Early identification of heat and UV-B stress in wheat based on the combination of hyperspectral technology and gas detection method

• A self-made gas detection system was used to detect CH 4 emitted in wheat. • ELM optimized with iPSO was used to model abiotic stress identification. • Heat stress plays a dominant role in the combined heat and UV-B stress in wheat. • Combined hyperspectral and gas detection effectively identify abiotic stress. Wheat is an important food crop, valued for its stable yield and substantial nutritional value. However, its growth and development during the tillering stage are significantly inhibited by heat stress (HS) and ultraviolet-B stress (UV-BS). Therefore, early identification of these stresses is essential to ensure the healthy growth of wheat. Hyperspectral technology has been widely used to identify abiotic stress in plants, yet the spectral similarity among different abiotic stresses could confound accurate identification. To address this issue, endogenous methane (CH 4 ), a critical response indicator of wheat under HS and UV-BS, was also employed as an important marker to differentiate between these stresses. In this paper, the early identification model of HS and UV-BS in wheat was effectively established through the combination of hyperspectral technology and the tunable diode laser absorption spectroscopy (TDLAS)-based gas detection method. The model was established using trend-based feature extraction methods and an extreme learning machine (ELM) algorithm optimized with improved particle swarm optimization (iPSO). The time-segmented HS and UV-BS identification models established solely based on spectral data exhibit an accuracy range of 80.36 % to 91.07 %. In contrast, models established by integrating spectral data with CH 4 concentration data have significantly improved their accuracy, achieving an accuracy range of 95.54 % to 98.88 %. Therefore, a multi-temporal integrated model for early identification of HS and UV-BS in wheat was established based on the integrated data, achieving an accuracy of 93.4 %. The results demonstrate that the combination of hyperspectral technology and gas detection method is effective for the early identification of HS and UV-BS in wheat.