Search

Abstract Hyperspectral data and machine learning offer great potential for identifying valuable open ecosystems. Due to the large volume of data, preprocessing of hyperspectral images must involve dimensionality reduction. The main goal of this study was to test the effectiveness of various types of feature reduction (feature selection and feature extraction) when performing classification using the Random Forest algorithm. A comparison was conducted between two ecosystems - heathlands and mires protected as Natura 2000 habitats. Two transformations of feature extraction were chosen, namely Minimum Noise Fraction (MNF) and Principal Component Analysis (PCA), while Linear Discriminant Analysis (LDA) was used as a feature selection method. It was proven that irrespective of the class type, accuracy is higher with the feature extraction method (mean F1 accuracy of 0.922) than with feature selection (mean F1 accuracy of 0.787). At the same time, no significant differences in accuracies were found between the MNF and PCA methods. Although LDA resulted in lower accuracies (0.816 for heathland and 0.750 for mires), the method could also be used due to relatively high F1 values. The effectiveness of the LDA method for feature reduction in open ecosystem identification was confirmed for the first time for open natural vegetation.

Presently remote sensing appears to be the only technology that makes it possible to conduct an inventory of tree taxa throughout the whole city. One of the main challenges during a remote sensing implementation project is the constructing of a map legend for tree taxa. The article presents the first comprehensive investigation into the construction of a legend for urban tree species mapping in the entire city, using machine learning and airborne data. The analyses were based on the CatBoost algorithm, which is relatively new in environmental research, as well as hyperspectral and LiDAR data acquired by instrument fusion. The analysis covered the entire city of Warsaw (517.24 km2). The main aim of the study was to compare three approaches to creating map legends of taxonomic diversity in city trees. This research sought to answer which of these three approaches to creating a map legend produces a result with the highest application potential for use by urban greenery managers. Two scenarios are based on different levels of taxonomic organization. The first one, based on a generic level, contained 42 classes and obtained an overall accuracy score (OA) of 0.809. The second one, on the species level, consisted of 81 classes, and the OA reached an accuracy of 0.728. The third scenario presents a mixed approach by grouping classes as different combinations of genus, species and variety. It contained 60 classes and its OA totaled to 0.798. The obtained results indicate that while in the classification process the generic scenario achieved the highest accuracy and the species scenario had the most complex legend, the mixed scenario emerged as the most useful for the greenery managers because it presented a compromise between the complexity of the legend and the high accuracy of the map. In the mixed scenario, accurate classification was achieved by separating high-accuracy individual taxa at the species level and grouping the rest at different taxonomic levels. One of the most significant novelties of our research is proving that remote sensing currently makes it possible to develop a map of tree species containing as many as 60 classes of legends and to achieve an accuracy that allows for practical application in urban greenery management. The study lists suggestions of practical strategies for enhancing map accuracy such as aggregating similar species into species groups, creating distinct groups for ornamental cultivars with unique leaf characteristics or combining species or genera with similar traits.

Invasive alien species are one of the biggest threats to biodiversity today. Identifying their locations are mandatory parts of the strategies being developed to control them. Remote sensing along with machine learning are already proven and effective tools for monitoring invasive species, especially trees, shrubs, and tall perennials. However, annual vine species are particularly difficult to map using remote sensing because of their dynamic plant growth and the movement of shoots during the growing season. Therefore, the phenological phase in which the data is acquired, and the synchronization of airborne data acquisition with on-ground reference data, may be key factors for correct plant classification. This research aimed to answer the following questions: (i) What is the impact of acquiring synchronized on-ground data and hyperspectral data in different phases of plants’ phenological development on the annual vine IAPS (Invasive Alien Plant Species) classification results? (ii) How does the lack of synchronization while obtaining hyperspectral and on-ground data collection impact annual vine IAPS mapping results? (iii) Does multitemporal image fusion improve the results of annual vine IAPS classification? For this purpose, research was carried out on Echinocystis lobata, an annual vine species considered highly invasive in Europe. The obtained results indicate that the phenological phase in which the data is acquired has a very strong influence on the quality of the classification result. The period of flowering (summer) with the greatest coverage of the area with shoots was optimal for the classification of Echinocystis lobata with F1 classification accuracy of 0.87 ± 0.04. The accuracy of classifications was significantly less for spring (F1 = 0.64 ± 0.04) and autumn (F1 = 0.75 ± 0.05). Obtaining on-ground reference data that is mismatched temporally with hyperspectral data causes a decrease in the accuracy of the result up to 0.08 (from F1 = 0.64 to 0.56) in relation to data obtained synchronously. In the multitemporal image fusion method, using hyperspectral data linked from different phases of plants’ development to classify an image had a minimal improvement in classification accuracy compared to classifications trained on images from one phenological stage. The main conclusion is that mapping an annual vine using remote sensing and machine learning is possible and highly effective, provided the remote sensing and on-ground data are obtained in strict synchronization and the appropriate phenological phase. For the most efficient classification results, a single data acquisition per year is enough, even in the case of annual vine IAPS. Further research is needed to explore the possibility of mapping Echinocystis lobata using, i.e. multispectral or hyperspectral satellite data (e.g. EnMAP).

