Mapping Tree Species Research Articles

Mapping dominant tree species of miombo woodlands in Western Tanzania using PlanetScope imagery

Mapping dominant tree species in miombo woodlands is essential for enhancing their monitoring and management. We evaluated PlanetScope imagery to map Julbernardia globiflora, Brachystegia spiciformis, and Pterocarpus tinctorius in Tongwe Forest Reserve, Tanzania. The study assessed the effectiveness of PlanetScope bands in discriminating tree species and investigated how different months/seasons influenced tree species classification. Optimal months (seasons) and spectral bands were selected using Principal Component loading, temporal pattern analysis, mean decrease in accuracy, and mean decrease Gini techniques. Random forest classification was employed for tree species classification, and accuracy was assessed using an error matrix. The study identified March, July, and September as optimal months for acquiring imagery, with effective bands including blue, green-1, green, yellow, red, and red-edge. July and September imagery in the dry season achieved higher overall accuracies of 65% and 69%, respectively, while March imagery in the wet season reached 55%. The highest overall accuracy of 72% was achieved using images from different seasons. Producer’s accuracy was highest for Brachystegia spiciformis (79%) and Julbernardia globiflora (95%), whereas Pterocarpus tinctorius had lower accuracy (25%). User’s accuracy varied with 74% for Brachystegia spiciformis, 70% for Julbernardia globiflora, and 67% for Pterocarpus tinctorius. Mapping accuracy was notably high for Brachystegia spiciformis and Julbernardia globiflora, reflecting their high sample size (dominance) and distinct phenology. The yellow and red bands were particularly effective in distinguishing miombo tree species demonstrating PlanetScope’s capability. Future research should focus on scaling up PlanetScope’s application for broad miombo tree species mapping.

Characterizing long-term tree species dynamics in Canada’s forested ecosystems using annual time series remote sensing data

Mapping tree species and changes in species distribution over time enables monitoring of successional dynamics and linking of realized species distributions to regional, climatic, and disturbance processes. Such knowledge is valuable for informing forest management activities regarding the role of climate change and various types of forest disturbances on tree species distributions and successional processes. In this study, we used time-series Landsat imagery to produce annual maps of dominant tree species (n = 37 tree species classes) from 1984 to 2022 at a 30-m spatial resolution for the 650 Mha of Canada’s forested ecosystems. The classification approach was based on spectral, geographic, climatic, and topographic descriptive metrics and was independently calibrated and validated with data from Canada’s National Forest Inventory. Preliminary results were informed by disturbance events and post-processed to ensure consistency of the annual species maps. Assessment of the resulting annual species maps using independent validation data resulted in an overall accuracy of 86.1 % ± 0.14 % (95 %-confidence interval). Over the study period, Canada’s treed area increased from 335 to 359 Mha. The area dominated by conifers increased in absolute terms but decreased in relative terms (from 86 % to 84.9 %), especially in western ecozones where wildfire is the main agent of disturbance. The area dominated by black spruce (Picea mariana), the most common tree species in Canada, remained stable, but its relative area decreased by 4.6 %. The area dominated by balsam fir (Abies balsamea, 1.5 %), trembling aspen (Populus tremuloides, 0.9 %), jack pine (0.5 %), and sugar maple (Acer saccharum, 0.4 %) increased. In burned areas, the prevalence of black spruce (74.7 %) and jack pine (Pinus banksiana, 14.6 %) was greater than their prevalence in Canada’s forested ecosystems overall (58.7 % and 3.8 %, respectively), while trembling aspen (2.4 % vs. 10 %) and subalpine fir (1.1 % vs. 4.6 %; Abies lasiocarpa) were underrepresented in burned areas. The area dominated by black spruce was underrepresented in harvested areas (46.8 % vs. 58.7 %) whereas balsam fir (6.9 % vs. 3 %), Engelmann spruce (5.5 % vs. 2 %; Picea engelmannii), red spruce (1.8 % vs. 0.4 %; Picea rubens), lodgepole pine (11.2 % vs. 5.6 %; Pinus contorta), Douglas-fir (2.7 % vs. 1.3 %; Pseudotsuga menziesii), and western hemlock (4.8 % vs. 2 %; Tsuga heterophylla) were found to be overrepresented. Both post-fire and post-harvest dynamics indicated a general trend of regeneration to the same pre-disturbance tree species, with post-harvest landscapes exhibiting a greater variety of different dominant tree species. These results provide valuable information for sustainable forest management and can inform conservation strategies by helping to better understand the short- and long-term effects of forest disturbances on species presence and distribution.

