Detecting and mapping individual trees accurately and automatically from remote sensing images is of great significance for precision forest management. Many algorithms, including classical methods and deep learning techniques, have been developed and applied for tree crown detection from remote sensing images. However, few studies have evaluated the accuracy of different individual tree detection (ITD) algorithms and their data and processing requirements. This study explored the accuracy of ITD using local maxima (LM) algorithm, marker-controlled watershed segmentation (MCWS), and Mask Region-based Convolutional Neural Networks (Mask R-CNN) in a young plantation forest with different test images. Manually delineated tree crowns from UAV imagery were used for accuracy assessment of the three methods, followed by an evaluation of the data processing and application requirements for three methods to detect individual trees. Overall, Mask R-CNN can best use the information in multi-band input images for detecting individual trees. The results showed that the Mask R-CNN model with the multi-band combination produced higher accuracy than the model with a single-band image, and the RGB band combination achieved the highest accuracy for ITD (F1 score = 94.68%). Moreover, the Mask R-CNN models with multi-band images are capable of providing higher accuracies for ITD than the LM and MCWS algorithms. The LM algorithm and MCWS algorithm also achieved promising accuracies for ITD when the canopy height model (CHM) was used as the test image (F1 score = 87.86% for LM algorithm, F1 score = 85.92% for MCWS algorithm). The LM and MCWS algorithms are easy to use and lower computer computational requirements, but they are unable to identify tree species and are limited by algorithm parameters, which need to be adjusted for each classification. It is highlighted that the application of deep learning with its end-to-end-learning approach is very efficient and capable of deriving the information from multi-layer images, but an additional training set is needed for model training, robust computer resources are required, and a large number of accurate training samples are necessary. This study provides valuable information for forestry practitioners to select an optimal approach for detecting individual trees.

Island ecosystems are particularly susceptible to climate change and human activities. The change of land use and land cover (LULC) has considerable impacts on island ecosystems, and there is a critical need for a free and open-source tool for detecting land cover fluctuations and spatial distribution. This study used Google Earth Engine (GEE) to explore land cover classification and the spatial pattern of major land cover change from 1990 to 2019 on Haitan Island, China. The land cover classification was performed using multiple spectral bands (RGB, NIR, SWIR), vegetation indices (NDVI, NDBI, MNDWI), and tasseled cap transformation of Landsat images based on the random forest supervised algorithm. The major land cover conversion processes (transfer to and from) between 1990 and 2019 were analyzed in detail for the years of 1990, 2000, 2007, and 2019, and the overall accuracies ranged from 88.43% to 91.08%, while the Kappa coefficients varied from 0.86 to 0.90. During 1990–2019, other land, cultivated land, sandy land, and water area decreased by 30.70%, 13.63%, 3.76%, and 0.95%, respectively, while forest and built-up land increased by 30.94% and 16.20% of the study area, respectively. The predominant land cover was other land (34.49%) and cultivated land (26.80%) in 1990, which transitioned to forest land (53.57%) and built-up land (23.07%) in 2019. Reforestation, cultivated land reduction, and built-up land expansion were the major land cover change processes on Haitan Island. The spatial pattern of forest, cultivated land, and built-up land change is mainly explained by the implementation of a ‘Grain for Green Project’ and ‘Comprehensive Pilot Zone’ policy on Haitan Island. Policy and human activities are the major drivers for land use change, including reforestation, population growth, and economic development. This study is unique because it demonstrates the use of GEE for continuous monitoring of the impact of reforestation efforts and urbanization in an island environment.

