Mapping Tree Species Research Articles

Mango Yield Mapping at the Orchard Scale Based on Tree Structure and Land Cover Assessed by UAV

In the value chain, yields are key information for both growers and other stakeholders in market supply and exports. However, orchard yields are often still based on an extrapolation of tree production which is visually assessed on a limited number of trees; a tedious and inaccurate task that gives no yield information at a finer scale than the orchard plot. In this work, we propose a method to accurately map individual tree production at the orchard scale by developing a trade-off methodology between mechanistic yield modelling and extensive fruit counting using machine vision systems. A methodological toolbox was developed and tested to estimate and map tree species, structure, and yields in mango orchards of various cropping systems (from monocultivar to plurispecific orchards) in the Niayes region, West Senegal. Tree structure parameters (height, crown area and volume), species, and mango cultivars were measured using unmanned aerial vehicle (UAV) photogrammetry and geographic, object-based image analysis. This procedure reached an average overall accuracy of 0.89 for classifying tree species and mango cultivars. Tree structure parameters combined with a fruit load index, which takes into account year and management effects, were implemented in predictive production models of three mango cultivars. Models reached satisfying accuracies with R2 greater than 0.77 and RMSE% ranging from 20% to 29% when evaluated with the measured production of 60 validation trees. In 2017, this methodology was applied to 15 orchards overflown by UAV, and estimated yields were compared to those measured by the growers for six of them, showing the proper efficiency of our technology. The proposed method achieved the breakthrough of rapidly and precisely mapping mango yields without detecting fruits from ground imagery, but rather, by linking yields with tree structural parameters. Such a tool will provide growers with accurate yield estimations at the orchard scale, and will permit them to study the parameters that drive yield heterogeneity within and between orchards.

Global mapping of potential natural vegetation: an assessment of machine learning algorithms for estimating land potential.

Potential natural vegetation (PNV) is the vegetation cover in equilibrium with climate, that would exist at a given location if not impacted by human activities. PNV is useful for raising public awareness about land degradation and for estimating land potential. This paper presents results of assessing machine learning algorithms—neural networks (nnet package), random forest (ranger), gradient boosting (gbm), K-nearest neighborhood (class) and Cubist—for operational mapping of PNV. Three case studies were considered: (1) global distribution of biomes based on the BIOME 6000 data set (8,057 modern pollen-based site reconstructions), (2) distribution of forest tree taxa in Europe based on detailed occurrence records (1,546,435 ground observations), and (3) global monthly fraction of absorbed photosynthetically active radiation (FAPAR) values (30,301 randomly-sampled points). A stack of 160 global maps representing biophysical conditions over land, including atmospheric, climatic, relief, and lithologic variables, were used as explanatory variables. The overall results indicate that random forest gives the overall best performance. The highest accuracy for predicting BIOME 6000 classes (20) was estimated to be between 33% (with spatial cross-validation) and 68% (simple random sub-setting), with the most important predictors being total annual precipitation, monthly temperatures, and bioclimatic layers. Predicting forest tree species (73) resulted in mapping accuracy of 25%, with the most important predictors being monthly cloud fraction, mean annual and monthly temperatures, and elevation. Regression models for FAPAR (monthly images) gave an R-square of 90% with the most important predictors being total annual precipitation, monthly cloud fraction, CHELSA bioclimatic layers, and month of the year, respectively. Further developments of PNV mapping could include using all GBIF records to map the global distribution of plant species at different taxonomic levels. This methodology could also be extended to dynamic modeling of PNV, so that future climate scenarios can be incorporated. Global maps of biomes, FAPAR and tree species at one km spatial resolution are available for download via http://dx.doi.org/10.7910/DVN/QQHCIK.

