Forest Mapping Research Articles

Assessing the Performance of a Handheld Laser Scanning System for Individual Tree Mapping—A Mixed Forests Showcase in Spain

The use of mobile laser scanning to survey forest ecosystems is a promising, scalable technology to describe the 3D structure of forests at a high resolution. We use a structurally complex, mixed-species Mediterranean forest to test the performance of a mobile Handheld Laser Scanning (HLS) system to estimate tree attributes within a forest patch in central Spain. We describe the different stages of the HLS approach: field position, ground data collection, scanning path design, point cloud processing, alignment between detected trees and measured reference trees, and finally, the assessment of main tree structural attributes diameter at breast height (DBH) and tree height considering species and tree size as control factors. We surveyed 418 reference trees to account for omission and commission error rates over a 1 ha plot divided into 16 sections and scanned using two different scanning paths. The HLS-based approach reached a high of 88 and 92% tree detection rate for the best combination of scanning path and point cloud processing modes for the HLS system. The root mean squared errors for DBH estimates varied between species: errors for Pinus pinaster were below 2 cm for Scan 02. Quercus pyrenaica, and Alnus glutinosa showed higher error rates. We observed good agreement between ALS and HLS estimates for tree height, highlighting differences to field measurements. Despite the complexity of the mixed forest area surveyed, our results show that HLS is highly efficient at detecting tree locations, estimating DBH, and supporting tree height measurements as confirmed with airborne laser data used for validation. This study is one of the first HLS-based studies conducted in the Mediterranean mixed forest region, where variability in tree allometries and spacing and the presence of natural regeneration pose challenges for the HLS approach. HLS is a feasible, time-efficient, scalable technology for tree mapping in mixed forests with potential to support forest monitoring programmes such as national forest inventories lacking three-dimensional, remote sensing data to support field measurements.

APPLICATION OF REMOTE SENSING AND GEOGRAPHICAL INFORMATION SYSTEM FOR FOREST RESOURCES BALANCE MAPPINGIN THE TARUSAN SUB-DAS, PESISIR SELATAN DISTRICT

This forest resource balance mapping aims to (1) determine changes in the area of ​​forest resources in the XI Tarusan sub-district, Pesisir Selatan Regency; (2) find out the forest balance in XI Tarusan District, Pesisir Selatan District; (3) determine the accuracy of Landsat 8 imagery for preparing forest resource balances. The method used in mapping the balance of forest resources is by carrying out a supervised classification for land use and land cover. Then the data is overlaid using a two-dimensional matrix approach. Furthermore, the data is analyzed using a Geographic Information System which provides spatial information to be able to present data in map form. The results of this study are changes in the area of ​​forest resources in XI Tarusan District, Pesisir Selatan District, which is 28.62 Ha.,and the accuracy of the Landsat 8 image obtained in the preparation of this forest resource is 91.66%.

UTILIZING LANDSAT 8 IMAGERY FOR MAPPING OF BURNED AREAS USING THE NORMALIZE DIFFERENCE VEGETATION INDEX (NDVI) AND NORMALIZE BURN RATIO (NBR) METHODS

This study aims (1) to map changes in the area of ​​forest land in the western part of Bengkalis Regency in 2016 and 2021, (2) to determine the distribution of the area of ​​forest burned in the western part of Bengkalis Regency, (3) to determine the severity of forest fires in the District of Bengkalis West Bengkalis.This study used the NDVI (Normalized Difference Vegetation Index) method by Huete et. Al by compositing band 5 (NIR) and band 4 (Red) on Landsat 8 imagery which was processed using ArcGIS software before and after a forest fire. As well as the NBR (Normalized Burn Ratio) and dNBR (Difference Normalized Burn Ratio) methods by Eidenshink et al by compositing band 5 (NIR) and band 7 (SWIR) on Landsat 8 images processed using QGIS software. For sampling using random sampling method and accuracy test using overall accuracy, user's accuracy, producer's accuracy, and kappa analysis. The results of this study are (1) the area of ​​forest land in Bengkalis Regency continues to decrease every year, in 2016 the area of ​​forest land903,920 ha and 2021 the total forest area is463,441 ha. (2)The area of ​​forest land burned due to forest fires in Bengkalis Regency, which burned the least was 267.43 ha, while it was 1468.93 ha and the most extensive was 2186.53 ha.(3) Based on one forest fire distribution map, it is divided into 7 fire severity classes, namely high post-fire regrowth, low post-fire regrowth, no burning, low, medium-high and very high and the most dominant forest fire level is low-high.