Mapping forest tree species and its biodiversity using EnMAP hyperspectral data along with Sentinel-2 temporal data: An approach of tree species classification and diversity indices

Forests play a crucial role in maintaining ecological balance and biodiversity, making the accurate mapping of tree species and assessment of biodiversity indices essential for informed management decisions. This study introduces an innovative methodology that integrates EnMAP (Environmental Mapping and Analysis Program) hyperspectral data with Sentinel-2 multitemporal data to classify tree species in the biodiverse landscapes of Kampinos National Park and its surrounding regions in Poland.We extract essential vegetation indices such as NDVI, NDMI, SAVI, and EVI from Sentinel-2 data to assess forest health and dynamics. The Sentinel-2 data is upscaled from 10 m to 30 m to align with EnMAP’s spatial resolution, followed by precise co-registration of the images using QGIS. Utilizing a rich dataset from the National Forest Inventory, we extract spectral signatures of nine distinct tree species from both data sources. We employ five machine learning algorithms—Support Vector Machines (SVM), Random Forest (RF), CatBoost (CAT), Gradient Boosting Classifier (GBC), and XGBoost (XGB)—to enhance classification accuracy.Through iterative experimentation with data reduction techniques and algorithm tuning, we achieve optimal performance across needle-leaved and broad-leaved species. The resulting tree species maps are validated through quantitative accuracy assessments against mixed-species polygons from the National Forest Inventory and ground truthing in the Kampinos National Park. Achieving an overall accuracy of 85% to 93%, our study demonstrates the efficacy of this integrated approach in tree species mapping. Furthermore, the tree species maps serve as a foundation for deriving key biodiversity indices—species richness, Shannon-Wiener Diversity Index, Simpson’s Diversity Index, and a composite Biodiversity Index—providing insights into spatial biodiversity patterns and informing targeted conservation strategies. This study exemplifies the potential of combining advanced remote sensing techniques with field validation to enhance our understanding of forest ecosystems and guide sustainable management practices.

Evaluating Tree Species Mapping: Probability Sampling Validation of Pure and Mixed Species Classes Using Convolutional Neural Networks and Sentinel-2 Time Series

Evaluating Tree Species Mapping: Probability Sampling Validation of Pure and Mixed Species Classes Using Convolutional Neural Networks and Sentinel-2 Time Series

Mapping tree species of wetlands using multispectral images of UAVs and machine learning: A case study of the Dong Rui Commune

Wetlands provide resources, regulate the environment, and stabilize shorelines; however, they are among the most vulnerable ecosystems in the world. The classification of mangrove species allows the determination of the habitat of each species, thereby serving as a basis for determining protection solutions and planning plans for mangrove conservation and restoration according to each environmental condition. We used Phantom 4 multispectral unmanned aerial vehicles (UAVs) to collect data from wetland areas in the Dong Rui Commune, which is one of the most diverse and valuable wetland ecosystems in northern Vietnam. A tree-species classification map was constructed through a combination of the object-based image analysis method and spectral reflectance values of each plant species, and the characteristic distributions of mangrove plants, including Bruguiera gymnorrhiza, Rhizophora stylosa, and Kandelia obovata, were determined with an overall accuracy of 91.11 % and a kappa coefficient (K) of 0.87. The overall accuracy for Rhizophora stylosa was the highest (94.23 %), followed by Kandelia obovata (93.61 %) and Bruguiera gymnorrhiza (85.50 %). An experiment was conducted to map plant taxonomy in the same area based only on a graph of spectral reflectance values at five single-spectral bands, and normalized difference vegetation index values were constructed, resulting in an overall accuracy of 78.22 % and a K of 0.67. The constructed map is useful for classifying, monitoring, and evaluating the structure of each group of mangroves, thereby serving as a basis for determining the distribution of each mangrove species according to natural conditions and contributing to the formulation of policies for afforestation and mangrove conservation in Dong Rui commune.