Abstract Aerial hyperspectral and multispectral satellite data are the two most commonly used datasets to identify natural and semi-natural vegetation. However, there is no documented analysis based on data from several areas concerning the difference in the classification accuracy of non-forest Natura 2000 habitat with the use of aerial hyperspectral and satellite multispectral data. Also, there is no recommendation, on which habitat can be classified with sufficient accuracy using free multispectral images. This study aimed to analyse the difference in classification accuracy of Natura 2000 habitats representing: meadows, grasslands, heaths and mires between data with different spectral resolutions and the results utility for nature conservation compared to conventional maps. The analysis was conducted in five study areas in Poland. The classification was performed on multispectral Sentinel-2 (S2) and hyperspectral HySpex (HS) images using the Random Forest algorithm. Based on the results, it can be stated that the use of HS data resulted in higher classification accuracy, on average 0.14, than using S2 images, regardless of the area of the habitat. However, the difference in accuracy was not constant, varying by area and habitat characterisation. Greater differences in accuracy were observed for areas where habitats were characterised by high α-diversity or β-diversity. The HS and S2 data make it possible to create maps that provide a great deal of new knowledge about the distribution of Natura 2000 habitats, which is necessary for the management of protected areas. The obtained results indicate that by using S2 images it is possible to identify, at a satisfactory level, alluvial meadows and grassland. For heaths and mires, using HS data improved the results, but it is also possible to acquire general distribution of these classes, whereas HS images are obligatory for mapping salt, Molinia and lowland hay meadows.

Studies conducted in recent years have demonstrated high application potential of thermal remote sensing data in environmental analyses. The main goal of our studies was to determine the variability of tree canopy temperatures using a new sensor which acquires data in the still rarely used thermal spectral range (3.6–4.9 μm). This study was conducted on five selected tree species growing in the forest and outside the forest: Alnus glutinosa, Pinus sylvestris, Quercus petraea, Quercus rubra and Robinia pseudoacacia. Thermal data were acquired on 9 June 2019, between 8:10 and 14:00 (CET). The findings were as follows: i) Trees growing in the forest are on average 0.4–0.7°C cooler than trees outside the forest; ii) The canopy temperatures of species under study differ statistically irrespective of data acquisition time. Alnus glutinosa, Quercus rubra and Quercus petraea are species with the lowest canopy temperatures, and Pinus sylvestris has the highest canopy temperature. The studies showed that the biggest variation between species in the canopy temperature occurs at noon (12:00–13:00); iii) A thermal spectral range of 3.6–4.9 μm registers the canopy temperature of tree species with a high accuracy, which supports its usage in remote sensing vegetation studies.