Time Ring Data: Definition and Application in Spatio-Temporal Analysis of Urban Expansion and Forest Loss

Remote sensing can provide spatio-temporal continuous Earth observation data and is becoming the main data source for spatial and temporal analysis. Remote sensing data have been widely used in applications such as meteorological monitoring, forest investigation, environmental health, urban planning, and water conservancy. While long-time-series remote sensing data are used for spatio-temporal analysis, this analysis is usually limited because of the large data volumes and complex models used. This study intends to develop an innovative and simple approach to reveal the spatio-temporal characteristics of geographic features from the perspective of remote sensing data themselves. We defined an efficient remote sensing data structure, namely time ring (TR) data, to depict the spatio-temporal dynamics of two common geographic features. One is spatially expansive features. Taking nighttime light (NTL) as an example, we generated a NTL TR map to exhibit urban expansion with spatial and temporal information. The speed and acceleration maps of NTL TR data indicated extraordinary expansion in the last 10 years, especially in coastal cities and provincial capitals. Beijing, Tianjin, Hebei Province, Shandong Province, and Jiangsu Province exhibited fast acceleration of urbanization. The other is spatially contractive features. We took forest loss in the Amazon basin as an example and produced a forest cover TR map. The speed and acceleration were mapped in two 10-year periods (2000–2010 and 2010–2020) in order to observe the changes in Amazon forest cover. Then, combining cropland TR data, we determined the consistency of the spatio-temporal variations and used a linear regression model to detect the association between the acceleration of cropland and forest. The forest TR map showed that, spatially, there was an apparent phenomenon of forest loss occurring in the southern and eastern Amazon basin. Temporally, the speed of forest loss was more drastic between 2000 and 2010 than that in 2010–2020. In addition, the acceleration of forest loss showed a dispersed distribution, except for in Bolivia, which demonstrated a concentrated regional acceleration. The R-squared value of the linear regression between forest and cropland acceleration reached 0.75, indicating that forest loss was closely linked to the expansion of cropland. The TR data defined in this study not only optimized the use of remote sensing data, but also facilitated their application in spatio-temporal integrative analysis. More importantly, multi-field TR data could be jointly applied to explore the driving force at spatial and temporal scales.

TreeSatAI Benchmark Archive: a multi-sensor, multi-label dataset for tree species classification in remote sensing

Abstract. Airborne and spaceborne platforms are the primary data sources for large-scale forest mapping, but visual interpretation for individual species determination is labor-intensive. Hence, various studies focusing on forests have investigated the benefits of multiple sensors for automated tree species classification. However, transferable deep learning approaches for large-scale applications are still lacking. This gap motivated us to create a novel dataset for tree species classification in central Europe based on multi-sensor data from aerial, Sentinel-1 and Sentinel-2 imagery. In this paper, we introduce the TreeSatAI Benchmark Archive, which contains labels of 20 European tree species (i.e., 15 tree genera) derived from forest administration data of the federal state of Lower Saxony, Germany. We propose models and guidelines for the application of the latest machine learning techniques for the task of tree species classification with multi-label data. Finally, we provide various benchmark experiments showcasing the information which can be derived from the different sensors including artificial neural networks and tree-based machine learning methods. We found that residual neural networks (ResNet) perform sufficiently well with weighted precision scores up to 79 % only by using the RGB bands of aerial imagery. This result indicates that the spatial content present within the 0.2 m resolution data is very informative for tree species classification. With the incorporation of Sentinel-1 and Sentinel-2 imagery, performance improved marginally. However, the sole use of Sentinel-2 still allows for weighted precision scores of up to 74 % using either multi-layer perceptron (MLP) or Light Gradient Boosting Machine (LightGBM) models. Since the dataset is derived from real-world reference data, it contains high class imbalances. We found that this dataset attribute negatively affects the models' performances for many of the underrepresented classes (i.e., scarce tree species). However, the class-wise precision of the best-performing late fusion model still reached values ranging from 54 % (Acer) to 88 % (Pinus). Based on our results, we conclude that deep learning techniques using aerial imagery could considerably support forestry administration in the provision of large-scale tree species maps at a very high resolution to plan for challenges driven by global environmental change. The original dataset used in this paper is shared via Zenodo (https://doi.org/10.5281/zenodo.6598390, Schulz et al., 2022). For citation of the dataset, we refer to this article.

