Soil organic carbon estimation using remote sensing data-driven machine learning.

Soil organic carbon (SOC) is a crucial component of the global carbon cycle, playing a significant role in ecosystem health and carbon balance. In this study, we focused on assessing the surface SOC content in Shandong Province based on land use types, and explored its spatial distribution pattern and influencing factors. Machine learning methods including random forest (RF), extreme gradient boosting (XGBoost), and support vector machine (SVM) were employed to estimate the surface SOC content in Shandong Province using diverse data sources like sample data, remote sensing data, socio-economic data, soil texture data, topographic data, and meteorological data. The results revealed that the SOC content in Shandong Province was 8.78 g/kg, exhibiting significant variation across different regions. Comparing the model error and correlation coefficient, the XGBoost model showed the highest prediction accuracy, with a coefficient of determination (R²) of 0.7548, root mean square error (RMSE) of 7.6792, and relative percentage difference (RPD) of 1.1311. Elevation and Clay exhibited the highest explanatory power in clarifying the surface SOC content in Shandong Province, contributing 21.74% and 13.47%, respectively. The spatial distribution analysis revealed that SOC content was higher in forest-covered mountainous regions compared to cropland-covered plains and coastal areas. In conclusion, these findings offer valuable scientific insights for land use planning and SOC conservation.

Multispectral versus texture features from ZiYuan-3 for recognizing on deciduous tree species with cloud and SVM models.

Tree species recognition accuracy greatly affects forest remote sensing mapping and forestry resource monitoring. The multispectral and texture features of the remote sensing images from the ZiYuan-3 (ZY-3) satellite at two phenological phases of autumn and winter (September 29th and December 7th) were selected for constructing and optimizing sensitive spectral indices and texture indices. Multidimensional cloud model and support vector machine (SVM) model were constructed by the screened spectral and texture indices for remote sensing recognition of Quercus acutissima (Q. acutissima) and Robinia pseudoacacia (R. pseudoacacia) on Mount Tai. The results showed that, the correlation intensities of the constructed spectral indices with tree species were preferable in winter than in autumn. The spectral indices constructed by band 4 showed the superior correlation compared with other bands, both in the autumn and winter time phases. The optimal sensitive texture indices for both phases were mean, homogeneity and contrast for Q. acutissima, and contrast, dissimilarity and second moment for R. pseudoacacia. Spectral features were found to have a higher recognition accuracy than textural features for recognizing on both Q. acutissima and R. pseudoacacia, and winter showing superior recognition accuracy than autumn, especially for Q. acutissima. The recognition accuracy of the multidimensional cloud model (89.98%) does not show a superior advantage over the one-dimensional cloud model (90.57%). The highest recognition accuracy derived from a three-dimensional SVM was 84.86%, which was lower than the cloud model (89.98%) in the same dimension. This study is expected to provide technical support for the precise recognition and forestry management on Mount Tai.

Inversion of Nitrogen Concentration in Apple Canopy Based on UAV Hyperspectral Images.

As the major nutrient affecting crop growth, accurate assessing of nitrogen (N) is crucial to precise agricultural management. Although improvements based on ground and satellite data nitrogen in monitoring crops have been made, the application of these technologies is limited by expensive costs, covering small spatial scales and low spatiotemporal resolution. This study strived to explore an effective approach for inversing and mapping the distributions of the canopy nitrogen concentration (CNC) based on Unmanned Aerial Vehicle (UAV) hyperspectral image data in a typical apple orchard area of China. A Cubert UHD185 imaging spectrometer mounted on a UAV was used to obtain the hyperspectral images of the apple canopy. The range of the apple canopy was determined by the threshold method to eliminate the effect of the background spectrum from bare soil and shadow. We analyzed and screened out the spectral parameters sensitive to CNC, including vegetation indices (VIs), random two-band spectral indices, and red-edge parameters. The partial least squares regression (PLSR) and backpropagation neural network (BPNN) were constructed to inverse CNC based on a single spectral parameter or a combination of multiple spectral parameters. The results show that when the thresholds of normalized difference vegetation index (NDVI) and normalized difference canopy shadow index (NDCSI) were set to 0.65 and 0.45, respectively, the canopy's CNC range could be effectively identified and extracted, which was more refined than random forest classifier (RFC); the correlation between random two-band spectral indices and nitrogen concentration was stronger than that of other spectral parameters; and the BPNN model based on the combination of random two-band spectral indices and red-edge parameters was the optimal model for accurately retrieving CNC. Its modeling determination coefficient (R2) and root mean square error (RMSE) were 0.77 and 0.16, respectively; and the validation R2 and residual predictive deviation (RPD) were 0.75 and 1.92. The findings of this study can provide a theoretical basis and technical support for the large-scale, rapid, and non-destructive monitoring of apple nutritional status.

