Using hyperspectral remote sensing technology to monitor leaf area index (LAI) in a timely, fast and non-destructive manner is essential for accurate quantitative crop management. The relationships between existing vegetation indices (VIs) and LAI usually tend to saturate under dense canopies in crop production. The purpose of this study was to propose a new VI in which the estimating saturation is greatly weakened, and prediction accuracy is improved under conditions of high LAI in winter wheat (Triticum aestivum L.). The quantitative relationship between ground-based canopy spectral reflectance and LAI in wheat was investigated. The results showed that the optimized band combination, namely, the form of non-linear vegetation index (NLI) was more sensitive to changes in LAI. When λ(x1) = 798 nm and λ(y2) = 728 nm, the band combination NLI (798,728) had the highest R² of 0.757. Among the common VIs, the modified triangular vegetation index 2 (MTVI2), the ratio spectral index [RSI (760,730)] and the 2-band enhanced vegetation index (EVI2) gave superior performance (R² > 0.710) in terms of LAI estimation, but were worse than NLI (798,728). Inspired by the modified non-linear vegetation index (MNLI), NLI (798,728) was further optimized to become a novel optimized non-linear vegetation index (ONLI), which can be calculated by the formula \({{\left( { 1 { + 0} . 0 5} \right) \, \times \, \left( { 0. 6\, \times \,R_{ 7 9 8}^{2} \, - \,R_{ 7 2 8} } \right)} \mathord{\left/ {\vphantom {{\left( { 1 { + 0} . 0 5} \right) \, \times \, \left( { 0. 6\, \times \,R_{ 7 9 8}^{2} \, - \,R_{ 7 2 8} } \right)} { \left( { 0. 6\, \times \,R_{ 7 9 8}^{2} \, + \,R_{ 7 2 8} { + 0} . 0 5} \right)}}} \right. \kern-0pt} { \left( { 0. 6\, \times \,R_{ 7 9 8}^{2} \, + \,R_{ 7 2 8} { + 0} . 0 5} \right)}}\). The unified ONLI model gave an R² of 0.779 and root mean square error (RMSE) of 1.013 across all datasets. These results indicate that the novel ONLI has strong adaptability to various cultivation conditions and can provide a good estimate of LAI in winter wheat.

The pathogenesis of wheat powdery mildew (WPM) is from the bottom to upper layers of the plant, and the vertical observation angle limits the early monitoring of WPM status. Multiangle remote sensing could effectively extract spatial structural information from different plant layers. The objectives of this study were to improve the monitoring accuracy of WPM severity and to screen suitable observation angles by developing a novel vegetation index. The monitoring capacities of the novel parameters [Normalized Powdery Mildew Index (NPMI) and Ratio Powdery Mildew Index (RPMI)] and 14 optimized traditional spectral parameters were compared at 13 observation angles and different angle ranges. The correlation between all spectral parameters and disease severity was superior in the forward observation direction than in the backward observation direction in the principal plane of the Sun, and the correlation between the two observation directions decreased with an increase in observation angle. The new spectral parameter suitable for inversion of disease index was RPMI (R ₇₄₄ /R ₇₆₂ - 0.5 × R ₇₁₀ ), and the best optimal observation angle was +10°, with a 0.852 coefficient, which was 31.74% higher than that of the two-band spectral parameter, R ₇₄₄ /R ₇₆₂ . The fitting accuracy of the new parameter in the range of 0° to +30° in the forward direction was 0.704. RPMI could not only improve the monitoring accuracy of powdery mildew severity at a single angle but also achieve a more stable monitoring accuracy in the 0° to +30° range in the forward direction, which significantly expands the application scope of remote sensing monitoring and enhances the flexibility of the technology in actual production environments.

The aim of this study was to verify the applicability of the critical nitrogen concentration dilution curve (Nc) of wheat grown under different irrigation conditions in the field, and discuss the feasibility of using the N nutrition index (NNI) to optimize N fertilizer application. The high-yield, medium-protein wheat varieties Zhoumai 27 and Zhoumai 22 were used in field experiments in two different locations (Zhengzhou and Shangshui) in Huang-Huai, China. Plants were grown under rainfed and irrigation conditions, with five N application rates. Nc models of the leaves, stems, and whole plant were constructed, followed by establishment of an NNI model and accumulative N deficit model (N_and). As previous research reported, our results also showed that the critical N concentration and biomass formed a power function relationship (N = aDW^−b). When the biomass was the same, the critical N concentration was higher under irrigation than rainfed treatment. Meanwhile, the fitting accuracy (R²) of the Nc model was also higher under irrigation than rainfed treatment in both sites, and was higher in the stems and whole plant. The NNI calculated using the Nc model increased with increasing N application, reflecting N deficiency. Moreover, there was a significant negative linear correlation between NNI and N_and, and both indices could be uniformly modeled between locations and water treatments. The accuracy of the N_and model was highest in the whole plant, followed by the leaves and stems. The models constructed in this paper provide a theoretical basis for accurate management of N fertilizer application in wheat production.

Relationships between vegetation indices (VIs) and leaf area index (LAI) tend to saturate in the nadir direction, and vary with crop canopy structure and view zenith angles (VZAs). The objective of this paper was to improve the monitoring accuracy and angular stability of VIs for estimating LAI using multiangular remote sensing data. The relationship between LAI and ground-based hyperspectral spectral reflectance was quantified in winter wheat (Triticum aestivum L.) exhibiting erectophile and planophile growth habits. To further reduce the saturation, species specificity, and angular sensitivity, we developed a saturation factor (SF), based on near-infrared and green bands. Multiplying all VIs by SF greatly improved the association with LAI across all VZAs (R ² = 0.73-0.82). Most VI × SF values, particularly optimized soil-adjusted VI × SF, were able to construct universal algorithms across VZAs for accurate estimation of LAI due to the sensitivity of SF to LAI in a dense canopy, and the insensitivity of SF to view effects with larger VZAs. This approach is also promising for the exploitation of multiangular satellite data for the design and calibration of nonview-angle-corrected spectral reflectance, for which the sensor is only deployed at fixed observation direction.

Accurate agricultural yield estimates are vital for effective food resource management, and the efficacy of yield remote-sensing models is influenced by factors such as geographic location and temporal variations, posing challenges to both the accuracy and transferability. This study used multispectral (MS), thermal infrared (TIR), red-green-blue (RGB) sensors, and LiDAR, alongside soil and precipitation data, over three locations for two years, resulting in six datasets. Three models, stacking ensemble learning (SEL), long short-term memory (LSTM), and LSTM-multi-head self-attention (LSTM-MH-SA), were compared for wheat yield estimation. The flowering stage showed the highest accuracy with SEL ( $R^{2} = 0.83$ ), but the LSTM-MH-SA model outperformed SEL when integrating multitemporal features. As more data types were introduced, precision improved. Specifically, the LSTM-MH-SA model using vegetation indices, texture features (TFs), geographical information, and climate data showed superior accuracy, enhancing $R^{2}$ by 29.58% and reducing errors compared to a basic vegetation indices model. Incorporating the joint distribution adaptation (JDA) method enabled model transfer between varying datasets, maintaining a stable accuracy ( $R^{2} = 0.81$ ) with just 4% target dataset supplementation. Combining deep learning (DL) with transfer learning provides an innovative method for more efficient agricultural yield prediction.