Search

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.

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.

Supervised classification methods, used for many applications, including vegetation mapping require accurate “ground truth” to be effective. Nevertheless, it is common for the quality of this data to be poorly verified prior to it being used for the training and validation of classification models. The fact that noisy or erroneous parts of the reference dataset are not removed is usually explained by the relatively high resistance of some algorithms to errors. The objective of this study was to demonstrate the rationale for cleaning the reference dataset used for the classification of heterogeneous non-forest vegetation, and to present a workflow based on the t-distributed stochastic neighbor embedding (t-SNE) algorithm for the better integration of reference data with remote sensing data in order to improve outcomes. The proposed analysis is a new application of the t-SNE algorithm. The effectiveness of this workflow was tested by classifying three heterogeneous non-forest Natura 2000 habitats: Molinia meadows (Molinion caeruleae; code 6410), species-rich Nardus grassland (code 6230) and dry heaths (code 4030), employing two commonly used algorithms: random forest (RF) and AdaBoost (AB), which, according to the literature, differ in their resistance to errors in reference datasets. Polygons collected in the field (on-ground reference data) in 2016 and 2017, containing no intentional errors, were used as the on-ground reference dataset. The remote sensing data used in the classification were obtained in 2017 during the peak growing season by a HySpex sensor consisting of two imaging spectrometers covering spectral ranges of 0.4–0.9 μm (VNIR-1800) and 0.9–2.5 μm (SWIR-384). The on-ground reference dataset was gradually cleaned by verifying candidate polygons selected by visual interpretation of t-SNE plots. Around 40–50% of candidate polygons were ultimately found to contain errors. Altogether, 15% of reference polygons were removed. As a result, the quality of the final map, as assessed by the Kappa and F1 accuracy measures as well as by visual evaluation, was significantly improved. The global map accuracy increased by about 6% (in Kappa coefficient), relative to the baseline classification obtained using random removal of the same number of reference polygons.

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.

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.