Hyperspectral Characteristics Of Rice Canopy And Grain Quality Under Different Nitrogen Levels And Lead-Cadmium Pollution

Rice is one of the main grain crops of China, and the quality and security of rice are of great significance to people’s life and health. For the purpose of real-time monitoring of rice growth dynamics with high yield, good quality, and safe security, the advantages of non-destructive determination of hyperspectral technology was employed to investigate the influences of nitrogen levels and lead-cadmium pollutions on hyperspectral characteristics and indexes of rice canopy and grain quality, and to provide basis for the high-yielding, high-quality and clean-production rice cultivation.Therefore, experiments were conducted at the agriculture experimental farms of Agriculture College of Yangzhou University in2010and2011, with the nitrogen level and the pollution of heavy metals lead and cadmium as the research objects, the canopy hyperspectral characteristics and physiological and ecological parameters of different rice varieties were measured in rice critical growth periods, in order to explore the intrinsic quantitative relationships of different nitrogen levels and heavy metal pollution degrees with rice plants, and to provide theoretical support for the remote sensing monitoring of pollution status and rice quality. The main results were as follows.(1) Spectral characteristics of Rice under different nitrogen levelsUsing the canopy reflectance spectra of the key rice growth periods attained through plot tests, feature analysis was done on the rice canopy spectra of different japonica rice varieties in different growth periods and under different nitrogen levels. Different rice varieties during the whole growing period showed the similar overall variation trends in canopy reflectance spectra under different nitrogen levels, high nitrogen treatment relating to high reflectivity in the near infrared region and low nitrogen treatment relating to light reflectance in the range of visible zones. Vegetation index RVI and NDVI of different varieties in whole growth period kept increasing with the growth process, and reached the maximum value at heading, while started decreasing after entering the maturity period. Rice nitrogen nutrition level had a remarkable influence on leaf spectral characteristics, for in the whole growth stage, leaf area index LAI was significantly related to NDVI and RVI, and rice nitrogen accumulation showed good correlativity with canopy spectral reflectance at810and870nm. The correlation of rice nitrogen content with spectral indexes RVI and NDVI was highest at heading with the determination coefficient being the largest. Nitrogen content was significantly related to spectral indexes RVI and NDVI in every growth periods, suggesting the feasibility and high accuracy of the regression models for parameters prediction constructed with RVI and NDVI.(2) Constructing Estimation Models through Hyperspectra for Rice Quality under different nitrogen levelsHow the hyperspectral data characteristics of rice grains of different cultivars under various fertilizer levels were related to their Amylase content, Protein content and Gel Consistency was analyzed by samples from field experiments. Having built various models on the relationship of NDI(x) and rice quality indexes(y) and compared the significance of predictions drawn from each model under different fertilization levels, we found that the polynomial models showed highly predictive value for rice quality indexes, whose general formula was y=Ax2+Bx+C, in which the factors A, B and C took different values depending on fertilization levels. Besides, on the survey of different genotypes, by comparing the related coefficients between Rl values and rice quality indexes, we inferred that the best band combinations were RI amylose=R783/R634, RI protein=R829/R646and RI gel consistency=R900/R670separately, and then reasoned the linear regression equation for rice quality indexes to be y=A x RI+B, where the factors A and B differed according to the genotypes. The results from this study showed that the method to estimate Amylase content, Protein and Gel Consistency of rice grains through hyperspectral data was practicable, thus providing the basis for rice quality monitoring by remote-sensing measurements.