Mapping Land Use/Land Cover Changes and Forest Disturbances in Vietnam Using a Landsat Temporal Segmentation Algorithm

Accurately mapping land use/land cover changes (LULCC) and forest disturbances provides valuable information for understanding the influence of anthropogenic activities on the environment at regional and global scales. Many approaches using satellite remote sensing data have been proposed for characterizing these long-term changes. However, a spatially and temporally consistent mapping of both LULCC and forest disturbances at medium spatial resolution is still limited despite their critical contributions to the carbon cycle. In this study, we examined the applicability of Landsat time series temporal segmentation and random forest classifiers to mapping LULCC and forest disturbances in Vietnam. We used the LandTrendr temporal segmentation algorithm to derive key features of land use/land cover transitions and forest disturbances from annual Landsat time series data. We developed separate random forest models for classifying land use/land cover and detecting forest disturbances at each segment and then derived LULCC and forest disturbances that coincided with each other during the period of 1988–2019. The results showed that both LULCC classification and forest disturbance detection achieved low accuracy in several classes (e.g., producer’s and user’s accuracies of 23.7% and 78.8%, respectively, for forest disturbance class); however, the level of accuracy was comparable to that of existing datasets using the same reference samples in the study area. We found relatively high confusion between several land use/land cover classes (e.g., grass/shrub, forest, and cropland) that can explain the lower overall accuracies of 67.6% and 68.4% in 1988 and 2019, respectively. The mapping of forest disturbances and LULCC suggested that most forest disturbances were followed by forest recovery, not by transitions to other land use/land cover classes. The landscape complexity and ephemeral forest disturbances contributed to the lower classification and detection accuracies in this study area. Nevertheless, temporal segmentation and derived features from LandTrendr were useful for the consistent mapping of LULCC and forest disturbances. We recommend that future studies focus on improving the accuracy of forest disturbance detection, especially in areas with subtle landscape changes, as well as land use/land cover classification in ambiguous and complex landscapes. Using more training samples and effective variables would potentially improve the classification and detection accuracies.

Extension of Scattering Power Decomposition to Dual-Polarization Data for Tropical Forest Monitoring

A new scattering power decomposition method is developed for accurate tropical forest monitoring that utilizes data in dual-polarization mode instead of quad-polarization (POLSAR) data. This improves the forest classification accuracy and helps to realize rapid deforestation detection because dual-polarization data are more frequently acquired than POLSAR data. The proposed method involves constructing scattering power models for dual-polarization data considering the radar scattering scenario of tropical forests (i.e., ground scattering, volume scattering, and helix scattering). Then, a covariance matrix is created for dual-polarization data and is decomposed to obtain three scattering powers. We evaluated the proposed method by using simulated dual-polarization data for the Amazon, Southeast Asia, and Africa. The proposed method showed an excellent forest classification performance with both user’s accuracy and producer’s accuracy at >98% for window sizes greater than 7 × 14 pixels, regardless of the transmission polarization. It also showed a comparable deforestation detection performance to that obtained by POLSAR data analysis. Moreover, the proposed method showed better classification performance than vegetation indices and was found to be robust regardless of the transmission polarization. When applied to actual dual-polarization data from the Amazon, it provided accurate forest map and deforestation detection. The proposed method will serve tropical forest monitoring very effectively not only for future dual-polarization data but also for accumulated data that have not been fully utilized.