Remote Sensing Classification and Mapping of Forest Dominant Tree Species in the Three Gorges Reservoir Area of China Based on Sample Migration and Machine Learning

Remote Sensing Classification and Mapping of Forest Dominant Tree Species in the Three Gorges Reservoir Area of China Based on Sample Migration and Machine Learning

Data homogeneity impact in tree species classification based on Sentinel-2 multitemporal data case study in central Sweden

ABSTRACT Spatial information on forest composition is invaluable for achieving scientific, ecological, and management objectives and for monitoring multiple changes in forest ecosystems. The increased flow of optical satellite data provides new opportunities to improve tree species mapping. However, the accuracy of such maps is affected by training data, and in particular on the homogeneity of individual classes. Thus, we evaluated the effect of data homogeneity in tree species classification. We performed tree species classification by considering different ways to partition data into tree species classes. The class sets considered were (i) only mixed coniferous and mixed deciduous forest classes, (ii) single-species classes, (iii) single-species, mixed coniferous and mixed deciduous classes, and (iv) single-species, mixed coniferous and mixed deciduous classes and a true mixed class. Using data from the Swedish National Forest Inventory, we varied the threshold that defined dominating species. Tree species were classified for a study area in central Sweden using Sentinel-2 data and two classification approaches: Bayesian inference and random forest (RF). Images were selected by class separability and the most informative images based on variable selection with RF. The most informative images tended to be selected by both methods. However, in forests with tree species of similar spectral behaviour, image selection on the basis of class separability was found to be more reliable. More accurate classification results were achieved as the number of classes decreased and the threshold of plot purity increased. The Bayesian classification approach of only mixed coniferous and mixed deciduous classes gave the highest OA, always greater than 90%. When discriminating between pure plots of Birch (Betula spp.), Spruce (Picea abies), Scots pine (Pinus sylvestris) and Lodgepole pine (Pinus contorta), the best OA values were 84% for Bayesian and 80% for RF. In more complicated scenarios, RF resulted in higher overall accuracies (OA).

Fusion of multi-temporal PlanetScope data and very high-resolution aerial imagery for urban tree species mapping

Detailed assessment of the ecosystem services provided by urban green spaces requires data on urban tree species. While many approaches for mapping of urban trees from remotely sensed data have been proposed, the fusion of multi-temporal satellite imagery with very high resolution orthophotos remains relatively underexplored. In this research, we assess the potential of a multimodal deep learning approach with intermediate data fusion for classifying common tree species found in the Brussels Capital Region. Our method combines two image sources: (a) multi-temporal PlanetScope data and (b) high-resolution aerial imagery. To evaluate the contribution of each image source, we separately train and assess each branch of the network. Both image sources demonstrate the ability to predict prevalent tree species with high accuracy. However, the fusion of the two image sources yields the best results, achieving an overall accuracy of 0.88 for the five most common tree species in the region. Our approach is compared to two conventional machine learning methods: a random forest (RF) and a support vector machine classifier (SVM) and outperforms both with a 11 percentage point increase in overall accuracy over RF and a 10 percentage point increase over SVM. Increasing the number of species from five to thirteen, including all species with more than 500 tree samples, results in a marginal decrease in accuracy (from 0.88 to 0.84). Overall, our deep learning approach demonstrates its efficacy in classifying common tree species in urban settings and provides a foundation for a comprehensive quantification of ecosystem services offered by urban trees through remote sensing data.

High-resolution mapping of tree species and associated uncertainty by combining aerial remote sensing data and convolutional neural networks ensemble

Mapping tree species diversity is essential for monitoring and managing forest ecosystems. Automating ecoforestry mapping using remote sensing images remains an important challenge due to the tremendous variability in forest covers and the conditions under which images used for classification are acquired. Deep learning algorithms have been increasingly used over the last few years to analyse remote sensing data for mapping tree species. However, most of these studies focus only on a small number of species or a limited area and avoid providing a spatially explicit representation of the uncertainty related to the predictions, rendering them unsuitable for operational use.In this study, we used an ensemble of convolutional neural networks to map forest species and land cover types across a 10,000 km2 area in Quebec, Canada, spanning mixed and boreal forests. We built a georeferenced label database to train and test nine models, which resulted from the combination of three training datasets and three-commonly used convolutional neural network architectures (VGG16, ResNet50v2, Densenet121). These models were trained on multiband aerial photographs and a canopy height model derived from airborne lidar point clouds and used to map the diversity and distribution of tree species and land cover types. The level of agreement among models was used to generate uncertainty maps. The performance of the super-ensemble using 1311 independent forest inventory plots and assessed the extent to which uncertainty maps could serve as a spatially explicit indicator of model performance.The super-ensemble achieved 90% global accuracy. Our results indicated that the performance of the super-ensemble surpassed that of all individual architectures while also showing a positive effect of the canopy height model on performance. The comparison of the super-ensemble map with the proportion of basal area measured in forest inventory plots confirmed the reliability of the model over the study area. Our results also indicated that mapping the inter-model agreement provides a reliable spatially explicit estimate of model performance. The robustness and reliability of the proposed approach support its use in an operational context while also providing a conceptual framework to evaluate the reliability of uncertainty maps.