Characterization of indicator tree species in neotropical environments and implications for geological mapping

Geobotanical remote sensing (GbRS) in the strict sense is an indirect approach to obtain geological information in heavily vegetated areas for mineral prospecting and geological mapping. Using ultra- and hyperspectral technologies, the goals of this research comprise the definition and mapping of Neotropical tree species that are associated with geological facies (here called geo-environments) as well as their spectral discrimination at leaf and crown scales. This work also aims to investigate the possible relationship between leaf and crown spectral and chemical properties. The study was developed at the Mogi-Guaçu Ecological Station, in the Cerrado domain, southeastern Brazil. Data from 70 sample units, such as sediment texture and species from inventories, were first analyzed through vectorial quantization using Self-Organizing Maps (SOM). Principal Component Analysis and Spearman's ranked correlation coefficients were used to define geo-environments and target-species, respectively. Biochemical and visible to shortwave infrared (VSWIR) point spectral data (350–2500 nm) were collected from the leaves of the target-species, during both rainy and dry seasons. Spectral data from target-species crowns were obtained from hyperspectral images (530–2.532 nm, ProSpecTIR-VS sensor) with 1 m spatial resolution, and acquired in the beginning of the dry season. These spectra were classified using Multiple Endmember Spectral Mixture Analysis (MESMA) with two endmembers (EMs). Based on the MESMA results with two EMs, the best dataset per target-species was chosen for pixel-based image unmixing with three EMs (target-species, other vegetation types and shade). From 121 species sampled in the field, two proved to be associated with floodplains (Alluvial Deposits sequence), two with hills and plateaus of the Aquidauna Formation (Carboniferous sedimentary rocks, Paraná Basin), and two more with a specific facies of the Aquidauna Formation that has a distinctive presence of coarse and very coarse sand. Five target-species were well discriminated at the leaf scale, reaching 90.0% and 85.0% of global accuracies in the rainy season and in the dry season, respectively. Accurate spectral discrimination appears to be linked to the considerable biochemical variability of their leaves in both seasons. Three species were discriminated at the crown scale, with 70.6% of global accuracy. When eight other landscape scale vegetation classes were included in the analyses, only Qualea grandiflora Mart. produced a satisfactory accuracy (61.1% and 100% of producer and user's accuracies, respectively). The spatial distribution of its fraction in the unmixed image, particularly, matches with the geological facies to which it was associated in the field. Ecological requirements for successfully mapping indicator species include broad and random distribution of the target-species' population, and singular physiological, phenological (and spectral) behavior at the imagery acquisition date. Our study shows that, even in tropical conditions, it is possible to use plant species mapping to support geological delineation, where rock exposures are typically rare.

Assessing the potential of multi-seasonal high resolution Pléiades satellite imagery for mapping urban tree species

This study evaluated the potential of five seasonal high resolution Pléiades satellite images for improving urban tree species classification in the City of Tampa, FL, USA. We assessed and compared the capabilities of individual and combined Pléiades images acquired during different seasons for classifying the urban tree species to understand the seasonal effect on tree species mapping accuracy. The seven species and groups included sand live oak (Quercus geminata), laurel oak (Q. laurifolia), live oak (Q. virginiana), pine (species group), palm (species group), camphor (Cinnamomum camphora), and magnolia (Magnolia grandiflora). A multi-level classification system was adopted to classify image objects of the tree species and groups. Species classification performance was compared between the five individual seasonal Pléiades images, between two combined dry-wet season images, and between the optimal single season and combined dry-wet season image data. Shade image objects were spectrally normalized to similar sunlit image objects. The tree species fraction features were extracted from the seasonal images using the Mixture Tuned Matching Filtering approach and used as additional features to a set of spectral and spatial/textural features. Random Forest, Support Vector Machine and Linear Discriminant Analysis classifiers were used to classify the seven tree species and groups with image objects features. The experimental results indicate significantly improved tree species mapping accuracies using late spring season (April) image compared to all other seasonal images (p < 0.01), and combined dry-wet season images performed even better. Results suggest a significant seasonal effect on tree species classification. The results also demonstrate that the Random Forest outperformed the Support Vector Machine and Linear Discriminant Analysis classifiers for tree species classification. Therefore, in practice, it is important to choose appropriate seasonal remote sensing data for mapping tree species.