The Role of C-Reactive Protein in the Prognosis of Prostate Cancer: A Meta-Analysis.

To investigate the role of C-reactive protein (CRP) in the prognosis of prostate cancer (PCa). The studies related to C-reactive protein and prostate cancer were searched by computer, including PubMed and Web of Science. The retrieval time was from the establishment of the database to August 2022. QUADAS score was employed to assess the studies' quality, funnel plot was employed to analyze the bias of the included studies, and RevMan and STATA statistical software programs were used to draw forest maps to represent the analysis results. In the initial examination, 432 articles were obtained. After removing the duplicate articles, reading the abstract and theme, and then reading the full text, 12 articles finally met the inclusion criteria. The results revealed that serum C-reactive protein (CRP) levels were associated with overall survival (OS) in patients with PCa (OR = 1.47 [1.19, 1.82], P < 0.05), and patients with high CRP levels had an increased risk of developing prostate cancer (HR = 0.26, 95% CI: 0.23, 0.29). However, there was no obvious difference in circulating CRP levels between patients with prostate cancer and healthy controls (P > 0.05). CRP levels are associated with PCa patients' OS. High CRP levels have an elevated incidence of PCa, but there was no obvious distinction in circulating CRP levels between patients with prostate cancer and healthy controls. Therefore, C-reactive protein has certain reference value for judging the prognosis of prostate cancer.

Mapping Nature across the Americas

Mapping Nature across the Americas

Semi- and weak-supervised learning for Norwegian tree species detection

Tree species mapping of Norwegian production forests is a time-consuming process as forest associations largely rely on manual interpretation of earth observation data. Deep learning based image segmentation techniques have the potential to improve automated tree species classification, but a major challenge is the limited quality and availability of training data. Semi-supervised techniques could alleviate the need for training label and weak supervision enables handling coarse-grained and noisy labels. In this study, we evaluated the added value of semi-supervised deep learning methods in a weakly supervised setting. Specifically, consistency training and pseudo-labeling are applied for tree species classification from aerial ortho imagery in Norway. The techniques are generic and relevant for the wider earth observation domain, especially for other land cover segmentation tasks. The results show that consistency training gives a significant performance increase. Pseudo-labeling on the other hand does not, potentially this is due to varying convergence speeds for different classes causing confirmation bias or a partial violation of the cluster assumption.

Effectiveness and safety of Inclisiran in hyperlipidemia treatment: An overview of systematic reviews.

This paper aimed to comprehensively evaluate the effectiveness and safety of Inclisiran in treating hyperlipidemia through an overview of systematic reviews (SRs). The Cochrane Library, EMBASE, PubMed, CNKI, WANGFANG database, VIP database, ClinicalTrials.gov, and ICRT were searched electronically to collect SRs and meta-analysis of Inclisiran in hyperlipidemia treatment from the establishment of the database till May 2022. Two researchers independently screened the relevant literature, then the assessment of multiple systematic reviews tool was made into assess the methodological quality of the included studies. Data extracted were used to perform the study through RevMan5.3 software. The grading of recommendations assessment, development, and evaluation tool was used to grade the quality of the evidence of the outcomes included in the SRs. Prospero ID: CRD 42022326845. A total of 10 relevant SRs were included, involving 7 randomized controlled trials. The assessment results of the assessment of multiple systematic reviews tool suggested that the quality of the SRs included needed to be improved. The reduced level of low-density lipoprotein cholesterol of the experimental group was lower than the control group, and the difference in the amount of effectiveness was statistically significant (MD = -50.13, 95%CI: -56.2 to -44.06, P < .00001). The grading of recommendations assessment, development, and evaluation results showed that out of 27 outcomes, 8 were high-quality, 3 were of medium quality, 6 were of low quality, and 10 were of the most inferior quality. 300mg Inclisiran with 2 injections a year has the best therapeutic effect, which can significantly reduce low-density lipoprotein cholesterol and total cholesterol, and increase high-density lipoprotein cholesterol levels in patients with hyperlipidemia. Inclisiran has a favorable safety profile, with no significant difference in the incidence of adverse reactions compared to a placebo. Most of the adverse effects were associated with the reaction on the injection site.

