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

Meichen He, He Ren, Xue Liu, Tingting He, C. L. Philip Chen, Lifan Zhang, Yanjie TangFrontiers in Plant Science 2026 · DOI

TL;DR

This study investigates how alfalfa's physiological and spectral responses change under simulated heat stress from spontaneous coal gangue combustion across different growth stages. They found that the budding stage is optimal for detection and that spectral vegetation indices, particularly those derived from first-derivative spectra, are highly effective for monitoring this stress.

What Problem Does This Paper Address?

Coal gangue dumps, a significant waste product of coal mining, pose a persistent risk of spontaneous combustion even after reclamation. This combustion releases harmful gases and threatens ecological restoration efforts. Early detection is crucial, but the vegetation growing on these dumps responds dynamically to the heat stress, with these responses varying across different growth stages. Existing monitoring methods often focus on surface temperature or gas emissions, but understanding how vegetation's physiological state and spectral signatures change over time under this specific stress is critical for developing effective, non-invasive early warning systems. This paper aims to fill this gap by detailing these stage-dependent responses and identifying key spectral indicators.

Methods

The researchers conducted a controlled indoor experiment simulating heat stress from spontaneous combustion by heating the soil around alfalfa plants. They measured key physiological indicators, including chlorophyll content (SPAD) and photosynthetic parameters (photosynthetic rate, stomatal conductance, intercellular CO2 concentration, and chlorophyll fluorescence parameters). Simultaneously, they collected canopy spectral data using a high-resolution field spectrometer.

To analyze the spectral data, they employed several techniques:

  • Correlation Analysis: To identify relationships between physiological indicators and spectral bands.
  • Successive Projections Algorithm (SPA): To select the most sensitive spectral bands and reduce redundancy.
  • Vegetation Indices (VIs): Including NDVI, DVI, FDNDVI, etc., calculated from both original and first-derivative spectra.
  • Triangular Spectral Parameters (TP): Another set of spectral features.
  • Machine Learning Models: Support Vector Regression (SVR), Random Forest Regression (RFR), and Partial Least Squares Regression (PLSR) were used to build predictive models for physiological indicators based on selected spectral features.

Datasets

The study utilized data collected from a controlled experiment involving alfalfa plants.

  • Experimental Setup: Alfalfa plants were grown in pots under two conditions: a control group (CK) at room temperature and a treatment group (T) subjected to simulated heat stress (120°C soil temperature at 30cm depth).
  • Data Collection: Physiological and spectral data were collected at four-day intervals across three growth stages: branching, budding, and flowering.
  • Model Training/Testing: A 3:1 ratio was used for training and testing the predictive models (60 samples for training, 20 for testing).

No new public datasets were introduced; the data was generated specifically for this study.

Key Insights and Ideas

  • Growth Stage Matters: Alfalfa's response to heat stress isn't static; it changes significantly across its life cycle. This highlights the need for stage-specific monitoring strategies.
  • Budding Stage is Key: The budding stage emerged as the most sensitive period for detecting heat stress impacts, offering the best predictive accuracy for physiological indicators.
  • Spectral Indices are Powerful: Vegetation indices, especially those derived from first-derivative spectra (like FDNDVI), proved to be more effective than raw spectral data or simple band selection for predicting physiological parameters.
  • SVR Dominates: The Support Vector Regression (SVR) model consistently outperformed RFR and PLSR in predicting physiological indicators across all growth stages, particularly for photosynthetic parameters.
  • Near-Infrared is Sensitive: The near-infrared (NIR) spectral region showed a more pronounced response to heat stress compared to the visible light spectrum.

Key Results

  • Physiological Trends: While both control and stressed alfalfa showed similar temporal patterns in SPAD and photosynthetic parameters (initial increase then decrease), heat stress consistently reduced these values and their variation amplitudes in the treatment group.
  • Sensitive Spectral Features: Features like OS2, FDS1, FDS2, TP, NDVI, and FDNDVI were identified as sensitive indicators of heat stress.
  • Model Performance: The SVR model achieved the highest prediction accuracy, with the budding stage showing the best performance. For instance, predicting SPAD in the treatment group during the budding stage yielded an R² of 0.77 and an RMSE of 3.50 using FDNDVI. For photosynthetic parameters, NDVI also showed high accuracy (e.g., for A, R²=0.88, RMSE=1.39).
  • Spectral Index Superiority: Vegetation indices derived from original spectra achieved the highest prediction accuracy, followed by first-derivative spectra, triangular parameters, and then original spectral bands.

Limitations and Future Work

  • Controlled Environment: The study was conducted under constant temperature conditions, which differs from the dynamic and variable temperature fluctuations found in actual field environments. Future research should incorporate dynamic temperature simulations and field observations to improve applicability.
  • Model Specificity: The predictive models were empirical. Further validation with larger datasets and potentially more advanced modeling techniques could enhance accuracy.
  • Broader Species and Conditions: While alfalfa is a representative species, future work could explore responses in other vegetation types and under varying soil conditions typical of coal gangue dumps.
  • Integration with Remote Sensing: The authors suggest that integrating these findings with remote sensing imagery could enable large-scale assessment of vegetation heat stress and early detection of spontaneous combustion risks.

Summary updated

Abstract

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.
BibTeX
@inproceedings{he2026study,
  author = {Meichen He and He Ren and Xue Liu and Tingting He and C. L. Philip Chen and Lifan Zhang and Yanjie Tang},
  title = {{Study on physiological indicators and spectral response characteristics of alfalfa under simulated spontaneous combustion of coal gangue dumps}},
  booktitle = {Frontiers in Plant Science},
  year = {2026},
  doi = {10.3389/fpls.2026.1745759},
}
Figure