Xujun Ye

Estimation of citrus yield from canopy spectral features determined by airborne hyperspectral imagery

Hyperspectral imagery has become increasingly available in recent years and this has necessitated the evaluation of its potential for crop monitoring and precision agriculture applications. The potential of using airborne hyperspectral imagery to develop yield prediction models for citrus fruits was examined in this paper. Hyperspectral images in 72 visible and near-infrared (NIR) wavelengths (407–898 nm) were acquired over a citrus orchard in Japan by an Airborne Imaging Spectrometer for Applications (AISA) Eagle system. The canopy spectral features of individual trees were identified using pixel-based average spectral reflectance values at various wavelengths from the acquired images, which were then used to develop yield prediction models. Yield prediction models were developed using three different techniques: (i) three commonly employed vegetation indices, i.e. the normalized difference vegetation index (NDVI), simple ratio (SR) and photochemical reflectance index (PRI); (ii) a few significant wavelengths; and (iii) partial least squares (PLS) regression factors. Greater prediction accuracy was obtained with PLS models than with the models based on NDVI, SR or PRI, or the significant wavelengths. PLS models showed a significant correlation between hyperspectral imagery data and actual citrus yield for data acquired in 2003 and 2004. These results confirmed the hypothesized correlation between canopy spectral features and citrus yield. This information is valuable for forecasting yields, planning harvest schedules and generating prescription maps for the application of tree-specific alternate bearing control measures and management practices.

Inter-Relationships Between Canopy Features and Fruit Yield in Citrus as Detected by Airborne Multispectral Imagery

The objective of this research was to examine the inter-relationships between canopy features and the fruit yield of citrus crops. Satsuma Mandarin (Citrus unshiu Marc.), a native citrus variety in southeastern Asia, grown in an orchard located at Nebukawa Agricultural Research Station, Kanagawa prefecture, Japan, was used for this preliminary analysis. Airborne multispectral images in the red, green, blue, and near-infrared (NIR) bands with a high spatial resolution of 0.2 ×0.2 m were acquired over the experimental site at four time periods in 2002 and 2003. Images based on normalized difference vegetation index (NDVI) were generated with ERDAS Imagine 8.6 software. From these images, thresholded pixel counts (TPCs), indicators of the relative leaf areas of several leaf types in each canopy, for 48 selected tree samples were extracted using a program developed in MATLAB R12. Pearsons correlation analysis was employed to examine the relationships between each of the TPCs and the fruit yields of citrus in 2002, 2003, and 2004. Results indicated that some TPCs showed a higher correlation with citrus fruit yield than those corresponding to the entire canopy size, particularly the TPCs extracted from the visible red, green, and blue wavelengths. The TPCs corresponding to the mature leaves before the fast vegetative growth (May) were found to be significantly correlated with the fruit yield of the same growing season, while those corresponding to the younger leaves during this period were more significantly correlated with the fruit yields of the previous and the following growing seasons. These results confirmed the inter-relationships between canopy features and the fruit yield of citrus crops. This information also implies an unmatched energy allocation dynamic between different leaf types within the canopy, which may lead to an unsynchronized leaf energy contribution, direct (mature leaves) or delayed (younger leaves), to the fruiting of citrus crops. In addition, the models based on the TPCs extracted from early seasons images demonstrated the potential of airborne multispectral imagery to forecast the fruit yield of citrus trees. The obtained yield estimates can provide valuable information for planning fruit harvest schedules and generating prescription maps for tree-specific management practices on an individual tree basis. However, further investigations are necessary before these models can be applied in a practical situation.

Limited and time-delayed internal resource allocation generates oscillations and chaos in the dynamics of citrus crops

Alternate bearing or masting is a yield variability phenomenon in perennial crops. The complex dynamics in this phenomenon have stimulated much ecological research. Motivated by data from an eight-year experiment with forty-eight individual trees, we explored the mechanism inherent to these dynamics in Satsuma mandarin (Citrus unshiu Marc.). By integrating high-resolution imaging technology, we found that the canopy structure and reproduction output of individual citrus crops are mutually dependent on each other. Furthermore, it was revealed that the mature leaves in early season contribute their energy to the fruiting of the current growing season, whereas the younger leaves show a delayed contribution to the next growing season. We thus hypothesized that the annual yield variability might be caused by the limited and time-delayed resource allocation in individual plants. A novel lattice model based on this hypothesis demonstrates that this pattern of resource allocation will generate oscillations and chaos in citrus yield.