Does fragmentation contribute to the forest crisis in Germany?

Intact forests contribute to the ecosystem functionality of landscapes by storing and sequestering carbon, buffering and cooling the microclimate, and providing a range of related ecosystem functions. Forest fragmentation not only poses a threat to many organisms but also reduces the resistance and resilience of the ecosystem, which is especially relevant to the ongoing climate crisis. The effects of recent extreme heat years on forests in Germany have not been studied in detail for the influence of fragmentation. We investigate the relation of forest fragmentation with temperature and vitality in Germany per ecoregion at the canopy level using satellite imagery at 1-km and 30-m resolution. We compiled and correlated forest maps for connectivity based on Thiessen polygons, canopy temperatures on the hottest days based on land surface temperature, and forest vitality based on the maximum normalized difference vegetation index per growing season. We differentiated between ecoregions and main forest types. In 2022, larger intact tree-covered areas that are less fragmented have relatively low temperatures on hot days and higher overall vitality. Nearly 98% of the almost 1.95 million forest fragments at 30-m resolution in Germany are smaller than 1 km2, which cover nearly 30% of the total forest area. To counteract the forest crisis, forest and landscape management should aim to reduce fragmentation and maintain tree biomass and forest cover in the landscape. Increasing the size of continuous forest fragments contributes to ecosystem-based adaptation to climate change.

Mapping China’s planted forests using high resolution imagery and massive amounts of crowdsourced samples

Tree planting has been suggested as a potentially effective solution for mitigating climate change. China has implemented the world’s largest afforestation and reforestation project since the 1970s, but high-resolution maps of China’s planted forests remain unavailable. In this study, we explored the use of multi-source remote sensing images and crowdsourced samples to produce the first high-resolution (30-m) map of China’s planted forests. We constructed a Google Earth Engine (GEE)-based mapping framework using spectral, temporal, structural, textural and topographic features derived from Landsat and Sentinel-1 time series imagery, Digital Elevation Model (DEM) and Chinese Forest Canopy Height (CFCH) data. Over 300,000 high-quality crowdsourced samples were collected for training the mapping pipeline. Validation against independent field samples indicated an accuracy of 84.93 % and an F1 score of 0.85. The uncertainty map of each pixel was also constructed and showed that the areas of low and medium uncertainties accounted for 38.27 % and 50.98 % of the total area, respectively, indicating the high estimation reliabilities of the planted forest map. We show that China’s planted forests in the year of 2020 had a total area of 769853.01 km2, accounting for 31.30 % of the world’s total planted forests. The majority (77.45 %) of China’s planted forests were located in the Eastern, Center-South, and Southwestern regions. By further assessing the performance of the image features used to map the planted forests, we found that temporal features are key to identifying the planted forests in East and Center-South of China, where they are mainly timber plantations. However, structural and textural features were more useful for locating the planted forests in North and Northeast of China, where are dominated by planted shelterbelts. Our study demonstrated that combining crowdsourced samples with high-resolution satellite images allows mapping planted forests with unprecedented resolution (30-m) across large areas. Our map could contribute to the sustainable management of China’s forests and a more accurate quantification of the carbon balance of China’s natural ecosystems.

