CNN-Based Individual Tree Species Classification Using High-Resolution Satellite Imagery and Airborne LiDAR Data

Deep learning (DL) has shown promising performances in various remote sensing applications as a powerful tool. To explore the great potential of DL in improving the accuracy of individual tree species (ITS) classification, four convolutional neural network models (ResNet-18, ResNet-34, ResNet-50, and DenseNet-40) were employed to classify four tree species using the combined high-resolution satellite imagery and airborne LiDAR data. A total of 1503 samples of four tree species, including maple, pine, locust, and spruce, were used in the experiments. When both WorldView-2 and airborne LiDAR data were used, the overall accuracies (OA) obtained by ResNet-18, ResNet-34, ResNet-50, and DenseNet-40 were 90.9%, 89.1%, 89.1%, and 86.9%, respectively. The OA of ResNet-18 was increased by 4.0% and 1.8% compared with random forest (86.7%) and support vector machine (89.1%), respectively. The experimental results demonstrated that the size of input images impacted on the classification accuracy of ResNet-18. It is suggested that the input size of ResNet models can be determined according to the maximum size of all tree crown sample images. The use of LiDAR intensity image was helpful in improving the accuracies of ITS classification and atmospheric correction is unnecessary when both pansharpened WorldView-2 images and airborne LiDAR data were used.

Forest Structural Estimates Derived Using a Practical, Open-Source Lidar-Processing Workflow

Lidar data is increasingly available over large spatial extents and can also be combined with satellite imagery to provide detailed vegetation structural metrics. To fully realize the benefits of lidar data, practical and scalable processing workflows are needed. In this study, we used the lidR R software package, a custom forest metrics function in R, and a distributed cloud computing environment to process 11 TB of airborne lidar data covering ~22,900 km2 into 28 height, cover, and density metrics. We combined these lidar outputs with field plot data to model basal area, trees per acre, and quadratic mean diameter. We compared lidar-only models with models informed by spectral imagery only, and lidar and spectral imagery together. We found that lidar models outperformed spectral imagery models for all three metrics, and combination models performed slightly better than lidar models in two of the three metrics. One lidar variable, the relative density of low midstory canopy, was selected in all lidar and combination models, demonstrating the importance of midstory forest structure in the study area. In general, this open-source lidar-processing workflow provides a practical, scalable option for estimating structure over large, forested landscapes. The methodology and systems used for this study offered us the capability to process large quantities of lidar data into useful forest structure metrics in compressed timeframes.

Mining Waste Volume Estimation Using Airborne Lidar Data and Historical Maps: A Case Study of Tailing Piles in Szklary, Lower Silesia

Reliable feasibility analysis of potential exploitation of a mining waste deposit poses a great challenge. One of the most crucial parts of this process is the approximation of the deposit volume. In this case study we propose a novel method of tailing pile volume estimation using open remote sensing and cartographic data. For selected piles, the difference between the proposed and classical approach reach 50% of the pile volume, which is a significant change in the potential value of the deposit.

Parameter-Free Half-Spaces Based 3D Building Reconstruction Using Ground and Segmented Building Points from Airborne LiDAR Data with 2D Outlines

This paper aims to automatically reconstruct 3D building models on a large scale using a new approach on the basis of half-spaces, while making no assumptions about the building layout and keeping the number of input parameters to a minimum. The proposed algorithm is performed in two stages. First, the airborne LiDAR data and buildings’ outlines are preprocessed to generate buildings’ base models and the corresponding half-spaces. In the second stage, the half-spaces are analysed and used for shaping the final 3D building model using 3D Boolean operations. In experiments, the proposed algorithm was applied on a large scale, and its’ performance was inspected on a city level and on a single building level. Accurate reconstruction of buildings with various layouts were demonstrated and limitations were identified for large-scale applications. Finally, the proposed algorithm was validated on an ISPRS benchmark dataset, where a RMSE of 1.31 m and completeness of 98.9 % were obtained.

Near-Field High-Resolution Maps of the Ridgecrest Earthquakes from Aerial Imagery

High-resolution maps of surface rupturing earthquakes are essential tools for quantifying rupture hazard, understanding the mechanics of rupture propagation, and interpreting evidence of past earthquakes in the landscape. We present highly detailed maps of five portions of the surface rupture of the 2019 Ridgecrest earthquakes, derived from 5 cm per pixel aerial imagery and 2–20 cm per pixel unmanned aerial vehicle imagery. Our high-resolution maps cover areas of complexity and distributed deformation, sections in which strain is very localized, and areas where the rupture breaks through sediment and bedrock, ensuring sampling of the diverse rupture styles of this earthquake sequence. These maps reveal the near-field deformation of the surface rupture with a high level of detail, resolving the extent of secondary fracturing, lateral spreading, and liquefaction features that are below the resolution of airborne lidar data, field mapping, and geodesy. These data may serve as a machine learning training dataset, and offer opportunities for detailed kinematic analysis and high-resolution probabilistic displacement hazard analysis.

Estimating Historically Cleared and Forested Land in Massachusetts, USA, Using Airborne LiDAR and Archival Records

In the northeastern United States, widespread deforestation occurred during the 17–19th centuries as a result of Euro-American agricultural activity. In the late 19th and early 20th centuries, much of this agricultural landscape was reforested as the region experienced industrialization and farmland became abandoned. Many previous studies have addressed these landscape changes, but the primary method for estimating the amount and distribution of cleared and forested land during this time period has been using archival records. This study estimates areas of cleared and forested land using historical land use features extracted from airborne LiDAR data and compares these estimates to those from 19th century archival maps and agricultural census records for several towns in Massachusetts, a state in the northeastern United States. Results expand on previous studies in adjacent areas, and demonstrate that features representative of historical deforestation identified in LiDAR data can be reliably used as a proxy to estimate the spatial extents and area of cleared and forested land in Massachusetts and elsewhere in the northeastern United States. Results also demonstrate limitations to this methodology which can be mitigated through an understanding of the surficial geology of the region as well as sources of error in archival materials.