CROWN-LEVEL TREE SPECIES CLASSIFICATION USING INTEGRATED AIRBORNE HYPERSPECTRAL AND LIDAR REMOTE SENSING DATA

Abstract. Mapping tree species is essential for sustainable planning as well as to improve our understanding of the role of different trees as different ecological service. However, crown-level tree species automatic classification is a challenging task due to the spectral similarity among diversified tree species, fine-scale spatial variation, shadow, and underlying objects within a crown. Advanced remote sensing data such as airborne Light Detection and Ranging (LiDAR) and hyperspectral imagery offer a great potential opportunity to derive crown spectral, structure and canopy physiological information at the individual crown scale, which can be useful for mapping tree species. In this paper, an innovative approach was developed for tree species classification at the crown level. The method utilized LiDAR data for individual tree crown delineation and morphological structure extraction, and Compact Airborne Spectrographic Imager (CASI) hyperspectral imagery for pure crown-scale spectral extraction. Specifically, four steps were include: 1) A weighted mean filtering method was developed to improve the accuracy of the smoothed Canopy Height Model (CHM) derived from LiDAR data; 2) The marker-controlled watershed segmentation algorithm was, therefore, also employed to delineate the tree-level canopy from the CHM image in this study, and then individual tree height and tree crown were calculated according to the delineated crown; 3) Spectral features within 3 × 3 neighborhood regions centered on the treetops detected by the treetop detection algorithm were derived from the spectrally normalized CASI imagery; 4) The shape characteristics related to their crown diameters and heights were established, and different crown-level tree species were classified using the combination of spectral and shape characteristics. Analysis of results suggests that the developed classification strategy in this paper (OA = 85.12 %, Kc = 0.90) performed better than LiDAR-metrics method (OA = 79.86 %, Kc = 0.81) and spectral-metircs method (OA = 71.26, Kc = 0.69) in terms of classification accuracy, which indicated that the advanced method of data processing and sensitive feature selection are critical for improving the accuracy of crown-level tree species classification.

Deep Learning for Fusion of APEX Hyperspectral and Full-Waveform LiDAR Remote Sensing Data for Tree Species Mapping

Deep learning has been widely used to fuse multi-sensor data for classification. However, current deep learning architecture for multi-sensor data fusion might not always perform better than single data source, especially for the fusion of hyperspectral and light detection and ranging (LiDAR) remote sensing data for tree species mapping in complex, closed forest canopies. In this paper, we propose a new deep fusion framework to integrate the complementary information from hyperspectral and LiDAR data for tree species mapping. We also investigate the fusion of either “single-band” or multi-band (i.e., full-waveform) LiDAR with hyperspectral data for tree species mapping. Additionally, we provide a solution to estimate the crown size of tree species by the fusion of multi-sensor data. Experimental results on fusing real APEX hyperspectral and LiDAR data demonstrate the effectiveness of the proposed deep fusion framework. Compared to using only single data source or current deep fusion architecture, our proposed method yields improvements in overall and average classification accuracies ranging from 82.21% to 87.10% and 76.71% to 83.45%, respectively.

Analysis of Vegetation Change and Mapping Tree Species in Mountainous Area Using Multi-Source Satellite Data: A Case Study of Djebel El Ouahch, Algeria

Monitoring vegetation cover change and mapping the distribution of forest tree species has been considered as a key issue in sustainable development policies, the purpose of this research is to detect vegetation cover changes over 30 years (1987- 2016), and perform forest type classification by using Multi-Source Satellite Data and field survey investigations in the massive of Djebel El Ouahch. Algeria. The vegetation cover changes were quantified using Normalized Difference Vegetation Index (NDVI) combined with supervised classification methods, based on time series Landsat images. Furthermore, the dominant tree species in the wooded area was estimated using Google Earth imagery, the fieldwork was done simultaneously with a visual interpretation of Google Earth images, the thematic maps were prepared and accuracy assessment results are considered satisfactory. The results revealed that the cropland and non-vegetated areas have increased by 22%, and 6.22% respectively. In contrast, significant spatial reduction in natural vegetation was observed. Both of Moderate and dense vegetation were decreased by 26.23% and 1.98%, respectively, the current results play an important role in any planning process and indicating the necessity to create a new strategy based on protecting the natural vegetation and plants diversity.