Background: Urban green space (UGS) has been shown to play an important role in mitigating urban heat island (UHI) effects. In the context of accelerating urbanization, a better understanding of the landscape pattern mechanisms affecting the thermal environment is important for the improvement of the urban ecological environment. Methods: In this study, the relationship between land surface temperature (LST) and the spatial patterns of green space was analyzed using a bivariate spatial autocorrelation and spatial autoregression model in three seasons (summer, transition season (spring), and winter) with different grid scales in Fuzhou city. Results: Our results indicated that the LST in Fuzhou City has a significant spatial autocorrelation. The percentage of landscape and patch density area were negatively correlated with surface temperature. The results of our indicators differed according to the season, with population density and distance to the water indicators not being significant in the winter. The coefficient of determination was higher at the 510 m grid scale on this study’s scale. Conclusion: This study extends our understanding on the influence of UHI effects after accounting for different seasonal and spatial scale factors. It also provides a reference for urban planners to mitigate heat islands in the future.

Island ecosystems are susceptible to the considerable impacts of increasing human activities, landscape reconstruction, and urban expansion, resulting in changes in the ecological environment and urban ecological security issues. Remote sensing techniques can achieve the near-real-time ecological environment monitoring of these rapidly changing areas. The remote sensing-based ecological index (RSEI), as a comprehensive remote sensing ecological environment index, was adopted to dynamically monitor urban ecological quality (EQ) over time in this study, combined with the Landsat-based detection of trends in disturbance and recovery (LandTrendr) algorithm. Annual composite images were generated using Landsat 5, Landsat 7, and Landsat 8 imagery to extract four metrics (Greenness, Moisture, Heat, and Dryness) to calculate RSEI from 1991 to 2021. The ecological quality in the study area was evaluated using a five-level classification (poor, inferior, medium, good, and excellent), and the changes in EQ on a pixel basis were identified by the LandTrendr algorithm. The results showed that (1) the average value of the RSEI ranged from 0.47 to 0.57 over 31 years, indicating that EQ was maintained at the medium level; (2) the distribution of different EQ levels had visible patterns, and an area of 47.87 km² was affected by a poor EQ at least once in 31 years; (3) 38.22 km² of this area experienced EQ poor disturbance once, and 3.05 km² of the area had poor disturbance twice. Urban expansion, forest degradation, and policy are the main factors causing the reduction of the RSEI. The results demonstrate that combining time series of RSEI and LandTrendr can effectively monitor the changes of EQ, which is helpful to identify the spatial–temporal variation patterns of EQ and provide valuable information for policymakers and protection.

Monitoring and assessing vegetation using deep learning approaches has shown promise in forestry applications. Sample labeling to represent forest complexity is the main limitation for deep learning approaches for remote sensing vegetation classification applications, and few studies have focused on the impact of sample labeling methods on model performance and model training efficiency. This study is the first-of-its-kind that uses Mask region-based convolutional neural networks (Mask R-CNN) to evaluate the influence of sample labeling methods (including sample size and sample distribution) on individual tree-crown detection and delineation. A flight was conducted over a plantation with Fokienia hodginsii as the main tree species using a Phantom4-Multispectral (P4M) to obtain UAV imagery, and a total of 2061 manually and accurately delineated tree crowns were used for training and validating (1689) and testing (372). First, the model performance of three pre-trained backbones (ResNet-34, ResNet-50, and ResNet-101) was evaluated. Second, random deleting and clumped deleting methods were used to repeatedly delete 10% from the original sample set to reduce the training and validation set, to simulate two different sample distributions (the random sample set and the clumped sample set). Both RGB image and Multi-band images derived from UAV flights were used to evaluate model performance. Each model’s average per-epoch training time was calculated to evaluate the model training efficiency. The results showed that ResNet-50 yielded a more robust network than ResNet-34 and ResNet-101 when the same parameters were used for Mask R-CNN. The sample size determined the influence of sample labeling methods on the model performance. Random sample labeling had lower requirements for sample size compared to clumped sample labeling, and unlabeled trees in random sample labeling had no impact on model training. Additionally, the model with clumped samples provides a shorter average per-epoch training time than the model with random samples. This study demonstrates that random sample labeling can greatly reduce the requirement of sample size, and it is not necessary to accurately label each sample in the image during the sample labeling process.