Hyperspectral Estimation Of Apple Tree Leaf Nitrogen,Chlorophyll And Water Contents

Nitrogen is the essential element for the apple trees’growth and the formation of yield and quality. The management of Nitrogen is one of the most important management measures of plant production which put the high yield and good quality as the goal.Nitrogen deficiency will directly affect the plant’s biosynthesis such as amino acid, protein, nucleic acid which will decrease apple trees’photosynthesis capacity and ultimate yield. Because of nitrogen’s leaching in soil, excess nitrogen is easy to cause the pollution of groundwater. As the environment problem is getting more and more attention, how to ensure high yield and good quality plants at the same time, to prevent or minimize the pollution to the environment in the plant production is the problem must be solved in the current.The traditional nutrition diagnosis of the apple trees used laboratory analysis test and its experience discrimination method based on the sensory characteristics. This method need samples, time-consuming, laborious and can’t take the nutrient dynamic monitoring of same sample. It will greatly limit the apple cultivation production and information management. Therefore, to explore the estimation methods of nitrogen content of the apple trees rapidly, accurately and nondestructively have an important practical significance in improving the trees’nutrition, further improving the apple yield and quality, avoiding pollution of the environment caused by excessive nitrogen, and promoting the sustainable development of apple industry in China.In recent years the rapid development of hyperspectral remote sensing technology has many advantages such as high spectral resolution, strong wavelength continuity, large amount of information and so on. It can record the subtle changes in plant’s growth and organs’ development, extract the detail information of biological physical and chemical parameters, and in a timely manner to understand the physiological and ecological changes and the nutritional status of plants. Chlorophyll and water is the main carrier of photosynthesis. Their content is good indicator of plant nutrition stress, photosynthetic ability and every stage of the development of aging. Plants’reflectivity is mainly affected by chlorophyll in the visible light band. The chlorophyll content and plant nitrogen content are closely related, so chlorophyll content is usually used to indicate the nitrogen content indirectly. Related research shows that there is obvious interaction between water and nitrogen, the nitrogen level under different moisture conditions or the same moisture levels under the different nitrogen conditions, the spectral reflectance of plants vary enormously. Therefore, to estimate chlorophyll content and plant moisture content quickly and accurately plays an important role in plant growth monitoring and implementing precision agriculture.This paper used apple orchards in MengYin County, Shandong Province as the experimental zone. The ground object high spectrum analysis technology was used to measure the spectral reflectance of the bottom, middle and mucro of the apple leaf respectively.The study aimed to explore best position of estimating the content of nitrogen, chlorophyll and moisture by relativity analyzing. The results showed that middle position of the leaf is the best estimation position.The study analyzed of characteristics of hyperspectral curve of the leaf. Transformed the original spectrum which was measured in best estimate location-the middle of the leaf by processing of first derivative, second derivative, red edge、blue edge and yellow edge position, vegetation index like LCI、WI and NDWI, wavelet transformation. Then made correlation analysis and regression analysis of these variables and the leaf content of nitrogen、 chlorophyll and moisture respectively to establish models for predicting the leaf content of nitrogen、chlorophyll and moisture and test to select the high fitting precision models.Results showed that the best estimation model for leaf nitrogen content is the model with the variables of the approximate coefficient of the db5wavelet in8layers decomposition of the original spectral reflectance in579nm,493nm and579nm. The model is N=3.161+1111.965*dbA(R579)+169.089*dbA(R493)-1299.704*dbA(R575). The determination coefficient R2is0.464, the root mean square error RMSE is0.166and the relative error RE%is5.77%.Best estimation model for leaf chlorophyll content is the model with the variables of the first derivative in741nm and731nm. The estimation model is Chl=47.577+23688.831*R’741-11346.646*R’731。The determination coefficient R2is0.364, the root mean square error RMSE is2.02and the relative error RE%is2.73%.Best estimation model for leaf moisture content is the model with the variables of the first derivative in1428nm. The estimation model is FMC=85.582+60.955.404*R’1428。The determination coefficient R2is0.398, root mean square error RMSE is4.59and the relative error RE%is6.69%.This paper studied the chlorophyll and moisture content’s influence to the nitrogen estimating model. The results showed that leaf water content has little effect to nitrogen estimation. The content of chlorophyll could improve the accuracy of the nitrogen estimation.The best estimation model for nitrogen content is the model with the variables of the approximate coefficient of the db5wavelet in8layers decomposition of the original spectral reflectance and the SPAD value. Its fitting precision is highest. The model is N=2.704+862.237*dbA(R579)+0.014*SPAD+147.980*dbA(R493)-1029.272*dbA(R575). The determination coefficient R2is0.512, the root mean square error RMSE is0.149and the relative error RE%is4.93%.