Inversion reflectance by apple tree canopy ground and unmanned aerial vehicle integrated remote sensing data.

To obtain accurate spatially continuous reflectance from Unmanned Aerial Vehicle (UAV) remote sensing, UAV data needs to be integrated with the data on the ground. Here, we tested accuracy of two methods to inverse reflectance, Ground-UAV-Linear Spectral Mixture Model (G-UAV-LSMM) and Minimum Noise Fraction-Pixel Purity Index-Linear Spectral Mixture Model (MNF-PPI-LSMM). At wavelengths of 550, 660, 735 and 790 nm, which were obtained by UAV multispectral observations, we calculated the canopy abundance based on the two methods to acquire the inversion reflectance. The correlation of the inversion and measured reflectance values was stronger in G-UAV-LSMM than MNF-PPI-LSMM. We conclude that G-UAV-LSMM is the better model to obtain the canopy inversion reflectance.

Estimating apple tree canopy chlorophyll content based on Sentinel-2A remote sensing imaging.

The remote sensing technology provides a new means for the determination of chlorophyll content in apple trees that includes a rapid analysis, low cost and large monitoring area. The Back-Propagation Neural Network (BPNN) and the Supported Vector Machine Regression (SVMR) methods were both frequently used method to construct estimation model based on remote sensing imaging. The aim of this study was to find out which estimation model of apple tree canopy chlorophyll content based on the vegetation indices constructed with visible, red edge and near-infrared bands of the sensor of Sentinel-2 was more accurate and stabler. The results were as follows: The calibration set coefficient of determination (R2) value of 0.729 and validation set R2 value of 0.667 of the model using the SVMR method based on the vegetation indices (NDVIgreen + NDVIred + NDVIre) were higher than those of the model using the BPNN method by 8.2% and 11.0%, respectively. The calibration set root mean square error (RMSE) of 0.159 and validation set RMSE of 0.178 of the model using the SVMR method based on the vegetation indices (NDVIgreen + NDVIred + NDVIre) were lower than those of the model using the BPNN method by 5.9% and 3.8%, respectively.

Evaluation of the nitrogen content during the new-shoot-growing stage in apple leaves using two-dimensional correlation spectroscopy.

The new-shoot-growing stage is an important period of apple tree nutrition distribution. The objective of this study is to provide technical support for apple tree nutrition diagnosis by constructing quantitative evaluation models between the apple leaf nitrogen content during the new-shoot-growing stage and characteristic spectral parameters. The correlation coefficients between the original spectral data and the nitrogen content were calculated. Then, the sensitive bands of the nitrogen content were selected using the theory of two-dimensional (2D) correlation spectroscopy. Finally, partial least squares regression (PLSR) and support vector machine (SVM) evaluation models were established using 2 parameters: Rx (maximum spectral reflectivity in the waveband) and Sx (total spectral reflectivity in the waveband). The results showed that the sensitive bands in the 2D correlation synchronous and asynchronous spectrograms were 537-560 nm and 708-719 nm. The PLSR model can be used to estimate the nitrogen content. Compared with PLSR, SVM provided better modeling and testing results, with a larger coefficient of determination (R2) and a smaller root-mean-square error (RMSE). The SVM model based on Sx was a good backup method. The calibration R2 of the model was 0.821, its RMSE was 0.710 g·kg-1, the validation R2 was 0.768, and its RMSE was 1.019 g·kg-1. The SVM model based on 2D correlation spectroscopy can be used to quantitatively estimate the nitrogen content in apple leaves.

Improving the Accuracy of the Hyperspectral Model for Apple Canopy Water Content Prediction using the Equidistant Sampling Method.