Comparison of Deep and Machine Learning Approaches for Quebec Tree Species Classification Using a Combination of Multispectral and LiDAR Data

Accessing tree species information is required for making appropriate decisions in forest management. 3D photo-interpretation using high spatial resolution aerial imagery is used to provide information on tree species in the province of Quebec. However, the shortage of qualified interpreters and the increasing costs of 3D photo-interpretation have affected the production of the forest inventory. In this study, we employed deep and machine learning models to classify nine tree species (i.e., paper birch, yellow birch, red maple, poplar, black spruce, white spruce, tamarack, jack pine, and balsam fir). We used a combination of spectral and vertical structural information extracted from 30-cm aerial imagery and airborne light detection and ranging (LiDAR) data in a 9 , 100 km 2 forested area in Quebec, Canada. The results indicated that Dense Convolution Network (DenseNet) achieved the best overall accuracy of 78%, outperforming machine learning methods by 5%. In addition, the models’ performance was independently assessed using permanent and temporary ground sample plots, acknowledging the superiority of DenseNet in terms of overall accuracy (73%) in predicting the dominant species. Our results suggest that the combination of aerial imagery and airborne LiDAR data, using deep learning approaches, can be applied to accurately map tree species.

Multi-Sensor Classification Framework of Urban Vegetation for Improving Ecological Services Management

Recent climatic changes have profoundly impacted the urban microclimate, exposing city dwellers to harsh living conditions. One effective approach to mitigating these events involves incorporating more green infrastructure into the cityscape. The ecological services provided by urban vegetation play a crucial role in enhancing the sustainability and livability of cities. However, monitoring urban vegetation and accurately estimating its status pose challenges due to the heterogeneous nature of the urban environment. In response to this, the current study proposes utilizing a remote sensing-based classification framework to enhance data availability, thereby improving practices related to urban vegetation management. The aim of the current research is to explore the spatial pattern of vegetation and enhance the classification of tree species within diverse and complex urban environments. This study combines various remote sensing observations to enhance classification capabilities. High-resolution colored rectified aerial photographs, LiDAR-derived products, and hyperspectral data are merged and analyzed using advanced classifier methods, specifically partial least squares-discriminant analysis (PLS-DA) and object-based image analysis (OBIA). The OBIA method demonstrates an impressive overall accuracy of 95.30%, while the PLS-DA model excels with a remarkable overall accuracy of 100%. The findings validate the efficacy of incorporating OBIA, aerial photographs, LiDAR, and hyperspectral data in improving tree species classification and mapping within the context of PLS-DA. This classification framework holds significant potential for enhancing management practices and tools, thereby optimizing the ecological services provided by urban vegetation and fostering the development of sustainable cities.

A novel framework combining band selection algorithm and improved 3D prototypical network for tree species classification using airborne hyperspectral images