(3) The Establishment of Hyperspectral Identification Models Rice under different Lead PollutionPot experiments were conducted to investigate the characteristics of lead accumulation in rice aboveground organs and relationships between canopy spectral characteristics and the concentration of plumbum(Pd) in each part of rice plants under different lead levels with the experimental materials of Nanjing44and Liangyoupeijiu. Results showed that Pb contents of the whole plants, stems, leaves and spikes were increasing along with raised Pb levels, among which the content of Pd in stems was always the highest. Significant differences were found in the wavelength band of red light in canopy reflectance spectra curves taken under various Pb stress levels. Interestingly, with the increasing of Pd stress level, the strength of spectral reflectance reduced, and the normalized value of reflectance after removing envelopes curved downward. Several types of relationship models between NVDI(x) and Pd content in rice plant organs were built. By comparing the prediction significance of models, the optimal prediction models were established for Pd stress remote-sensing monitoring of each part of the studied cultivars, respectively. The Pb contents of each part in Nanjing44were defined as:y whoie-piam=2270.4x2-2292.8x+577.35, y stem=4260.9x2-4294x+1077.8, y leaf=2780.8x2-2777.9x+690.71, and y spike=309.31x2-306.07x+75.369, while for Liangyoupeijiu, y whole-plant=524269e-25.557x, y stem=1E+07e-31.65x, y leaf=2E+07e-34.056x, and y spike=14320e-21.756x. From these results, it could be concluded that the method of field hyperspectral remote sensing responded fairly well to Pd stress of rice plants and its stress intensity as well, thus the fast and undamaging probe of Pd pollution of rice could be achieved by the Difference Analysis of rice canopy spectra.(4) Rice Canopy Spectral Characteristics and its Forecast Evaluation under Cadmiun StressRelationships between rice canopy spectral characteristics and the concentration of cadmium(Cd) in each part of rice plants under different treatments of Cd levels were investigated quantificationally through pot experiments using Nanjing44and Liangyoupeijiu as materials. The Cd contents in whole plants, stems, leaves and spikes of both rice cultivars conformably increased with the Cd stress level, among which the content of Cd in stems was always the highest. Besides, significant differences were found in the wavelength band of red light of canopy reflectance spectra curves taken under various stress levels. As Cd stress level increased, the strength of spectral reflectance reduced, the normalized value of reflectance after removing envelopes curved downward, and the red edge positions showed apparent’red-shift’phenomena of varying degrees in these two cultivars. Multiple relationship models were built for NVDI(x) and Cd(y) contents in rice plant organs, the significances of their prediction were compared, and eventually the prediction models for each part of the studied cultivars were formulated for use of Cd stress remote-sensing monitoring. For Nanjing44, y Wholc-plant=86.207x2-56.633x+9.7361, y slem=157.65x2-101.89x+17.317, y leaf=21.619x2-14.192x+2.5283, and y spikc=4.7011x2-1.4549x+0.1628. For Liangyoupeijiu, y wholc-plant=41.495x2-29.34x+5.1829, y stem=53.364x236.778x+6.3612, y leaf=22.981x2-15.768x+2.7588, and y spike=36.347x2-25.477x+4.4473. Given these results, we conclude that field hyperspectral remote sensing is capable of responding fairly well to Cd stress of rice plants and also to the stress levels, enabling a method of fast and undamaging probe of rice Cd pollution through difference analysis of rice canopy spectra.(5) Spectral analysis for Rice under different concentrations of combined lead-cadmium pollutionThrough pot experiments, using Nanjing44and Liangyoupeijiu under different concentrations of compound pollution by heavy metals lead and cadmium as materials, with the rice canopy reflectance spectra obtained with spectroscopy, and lead and cadmium contents in organs simultaneously determined, the enrichment characteristics effects in rice aboveground organs under different concentrations of lead and cadmium pollution were investigated, and the quantitative relationships of canopy spectral characteristics with lead and cadmium contents in plant organs were studied. The canopy reflectance spectra were processed with the methods of envelope and first derivative. The red edge position changes were studied through the first derivative, and it was found that, with different cadmium pollution concentrations, the red edge position gradually moved to the longer wavelength, i.e.," red shift" appeared, and the with the concentration increased the movement degree differed. By constructing general linear models y=Ax+B for RVI, NDVI and the heavy metals contents in rice organs at maturity, and comparing the model predictions significance, remote sensing prediction models were put forward for the two rice varieties respectively, which could apply to different rice organs at maturity grown under lead-cadmium combination stress, providing the remote sensing forecasting of lead and cadmium contents. This research shows that, through the rice canopy spectral difference analysis, rice stress level of lead and cadmium pollution can be clarified, thus proposing a new way for the rapid, nondestructive detection of rice cadmium pollution.