Application of airborne hyperspectral imagery to estimating fruit yield in citrus

This study investigated the applicability of airborne hyperspectral imagery to the estimation of fruit yield in citrus. Hyperspectral images in 72 visible and near-infrared (NIR) wavelengths (from 407 to 898 nm) were acquired over a citrus orchard in Japan by an Airborne Imaging Spectrometer for Applications (AISA) Eagle system. The canopy features of individual trees were identified using pixel-based average spectral reflectance values at various wavelengths from the acquired images. Fruit yields on 48 individual trees were recorded and the yield prediction models were developed using different prediction variables — (i) several commonly used vegetation indices (VIs), (ii) the newly derived two band vegetation index (TBVI) and (iii) principal components (PCs) and partial least square regression (PLS) factors obtained by chemometrics analysis. In spite of the variations of prediction accuracies among different models, this study confirmed the potential of airborne hyperspectral imagery to predict the fruit yield in citrus. Yield estimates can provide valuable information for forecasting yields, planning harvest schedules and generating prescription maps for tree-specific application of alternate bearing control measures and other management practices.

Estimation and visualizaion of nitrogen content in citrus canopy using hyperspectral imagery

This study investigated the capability of hyperspectral imagery for estimating and visualizing the nitrogen content in citrus canopy. Fresh citrus leaf samples including the new, medium aged and old leaves were collected from a citrus orchard during the plants vigorous vegetative growing season. Hyperspectral imageries were obtained for leaf samples in laboratory as well as for the whole canopy in the field with ImSpector V10E (Spectral Imaging Ltd., Oulu, Finland). The average spectral data for each leaf sample were extracted with ENVI software. The nitrogen content in each leaf sample was measured by the Dumas combustion method with the rapid N cube (Elementar Analytical, Germany). Simple correlation analysis and the two band vegetation index (TBVI) were used to develop the spectra data-based nitrogen content prediction models. Results indicated that the model with the two band vegetation index (TBVI) based on the wavelengths 811 nm and 856 nm achieved the optimal estimation of nitrogen content in citrus leaves (R2=0.6692). The canopy image for the identified TBVI was calculated, and the nitrogen content of the canopy was visualized by incorporating the model into the TBVI image. The results suggest the potential of hyperspectral imagery for the detection and diagnosis of nitrogen status in citrus canopy. This would provide valuable information for the implementation of individual tree-based fertilization schemes in precision orchard management practices.

Application of portable hyper-spectral camera in andisols soil nitrogen assessment

Soil nitrogen content has very important influence on soil parameters and crop productions, so the soil nitrogen monitoring and management are urgently required to satisfy the demand for precision agriculture. This research is aimed to develop the simple and low cost monitoring methods to determine the soil nitrogen pools using hyper spectral camera in Andisols soil. Partial least square regression (PLSR) with cross-validation was used to calibrate the spectral data. The research estimated the potential of soil nitrogen qualitative analysis using visible (VIS) spectrographs, near infrared (NIR) spectrographs and combination of these two spectrographs. Results showed that model assessment accuracy by V10 (360nm-1010nm) spectrograph is higher than by N17E (900nm-1700nm) spectrograph. In addition, combining these two spectrographs increased the model prediction accuracy. Evaluation index (EI) test was employed in this research, and the results showed that prediction is acceptable to use of practical to predict soil nitrogen content.

Fruit Yield Estimation Through Multispectral Imaging

Recent advances in spectral imaging technology have enabled the development of models that estimate various crop parameters from spectral imagery data. The present research investigated the alternate bearing dynamics as well as the estimation of fruit yield in citrus crops using multispectral imaging technology. Canopy features of individual trees were extracted from the multispectral images and were then used to relate to the fruit yield of citrus trees through various modelling techniques. Results showed that the alternate bearing behaves more significantly in terms of the fruit density rather than the total fruit yield on individual trees. The normalized difference vegetation index (NDVI) demonstrated greater relevance than other multiple wavebands in predicting the fruit yield on individual citrus trees. Analysis results confirmed the interrelationships between canopy features and the fruit yield of citrus crops and implied the unmatched energy allocation mechanisms between different leaf types within the canopy of citrus crops. Effective models were developed for fruit yield estimation of citrus from multispectral images acquired before the fruit-growing season.