One of the key issues in wetland monitoring is mapping vegetation. Remote sensing and machine learning are used to acquire vegetation maps, which, due to the development of sensors and data processing and analysis methods, have increasingly high accuracy. The objectives of this study were to test: (i) which of the textural information (TI) features have the highest information potential for identifying wetland communities; and (ii) whether the use of TI improves the accuracy of wetland communities mapping using hyperspectral (HS) and Airborne Laser Scanning (ALS) data. The analysis indicated that the mean and entropy features of the Gray Level Co-occurrence Matrix had the highest potential to differentiate between various wetland communities. Adding these features to the dataset resulted in a small increase (0.005) in average F1 accuracy based on HS data and 0.011 for HS and ALS scenarios in wetland communities classification, and adding TI improved the delineation of patch boundaries. A higher increase was noted for forest and scrub vegetation (by 0.019 for the HS scenario and 0.022 for the HS and ALS scenario) and rushes (only for the HS and ALS scenario 0.017). It can be concluded that it is reasonable to use textural information for mapping wetland communities, especially for areas with a high proportion of scrub and forest and rushes vegetation included in the analysis.

Habitat mapping is essential for the management and monitoring of Natura 2000 sites. Time-consuming field surveys are still the most frequently used solution for the implementation of the European Habitats Directive, but the use of remote sensing tools for this is becoming more common. The high temporal resolution of Sentinel-2 data, registering the visible, near, and shortwave infrared ranges of the electromagnetic spectrum, makes them valuable material in this context. In this study, we aimed to use multitemporal Sentinel-2 data for mapping three grassland Natura 2000 habitats in Poland. We performed the classification based on spectro-temporal features extracted from data collected from eight different terms within the year 2017 using Convolutional Neural Networks (CNNs), and we also tested other widely used machine learning algorithms for comparison, such as Random Forests (RFs) and Support Vector Machines (SVMs). Based on ground truth data, we randomly selected training and validation polygons and then performed the evaluation iteratively (100 times). The best resulting median F1 accuracies that we obtained for habitats were as follows: 6210, 0.85; 6410, 0.80; and 6510, 0.84 (with SVM). Finally, we concluded that the accuracy of the results was comparable, but we obtained the best results using SVM (median OA = 88%, with 86% for RF and 84% for CNNs). In this work, we confirmed the usefulness of the spectral dimension of Sentinel-2 time series data for mapping grassland habitats, and researchers of future work can further develop the use of CNNs for this purpose.

The aim of this study is to evaluate the effectiveness of the identification of Natura 2000 wetland habitats (Alkaline fens—code 7230, and Transition mires and quaking bogs—code 7140) depending on various remotely sensed (RS) data acquired from an airborne platform. Both remote sensing data and botanical reference data were gathered for mentioned habitats in the Lower (LB) and Upper Biebrza (UB) River Valley and the Janowskie Forest (JF) in different seasonal stages. Several different classification scenarios were tested, and the ones that gave the best results for analyzed habitats were indicated in each campaign. In the final stage, a recommended term of data acquisition, as well as a list of remote sensing products, which allowed us to achieve the highest accuracy mapping for these two types of wetland habitats, were presented. Designed classification scenarios integrated different hyperspectral products such as Minimum Noise Fraction (MNF) bands, spectral indices and products derived from Airborne Laser Scanning (ALS) data representing topography (developed in SAGA), or statistical products (developed in OPALS—Orientation and Processing of Airborne Laser Scanning). The image classifications were performed using a Random Forest (RF) algorithm and a multi-classification approach. As part of the research, the correlation analysis of the developed remote sensing products was carried out, and the Recursive Feature Elimination with Cross-Validation (RFE-CV) analysis was performed to select the most important RS sub-products and thus increase the efficiency and accuracy of developing the final habitat distribution maps. The classification results showed that alkaline fens are better identified in summer (mean F1-SCORE equals 0.950 in the UB area, and 0.935 in the LB area), transition mires and quaking bogs that evolved on/or in the vicinity of alkaline fens in summer and autumn (mean F1-SCORE equals 0.931 in summer, and 0.923 in autumn in the UB area), and transition mires and quaking bogs that evolved on dystrophic lakes in spring and summer (mean F1-SCORE equals 0.953 in spring, and 0.948 in summer in the JF area). The study also points out that the classification accuracy of both wetland habitats is highly improved when combining selected hyperspectral products (MNF bands, spectral indices) with ALS topographical and statistical products. This article demonstrates that information provided by the synergetic use of data from different sensors can be used in mapping and monitoring both Natura 2000 wetland habitats for its future functional assessment and/or protection activities planning with high accuracy.