Biomass Estimation of Subtropical Arboreal Forest at Single Tree Scale Based on Feature Fusion of Airborne LiDAR Data and Aerial Images

Low-cost UAV aerial photogrammetry and airborne lidar scanning have been widely used in forest biomass survey and mapping. However, the feature dimension after multisource remote sensing fusion is too high and screening key features to achieve feature dimension reduction is of great significance for improving the accuracy and efficiency of biomass estimation. In this study, UAV image and point cloud data were combined to estimate and map the biomass of subtropical forests. Firstly, a total of 173 dimensions of visible light vegetation index, texture, point cloud height, intensity, density, canopy, and topographic features were extracted as variables. Secondly, the Kendall Rank correlation coefficient and permutation importance (PI) index were used to identify the key features of biomass estimation among different tree species. The random forest (RF) model and XGBoost model finally were used to compare the accuracy of biomass estimation with different variable sets. The experimental results showed that the point cloud height, canopy features, and topographic factors were identified as the key parameters of the biomass estimate, which had a significant influence on the biomass estimation of the three dominant tree species in the study area. In addition, the differences in the importance of characteristics among the tree species were discussed. The fusion features combined with the PI index screening and RF model achieved the best estimation accuracy, the R2 of 0.7356, 0.8578, and 0.6823 were obtained for the three tree species, respectively.

Mapping Large-Scale Bamboo Forest Based on Phenology and Morphology Features

Bamboo forest is a unique forest landscape that is mainly composed of herbal plants. It has a stronger capability to increase terrestrial carbon sinks than woody forests in the same environment, thus playing a special role in absorbing atmospheric CO2. Accurate and timely bamboo forest maps are necessary to better understand and quantify their contribution to the carbon and hydrological cycles. Previous studies have reported that the unique phenology pattern of bamboo forests, i.e., the on- and off-year cycle, can be detected with time-series high spatial resolution remote sensing (RS) images. Nevertheless, this information has not yet been applied in large-scale bamboo mapping. In this study, we innovatively incorporate newly designed phenology features reflecting the aforementioned on- and off-year cycles into a typical end-to-end classification workflow, including two features describing growing efficiency during the green-up season and two features describing the difference between annual peak greenness. Additionally, two horizonal morphology features and one tree height feature were employed, simultaneously. An experiment in southeast China was carried out to test the method’s performance, in which seven categories were focused. A total of 987 field samples were used for training and validation (70% and 30%, respectively). The results show that combining the time-series features based on spectral bands and vegetation indices and newly designed phenology and morphology patterns can differentiate bamboo forests from other vegetation categories. Based on these features, the classification results exhibit a reasonable spatial distribution and a satisfactory overall accuracy (0.89). The detected bamboo area proportion in 82 counties agrees with the statistics from China’s Third National Land Survey, which was produced based on high resolution images from commercial satellites and human interpretation (correlation coefficient = 0.69, and root mean squared error = 5.1%). This study demonstrates that the new scheme incorporating phenology features helps to map bamboo forests accurately while reducing the sample size requirement.

Improving GEDI Forest Canopy Height Products by Considering the Stand Age Factor Derived from Time-Series Remote Sensing Images: A Case Study in Fujian, China

Forest canopy height plays an important role in forest resource management and conservation. The accurate estimation of forest canopy height on a large scale is important for forest carbon stock, biodiversity, and the carbon cycle. With the technological development of satellite-based LiDAR, it is possible to determine forest canopy height over a large area. However, the forest canopy height that is acquired by this technology is influenced by topography and climate, and the canopy height that is acquired in complex subtropical mountainous regions has large errors. In this paper, we propose a method for estimating forest canopy height by combining long-time series Landsat images with GEDI satellite-based LiDAR data, with Fujian, China, as the study area. This approach optimizes the quality of GEDI canopy height data in topographically complex areas by combining stand age and tree height, while retaining the advantage of fast and effective forest canopy height measurements with satellite-based LiDAR. In this study, the growth curves of the main forest types in Fujian were first obtained by using a large amount of forest survey data, and the LandTrendr algorithm was used to obtain the forest age distribution in 2020. The obtained forest age was then combined with the growth curves of each forest type in order to determine the tree height distribution. Finally, the obtained average tree heights were merged with the GEDI_V27 canopy height product in order to create a modified forest canopy height model (MGEDI_V27) with a 30 m spatial resolution. The results showed that the estimated forest canopy height had a mean of 15.04 m, with a standard deviation of 4.98 m. In addition, we evaluated the accuracy of the GEDI_V27 and the MGEDI_V27 using the sample dataset. The MGEDI_V27 had a higher accuracy (R2 = 0.67, RMSE = 2.24 m, MAE = 1.85 m) than the GEDI_V27 (R2 = 0.39, RMSE = 3.35 m, MAE = 2.41 m). R2, RMSE, and MAE were improved by 71.79%, 33.13%, and 22.53%, respectively. We also produced a forest age distribution map of Fujian for the year 2020 and a forest disturbance map of Fujian for the past 32 years. The research results can provide decision support for forest ecological protection and management and for carbon sink analysis in Fujian.

