The goal of this review is to present leading examples of current methodologies for extracting forest characteristics from full-waveform LiDAR data. Four key questions are addressed: (i) does ...full-waveform LiDAR provide advantages over discrete-return laser sensors; (ii) will full-waveform LiDAR provide valid results in support of forest inventory operations and allow for a decrease in ground sampling efforts; (iii) is the use of full-waveform LiDAR data cost effective; and (iv) what is the scope of the applied methods (i.e., is full-waveform LiDAR accurate for different forest compositions, structures, and densities, and is it sensitive to leaf-off/leaf-on conditions)? Key forest structure characteristics can be estimated with significant accuracy using full-waveform metrics, although methodologies and their corresponding accuracies differ. For example, some processing methods are valid at the plot scale, whereas other procedures perform well at the regional scale; to be effective, certain LiDAR data analyses require a minimum point density, whereas other methods perform well using large-footprint sensors. Therefore, it is important to match processing methods with the appropriate scale and scope. The aim of this paper is to provide the forest research community and remote sensing technology developers with an overview of existing methods for inferring key forest characteristics, including their applicability and performance.
Sentinel-5P (S5P) data provide information on atmospheric pollutants daily, and, for higher latitudes, consequent orbits partially overlap the same day. Provided clear atmospheric conditions, these ...data can provide insights on emission hotspots and on spatial distribution of critical air quality issues. The purpose of this work is to analyse several aspects of NO2 data from S5P over the years 2019, 2020 and 2021, in particular: (i) yearly average values between S5P data and 624 ground measurement stations were tested for correlation; (ii) 387 pairs of images from overlapping orbits on the same day were used to test for correlation on consecutive images with four different methods – simple linear regression over all valid cell values across the two images, over a subset with a low cloud fraction, and linear and tree-based methods using multiple predictors; (iii) local maxima values extracted from yearly NO2 emission maps were analysed to check potential hotspots of NO2 emissions.Results show that ground measurements correlate with S5P values, with r-squared values of 0.37 and 0.43 and RMSE of 7.4 and 8.6 µmol/m2 respectively for 2019 and 2020. Simple linear regression of overlapping consequent images returned average and standard deviation (sd) on r-squared respectively of 0.50(sd=0.21) and for RMSE of 11.3(sd=4.2) µmol/m2. Points from local maxima clearly detected 19 specific positions in large cities or nearby industrial areas, mostly in the north of Italy, with average NO2 values above 90 µmol/m2 in some cases consistently over the three years, proving that S5P imagery is a valid index for spatial distribution of NO2 concentration and air quality.
The human pressure over coastal areas is becoming increasingly relevant, due to the combinations of resource depletion, climate change effects and ocean eutrophication. Coastal ecosystems are so ...exposed to a huge number of stress factors that endanger their ecosystem services, like carbon uptake and biodiversity maintenance, that can be crucial in facing the effects of climate changes. With a particular focus on seaweeds, these ecosystems are becoming rapidly relevant both for carbon sinks and as a source of high value products, for example thanks to cosmetic and food industries that produce high added values products.In this contest the capability of conducting efficient monitoring is crucial to monitor environmental dynamics and resources trends. Traditionally seaweed monitoring was carried out with on field surveys that could be based on botanic analysis combined with genetic study, depending on the aims. Recently Remote Sensing techniques, combined with Artificial Intelligence ones, gave a new perspective to seaweed monitoring, introducing tools that are always more efficient.In this contest the present work aims to test the potentiality of remote sensing and artificial intelligence techniques for seaweed monitoring along the Irish west coast, building the basis for a fully automated tool for monitoring. The results showed that, with a supervised classification approach, it is possible to train Random Forest (RF) to perform very precise classification over the entire West Coast of Ireland. In particular, with all the RF configurations tested the Overall Accuracy (OA) was greater than 98.61, with the best performance obtained with the configuration Ntree = 600 and mtry = 2 that produced an OA = 98.87.