The study was focused on a plant native to Poland, the European dewberry Rubus caesius L., which is a species with the ability to become excessively abundant within its original range, potentially causing significant changes in ecosystems, including biodiversity loss. Monitoring plant distributions over large areas requires mapping that is fast, reliable, and repeatable. For Rubus, different types of data were successfully used for classification, but most of the studies used data with a very high spectral resolution. The aim of this study was to indicate, using hyperspectral and Light Detection and Ranging (LiDAR) data, the main functional trait crucial for R. caesius differentiation from non-Rubus. This analysis was carried out with consideration of the seasonal variability and different percentages of R. caesius in the vegetation patches. The analysis was based on hyperspectral HySpex images and Airborne Laser Scanning (ALS) products. Data were acquired during three campaigns: early summer, summer, and autumn. Differentiation based on Linear Discriminate Analysis (LDA) and Non-Parametric Multivariate Analysis of Variance (NPMANOVA) analysis was successful for each of the analysed campaigns using optical data, but the ALS data were less useful for identification. The analysis indicated that selected spectral ranges (VIS, red-edge, and parts of the NIR and possibly SWIR ranges) can be useful for differentiating R. caesius from non-Rubus. The most useful indices were ARI1, CRI1, ARVI, GDVI, CAI, NDNI, and MRESR. The obtained results indicate that it is possible to classify R. caesius using images with lower spectral resolution than hyperspectral data.

Knowledge of the vegetation and the monitoring of its changes in preserved areas is an essential part of effective conservation policy and management. The aim of this study was to assess the effectiveness of traditional methods of conservation of silver fir forests. The study analyses the changes in the structure and species composition of a temperate forest excluded from the commercial silvicultural management for 50 years, and since then protected as a nature reserve. The study is based on a comparative analysis of phytosociological reléves made on permanent plots in 1961, 1982, 1994 and 2011. PCA and ecological indicator values were analyzed, as well as characteristic species based on an indicator value (IndVal) index. Results revealed significant and dynamic changes in the forest structure and composition. The mixed coniferous-broadleaved forest with Abies alba and diverse ground flora, considered in the 1960s as valuable and worthy of conservation, was found to have been anthropogenically transformed and unstable. Significant reduction in the human impact was followed by spontaneous regeneration of oak–hornbeam forest. However, the directional process of changes in vegetation was modified by such silvicultural treatments as selective cutting of trees and gap creation, all intended for silver fir maintenance. The results show that Carpinus betulus effectively outcompeted Pinus sylvestris, Picea abies, Quercus robur and A. alba. Changes in the forest overstory and understory caused temporal changes in the habitat conditions reflected in changes in the ground vegetation composition. The proportion of light-demanding and oligotrophic species significantly decreased, while the contribution of species with a wide ecological amplitude, i.e. more shade-tolerant and nutrient-demanding – increased. The share of A. alba was reduced. Species defined in this study as most valuable, should be actively protected, or selection of conservation targets should be re-evaluated.

Some alien woody species used in commercial forestry become invasive and, as invaders, cause major problems in natural and semi-natural ecosystems. However, the deliberate introduction of aliens can bring unintended negative changes also within areas of their cultivation. This paper presents the effects of the intentional introduction of the North-American Quercus rubra in European mixed Scots pine-Pedunculate oak forests (POFs): Querco roboris-Pinetum (W. Mat. 1981) J. Mat. 1988. Phytosociological data from field research combined with GIS data analysis of the current distribution of Northern Red oak in the studied habitat were used to determine the composition and structure of forest communities in plots with and without Q. rubra participation. The results show that Q. rubra significantly reduces native species richness and abundance, both in old-growth and in secondary (post-agricultural) forests. Not one resident vascular plant benefits from the introduction of Northern Red oak and only a few are able to tolerate its co-occurrence. The natural restocking of all native woody species is also strongly limited by this alien tree. The introduction of Northern Red oak significantly limits the environmental functions of the POF ecosystem and weakens its economic and social aspects. However, its further cultivation is justified from an economic point of view, as the essential function of the studied forests is commercial timber production, and the introduction of this fast growing alien tree supports the provisioning ecosystem services. A clear description of the level of trade-off between the accepted negative and positive effects of the introduction of Q. rubra on forest ecosystem services requires further interdisciplinary studies.