The influence of the equidistant sampling method was explored in a hyperspectral model for the accurate prediction of the water content of apple tree canopy. The relationship between spectral reflectance and water content was explored using the sample partition methods of equidistant sampling and random sampling, and a stepwise regression model of the apple canopy water content was established. The results showed that the random sampling model was Y = 0.4797 - 721787.3883 × Z3 - 766567.1103 × Z5 - 771392.9030 × Z6; the equidistant sampling model was Y = 0.4613 - 480610.4213 × Z2 - 552189.0450 × Z5 - 1006181.8358 × Z6. After verification, the equidistant sampling method was verified to offer a superior prediction ability. The calibration set coefficient of determination of 0.6599 and validation set coefficient of determination of 0.8221 were higher than that of the random sampling model by 9.20% and 10.90%, respectively. The root mean square error (RMSE) of 0.0365 and relative error (RE) of 0.0626 were lower than that of the random sampling model by 17.23% and 17.09%, respectively. Dividing the calibration set and validation set by the equidistant sampling method can improve the prediction accuracy of the hyperspectral model of apple canopy water content.

[Hyperspectral Estimation of Apple Tree Canopy LAI Based on SVM and RF Regression].

Leaf area index (LAI) is the dynamic index of crop population size. Hyperspectral technology can be used to estimate apple canopy LAI rapidly and nondestructively. It can be provide a reference for monitoring the tree growing and yield estimation. The Red Fuji apple trees of full bearing fruit are the researching objects. Ninety apple trees canopies spectral reflectance and LAI values were measured by the ASD Fieldspec3 spectrometer and LAI-2200 in thirty orchards in constant two years in Qixia research area of Shandong Province. The optimal vegetation indices were selected by the method of correlation analysis of the original spectral reflectance and vegetation indices. The models of predicting the LAI were built with the multivariate regression analysis method of support vector machine (SVM) and random forest (RF). The new vegetation indices, GNDVI527, ND-VI676, RVI682, FD-NVI656 and GRVI517 and the previous two main vegetation indices, NDVI670 and NDVI705, are in accordance with LAI. In the RF regression model, the calibration set decision coefficient C-R2 of 0.920 and validation set decision coefficient V-R2 of 0.889 are higher than the SVM regression model by 0.045 and 0.033 respectively. The root mean square error of calibration set C-RMSE of 0.249, the root mean square error validation set V-RMSE of 0.236 are lower than that of the SVM regression model by 0.054 and 0.058 respectively. Relative analysis of calibrating error C-RPD and relative analysis of validation set V-RPD reached 3.363 and 2.520, 0.598 and 0.262, respectively, which were higher than the SVM regression model. The measured and predicted the scatterplot trend line slope of the calibration set and validation set C-S and V-S are close to 1. The estimation result of RF regression model is better than that of the SVM. RF regression model can be used to estimate the LAI of red Fuji apple trees in full fruit period.

[Using atomic fluorescence spectrometry to study the spatial distribution of As and Hg in orchard soils].

Aqua regia digestion, double channels-atomic fluorescence spectrometry method was used to determine the concentrations of As and Hg in orchard soils of Qixia City - the main apple production area of Shandong province. Validate The detection limitation, accuracy and precision of the method were validated, the spatial distribution was analyzed, and the characteristics of As and Hg pollution in Qixia orchard soils were assessed. The results showed that the range of As concentration in Qixia soils is between 2.79 and 20.93 mg x kg(-1), the average concentration is 10.59 mg x kg(-1), the range of Hg concentration in Qixia soil is between 0.01 and 0.79 mg x kg(-1), the average concentration is 0.12 mg x kg(-1). The variation of As concentration in soils is small, whereas that of Hg concentration is large. Frequency distribution graphics of As and Hg showed that the concentration of As in soils is according with the normal distribution approximately and the concentrations are mostly between 7 and 15 mg x kg(-1), the concentration of Hg in soil isn't according with the normal distribution and the concentrations are mostly between 0.03 and 0.21 mg x kg(-1). The correlations between the concentrations of As or Hg in soils and the nutrient are not significant and there is no significant correlation even between As and Hg. Based on the environmental technical terms for green food production area, the As concentration in orchard soil of Qixia City is at clean level, but there are 4.76% of sample points with Hg pollution index exceeding 1, and this should be attracted the attention of the administrators.

[Estimation of chlorophyll content in apple tree canopy based on hyperspectral parameters].

The hyperspectral reflectance of apple tree canopy during spring shoots stopping growth period was measured using ASD FieldSpec3 field spectrometer. Original spectral data were processed in deviation forms, and significant spectrum parameters correlated with chlorophyll content were found out with correlation analysis. The best vegetation indices were chosen and the apple canopy chlorophyll content estimation model was established by analyzing vegetation index of two-band combination in the sensitive region 400-1 350 nm. The result showed that (1) The sensitive band region of apple canopy chlorophyll content is 400-1 350 nm. (2) The vegetation index CCI(D(794)/D(763)) can commendably estimate the apple canopy chlorophyll content. (3) The model with CCI(D(794)/D(763)) as the independent variables was determined to be the best for chlorophyll content prediction of apple tree canopy. Therefore, using hyperspectral technology can estimate apple canopy chlorophyll content more rapidly and accurately, and provides a theoretical basis for rapid apple tree canopy nutrition diagnosis and growth monitoring.