Comparison between Parametric and Non-Parametric Supervised Land Cover Classifications of Sentinel-2 MSI and Landsat-8 OLI Data

The present research aims at verifying whether there are significant differences between Land Use/Land Cover (LULC) classifications performed using Landsat 8 Operational Land Imager (OLI) and Sentinel-2 Multispectral Instrument (MSI) data—abbreviated as L8 and S2. To comprehend the degree of accuracy between these classifications, both L8 and S2 scenes covering the study area located in the Basilicata region (Italy) and acquired within a couple of days in August 2017 were considered. Both images were geometrically and atmospherically corrected and then resampled at 30 m. To identify the ground truth for training and validation, a LULC map and a forest map realized by the Basilicata region were used as references. Then, each point was verified through photo-interpretation using the orthophoto AGEA 2017 (spatial resolution of 20 cm) as a ground truth image and, only in doubtful cases, a direct GPS field survey. MLC and SVM supervised classifications were applied to both types of images and an error matrix was computed using the same reference points (ground truth) to evaluate the classification accuracy of different LULC classes. The contribution of S2′s red-edge bands in improving classifications was also verified. Definitively, ML classifications show better performance than SVM, and Landsat data provide higher accuracy than Sentinel-2.

Mapping Forest Growing Stem Volume Using Novel Feature Evaluation Criteria Based on Spectral Saturation in Planted Chinese Fir Forest

Forest growing stem volume (GSV) is regarded as one of the most important parameters for the quality evaluation and dynamic monitoring of forest resources. The accuracy of mapping forest GSV is highly related to the employed models and involved remote sensing features, and the criteria of feature evaluation severely affect the performance of the employed models. However, due to the linear or nonlinear relationships between remote sensing features and GSV, widely used evaluation criteria inadequately express the complex sensitivity between forest GSV and spectral features, especially the saturation levels of features in a planted forest. In this study, novel feature evaluation criteria were constructed based on the Pearson correlations and optical saturation levels of the alternative remote sensing features extracted from two common optical remote sensing image sets (GF-1 and Sentinel-2). Initially, the spectral saturation level of each feature was quantified using the kriging spherical model and the quadratic model. Then, optimal feature sets were obtained with the proposed criteria and the linear stepwise regression model. Finally, four widely used machine learning models—support vector machine (SVM), multiple linear stepwise regression (MLR), random forest (RF) and K-neighborhood (KNN)—were employed to map forest GSV in a planted Chinese fir forest. The results showed that the proposed feature evaluation criteria could effectively improve the accuracy of estimating forest GSV and that the systematic distribution of errors between the predicted and ground measurements in the range of forest GSV was less than 300 m3/hm2. After using the proposed feature evaluation criteria, the highest accuracy of mapping GSV was obtained with the RF model for GF-1 images (R2 = 0.49, rRMSE = 28.67%) and the SVM model for Sentinel-2 images (R2 = 0.52, rRMSE = 26.65%), and the decreased rRMSE values ranged from 1.1 to 6.2 for GF-1 images (28.67% to 33.08%) and from 2.3 to 6.8 for Sentinel-2 images (26.85% to 33.28%). It was concluded that the sensitivity of the optimal feature set and the accuracy of the estimated GSV could be improved using the proposed evaluation criteria (less than 300 m3/hm2). However, these criteria were barely able to improve mapping accuracy for a forest with a high GSV (larger than 300 m3/hm2).