The process of secondary succession is one of the most significant threats to non-forest (natural and semi-natural open) Natura 2000 habitats in Poland; shrub and tree encroachment taking place on abandoned, low productive agricultural areas, historically used as pastures or meadows, leads to changes to the composition of species and biodiversity loss, and results in landscape transformations. There is a perceived need to create a methodology for the monitoring of vegetation succession by airborne remote sensing, both from quantitative (area, volume) and qualitative (plant species) perspectives. This is likely to become a very important issue for the effective protection of natural and semi-natural habitats and to advance conservation planning. A key variable to be established when implementing a qualitative approach is the remote sensing data acquisition date, which determines the developmental stage of trees and shrubs forming the succession process. It is essential to choose the optimal date on which the spectral and geometrical characteristics of the species are as different from each other as possible. As part of the research presented here, we compare classifications based on remote sensing data acquired during three different parts of the growing season (spring, summer and autumn) for five study areas. The remote sensing data used include high-resolution hyperspectral imagery and LiDAR (Light Detection and Ranging) data acquired simultaneously from a common aerial platform. Classifications are done using the random forest algorithm, and the set of features to be classified is determined by a recursive feature elimination procedure. The results show that the time of remote sensing data acquisition influences the possibility of differentiating succession species. This was demonstrated by significant differences in the spatial extent of species, which ranged from 33.2% to 56.2% when comparing pairs of maps, and differences in classification accuracies, which when expressed in values of Cohen’s Kappa reached ~0.2. For most of the analysed species, the spring and autumn dates turned out to be slightly more favourable than the summer one. However, the final recommendation for the data acquisition time should take into consideration the phenological cycle of deciduous species present within the research area and the abiotic conditions.

Accurately identifying Natura 2000 habitat areas with the support of remote sensing techniques is becoming increasingly feasible. Various data types and methods are used for this purpose, and the fusion of data from various sensors and temporal periods (terms) within the phenological cycle allows natural habitats to be precisely identified. This research was aimed at selecting optimal datasets to classify three grassland Natura 2000 habitats (codes 6210, 6410 and 6510) in the Ostoja Nidziańska Natura 2000 site in Poland based on hyperspectral imagery and botanical on-ground reference data acquired in three terms during one vegetative period in 2017 (May, July and September), as well as a digital terrain model (DTM) obtained by airborne laser scanning (ALS). The classifications were carried out using a random forest (RF) algorithm on minimum noise fraction (MNF) transform output bands obtained for single terms, as well as data fusion combining the topographic indices (TOPO) calculated from the DTM, multitemporal hyperspectral data, or a combination of the two. The classification accuracy statistics were analysed in various combinations based on the datasets and their terms of acquisition. Topographic indices improved the classification accuracy of habitats 6210 and 6410, with the greatest impact noted in increased classification accuracy of xerothermic grasslands. The best terms for identifying specific habitats were autumn for 6510 and summer for 6210 and 6410, while the best results overall were obtained by combining data from all terms. The highest obtained values of the F1 coefficient were 84.5% for habitat 6210, 83.2% for habitat 6410, and 69.9% for habitat 6510. Comparing the data fusion results for habitats 6210 and 6410, greater accuracy was obtained by adding topographic indices to multitemporal hyperspectral data, while for habitat 6510, greater accuracy was obtained by fusing only multitemporal hyperspectral data.

