. Subsequently, support vector regression (SVR), random forest regression (RFR), and partial least squares regression (PLSR) models were developed (training set: test set = 3:1, n = 60:20) to determine optimal combinations of growth stage, indicators, and spectral features for monitoring spontaneous-combustion-induced heat stress.

(1) during branching, budding, and flowering stages, the SPAD values and photosynthetic parameters of treatment (T) and control (CK) groups exhibited consistent trends; (2) sensitive features included OS2, FDS1, FDS2, TP, NDVI, and FDNDVI; and (3) vegetation indices based on original spectra achieved the highest prediction accuracy, followed by those derived from first-derivative spectra, triangular parameters, first-derivative reflectance, and original spectral bands. Among all growth stages, the budding stage was identified as the optimal observation period, and the SVR model showed the best performance (FDNDVI: R² = 0.77, RMSE = 3.50).

This study reveals growth-stage-dependent physiological and spectral responses of vegetation to spontaneous combustion heat stress and provides a theoretical basis and technical reference for ecological monitoring and early warning of spontaneous combustion in reclaimed coal gangue dumps.