Fine-grained identification of forest types and tree species represents a critical aspect of forest resource inventory and monitoring. The use of airborne hyperspectral remote sensing imagery stands out for its ability to finely differentiate among tree species, leveraging its exceptional spatial resolution and rich spectral details. However, this approach is limited by several challenges (e.g., high spectral correlation and information redundancy). In accordance, the adoption of a lightweight deep learning approach in the form of a few-shot learning model can effectively resolve the challenges of multi-forest tree species classification. Therefore, integrating a data dimensionality reduction algorithm with a few-shot classification model presents a promising avenue for resolving the fine-grained classification of forest tree species. In this study, we propose the innovative classification framework FAST 3D-CNN P-Net. This framework utilizes CNN for band selection, enhances the fine-grained identification process in hyperspectral data, and integrates an optimized FAST 3D-CNN into the P-Net classifier (a few-shot classifier). First, a CNN-based band selection method is employed to learn the nonlinear dependencies between spectral bands, assign weights to rank the bands, and reconstruct the global spectral information using the most informative bands. It then constructs a novel classification model, designated FAST 3D-CNN P-Net, through the integration of an optimal 3D-CNN with a prototypical network. To enhance classification performance, the FAST 3D-CNN P-Net utilizes reconstructed hyperspectral images derived from the band selection results as input. The effectiveness of the proposed framework was assessed with the airborne GFF dataset and the widely accessible medium-resolution hyperspectral datasets, Indian Pines (IP) and Kennedy Space Center (KSC). The overall classification accuracy reached 98.33 % for the GFF dataset and 97.21 % and 99.43 % for the IP and KSC, respectively, exhibiting performance superiority compared to the standalone 3D-CNN classification network. This classification framework demonstrates efficiency in selecting a subset of hyperspectral bands with minimal redundancy, empowering the rapid and accurate classification and mapping of tree species in complicated, multi-species forest stands, even with a limited quantity of labeled samples.

Mapping tree species diversity in a typical natural secondary forest by combining multispectral and LiDAR data

Accurate monitoring of tree species diversity is crucial for understanding the dynamic changes in tree species diversity and its relationships with other services and functions in forest ecosystems. Traditional optical remote sensing data have been widely used for monitoring tree species diversity based on the spectral variation hypothesis (SVH). However, this method cannot capture the three-dimensional structural variations in complex species compositions under different stand conditions. In this study, we modeled tree species diversity in terms of spectral variation and stand structural complexity in a typical natural secondary forest in Northeast China by combining Sentinel-2 data and UAV-borne light detection and ranging (LiDAR) point cloud data. First, species diversity indices (including the Shannon index H' and Simpson index D1) were derived from 60 field-measured plots. Second, recursive feature elimination (RFE) was utilized for feature filtering of ten spectral bands and four vegetation indices extracted from Sentinel-2 data and Rao's Q index, as well as eleven features extracted from LiDAR point clouds reflecting the complexity of the stand structure. Subsequently, the random forest method was utilized to fit and predict the relationship between the remote sensing feature set and tree species diversity. The results showed that the use of multisource remote sensing feature set to estimate tree species diversity had the highest accuracy (R2 = 0.44, RMSE = 0.28 for H') compared to the use of only one data source. Moreover, when using a single remote sensing feature set, the estimation accuracy of the optical remote sensing feature set is higher than that of the LiDAR feature set for H' and D1, and the NIRv is the most influential spectral feature. This study clarified the value of spectral variation and productivity heterogeneity embodied in optical remote sensing features for monitoring tree species diversity, as well as evaluating the shortcomings and possibilities of using LiDAR point cloud features independently, and fully confirmed the positive significance of complementary effects between multisource remote sensing feature sets.

Evaluating the Potential of Sentinel-2 Time Series Imagery and Machine Learning for Tree Species Classification in a Mountainous Forest

Accurate and reliable information on tree species composition and distribution is crucial in operational and sustainable forest management. Developing a high-precision tree species map based on time series satellite data is an effective and cost-efficient approach. However, we do not quantitatively know how the time scale of data acquisitions contributes to complex tree species mapping. This study aimed to produce a detailed tree species map in a typical forest zone of the Changbai Mountains by incorporating Sentinel-2 images, topography data, and machine learning algorithms. We focused on exploring the effects of the three-year time series of Sentinel-2 within monthly, seasonal, and yearly time scales on the classification of ten dominant tree species. A random forest (RF) and support vector machine (SVM) were compared and employed to map continuous tree species. The results showed that classification with monthly datasets (overall accuracy (OA): 83.38–87.45%) outperformed that with seasonal and yearly datasets (OA:72.38–85.91%), and the RF (OA: 81.70–87.45%) was better than the SVM (OA: 72.38–83.38%) at processing the same datasets. Short-wave infrared, the normalized vegetation index, and elevation were the most important variables for tree species classification. The highest classification accuracy of 87.45% was achieved by combining RF, monthly datasets, and topography information. In terms of single species’ accuracy, the F1 scores of the ten tree species ranged from 62.99% (Manchurian ash) to 97.04% (Mongolian Oak), and eight of them obtained high F1 scores greater than 87%. This study confirmed that monthly Sentinel-2 datasets, topography data, and machine learning algorithms have great potential for accurate tree species mapping in mountainous regions.

