In recent years, the rapid development of unmanned aerial vehicle (UAV) remote sensing technology has provided a new means to efficiently monitor forest resources and effectively prevent and control pests and diseases. This study aims to develop a detection model to study the damage caused to Moso bamboo forests by Pantana phyllostachysae Chao (PPC), a major leaf-eating pest, at 5 cm resolution. Damage sensitive features were extracted from multispectral images acquired by UAVs and used to train detection models based on support vector machines (SVM), random forests (RF), and extreme gradient boosting tree (XGBoost) machine learning algorithms. The overall detection accuracy (OA) and Kappa coefficient of SVM, RF, and XGBoost were 81.95%, 0.733, 85.71%, 0.805, and 86.47%, 0.811, respectively. Meanwhile, the detection accuracies of SVM, RF, and XGBoost were 78.26%, 76.19%, and 80.95% for healthy, 75.00%, 83.87%, and 79.17% for mild damage, 83.33%, 86.49%, and 85.00% for moderate damage, and 82.5%, 90.91%, and 93.75% for severe damage Moso bamboo, respectively. Overall, XGBoost exhibited the best detection performance, followed by RF and SVM. Thus, the study findings provide a technical reference for the regional monitoring and control of PPC in Moso bamboo.

Effectively monitoring Pantana phyllostachysae Chao (PPC) is essential for the sustainable development of the bamboo industry. However, the morphological similarity between damaged and off-year bamboo imposes challenges in the monitoring. The knowledge on whether the severity of this pest could be effectively monitored by using remote sensing methods is very limited. To fill this gap, this study aimed to identify the PPC damage of moso bamboo leaves using hyperspectral data. Specifically, we investigated differences in relative chlorophyll content (RCC), leaf water content (LWC), leaf nitrogen content (LNC), and hyperspectral spectrum among healthy, damaged (mildly damage, moderately damage, severely damage), and off-year bamboo leaves. Then, the hyperspectral indices sensitive to pest damage were selected by recursive feature elimination (RFE). The PPC damage identification model was constructed using the light gradient boosting machine (LightGBM) algorithm. We designed two different scenarios, without (A) and with (B) off-year samples, to evaluate the impact of off-year leaves on identification results. The RCC, the LWC, and the LNC of damaged leaves generally showed clear declined trends with the deterioration of damaged severity. The RCC and the LNC of off-year leaves were significantly lower than those of healthy and damaged leaves, whereas the LWC of off-leaves was significantly different from that of damaged leaves. The pest infestation caused noticeable distortion of leaf spectrum, increases in red and shortwave infrared bands, and decreases in green and near-infrared bands. The magnitude of reflectance change increased with the pest severity. The reflectance of off-year leaves in visible and near-infrared regions was distinguishably higher than that of healthy and damaged leaves. The overall accuracy (OA) of the constructed model for the identification of leaves with different degrees of damage severity reached 81.51%. When off-year, healthy, and damaged leaves were lumped together, the OA of the constructed model decreased by 5%. About half of the off-year leaf samples were misclassified into the damaged group. The identification of off-year leaves is a challenge for monitoring PPC damage using hyperspectral data. These results can provide practical guidance for monitoring PPC using remote sensing methods.

Leaf area index (LAI) and chlorophyll content are crucial variables in photosynthesis, respiration, and transpiration, playing a vital role in monitoring vegetation stress, estimating productivity, and evaluating carbon cycling processes. Currently, physical models are widely adopted for estimating LAI and canopy chlorophyll content (CCC). However, the main challenges of physical model-based methods for estimating LAI and CCC are the high computational cost and the fact that different combinations of canopy variables result in similar spectral reflectance for local minima. To address this limitation, a hybrid model was proposed to invert the LAI and CCC in Moso bamboo (Phyllostachys pubescens) forests. This approach utilized the PROSAIL canopy radiation transfer model, established look-up table (LUT) for LAI and CCC, and employed the Stacking ensemble learning framework. Compared to the PROSAIL LUT method, the hybrid model demonstrated higher performance in predicting LAI and CCC by incorporating the strengths of different models within the hybrid framework. The R2 values between predicted and measured values were improved by 3.28% and 7.15%, while the RMSE values were reduced by 19.71% and 16.14%, respectively. Moreover, the hybrid model based on Stacking ensemble learning achieved an 86% reduction in running time. Therefore, the hybrid model, which integrates the PROSAIL model with the Stacking ensemble learning framework, offers a more efficient and accurate approach for remotely estimating the LAI and CCC in Moso bamboo forests. The high efficiency of this method makes it promising and suitable for application to other types of vegetation.