Automated Recognition of Tree Species Composition of Forest Communities Using Sentinel-2 Satellite Data

Information about the species composition of a forest is necessary for assessing biodiversity in a particular region and making economic decisions on the management of forest resources. Recognition of the species composition, according to the Earth’s remote sensing data, greatly simplifies the work and reduces time and labor costs in comparison with a traditional inventory of the forest, conducted through ground-based observations. This study analyzes the possibilities of tree species discrimination in coniferous–deciduous forests according to Sentinel-2 data using two automated recognition methods: random forest (RF) and generative topographic mapping (GTM). As remote sensing data, Sentinel-2 images of the Raifa section of Volga-Kama State Reserve in the Tatarstan Republic, Russia used: six images for the vegetation period of 2020. The analysis was carried out for the main forest-forming species. The training sample was created based on the cadastral data of the forest fund. The recognition quality was assessed using the F1-score, precision, recall, and accuracy metrics. The RF method showed a higher recognition accuracy. The accuracy of correct recognition by the RF method on the training sample reaches 0.987, F1-score = 0.976, on the control sample, accuracy = 0.764, F1-score = 0.709.

Strategy for mapping soil salt contents during the bare soil period through a satellite image: Optimal calibration set combined with random forest

Salt is a basic soil parameter that is especially sensitive to land degradation and global climate warming. Machine learning combined with satellite images offers the potential for effectively surveying soil salt contents (SSC) in the spatial and temporal dimensions. However, outlier spectral samples caused by environmental factors (e.g., particle size distribution, moisture, and vegetation) have interfered with the determination of the mapping relationship between ground-truth SSC and corresponding spectral features, resulting in a random forest (RF) mapping model with low generalization capability. Therefore, a spatial association module was introduced into the RF to remove the outlier spectral samples for building an optimal sample set in order to calibrate an improved random forest (IRF) model for SSC mapping. A total of 233 soil samples and a simultaneous Sentinel-2B multispectral image were acquired for the Weibei Plain of China. The spectral absorption features of SSC were characterized by three indices (band difference, band ratio, and band normalized). Then a spatial association function was developed to identify the outlier spectral samples for optimizing the calibration samples. Lastly, the SSC-responding bands and SSC-related indices were used as input variables to build the improved random forest (IRF) model for regional SSC mapping. Results showed that: (1) bands 3, 8, and 11 in the Sentinel-2B image were the SSC-responding bands, and band 11 had a significant correlation with SSC; (2) the band ratio index synthesized the multi-band spectral information, and effectively enhanced the spectral absorption signal of soil salt; the optimal SSC-related indices includedRI34 (reflectance ratio of band 3 and band 4),RI711,NDI611 (normalized reflectance values of band 6 and band 11), andDI45 (reflectance difference value of band 4 and band 5); (3) when the SSC-responding bands and optimal SSC-related indices were input into the IRF remote sensing estimation model, the accuracy parameters R2v and RPIQ were greater (0.89 and 3.52, respectively) than the comparable values with the RF model (R2v = 0.69; RPIQ = 2.90), with accuracy improved by 28.99 % and 21.38 %, respectively, indicating that satisfactory results for regional SSC mapping had been obtained; (4) SSC distribution showed that the values in the central part of the study area were slightly higher than in the southern and northern parts; the highest SSC area was mainly related to the distribution of saltern; additionally, agricultural activities and microtopography contributed significantly to the distribution of SSC at the farm scale. The IRF model based on the SSC-responding bands and optimal SSC-related indices provides a strategy for supporting regional mapping of SSC, and will be useful for environmental policy-making.