Remote sensing (RS) is currently regarded as one of the standard tools used for mapping invasive and expansive plants for scientific purposes and it is increasingly widely used in nature conservation management. The applicability of RS methods is determined by its limitations and requirements. One of the most important limitations is the species percentage cover at which the classification result is correct and useful for nature conservation. The primary objective, carried out in 2017 in three areas of Poland, was to determine the minimum percentage cover from which it is possible to identify a target species by RS methods. A secondary objective of this research, related to the requirements of the method, was to optimize the set of training polygons for a target species in terms of the number of polygons and abundance percentage cover of the target species. Our method has to be easy to use, effective, and applicable, therefore the analysis was carried out using the basic set of rasters—the first 30 channels after the Minimum Noise Fraction (MNF) transformation (the mosaic of hyperspectral data from HySpex sensors with spectral range 0.4−2.5 µm) and commonly used Random Forest algorithm. The analysis used airborne hyperspectral data with a spatial resolution of 1 m to perform classification of one invasive and three expansive plants—two grasses and two large perennials. On-ground training and validation data sets were collected simultaneously with airborne data collection. When testing different classification scenarios, only the set of training polygons for a target species was changed. Classification results were evaluated based on three methods: accuracy measures (Kappa and F1), true-positive pixels in subclasses with different species cover and compatibility with field mapping. The classification results indicate that to classify the target plant species at the accepted level, the training dataset should contain polygons with a species cover ranging from 80−100%. Training performed only using polygons with a species characterized by a variable, but lower, cover (20−70%) and missing samples in the 80−100% range, led to a map which was not acceptable because of a high overestimation of target species. We achieved effective identification of species in areas where the species cover is above 50%, considering that ecosystems are heterogeneous. The results of these studies developed a methodology of field data acquisition and the necessity of synchronization in the acquisition of airborne data, and training and validation of on-ground sampling.

Secondary succession is considered a threat to non-forest Natura 2000 habitats. Currently available data and techniques such as airborne laser scanning (ALS) data processing can be used to study this process. Thanks to these techniques, information about the spatial extent and the height of research objects—trees and shrubs—can be obtained. However, only archival aerial photographs can be used to conduct analyses of the stage of succession process that took place in the 1960s or 1970s. On their basis, the extent of trees and shrubs can be determined using photointerpretation, but height information requires stereoscopic measurements. State-of-the-art dense image matching (DIM) algorithms provide the ability to automate this process and create digital surface models (DSMs) that are much more detailed than ones obtained using image matching techniques developed a dozen years ago. This research was part of the HabitARS project on the Ostoja Olsztyńsko-Mirowska Natura 2000 protected site (PLH240015). The source data included archival aerial photographs (analogue and digital) acquired from various phenological periods from 1971–2015, ALS data from 2016, and data from botanical campaigns. First, using the DIM algorithms, point clouds were generated and converted to DSMs. Heights interpolated from the DSMs were compared with stereoscopic measurements (1971–2012) and ALS data (2016). Then, the effectiveness of tree and shrub detection was analysed, considering the relationship between the date and the parameters of aerial images acquisition and DIM effects. The results showed that DIM can be used successfully in tree and shrub detection and monitoring, but the source images must meet certain conditions related to their quality. Based on the extensive material analysed, the detection of small trees and shrubs in aerial photographs must have a scale greater than 1:13,000 or a 25 cm GSD (Ground Sample Distance) at most, an image acquisition date from June–September (the period of full foliage in Poland), and good radiometric quality.

Fusion of remote sensing data often improves vegetation mapping, compared to using data from only a single source. The effectiveness of this fusion is subject to many factors, including the type of data, collection method, and purpose of the analysis. In this study, we compare the usefulness of hyperspectral (HS) and Airborne Laser System (ALS) data fusion acquired in separate flights, Multiple Flights Data Fusion (MFDF), and during a single flight through Instrument Fusion (IF) for the classification of non-forest vegetation. An area of 6.75 km2 was selected, where hyperspectral and ALS data was collected during two flights in 2015 and one flight in 2017. This data was used to classify three non-forest Natura 2000 habitats i.e., Xeric sand calcareous grasslands (code 6120), alluvial meadows of river valleys of the Cnidion dubii (code 6440), species-rich Nardus grasslands (code 6230) using a Random Forest classifier. Our findings show that it is not possible to determine which sensor, HS, or ALS used independently leads to a higher classification accuracy for investigated Natura 2000 habitats. Concurrently, increased stability and consistency of classification results was confirmed, regardless of the type of fusion used; IF, MFDF and varied information relevance of single sensor data. The research shows that the manner of data collection, using MFDF or IF, does not determine the level of relevance of ALS or HS data. The analysis of fusion effectiveness, gauged as the accuracy of the classification result and time consumed for data collection, has shown a superiority of IF over MFDF. IF delivered classification results that are more accurate compared to MFDF. IF is always cheaper than MFDF and the difference in effectiveness of both methods becomes more pronounced when the area of aerial data collection becomes larger.