Pantana phyllostachysae Chao (PPC) is one of the deadliest defoliators of Moso bamboo. Accurately locating and evaluating PPC damage is essential for the management of bamboo forests. Moso bamboo has a unique biennial growth cycle, consisting of the on-year period (bamboo shoots are incubated and then produced) and the off-year period (old leaves are dropped and then new leaves are grown, and no bamboo shoots are produced in the coming year). The similar physiological characteristics of off-year bamboo and damaged on-year bamboo create difficulties in monitoring PPC damage using remote sensing data. In this study, we synergistically used Sentinel-1, Sentinel-2, and field inventory data to construct machine learning (extreme gradient boosting, XGBoost) models monitoring PPC damage. The results show that the single-time observation feature-based model (using images from October) outperformed the double-time observation feature-based model (using the differences between remote sensing signals from October and February or April) due to the interference from other disturbance agents (e.g., logging and weeding). The overall accuracy (OA) values of the single-time observation feature-based model were at least 3% and 10% higher than those for double-time observation feature-based models for on- and off-year samples, respectively. With the consideration of the on- and off-year phenological differences, OA was improved by over 4%. The model without differentiation of the phenological difference tended to underestimate the damaged area of on-year bamboo and overestimate that of off-year bamboo. We also found that the responses of optical and SAR (synthetic aperture radar) features to PPC damage were different. The optical features increased or decreased with increasing damage severity. SAR features decreased significantly at the initial stage of PPC damage and then changed marginally with the increase in damage severity. The addition of SAR features to optical features improved the model performance, mainly for healthy and mildly damaged samples. The methodology developed in this study provides technical and theoretical support for the pest monitoring of bamboo forests using remote sensing data.

Leaf water content (LWC) is very important in the growth of vegetation. LWC and leaf spectra change when the leaves are under pest stress; exploring the change mechanism between LWC, leaf spectra, and pest stress can lay the foundation for pest detection. In this study, we measured the LWC and leaf spectra of moso bamboo leaves under different damage levels, used the Pearson–Lasso method to screen the features, and established a multiple linear regression (MLR) and random forest regression (RFR) model to estimate the LWC. We analyzed the relationship between LWC and spectral features of moso bamboo leaves under Pantana phyllostachysae Chao (PPC) stress and their changes. The results showed that: (1) the LWC showed a decreasing trend as the pest level increased. (2) The spectra changed substantially when the leaves were under pest stress. (3) The number and significance of response features associated with the LWC were diverse under different damage levels. (4) The estimation of LWC under different damage levels differed significantly. LWC, leaf spectra, response features, and the model estimation effect were diverse under different damage levels. The correlation between LWC and features was higher for healthy leaves than for damaged and off-year leaves. The two models were more effective in estimating the LWC of healthy leaves but less effective for damaged and off-year leaves. This study provides theoretical support for the prediction of PPC stress and lays the foundation for remote sensing monitoring.

Nitrogen is an important indicator of vegetation health, but the relationship between changes in the leaf nitrogen content of Moso bamboo leaves under Pantana phyllostachysae Chao (PPC) stress and leaf spectra remains unclear. We analyzed the relationship between the leaf nitrogen content and leaf spectra of Moso bamboo leaves under PPC stress to investigate whether the relationship could be used to detect pests and prevent their spread. We measured the nitrogen content and leaf spectra of Moso bamboo leaves under different damage levels, identified spectral indicators that were correlated with leaf nitrogen content (by removing the envelope and first-order differentiation of the raw spectra), and estimated leaf nitrogen content from the spectral data using regression models. Leaf nitrogen content decreased with increasing pest damage, and the leaf spectral curves changed, with the “green peak” and “red valley” in the visible range disappearing and the slope of the spectral curve decreasing. The wavelength region with the strongest correlation between the nitrogen content and spectral characteristics changed significantly with increasing pest damage, and the correlation in the red-edge region gradually decreased. The fits of nitrogen-content estimation models tended to decrease and then increase with increasing pest damage and were worst among leaves in the moderate damage state (Mo). A disordered relationship between nitrogen content and spectral characteristics indicated possible PPC damage. The degree of disorder was greatest in the Mo state. This study provides theoretical support for remote sensing monitoring of PPC hazards.