Classification of Tree Species in a Diverse African Agroforestry Landscape Using Imaging Spectroscopy and Laser Scanning

Airborne imaging spectroscopy (IS) and laser scanning (ALS) have been explored widely for tree species classification during the past decades. However, African agroforestry areas, where a few exotic tree species are dominant and many native species occur less frequently, have not yet been studied. Obtaining maps of tree species would provide useful information for the characterization of agroforestry systems and detecting invasive species. Our objective was to study tree species classification in a diverse tropical landscape using IS and ALS data at the tree crown level, with primary interest in the exotic tree species. We performed multiple analyses based on different IS and ALS feature sets, identified important features using feature selection, and evaluated the impact of combining the two data sources. Given that a high number of tree species with limited sample size (499 samples for 31 species) was expected to limit the classification accuracy, we tested different approaches to group the species based on the frequency of their occurrence and Jeffries–Matusita (JM) distance. Surface reflectance at wavelengths between 400–450 nm and 750–800 nm, and height to crown width ratio, were identified as important features. Nonetheless, a selection of minimum noise fraction (MNF) transformed reflectance bands showed superior performance. Support vector machine classifier performed slightly better than the random forest classifier, but the improvement was not statistically significant for the best performing feature set. The highest F1-scores were achieved when each of the species was classified separately against a mixed group of all other species, which makes this approach suitable for invasive species detection. Our results are valuable for organizations working on biodiversity conservation and improving agroforestry practices, as we showed how the non-native Eucalyptus spp., Acacia mearnsii and Grevillea robusta (mean F1-scores 76%, 79% and 89%, respectively) trees can be mapped with good accuracy. We also found a group of six fruit bearing trees using JM distance, which was classified with mean F1-score of 65%. This was a useful finding, as these species could not be classified with acceptable accuracy individually, while they all share common economic and ecological importance.

Mapping urban tree species using integrated airborne hyperspectral and LiDAR remote sensing data

Mapping tree species within urban areas is essential for sustainable urban planning as well as to improve our understanding of the role of urban vegetation as an ecological service. Urban trees contribute significantly in mitigating the urban heat island effect and supporting biodiversity. However, accurate and up-to-date mapping of urban tree species is difficult because of the time-consuming nature of field sampling, fine-scale spatial variation, and potentially high species diversity. Advanced remote sensing data such as airborne Light Detection and Ranging (LiDAR) with high pulse density (25point/m2) and hyperspectral imagery offer two different yet complementary approaches to estimating crown structure and canopy physiological information at the individual crown scale, which can be useful for mapping tree species. In this paper, we evaluate the potential of these technologies to map 15 common urban tree species using a Random Forest (RF) classifier in the City of Surrey, British Columbia, Canada. LiDAR-derived crown structural information was combined with hyperspectral-derived spectral vegetation indices for species classification. Results indicate an overall accuracy of 51.1%, 61.0%, and 70.0% using hyperspectral, LiDAR and the combined data respectively. The overall accuracy for the two most important and iconic native coniferous species improved markedly from 78.3% up to 91% using the combined data. The results of this research highlight that (1) the combination of structural and spectral information provided an improved classification accuracy than when used separately, and variables derived from LiDAR data contributed more to the accurate prediction of species than hyperspectral features; (2) higher classification accuracies were observed for evergreen species, species with distinguishable crown structure, and species undergoing flowering; (3) and finally the anthocyanin content index and photochemical reflectance index were the most important hyperspectral features for the discrimination of tree species in the spring budburst stage.

Production of a Large-Area Individual Tree Species Map for Forest Inventory in a Complex Forest Setting and Lessons Learned

A large-area trial of automated individual tree crown isolation and species classification on a 228-km2 site in central-eastern Ontario using 40 cm multispectral imagery gave insight into the complications and effectiveness of such approaches when applied to a complex mixed forest setting. Tree isolation was reasonably effective with few omissions and 77% of automated tree isolations (isols) were considered a good correspondence to ground truth delineations. There were issues with grouping several trees within 1 delineation, poor isols at the edges of stands and minor splitting (multiple isols per tree). Spectral characteristics of 18 species indicated considerable variability within species and overlap of signatures. To circumvent this, several spectral subclasses were created for certain species and the site was divided into broad and localized zones, which for ecological reasons had a reduced number of species. Mapping accuracy and classification accuracy of manually delineated trees for a main classification zone of 14 species, both typically ranged from 40% to 85%. For a simpler zone of 8 species, manual tree class accuracy averaged 76%. Possible improvements are discussed. Regardless of sophistication of techniques used, production of individual tree species maps in complex forests will require judicious use of human judgment and intervention.