Getting to Know the Białowieża Forest stands: Comprehensive Mapping of Living and Dead Tree Species Using Multi-Temporal ALS and CIR Data

Getting to Know the Białowieża Forest stands: Comprehensive Mapping of Living and Dead Tree Species Using Multi-Temporal ALS and CIR Data

A Hybrid convolution neural network for the classification of tree species using hyperspectral imagery.

In recent years, the advancement of hyperspectral remote sensing technology has greatly enhanced the detailed mapping of tree species. Nevertheless, delving deep into the significance of hyperspectral remote sensing data features for tree species recognition remains a challenging endeavor. The method of Hybrid-CS was proposed to addresses this challenge by synergizing the strengths of both deep learning and traditional learning techniques. Initially, we extract comprehensive correlation structures and spectral features. Subsequently, a hybrid approach, combining correlation-based feature selection with an optimized recursive feature elimination algorithm, identifies the most valuable feature set. We leverage the Support Vector Machine algorithm to evaluate feature importance and perform classification. Through rigorous experimentation, we evaluate the robustness of hyperspectral image-derived features and compare our method with other state-of-the-art classification methods. The results demonstrate: (1) Superior classification accuracy compared to traditional machine learning methods (e.g., SVM, RF) and advanced deep learning approaches on the tree species dataset. (2) Enhanced classification accuracy achieved by incorporating SVM and CNN information, particularly with the integration of attention mechanisms into the network architecture. Additionally, the classification performance of a two-branch network surpasses that of a single-branch network. (3) Consistent high accuracy across different proportions of training samples, indicating the stability and robustness of the method. This study underscores the potential of hyperspectral images and our proposed methodology for achieving precise tree species classification, thus holding significant promise for applications in forest resource management and monitoring.

Impact of GEDI-Derived Forest Vertical Structure Characteristics on the Accuracy Gains in Regional Dominant Tree Species Mapping

Impact of GEDI-Derived Forest Vertical Structure Characteristics on the Accuracy Gains in Regional Dominant Tree Species Mapping

Merging multiple sensing platforms and deep learning empowers individual tree mapping and species detection at the city scale

The precise estimation of the number of trees, their individual tree locations, along with species information, is crucial for enhancing ecosystem services in urban areas. Previous studies largely used satellite or airborne images for mapping trees, but they were insufficient to generate a large number of species distributions. Ground-level data also led to inaccurate positional information of individual trees, although they provided reliable species detection results. In this study, we propose a novel framework that fully explores the complementary strengths of air- and ground-level sensing platforms by leveraging various deep neural networks to generate a detailed tree maps at a city-wide scale. Our strategy includes individual tree mapping and tree species detection using three-color channel images acquired from multiple sensing platforms. Through publicly available airborne imagery, we estimate the presence of over 1.2 million trees in Suwon city of South Korea spanning 121.04 km2 (R2 of 0.95 and relative bias of −1.9 %), achieving more accurate individual tree positions (positional uncertainty around 2.0 m) than conventional methods. Our comprehensive experiments also demonstrate the effectiveness of utilizing tree bark photos and street-level imagery taken by citizens and vehicles to identify urban tree species, with accuracy rates of over 80 % for citizen-sensed tree species maps and 66 % for vehicle-sensed tree species map. Along with the proliferation of web-based airborne images, the widespread use of smartphones, and the advancements in vehicle-mounted sensors, this study can facilitate efficient and accurate management of urban trees across scales. For reproducibility of the study, we share the source code and datasets at https://github.com/landkwon94/ryoungseob-master.

