Effect Of Different Water Situation On Hyperspectral Characteristic And Physiological-biochemical Indices For Tobacco

Water plays an important role in the growth, yeild and quality of tobacco, therefore, reasonable irrigation is necessary during the growth stage. It is nacessary to adopt a non-destructive, quick, convenient technique to monitor crop water status by employing hyperspectral remote sensing. In this work, terrestrial remote sensing platform was employed to research canopy and leaf physiological-biochemical indices and hyperspectral reflectance characteristics under different soil water content treatments. We analysised the relationships between the tobacco canopy, fresh tobacco leaf and flue-cured leaf and the corresponding hyperspectral parameters, and used stepwise regression to establish estimation model. Samples were randomly selected in order to evaluate the predict value and the test value, results showed that the model has a good estimation and prediction ability. Details as below:1. The hyperspectral reflectance curves of tobacco canopy and leaf both have the commen reflection characteristics of plants, further, the reflectivity of tobacco canopy, leaf under different water treatments has significant difference, their red edge parameters also changes accompany with the variation of soil water contention, and an obvious red shift and blue shift can be observed. The first derivative spectra of canopy red edge has a "double peak", which becomes more and more evident with the increasing of the soil water content. These results can suply the theoretical basis for the water govern of formland.2. The effects of different water treatments on the two kinds of tobacco canopy, fresh tobacco leaf and flue-cured leaf physiological-biochemical indices showed consistent patterns, and the difference among different treatments is significant. The content of Chlorophyll a, Chlorophyll b, carotenoid, and the total Chlorophyll all becamed higher with the soil water content increased in root elongation stage and vigorous stage, while it showed 65% than 45%, 45% than 85% in the maturity stage. LAI, leaf fresh weight, leaf dry weight, ground fresh weight, ground dry weight leaf water content, plant water content all becamed higher with the soil water content increased. The carbohydrates like starch, reducing sugar and total sugar all raised with the soil water content increased. While the nitrogen-containing compounds like total nitrogen and nicotine reduced with the soil water content increased.3. Through the observation of tobacco canopy reflectance on the farm, it was found that the most relevant characteristic variables of Chlorophyll a, Chlorophyll b, carotenoid, total Chlorophyll content are the blue-edge position(λb). The most relevant characteristic variables of LAI, leaf fresh weight, ground fresh weight, ground dry weight are the red-edge position(λr). The most relevant characteristic variables of are the blue edge amplitude(Db).The most relevant characteristic variables of tobacco canopy reducing sugar, total nitrogen are the red-edge position(λr). The most relevant characteristic variables of starch, total sugar, nicotineare the red edge area(SDr).The most relevant characteristic variables of roasted canopy tobacco starch, reducing sugar, total nitrogen, nicotine are LCI. And total sugar is the red edge area(SDr).Through the observation of tobacco leaf reflectance in the room, it was found that the most relevant hyperspectral characteristic variables of Chlorophyll a content, Chlorophyll b content, total Chlorophyll are the red edge area(SDr), and the most relevant hyperspectral characteristic variables of leaf fresh weight, leaf dry weight, leaf water content are the red-edge position(λr). The most relevant hyperspectral characteristic variables of carotenoid content, leaf area, leaf dry weight are LCI. And for leaf dry weight, red edge area(SDr) and red-edge position(λr) are the same.The most relevant hyperspectral characteristic variable of starch in leaves is yellow edge area(SDy), and the most relevant hyperspectral characteristic variables of total sugar are blue edge amplitude(Db) and yellow edge area(SDy), correlation coefficients were 0.54,-0.54, respectively. The most relevant hyperspectral characteristic variable of both reducing sugar and total nitrogen is the normalized ratio of red edge area and yellow edge area[(SDr-SDy)/(SDr+SDy)]. The most relevant hyperspectral characteristic variable of nicotine is blue-edge position(λb).The most relevant hyperspectral characteristic variable of roasted tobacco leaf starch is the ratio of green peak and red valley (Rg/Rr) and the normalized ratio of green peak and red valley[(Rg-Rr)/(Rg+Rr)], correlation coefficients were both 0.59. The most relevant hyperspectral characteristic variable of reducing sugar is the ratio of green peak and red valley (Rg/Rr). The most relevant hyperspectral characteristic variables of total sugar, total nitrogen, nicotine are all the normalized ratio of red edge area with the yellow edge area [(SDr-SDy)/(SDr+SDy)]. The correlation coefficients of red edge amplitude (Dr), the normalized ratio of red edge area and yellow edge area are the same with the ralavent of nicotine, both of them are -0.71.4. Use physiological and biochemical indices of tobacco canopy, fresh tobacco leaf and flue-cured tobacco leaf as the dependent variable, spectral parameters of 21 species of leaf as the independent variables, the stepwise regression prediction model were established. The significance test results of the model's complex determination coefficient and independent variable regression coefficient via the random samples showed that the the data reched significant. The results further proved that spectral parameters can be used to estimate the physiological and biochemical indices of tobacco. A new method will be supplied to the water management for the tobacco farmland, which can preodict the tobacco growth condition and quality under different water content, soil water contention easily and fastly.