Predicting tree species richness in urban forests

There has been an increasing interest in urban forests and the levels of biodiversity they contain. Currently there are no spatially explicit maps of tree species richness in urban areas. This research tests and identifies GIS and remote sensing metrics (climate, area, productivity, three-dimensional structure) hypothesized to be associated with species richness in native forests and identifies methods that can be applied to predict and map tree species richness in cities. We quantified tree species richness, floristic composition, and structure in 28 1-ha plots in the city of Los Angeles. Climate and remote sensing metrics from high-resolution aerial imagery (10 cm), QuickBird (60 cm), Landsat (30 m), MODIS (250 m), and airborne lidar (2 m) were collected for each plot. There were 1208 individual stems and 108 trees identified to species. Species richness ranged from 2 to 31 species per ha and averaged 17 species per ha. Tree canopy cover from QuickBird explained the highest portion of variance (54%) in tree species richness followed by NDVI from Landsat (42%). Tree species richness can be higher in residential urban forests than native forests in the United States. Spatially explicit species richness maps at 1 ha can be created and tested for cities in order to identify both hotspots and coldspots of tree species richness and changes in species richness over time.

Review of studies on tree species classification from remotely sensed data

Abstract Spatially explicit information on tree species composition of managed and natural forests, plantations and urban vegetation provides valuable information for nature conservationists as well as for forest and urban managers and is frequently required over large spatial extents. Over the last four decades, advances in remote sensing technology have enabled the classification of tree species from several sensor types. While studies using remote sensing data to classify and map tree species reach back several decades, a recent review on the status, potentials, challenges and outlooks in this realm is missing. Here, we search for major trends in remote sensing techniques for tree species classification and discuss the effectiveness of different sensors and algorithms based on a literature review. This review demonstrates that the number of studies focusing on tree species classification has increased constantly over the last four decades and promising local scale approaches have been presented for several sensor types. However, there are few examples for tree species classifications over large geographic extents, and bridging the gap between current approaches and tree species inventories over large geographic extents is still one of the biggest challenges of this research field. Furthermore, we found only few studies which systematically described and examined the traits that drive the observed variance in the remote sensing signal and thereby enable or hamper species classifications. Most studies followed data-driven approaches and pursued an optimization of classification accuracy, while a concrete hypothesis or a targeted application was missing in all but a few exceptional studies. We recommend that future research efforts focus stronger on the causal understanding of why tree species classification approaches work under certain conditions or – maybe even more important - why they do not work in other cases. This might require more complex field acquisitions than those typically used in the reviewed studies. At the same time, we recommend reducing the number of purely data-driven studies and algorithm-benchmarking studies as these studies are of limited value, especially if the experimental design is limited, e.g. the tree population is not representative and only a few sensors or acquisition settings are simultaneously investigated.

FOREST TREE SPECIES DISTRIBUTION MAPPING USING LANDSAT SATELLITE IMAGERY AND TOPOGRAPHIC VARIABLES WITH THE MAXIMUM ENTROPY METHOD IN MONGOLIA

