Abstract: The North China Plain is the main production zone of winter wheat (Triticum aestivum L.), where wheat is affected seriously by later spring frostbite. Hyperspectral remote sensing is proved feasible in monitoring crop growth, especially after stresses. But the relationship between hyperspectral feature and crop growth characteristics after frostbite has been rarely studied so far. So we conducted a pot experiment with the artificial frost box to simulate frostbite (-5℃ and -7℃ respectively) in 2006–2007 winter wheat growing season. The wheat seedlings were treated in the box for 4 and 8 hours at jointing stage (treated only once either on April 9 or 15, 2007) respectively. The canopy hyperspectral reflectance curves were measured with ASD FieldSpec, and simultaneously, chlorophyll content (Chl) was tested with SPAD502. The result showed that chlorophyll content in control (CK) was 2–5 units more than that in the treatments after frostbite. In the early stage after frostbite, the reflectance and shape of hyperspectral curves were similar between treatments. After 20 days, frostbitten leaves turned into yellow in advance of CK, hyperspectral reflectance increased distinctly in yellow and red regions, the ‘red well’ (RW) was not distinctive, the shape of curves became plane. The relationship between chlorophyll content and hyperspectral reflectance ρ, logarithm of the reciprocal of ρ [log10 (1/ρ)], the first derivative of ρ (dρ/dλ) were analyzed, results indicated that ‘red edge’ (RE) shifted to shorter wavelength and RW to near infrared after frostbite. There were higher correlation between values of hyperspectral curve at special positions and frostbite stress levels, the correlation coefficients of frostbite degree with green peaks (ρg), RE, RW and the spectral absorbance index (SAI) were 0.36*, 0.69** , 0.42**, and 0.33* respectively, with the sum of reflectance in wide bands of blue(SBρ), green (SGρ), yellow (SYρ), and red(SRρ) were 0.34*, 0.43**, 0.45**, and 0.44* respectively, with the normalized difference vegetation index (NDVIa) between RE and RW was -0.33*, with the NDVIb between sum of reflectance at near infrared (SNirρ) and SRρ was -0.39**, with the chlorophyll content was -0.49**. In terms of distinct mean difference analysis, it is feasible to identify the frostbite injury and to estimate the injury degree based on the difference of the NDVIb.

Key words: Winter wheat, Frostbite, Hyperspectral, Correlation analysis