The study aims to compare land use land cover (LULC) change between Bangladesh and Indian Sundarbans from 1975 to 2020 using Landsat Satellite images. We performed supervised maximum likelihood (ML) ...to classify the study area at four time periods over 45 years (1975, 1990, 2005, and 2020). The classification was assigned to five classes: dense forest, moderate forest, sparse forest, barren land, and water body. Accuracy assessment of the classified images was completed with 250 control points for each year. The findings of our study revealed that the dense forest cover of Bangladesh and Indian parts was 54% and 31%, respectively, whereas, for the whole Sundarbans, it was 45% in 1975. However, the dense forest of Bangladesh and Indian Sundarbans decreased by an annual rate of 1.20% and 1.60%, respectively, from 1975 to 2020. From 1990 to 2005, Bangladesh Sundarbans slightly increased the dense forest cover by an annual rate of 0.68%, while the Indian Sundarbans decreased by an annual rate of 0.63%. The moderate dense forest of Bangladesh and Indian Sundarbans increased by giving almost the same annual rate of 3.62% and 3.59% from 1975 to 2020, whereas the increasing rate of the sparse forest was much higher for Bangladesh (8.36%) Sundarbans than Indian (3.36%) parts. The water bodies of Bangladesh and Indian Sundarbans increased by giving an annual rate of 0.48% and 0.71%, respectively, from 1975 to 2020. Our study found that most of the barren lands were located near the boundary between forest and human settlement of Indian Sundarbans compared to Bangladesh. The findings of the comparative assessment between these two countries can support sustainable forest management and planning by considering the best policy options.
Active remote sensing systems orbiting the Earth are only a small portion of the current constellation of satellites and will increase in number and advance in technology in the future. The launch of ...the GEDI sensor in December 2018, for an expected life-span period of about 2 years, is a fundamental step of this revolution, as it is the first spaceborne full-waveform lidar specifically designed for measuring the structure of ecosystems, providing information of the vertical profile of forests.Accuracy assessment of GEDI height metrics in the context of an Alpine forest environment in steep terrain scenarios has been conducted in this study. We used discrete return lidar from a recent aerial laser scanner survey as reference to analyse differences of heights of terrain elevation and maximum canopy height of the vegetation detected in each GEDI footprint. The height metrics differences between the discrete lidar and the GEDI data were then analysed to verify any correlation with the following factors: morphology (terrain slope), land cover (land cover type, fraction of canopy cover, vegetation density), GEDI laser beam characteristics (day/night-time acquisition, full power vs coverage laser beam, beam ID, laser sensitivity). Further analysis involved shifting the footprints’ location in 8 different direction and 4 distances to assess the impact of geolocation errors on accuracy and precision.Results show that what most influences accuracy in this study is the terrain slope, very likely linked to the uncertainty of geolocation of the GEDI footprints, suggesting caution in using single GEDI footprints if located in steep environments. Other than slope, terrain height accuracy varies mostly with forest type (conifer vs broadleaves), but not significantly with other factors. Canopy height instead is affected by most factors; high vegetation canopy is overestimated by ∼3 m in GEDI, and underestimated by 3 m over heath and bushes (median difference). Higher sensitivity pulses and night-time pulses provide better accuracy. Laser beams with full power also have better accuracy; beams with id 1000 and 1011 provide the most accurate canopy heights. Shifting the footprint position decreased accuracy except at 15 m and 270° with respect to orbit direction (left-looking).
In this work, an ensemble of machine learning algorithms was trained using stratified sampling from an existing European-scale biomass map from 2018 to predict an updated version for 2020. The ...objective of stratification is to make sure that the full range of biomass values is represented. The sampled biomass values from 2018 were filtered to remove areas that did were subject to forest disturbances between 2018 and 2020. This information was available from forest cover/loss/gain maps derived from satellite imagery. We train using a total of 49 features derived from the following sources: bioclimatic data, maps of land-cover, tree cover, tree height, annual composites of vegetation indices per pixel (EVI and NDVI) obtained from Sentinel-2, radar backscatter median annual values from Sentinel-1 and ALOS-2, and the ALOS DSM (3D) elevation grid. A model was created dividing Europe into 19 tiles to limit variability due to very different bioclimatic zones. The result is a raster with 100 m × 100 m resolution and an estimated value of biomass (Mg ha−1) at each node. Overall results on validation data over Europe report a root mean square error (RMSE) of 32.4 Mg ha−1 and a mean absolute error (MAE) of 21.5 Mg ha−1; when considering single tiles, the largest RMSE was 54.7 Mg ha−1 in tile D2, which can be explained by the very high variance of climate, environment, terrain topography and biomass values as the tile enclosed the Alpine region and the western part of Eastern Europe.
