Monitoring Physiological Activity Of Winter Wheat Under Low Temperature Stress Using Hyperspectral Technology

In order to study the feasibility of hyperspectral monitoring changes of physiological activity of winter wheat under low temperature stress, two varieties of winter wheat were planted in plots. In this experiment, the winter wheat of Jintai 182 and Ji 22 was strong winter wheat and semi-winter wheat, respectively. A plant growth chamber was used to given winter wheat the low temperature stress treatment in the jointing stage. We have choosed 9 treatments by two factors, time and temperature of stress. The time of low temperature stress were 4 h, 8 h, 12 h and the temperature were -2 ℃,-4 ℃,-6 ℃.Then the change of physiological indicators (POD, SOD, MDA) were determined at different growth stages of winter wheat. At the same time, spectral reflectances of winter wheat were obtained. Then the models of POD, SOD, MDA based on the spectral characteristic parameters were constructed. The results showed that:1. In the whole growth stages of CK treatment, POD activity increased from the jointing stage to grain-filling stage, while SOD and MDA were stable. After the low temperature stress, the change of POD activities for stress treatments were as the same with CK, but the strong winter wheat with good cold resistance and the semi-winter wheat with good recovery ability. Moreover, SOD activities were more sensitivity to the time than temperature stress. The accumulation content of MDA did’t obviously changes.2. From the perspective of chemical-dimensional, there were broad ranges for the value of POD, SOD and MDA, and the validation set was included within the scope of the modeling set. The Offset amplitudes of modeling set and validation set were also available, indicated that the modeling set and validation set were representative.3. From the perspective of spectral analysis, it was feasible to monitor the activities of POD, SOD and MDA using hyperspectral technology and the best models for three activities indicators were SG-MLR model.Generally, the characteristic bands selected from each period mostly concentrated in the visible light region, particularly in the 400-450 nm range. There were 80% of characteristic bands located in visible region from jointing stage to grain-filling stage, and about 50% of characteristic bands were concentrated at the range of 400-450 nm, except for heading stage. So we believed that this region was more sensitive to the three physiological indicators, and more effective information were in the region.4. Through the comparation and analysis of modeling set with the whole spectra and characteristic bands, Rc, Rp, RMSEc and RMSEp of SPA-MLR and SG-MLR had little difference (except for the heading stage of winter wheat), while the hyperspectral data were compressed to 5% using the characteristic bands which greatly overcame the collinear problem, reduce redundant information and optimize the models. So the results turned out that it was available way to constructe the montor model with the characteristic bands instead of the whole spectral bands.