[Hyperspectral estimation of kalium content in apple florescence canopy based on fuzzy recognition].

The objective of the present paper is fast and nondestructive estimate of kalium content using ASD FieldSpec3 spectrometer determined hyperspectral data in apple florescence canopy. According to detection of hyperspectral data of the apple florescence canopy and kalium content data at laboratory in Qixia city of experimental orchards in 2008 and 2009, the correlation analysis of hyperspectral reflectance and its eleven transforms with kalium content was proceeded. The biggest correlation coefficient as independent variable and the estimation model of kalium content were established based on fuzzy recognition algorithms. The model was tested by sample inspection in 2008 and verified by data in 2009. The results showed that the correlation is less for the original spectral reflectance (R) and its reciprocal(1/R), logarithm (lgR), square root (R1/2) and the kalium content, but it is enhanced obviously for their first derivative and second derivative. The correlation coefficient(r) of kalium content estimating model y = 11.344 5h + 1.309 7 is 0.985 1, the total root mean square difference (RMSE) is 0.355 7 and F statistics is 3 085.6. The average relative error of measured values and estimated values for 24 inspection sample is 9.8%, estimation accuracy is 90.2% and verification accuracy is 83.3% utilizing test data in 2009. It was showed that this model is more stable by estimating apple florescence canopy of kalium content and the model precision is able to meet the needs of production.

[Satellite remote sensing retrieval of canopy nitrogen nutritional status of apple trees at blossom stage].

Taking Qixia City of Shandong, China as the study area, and based on the Landsat-5 TM and ALOS AVNIR-2 images, the canopy retrieval reflectance of apple trees at blossom stage was acquired. In combining with the measured reflectance of sample trees, the nitrogen-sensitive spectral indices were constructed and selected. By using the sensitive spectral indices as the independent variables, the nitrogen retrieval models were established, and the model with the best accuracy was used for spatial retrieve. The correlations between the spectral indices and the nitrogen nutritional status were in the order of canopy > leaf > flower. The sensitive indices were mainly composed of green, red, and near infrared bands. The accuracy of the retrieval models was in the order of support vector regression > multi-variable stepwise regression > one-variable regression. The retrieval results based on different images were similar, and showed that the leaf nitrogen content was mainly of grades 3-4 (27-33 g x kg(-1)), and the canopy nitrogen nutrient indices were mainly of grades 2-4 (TM: 38-47 g x kg(-1); ALOS: 32-41 g x kg(-1)). The spatial distribution of the retrieval nitrogen nutritional status based on different images also showed the similar trend, i. e., the nitrogen nutritional status was higher in the north and south than that in the middle part of the study area, and the areas with the high grades of leaf nitrogen and canopy nitrogen were mainly located in Sujiadian Town and Songshan subdistrict in the northwest, Zangjiazhuang Town and Tingkou Town in the northeast, and Shewopo Town in the south, which were consistent with the distribution of the key towns for apple production in Qixia City. This study provided a feasible method for the acquisition of nitrogen nutritional status of apple trees on macroscopic scale, and also, provided reference for other similar remote sensing retrievals.

[Quantitative remote sensing retrieval of apple tree canopy reflectance at blossom stage in hilly area].

By using the TM and ALOS images with different resolutions at the prosperous blossom stage of apple trees in Qixia City of Shandong Province, and taking the slope aspect coefficient and the ratio of canopy flower to leaf into account, the ground surface reflectance was retrieved through radiometric correction. The canopy reflectance of the apple trees was further retrieved by pixel unmixing method, and the retrieval effect and accuracy were assessed by the comparison of the retrieved reflectance with the measured canopy reflectance and apparent reflectance of 30 sample apple orchards. The results showed that radiometric correction effectively weakened the effects of atmosphere and topography, recovered the ground objects in the shadows, and obviously enhanced the analytical ability of ground surface retrieval reflectance images. Either TM or ALOS images, both the absolute and relative errors between retrieval reflectance and measured reflectance of apple tree canopy were the smallest. The relative errors of all bands were consistent, and its variation trend among the 30 sample apple orchards was also consistent with the measured reflectance, which showed the necessary of pixel unmixing. Moreover, the changes of the reflectance among the sample apple orchards showed similar characteristics when the retrieval method was used for different resolution images. The images with high resolution were more superior, but, because of band limitation, it would be better to integrate the high resolution images with moderate resolution images.