Remote sensing via orbiting satellite sensors is today a common tool to monitor numerous aspects related to the Earth surface and the atmosphere. The amount of data from imagery have increased ...tremendously since the past years, due to the increase in space missions and public and private agencies involved in this activity. A lot of these data are open-data, and academics and stakeholders in general can freely download and use it for any type of application. The bottle-neck is often not data availability anymore, but the processing resources and tools to analyse it. In particular multi-temporal analysis requires stacks of images thus digital space for storage and processing workflows that are tested and validated. Processing image by image is often not a viable approach anymore. Basic tools for multi-temporal analysis are provided via the same web interface, allowing the user to also apply parallel processing for a faster data extraction. A study case over burned areas in the north-eastern region of Italy are reported, to show how the multi-temporal analysis tools provided can be a valid source of data for further analysis. Multitemporal data consisting on the index values of each pixel inside user-defined areas can be downloaded in a spreadsheet that provides the values, the cell ids, the timestamp and the cloud and snow percentage. Also the full-resolution raster with index values that are rendered on screen can be downloaded as GeoTIFF at each specific date.
In this investigation a comparison between two machine learning (ML) models for semantic classification of an aerial laser scanner point cloud is presented. One model is Random Forest (RF), the other ...is a multi-layer neural network, TensorFlow (TF). Accuracy results were compared over a growing set of training data, using a stratified independent sampling over classes from 5% to 50% of the total dataset. Results show RF to have average F1 = 0.823 for the 9 classes considered, whereas TF had average F1 = 0.450. F1 values where higher for RF than TF, due to complexity in the determination of a suitable composition of the hidden layers of the neural network in TF, and this can likely be improved to reach higher accuracy values. Further study in this sense is planned.
Using multiple sources of 3D information over buildings to go from building footprints (LOD0) to higher LODs in CityGML models is a widely investigated topic. In this investigation we propose to use ...a very common 2.5D product, i.e. digital terrain and surface models (DTMs and DSMs), to test how much they can contribute to improve a CityGML model. The minimal information required to represents a 3 dimensional space in an urban environment is the combination of a DTM, the footprints of buildings and their heights; in this way a representation of urban environment to define LOD1 CityGML is guaranteed. In this paper we discuss the following research questions: can DTMs and DSMs provide significant information for modelling buildings at higher LODs? What characteristics can be extracted depending on the ground sampling distance (GSD) of the DTM/DSM? Results show that the used DTM/DSM at 1 m GSD provides potential significant information for higher LODs and that the conversion of the unstructured point cloud to a regular grid helps in defining single buildings using connected component analysis. Regularization of the original point cloud does loose accuracy of the source information due to smoothing or interpolation, but has the advantage of providing a predictable distance between points, thus allowing to join points belonging to the same building and provide initial primitives for further modelling.
Massive point clouds have now become a common product from surveys using passive (photogrammetry) or active (laser scanning) technologies. A common question is what is the difference in terms of ...accuracy and precision of different technologies and processing options. In this work four ultra-dense point-clouds (PCs) from drone surveys are compared. Two PCs were created from imagery using a photogrammetric workflow, with and without ground control points. The laser scanning PCs were created with two drone flights with Riegl MiniVUX-3 lidar sensor, resulting in a point cloud with ~300 million points, and Riegl VUX-120 lidar sensor, leading to a point cloud with ~1 billion points. Relative differences between pairs from permutations of the four PCs are analysed calculating point-to-point distances over nearest neighbours. Eleven clipped PC subsets are used for this task. Ground control points (GCPs) are also used to assess residuals in the two photogrammetric point clouds in order to quantify the improvement from using GCPs vs not using GCPs when processing the images.Results related to comparing the two photogrammetric point clouds with and without GCPs show an improvement of average absolute position error from 0.12 m to 0.05 m and RMSE from 0.03 m to 0.01 m. Point-to-point distances over the PC pairs show that the closest point clouds are the two lidar clouds, with mean absolute distance (MAD), median absolute distance (MdAD) and standard deviation of distances (RMSE) respectively of 0.031 m, 0.025 m, 0.019 m; largest difference is between photogrammetric PC with GCPs, with 0.208 m, 0.206 m and 0.116 m, with the Z component providing most of the difference. Photogrammetry without GCP was more consistent with the lidar point clouds, with MAD of 0.064 m, MdAD of 0.048 m and RMSE value of 0.114 m.