Cucumber (Cucumis sativus L.) is a widely cultivated and economically profitable crop. However, Fusarium wilt disease can seriously affect cucumber yields, as it is difficult to prevent and eliminate. Therefore, a reliable method is needed for the rapid and early detection of Fusarium infection in cucumbers, which could be provided via the kinetic imaging of chlorophyll fluorescence (ChlF). In this study, ChlF imaging and kinetic parameters were utilized with gray and radial basis function models to monitor cucumber Fusarium wilt disease. The results indicate that the disease can be detected and predicted using this imaging technique before symptoms become visible.

The size distribution and concentration of atmospheric aerosol particles are important parameters that characterize the physical and optical properties of aerosols. Understanding the spectral distribution of atmospheric particles is fundamental for studying particulate pollution. Based on observational data acquired by an aerodynamic particle sizer APS-3321 and a laser dust meter LD-5 from September 2017 to July 2018, the temporal evolution of particle size distribution and concentration of aerosols were studied at the North Minzu University in Yinchuan (38°29’52"N 106°06’28"E), China. By cluster analysis of particle surface area distribution in different weathers, seven clusters were obtained and discussed. The experimental results showed that the particle number and mass concentrations were different in the four seasons of the year. Compared with the unimodal distribution of the particle number concentration and the doublet mass concentration distribution in sunny weather conditions, there were great differences in the particle spectrum distribution in haze, snowfall, and dusty weather conditions. The particle surface area distributions of the seven clusters were clearly different from each other and influenced by meteorological conditions. Additionally, the PM2.5 value has a relationship with the meteorological elements. The research is helpful to reveal the formation mechanism and source of aerosols in Yinchuan, China.

Many plant diseases have significant effects on the quality and yield of barbary wolfberry (Lycium barbarum L.) plants. The Barbary wolfberry leaf blight disease is a new disease that has emerged in recent years. This disease has a high infection rate and can destroy the entire barbary wolfberry tree if not detected and addressed in a timely manner. Therefore, to minimize disease-related losses, reduce resource waste, and improve food security and the sustainable development of barbary wolfberry, it is crucial to identify and control disease in its early stages. Chlorophyll fluorescence (ChlF) can reflect changes in the photosynthetic physiology of crops with great sensitivity, and thus has been widely used in the early detection of crop diseases. Aiming to improve diagnostic accuracy, this paper proposes a machine learning-based algorithm for the early detection of barbary wolfberry leaf blight disease based on ChlF dynamics images. First, the ChlF dynamics images of 100 normal leaves, 100 pre-diseased leaves, and 100 leaf-blight leaves are collected, and the corresponding fluorescence kinetics curves are plotted according to the acquired time-series data. Three key parameters, namely, the slope of the fluorescence attenuation fitting line (K), envelope area of the kinetics curve (Area), and steady-state fluorescence value (S) are extracted. The random forest (RF) algorithm, support vector machine (SVM) algorithm, and K-nearest neighbor (KNN) algorithm are used to construct diagnosis models of the barbary wolfberry leaf blight disease. The results show that the average classification accuracies of the models based on the RF, SVM, and KNN algorithms for the three types of leaves are 90.7, 85.3, and 84.0%, respectively. Based on the results, the method presented in this study is effective in the diagnosis and prediction of barbary wolfberry leaf blight disease.

Lycium barbarum sprouts, as a green food with significant development potential, are vulnerable to pests and diseases, particularly mite infestation that can cause leaf distortion and deformation, severely impacting yield and quality. Hyperspectral technology, with its exceptional spectral and spatial resolution, has demonstrated great potential in the monitoring of crop pests and diseases. However, its high data dimensionality and feature similarity pose challenges for analysis. In this study, hyperspectral data from Lycium barbarum sprout leaves were collected to establish a dataset representing four levels of pest severity. KNN, RF, SVM, and 1D‐CNN classification models were then employed for modeling and predicting pest levels. Experimental results indicated that the 1D‐CNN model in deep learning achieved the highest classification accuracy, reaching 96.08%. Traditional machine learning algorithms, such as random forest (RF), also exhibited high accuracy, with 94.12%. In particular, when classifying severe and moderate pest infestations, the 1D‐CNN demonstrates superior precision, recall, and F1‐score. The study further revealed that hyperspectral technology can effectively detect changes in leaf physiological structure, offering a theoretical foundation for early pest detection. This study presents an efficient, nondestructive technical method for the early detection of pests and diseases in Lycium barbarum sprouts. Additionally, it allows for the determination of varying levels of pest infestation. This will guide farmers in adopting effective control measures, reducing the reliance on chemical pesticides, and enhancing the yield and quality of Lycium barbarum sprouts. The research methodology of this paper provides a scientific foundation for precision agriculture management.

The complex refraction index (CRI) and particle size distribution (PSD) are the most important microphysical properties of atmospheric aerosol. They play an important role in the light scattering characteristics of aerosol and atmospheric radiative forcing. Therefore, accurate and rapid inversion of the CRI and PSD of aerosols is very important to study the microphysical properties of aerosols and their effects on climate. Since most of the aerosol particles are non-spherical, and the interaction between aerosol CRI and PSD makes it very difficult to invert aerosol CRI and PSD simultaneous. In order to solve the above problems, a method of simultaneous retrieval of aerosol CRI and PSD using Latin hypercube sampling (LHS) is proposed in this study. The T-matrix method is used to calculate the light scattering parameters of non-spherical aerosol particles, and the multi-wavelength lidar is used to detect the extinction and backscattering parameters of aerosols. In this paper, the LHS algorithm is used to sample the real and imaginary parts of the aerosol CRI, and the total number of particles, the median radius and the mode width of the particle PSD. According to the sampling results, the aerosol optical parameters calculated by T-matrix are compared with those measured by lidar system, and the root-mean square error between them is calculated. The CRI and PSD of the actual aerosol are obtained by successive approximation method. In this paper, field experiments and sampling inversion are carried out at three different weather conditions. The result shows that, the inversion of CRI is in accordance with the CRI of local typical aerosol types. The error between the inversion PSD and the measured PSD by particle spectrometer are all within 10%. The method proposed in this paper for simultaneous inversion of aerosol CRI and PSD is of great practical value for further study of aerosol microphysical properties and their effects on climate.