[Hyperspectral estimation models of chlorophyll content in apple leaves].

The present study chose the apple orchard of Shandong Agricultural University as the study area to explore the method of apple leaf chlorophyll content estimation by hyperspectral analysis technology. Through analyzing the characteristics of apple leaves' hyperspectral curve, transforming the original spectral into first derivative, red edge position and leaf chlorophyll index (LCI) respectively, and making the correlation analysis and regression analysis of these variables with the chlorophyll content to establish the estimation models and test to select the high fitting precision models. Results showed that the fitting precision of the estimation model with variable of LCI and the estimation model with variable of the first derivative in the band of 521 and 523 nm was the highest. The coefficients of determination R2 were 0.845 and 0.839, the root mean square errors RMSE were 2.961 and 2.719, and the relative errors RE% were 4.71% and 4.70%, respectively. Therefore LCI and the first derivative are the important index for apple leaf chlorophyll content estimation. The models have positive significance to guide the production of apple cultivation.

[Remote sensing techniques of apple orchard information extraction based on linear spectral unmixing with measured data].

Taking Qixia City, Shandong Province of China as the research region, and by using pixel unmixing for the TM image at apple flowering stage, the apple orchard information was extracted. Based on the measured spectral end-members, wavelet transform was adopted to improve the linear unmixing model. The improved linear spectral unmixing model, measured end-member based linear spectral unmixing model, and TM image end-member based linear spectral unmixing model were employed to extract the apple orchard information, and the ALOS data were used for accuracy estimation. After the accurate atmospheric and topographic correction, it was feasible to use the measured spectral end-members for pixel unmixing, and the area precision of apple orchard information acquisition was greater than 97%. The regression analysis on the NDVI of abundance image and the average NDVI of ALOS data showed that the R2 was higher than 0.8. Therefore, using wavelet transform to improve the linear spectral unmixing model could improve the unmixing accuracy to a certain degree.

[Hyper-spectral estimation of soil organic matter content based on wavelet transformation].

A total of 60 soil samples with approximate contents of N, P, and K and greatly different content of organic matter were selected by statistical analysis. Through hyper-spectral detection and analysis, the first derivative spectrum of the soil logarithmic reflectance was obtained, and was decomposed by the Bior 1.3 wavelet function. The approximative signal of the lowest frequency and the noise signal of the highest frequency were removed from the input spectrum so as to obtain the characteristic spectrum corresponding to soil physical and chemical parameters. The sensitive bands of soil organic matter were selected by correlation analysis, and the forecasting models were built by multiple regression analysis, based on the sensitive bands and the characteristic spectrum, respectively. Through comparison analysis, the optimal wavelet decomposing resolution for extracting the characteristic spectrum of soil organic matter was ascertained, and the best forecasting model was established. The best wavelet decomposing resolution was 9, followed by 8 and 10. Based on the characteristic spectrum of wavelet decomposing of 9 resolutions, the model R2 reached 0.89, which was increased by 0.31 as compared to the model based on sensitive bands, and increased by 0.10 as compared to the model based on the original spectrum.

[Quantitative models between canopy hyperspectrum and its component features at apple tree prosperous fruit stage].

Hyperspectral technique has become the basis of quantitative remote sensing. Hyperspectrum of apple tree canopy at prosperous fruit stage consists of the complex information of fruits, leaves, stocks, soil and reflecting films, which was mostly affected by component features of canopy at this stage. First, the hyperspectrum of 18 sample apple trees with reflecting films was compared with that of 44 trees without reflecting films. It could be seen that the impact of reflecting films on reflectance was obvious, so the sample trees with ground reflecting films should be separated to analyze from those without ground films. Secondly, nine indexes of canopy components were built based on classified digital photos of 44 apple trees without ground films. Thirdly, the correlation between the nine indexes and canopy reflectance including some kinds of conversion data was analyzed. The results showed that the correlation between reflectance and the ratio of fruit to leaf was the best, among which the max coefficient reached 0.815, and the correlation between reflectance and the ratio of leaf was a little better than that between reflectance and the density of fruit. Then models of correlation analysis, linear regression, BP neural network and support vector regression were taken to explain the quantitative relationship between the hyperspectral reflectance and the ratio of fruit to leaf with the softwares of DPS and LIBSVM. It was feasible that all of the four models in 611-680 nm characteristic band are feasible to be used to predict, while the model accuracy of BP neural network and support vector regression was better than one-variable linear regression and multi-variable regression, and the accuracy of support vector regression model was the best. This study will be served as a reliable theoretical reference for the yield estimation of apples based on remote sensing data.

