Search

Abstract Tree species’ composition of forests is essential in forest management and nature conservation. We aimed to identify the tree species structure of a floodplain forest area using a hyperspectral image. We proposed an efficient novel strategy including the testing of three dimension reduction (DR) methods: Principal Component Analysis, Minimum Noise Fraction (MNF) and Indipendent Component Analysis with five machine learning (ML) algorithms (Maximum Likelihood Classifier, Support Vector Classification, Support Vector Machine, Random Forest and Artificial Neural Network) to find the most accurate outcome; altogether 300 models were calculated. Post-classification was applied by combining the multiresolution segmentation and filtering. MNF was the most efficient DR technique, and at least 7 components were needed to gain an overall accuracy (OA) of > 75%. Forty-five models had > 80% OAs; MNF was 43, and the Maximum Likelihood was 19 times among these models. Best classification belonged to MNF with 10 components and Maximum Likelihood classifier with the OA of 83.3%. Post-classification increased the OA to 86.1%. We quantified the differences among the possible DR and ML methods, and found that even > 10% worse model can be found using popular standard procedures related to the best results. Our workflow calls the attention of careful model selection to gain accurate maps.

Mapping and quantifying 3D vegetation structure is essential for assessing and monitoring ecosystem structure and function within wetlands. Airborne Laser Scanning (ALS) is a promising data source for developing indicators of 3D vegetation structure, but derived metrics are often not compared with 3D structural field measurements and the acquisition of ALS data is rarely standardized across different remote sensing surveys. Here, we compare a set of Light Detection And Ranging (LiDAR) metrics derived from ALS datasets with varying characteristics to a standardized set of field measurements of vegetation height, biomass and Leaf Area Index (LAI) across three Hungarian lakes (Lake Balaton, Lake Fertő and Lake Tisza). The ALS datasets differed in whether the recording type was full waveform (FWF) or discrete return, and in their point density (4 pt/m2 and 21 pt/m2). A total of eight LiDAR metrics captured radiometric information as well as descriptors of vegetation cover, height and vertical variability. Multivariate regression models with field-based measurements of vegetation height, biomass or LAI as response variable and LiDAR metrics as predictors showed major differences between ALS recording types, and were affected by differences in spatial resolution, temporal offset and seasonality between field and ALS data acquisition. Vegetation height could be estimated with high to intermediate accuracy (FWF ALS data only: R2 = 0.84; combination of ALS datasets: R2 = 0.67), demonstrating its potential as a robust indicator of 3D vegetation structure across different ALS datasets. In contrast, the estimation of biomass and LAI in these wetlands was sensitive to variation in ALS characteristics and to the discrepancies between field and ALS data in terms of spatial resolution, temporal offset and seasonality (biomass: R2 = 0.20–0.22; LAI: R2 = 0.08–0.30). We recommend the use of FWF ALS data within wetlands because it captures more vegetation structural details in dense reed and marshland vegetation. We further suggest that ecologists and remote sensing scientist should better coordinate the simultaneous and standardized acquisition of field and ALS data for testing the robustness of quantitative descriptors of vegetation cover, height and vertical variability within wetlands. This is important for establishing operational and spatially contiguous ALS-based indicators of 3D ecosystem structure across wetlands.

We analyzed the Corine Land Cover 2018 (CLC2018) dataset to reveal the correspondence between land cover categories of the CLC and the spectral information of Landsat-8, Sentinel-2 and PlanetScope images. Level 1 categories of the CLC2018 were analyzed in a 25 km × 25 km study area in Hungary. Spectral data were summarized by land cover polygons, and the dataset was evaluated with statistical tests. We then performed Linear Discriminant Analysis (LDA) and Random Forest classifications to reveal if CLC L1 level categories were confirmed by spectral values. Wetlands and water bodies were the most likely to be confused with other categories. The least mixture was observed when we applied the median to quantify the pixel variance of CLC polygons. RF outperformed the LDA’s accuracy, and PlanetScope’s data were the most accurate. Analysis of class level accuracies showed that agricultural areas and wetlands had the most issues with misclassification. We proved the representativeness of the results with a repeated randomized test, and only PlanetScope seemed to be ungeneralizable. Results showed that CLC polygons, as basic units of land cover, can ensure 71.1–78.5% OAs for the three satellite sensors; higher geometric resolution resulted in better accuracy. These results justified CLC polygons, in spite of visual interpretation, can hold relevant information about land cover considering the surface reflectance values of satellites. However, using CLC as ground truth data for land cover classifications can be questionable, at least in the L1 nomenclature.