Forest is a very important ecosystem and natural resource for living things. Based on forest inventories, government is able to make decisions to converse, improve and manage forests in a sustainable way. Field work for forestry investigation is difficult and time consuming, because it needs intensive physical labor and the costs are high, especially surveying in remote mountainous regions. A reliable forest inventory can give us a more accurate and timely information to develop new and efficient approaches of forest management. The remote sensing technology has been recently used for forest investigation at a large scale. To produce an informative forest inventory, forest attributes, including tree species are unavoidably required to be considered. &amp;lt;br&amp;gt;&amp;lt;br&amp;gt; In this study the aim is to classify forest tree species in Erdenebulgan County, Huwsgul province in Mongolia, using Maximum Entropy method. The study area is covered by a dense forest which is almost 70% of total territorial extension of Erdenebulgan County and is located in a high mountain region in northern Mongolia. For this study, Landsat satellite imagery and a Digital Elevation Model (DEM) were acquired to perform tree species mapping. The forest tree species inventory map was collected from the Forest Division of the Mongolian Ministry of Nature and Environment as training data and also used as ground truth to perform the accuracy assessment of the tree species classification. Landsat images and DEM were processed for maximum entropy modeling, and this study applied the model with two experiments. The first one is to use Landsat surface reflectance for tree species classification; and the second experiment incorporates terrain variables in addition to the Landsat surface reflectance to perform the tree species classification. All experimental results were compared with the tree species inventory to assess the classification accuracy. Results show that the second one which uses Landsat surface reflectance coupled with terrain variables produced better result, with the higher overall accuracy and kappa coefficient than first experiment. The results indicate that the Maximum Entropy method is an applicable, and to classify tree species using satellite imagery data coupled with terrain information can improve the classification of tree species in the study area.

FOREST TREE SPECIES DISTRIBUTION MAPPING USING LANDSAT SATELLITE IMAGERY AND TOPOGRAPHIC VARIABLES WITH THE MAXIMUM ENTROPY METHOD IN MONGOLIA

Abstract. Forest is a very important ecosystem and natural resource for living things. Based on forest inventories, government is able to make decisions to converse, improve and manage forests in a sustainable way. Field work for forestry investigation is difficult and time consuming, because it needs intensive physical labor and the costs are high, especially surveying in remote mountainous regions. A reliable forest inventory can give us a more accurate and timely information to develop new and efficient approaches of forest management. The remote sensing technology has been recently used for forest investigation at a large scale. To produce an informative forest inventory, forest attributes, including tree species are unavoidably required to be considered. In this study the aim is to classify forest tree species in Erdenebulgan County, Huwsgul province in Mongolia, using Maximum Entropy method. The study area is covered by a dense forest which is almost 70% of total territorial extension of Erdenebulgan County and is located in a high mountain region in northern Mongolia. For this study, Landsat satellite imagery and a Digital Elevation Model (DEM) were acquired to perform tree species mapping. The forest tree species inventory map was collected from the Forest Division of the Mongolian Ministry of Nature and Environment as training data and also used as ground truth to perform the accuracy assessment of the tree species classification. Landsat images and DEM were processed for maximum entropy modeling, and this study applied the model with two experiments. The first one is to use Landsat surface reflectance for tree species classification; and the second experiment incorporates terrain variables in addition to the Landsat surface reflectance to perform the tree species classification. All experimental results were compared with the tree species inventory to assess the classification accuracy. Results show that the second one which uses Landsat surface reflectance coupled with terrain variables produced better result, with the higher overall accuracy and kappa coefficient than first experiment. The results indicate that the Maximum Entropy method is an applicable, and to classify tree species using satellite imagery data coupled with terrain information can improve the classification of tree species in the study area.

Individual Tree Species Classification From Airborne Multisensor Imagery Using Robust PCA

Remote sensing of individual tree species has many applications in resource management, biodiversity assessment, and conservation. Airborne remote sensing using light detection and ranging (LiDAR) and hyperspectral sensors has been used extensively to extract biophysical traits of vegetation and to detect species. However, its application for individual tree mapping remains limited due to the technical challenges of precise coalignment of images acquired from different sensors and accurately delineating individual tree crowns (ITCs). In this study, we developed a generic workflow to map tree species at ITC level from hyperspectral imagery and LiDAR data using a combination of well established and recently developed techniques. The workflow uses a nonparametric image registration approach to coalign images, a multiclass normalized graph cut method for ITC delineation, robust principal component analysis for feature extraction, and support vector machine for species classification. This workflow allows us to automatically map tree species at both pixel- and ITC-level. Experimental tests of the technique were conducted using ground data collected from a fully mapped temperate woodland in the UK. The overall accuracy of pixel-level classification was 91%, while that of ITC-level classification was 61%. The test results demonstrate the effectiveness of the approach, and in particular the use of robust principal component analysis to prune the hyperspectral dataset and reveal subtle difference among species.