Local-scale mapping of tree species in a lower mountain area using Sentinel-1 and -2 multitemporal images, vegetation indices, and topographic information

IntroductionMapping tree species is an important activity that provides the information necessary for sustainable forest management. Remote sensing is a effective tool that offers data at different spatial and spectral resolutions over large areas. Free and open acces Sentinel satellite imagery and Google Earth Engine, which is a powerful cloud computing platform, can be used together to map tree species.MethodsIn this study we mapped tree species at a local scale using recent Sentinel-1 (S-1) and Sentinel-2 (S-2) time-series imagery, various vegetation indices (Normalized Difference Vegetation Index - NDVI, Enhanced Vegetation Index - EVI, Green Leaf Index - GLI, and Green Normalized Difference Vegetation Index - GNDVI) and topographic features (elevation, aspect and slope). Five sets of data were used, in different combinations, together with the Random Forest classifier in order to determine seven tree species (spruce, beech, larch, fir, pine, mixed, and other broadleaves [BLs]) in the studied area.Results and discussionDataset 1 was a combination of S-2 images (bands 2, 3, 4, 5, 6, 7, 8, 8a, 11 and 12), for which an overall accuracy of 76.74% was obtained. Dataset 2 comprised S-2 images and vegetation indices, leading to an overall accuracy of 78.24%. Dataset 3 included S-2 images and topographic features, which lead to an overall accuracy of 89.51%. Dataset 4 included S-2 images, vegetation indices, and topographic features, that have determined an overall accuracy of 89.36%. Dataset 5 was composed of S-2 images, S-1 images (VV and VH polarization), vegetation indices, and topographic features that lead to an overall accuracy of 89.68%. Among the five sets of data, Dataset 3 produced the most significant increase in accuracy, of 12.77%, compared to Dataset 1. Including the vegetation indices with the S-2 images (Dataset 2) gave an accuracy increase of only 1.50%. By combining the S-1 and S-2 images, vegetation indices and topographic features (Dataset 5) there was an accuracy increase of only 0.17%, compared with the S-2 images plus topographic features combination (Dataset 3). However, the input brought by the S-1 images was apparent in the increase in classification accuracy for the mixed and other BL species that were mostly found in hilly locations. Our findings confirm the potential of S-2 images, used together with other variables, for classifying tree species at the local scale.

A spectral-temporal constrained deep learning method for tree species mapping of plantation forests using time series Sentinel-2 imagery

Plantation forests provide critical ecosystem services and have experienced worldwide expansion during the past few decades. Accurate mapping of tree species through remote sensing is critical for managing plantation forests. The typical temporal behaviors and traits of tree species in satellite image time series (SITS) generate temporal and spectral features in multiple phenological stages that are critical to improve tree species mapping. However, the diverse input features, sequential relations and complex structures in SITS drastically increase the dimension and difficulty of spectral-temporal feature extraction, which challenges the capacity of many general classifiers not explicitly adapted for spectral-temporal learning. As a result, there is still a lack of a method that could automatically extract spectral-temporal features with high separability and regional adaptability from high-dimensional SITS for tree species mapping of plantation forests. Moreover, the effects of varying temporal resolution and feature combination on the plantation tree species mapping are under-explored. Here, we developed a multi-head attention-based method for automatically extracting spectral-temporal features with high separability based on a modified Transformer network (Transformer4SITS) for improved plantation tree species mapping. The end-to-end network model consists of a feature extraction module to learn deep spectral-temporal features from SITS and a fusion module to combine multiple features for improving mapping accuracy. We applied this method to two representative plantation forests in southern and northern China for tree species mapping. The results show that: (1) Transformer4SITS method could self-adaptively extract typical spectral-temporal features of key phenological stages (e.g., greenness rising and falling) from SITS, and achieved significantly improved accuracies by at most 15% in comparison with all four baseline methods (i.e., long short-term memory, harmonic analysis, time-weighted dynamic time warping, linear discriminant analysis); (2) time series with higher temporal resolution tended to produce more accurate species maps consistently across two sites, with their overall accuracies (OA) respectively increasing from 91.05% and 84.33% (60-day) to 94.88% and 88.72% (5-day), but the effect of high temporal resolution respectively leveled off around 90-day and 50-day resolution across two sites; (3) the mapping results using all available bands and two-band spectral indices outperformed the results using a subset of them, but with only modest increase in the accuracy (i.e., OA increased from 93.63% and 86.01% to 94.88% and 88.72%. This study thus provides a state-of-the-art deep learning-based method for improved tree species mapping, which is critical for sustainable management and biodiversity monitoring of plantation forests across large scales.