[Standardization of technical methods for apple fluorescence canopy spectral detection].

Aiming at spectral detection of apple fluorescence canopy, the present paper carried out spectral detection tests under different weather conditions, different detection times, and different detection heights and angles to apple canopy in the two years of 2008 and 2009, so as to analyze impacts of these factors on apple canopy spectral characteristics and explore standardized spectral detection methods for apple fluorescence canopy. The results indicated the regularity in spectral reflectance of apple fluorescence canopy to a certain degree under different conditions, especially in the 760-1 350 nm near-infrared bands. The authors found that canopy spectral reflectance declined along with the decrease in sunshine and it is appropriate to detect canopy spectrum in sunny days with few clouds. In addition, spectral reflectance tended to be stable when the wind scale was below grade 2. The discrepancy of canopy spectra is small during the time period from 10:00 to 15:00 of a day compared to that of other times. For maintaining stable spectral curves, the height of detector to apple canopy needed to be adjusted to cover the whole canopy within the field of view according to detection angle of the detector. The vertical or approximately vertical detection was the best for canopy spectral reflectance acquisition. The standardization of technical methods of spectral detection for apple fluorescence canopy was proposed accordingly, which provided theoretical references for spectral detection and information extraction of apple tree canopy.

[Hyperspectrum based prediction model for nitrogen content of apple flowers].

The present paper aims to quantitatively retrieve nitrogen content in apple flowers, so as to provide an important basis for apple informationization management. By using ASD FieldSpec 3 field spectrometer, hyperspectral reflectivity of 120 apple flower samples in full-bloom stage was measured and their nitrogen contents were analyzed. Based on the apple flower original spectrum and first derivative spectral characteristics, correlation analysis was carried out between apple flowers original spectrum and first derivative spectrum reflectivity and nitrogen contents, so as to determine the sensitive bands. Based on characteristic spectral parameters, prediction models were built, optimized and tested. The results indicated that the nitrogen content of apple was very significantly negatively correlated with the original spectral reflectance in the 374-696, 1 340-1 890 and 2 052-2 433 nm, while in 736-913 nm they were very significantly positively correlated; the first derivative spectrum in 637-675 nm was very significantly negatively correlated, and in 676-746 nm was very significantly positively correlated. All the six spectral parameters established were significantly correlated with the nitrogen content of apple flowers. Through further comparison and selection, the prediction models built with original spectral reflectance of 640 and 676 nm were determined as the best for nitrogen content prediction of apple flowers. The test results showed that the coefficients of determination (R2) of the two models were 0.825 8 and 0.893 6, the total root mean square errors (RMSE) were 0.732 and 0.638 6, and the slopes were 0.836 1 and 1.019 2 respectively. Therefore the models produced desired results for nitrogen content prediction of apple flowers with average prediction accuracy of 92.9% and 94.0%. This study will provide theoretical basis and technical support for rapid apple flower nitrogen content prediction and nutrition diagnosis.

[Study on hyperspectral characteristics of apple florescence canopy].

The present study aims to systematically analyze the hyperspectral characteristics of apple florescence canopy and explore the sensitive spectra to provide the theoretical basis for large area apple information extracting and remote sensing retrieval for nutrition diagnosis. Based on the 120 hyperspectral data of apple florescence canopy acquired with ASD Field Spec 3 portable object spectrometer, the effects of different sample numbers on hyperspectral characteristics were analyzed. Using variance analysis method, the hyperspectral characteristics of apple florescence canopy and the sensitive wave bands were obtained. The results showed that with the increase in cumulative sample numbers, the hyperspectrum curves of apple florescence became stable and smooth. At the 550 nm green peak and the 760-1,300 nm reflection plateau, the reflection rate reduced with the increase in flowering amount, while in the red valley of 670 nm, the reflection rate increased with the increase in flowering amount; At the wave bands of 350-500, 600-680 and 760-1,300 nm, the variance analysis results showed very significant differences, indicating that they were sensitive wave bands of florescence canopy. With the increase in flowering amount, the red-edge position, the red-edge slope and red edge area tended to decrease gradually.