SVM-based soft classification of urban tree species using very high-spatial resolution remote-sensing imagery

ABSTRACTClassification from remote-sensing data has proved to be fast and effective for mapping tree species. However, the interference from background noise and the weak spectral separability among most tree species have negative impacts on classification accuracy and may make conventional classification methods inappropriate. To solve this problem, a soft classification approach is presented. It consists of the technologies of soft partition, defuzzifying, and feature-location analyses (referred to here as F-L analyses). The soft partition is conducted by combined binary support vector machines, enabling a natural prototype of fuzzy clusters to be formed. Unsure members (members means both patches and pixels) in the clustering prototype can be defuzzified afterwards. F-L analyses run through the entire process in two ways: (1) by adding density description to feature space for classification and (2) through density-constrained defuzzifying. The former can enrich the feature space, thus making the soft partition perform better. By using the latter, the credibility of a class label is indicated by several weighted fuzzy measures in the form of hierarchical densities, thereby reducing the uncertainty of the defuzzifying. Experiments indicate that only 63% of members on average are surely allocated in a clustering prototype and the remainder can be ensured by the explored defuzzifying approach. The mean overall accuracy (OA) of the defuzzified members is very close to that of the originally allocated sure ones. A comparative assessment shows that the mean OA and mean kappa statistic value of members defuzzified using the explored fuzzy measure are, on average, 3.84 and 4.89% respectively higher than those using conventional maximum membership measures. Experiments also show that the former is superior to the latter in regard to the formation of complete tree crown objects.

Mapping tree species in tropical seasonal semi-deciduous forests with hyperspectral and multispectral data

Abstract Accurately mapping the spatial distribution of tree species in tropical environments provides valuable insights for ecologists and forest managers. This process may play an important role in reducing fieldwork costs, monitoring changes in canopy biodiversity, and locating parent trees to collect seeds for forest restoration efforts. However, mapping tree species in tropical forests with remote sensing data is a challenge because of high floristic and spectral diversity. In this research, we discriminated and mapped tree species in tropical seasonal semi-deciduous forests (Brazilian Atlantic Forest Biome) by using airborne hyperspectral and simulated multispectral data in the 450 to 2400 nm wavelength range. After quantifying the spectral variability within and among individual tree crowns of eight species, three supervised machine learning classifiers were applied to discriminate the species at the pixel level. Linear Discriminant Analysis outperformed Support Vector Machines with Linear and Radial Basis Function (RBF-SVMs) kernels and Random Forests in almost all the tested cases. An average classification accuracy of 70% was obtained when using the visible/near-infrared (VNIR, 450–919 nm) bands. The inclusion of shortwave infrared bands (SWIR, 1045–2400 nm) increased the accuracy to 84%. Narrow-band vegetation indices (VIs) were also tested and increased the classification accuracy by up to 5% when combined with VNIR features. Furthermore, the spectral bands of the WorldView-3 (WV-3) satellite sensor were simulated for classification purposes. WV-3 VNIR bands provided an accuracy of 57.4%, which increased to 74.8% when using WV-3 SWIR bands. We also tested the production of species maps by using an object-oriented approach that integrated a novel segmentation algorithm that was tailored to delineate tree crowns and label high class membership pixels inside each object. In this scenario, RBF-SVMs produced the best species maps, correctly identifying 84.9% of crowns with hyperspectral data and 78.5% with simulated WV-3 data. The use of a reduced set of hyperspectral bands, which were selected with stepwise regression, did not significantly affect the classification accuracies but allowed us to depict the most important wavelengths to discriminate the species. These wavelengths were located around the green reflectance peak (550 nm), at the red absorption feature (650 nm) and in the SWIR range at 1200, 1700, 2100 and 2300 nm. These encouraging results suggest the feasibility of the proposed approach for mapping pioneering and climax tree species in the Brazilian Atlantic Forest Biome, highlighting its potential use in forest recovery and inventory initiatives.

Urban Tree Species Mapping Using Airborne LiDAR and Hyperspectral Data

In this study the high-spatial resolution (0.5 m) hyperspectral imagery with LiDAR data were fused at tree crown object level to classify 8 common tree species in Anyang, Henan, China. First, vertical forest features were extracted from LiDAR point clouds resulting in a canopy height model (CHM), followed by the acquisition of tree crown object (TCO) information from the CHM using a mean shift algorithm. Then, the CHM was combined with a minimum noise fraction transformation (MNF) and enhanced vegetation index (EVI), which were extracted from hyperspectral images. These combined features were used as the input to the SVM to produce a rough classification scheme for different tree species. Finally, a majority voting method was applied to the TCO to produce the final tree species map. The experiment showed that a combination of CHM-spatial-spectral features to classify tree species led to higher accuracy when compared to using only MNF features in the pixel-wise classification. However, the CHM and EVI features had their own limitations, largely depending on different characteristics of the different tree species.

Mapping tree species diversity of a tropical montane forest by unsupervised clustering of airborne imaging spectroscopy data

With the ongoing global biodiversity loss, approaches to measuring and monitoring biodiversity are necessary for effective conservation planning, especially in tropical forests. Remote sensing has much potential for biodiversity mapping, and high spatial resolution imaging spectroscopy (IS) allows for direct prediction of tree species diversity based on spectral reflectance. The objective of this study was to test an approach for mapping tree species alpha diversity that takes advantage of an unsupervised object-based clustering. Tree species diversity of a tropical montane forest in the Taita Hills, Kenya, was mapped based on spectral variation of high spatial resolution IS data.Airborne IS data and species data from 31 field plots were collected in the study area. Species diversity measures were obtained from the IS data by clustering spectrally similar image segments representing tree crowns. In order to do this, the image was segmented to objects that represented tree crowns. Three measures of species diversity were calculated based on the field data and on the clustering results, and the relationships were statistically analyzed.According to the results, the approach succeeded well in revealing tree species diversity patterns. Especially, tree species richness was well predicted (RMSE=3 species; r2=0.50) directly based on the clustering results. The optimal number of clusters was found to be close to the estimated number of tree species in the forest. Minimum tree size was an important determinant of the relationships, because only part of the trees are visible to the airborne sensor in the multi-layered closed canopy forest.In general, the object-based approach proved to be a viable alternative to a pixel-based clustering. The approach takes advantage of the capability of IS to detect spectral differences among tree crowns, but without the need for spectral training data, which is expensive to collect. With further development, the approach could be applied also for estimating beta diversity.

Object-Based Urban Tree Species Classification Using Bi-Temporal WorldView-2 and WorldView-3 Images

Urban tree species mapping is an important prerequisite to understanding the value of urban vegetation in ecological services. In this study, we explored the potential of bi-temporal WorldView-2 (WV2, acquired on 14 September 2012) and WorldView-3 images (WV3, acquired on 18 October 2014) for identifying five dominant urban tree species with the object-based Support Vector Machine (SVM) and Random Forest (RF) methods. Two study areas in Beijing, China, Capital Normal University (CNU) and Beijing Normal University (BNU), representing the typical urban environment, were evaluated. Three classification schemes—classification based solely on WV2; WV3; and bi-temporal WV2 and WV3 images—were examined. Our study showed that the single-date image did not produce satisfying classification results as both producer and user accuracies of tree species were relatively low (44.7%–82.5%), whereas those derived from bi-temporal images were on average 10.7% higher. In addition, the overall accuracy increased substantially (9.7%–20.2% for the CNU area and 4.7%–12% for BNU). A thorough analysis concluded that near-infrared 2, red-edge and green bands are always more important than the other bands to classification, and spectral features always contribute more than textural features. Our results also showed that the scattered distribution of trees and a more complex surrounding environment reduced classification accuracy. Comparisons between SVM and RF classifiers suggested that SVM is more effective for urban tree species classification as it outperforms RF when working with a smaller amount and imbalanced distribution of samples.