Places of past conflicts and persistent objects that reflect such events often attract the attention of archaeological prospection which facilitates the construction of conflict narratives. Field prospection as a precise method for localization of aerial bombing craters (as an example of such persistent features) is a highly time- and resource-consuming task. Therefore, methods for automatic identification of such features are evolving. We present a comparison of three methods for possible automatic crater detection based on (a) extraterrestrial crater detection algorithms, (b) geomorphology-based edge extraction, and (c) image pattern recognition via a state-of-the-art convolutional neural network (CNN). All methods were preliminarily tested on a case study of eight Second World War (WWII) aerial bombing crater sites in NW Czechia via Airborne Laser Scanned LiDAR-derived digital terrain models with different spatial resolutions. We found that extraterrestrial crater detection algorithms and geomorphology-based edge extraction methods yield worse results given the standard indices of precision and recall. By comparison, the CNN method utilized for a particular task achieved satisfying results, predominantly with 0.5 m/px resolution (which is often available at the country level) of the input raster. Nevertheless, overall performance with this resolution varies significantly among the sites. Therefore, the quality and readability of the input data are crucial factors for the successful acquisition of precise ordinance location identification.

Cutting-edge techniques for analyzing archival aerial survey images are crucial in enhancing terrain analysis and landscape modeling. This research study is centered on examining the effects of different image parameters, namely contrast, brightness, and smoothing, on the terrain model creation process. The goal is to determine the ideal combination of these factors that will optimize the quality of the produced models and improve their precision for water management applications.The methodology includes collecting and preparing the images (by cropping) and conducting experiments to adjust the image parameters before creating terrain models to identify the best combination of adjustments.The study's results reveal a significant influence of modified parameters on the final terrain models. Boosting contrast or brightness in images can enhance model intricacy, but excessive tweaking of these parameters may decrease accuracy. Image smoothing has emerged as a crucial component in noise reduction and obtaining smoother terrain models.

During 2022, Bohemian Switzerland NP was affected by the largest wildfire in the Czech Republic throughout its modern history. The NP is formed by sandstone towers, deep narrow valleys and dense forests. From the 19th century onwards, Norway spruce and non-native Weymouth pine were massively planted here. A series of weather extremes in the last years caused an exceptional drought and consequent massive bark beetle outbreak and spruce die off, followed by the catastrophic event. Wildfires of such a dimension are rather uncommon in Central Europe, and this event therefore serves as a perfect model situation to study the role of species composition, bark beetle and water availability on the fire dynamics as well as the changes in biodiversity and natural succession after the disaster. Before the fire, the area was dominated by conifers, mostly standing dry after the bark beetle attack except along the water courses, and further formed of clear cuts, healthy deciduous beech forests and rocky outcrops. Pre-fire vegetation state, fire severity and post-fire regeneration were assessed using a combination of remote sensing sources. In particular, we used pre- and post-fire series of Sentinel-2 satellite MSS imagery, and acquisition of multispectral (MSS) and LIDAR data. The whole area was sampled from small aircraft TL232 Condor by three sensors - photogrammetric camera Hasselblad A6D-100c (ground sampling distance - GSD - 5 cm), MSS sensor MicaSense Altum (GSD 32 cm) and LIDAR RIEGL VUX 1-LR (13 points/m), and detailed sites were sampled using drone mounted sensors - MSS (MicaSense Altum, GSD 5 cm) and LIDAR (DJI L1). Forest composition and changes in health status were derived using a range of spectral indices and supervised classification. Fire severity and forest structure were derived using a combination of Lidar and optical point cloud, fisheye camera, ground sampling, and analysis of optical data (supervised classification, vegetation indices). Our research revealed that fire disturbance was low or none at native deciduous tree stands and waterlogged sites. On the opposite, it was more severe at dry bark-beetle clearings covered by a thick layer of litter as compared to standing dead spruce. We can infer that in places where many stems were only partly burned or the trees postponed the die-off, the fire went faster and the severity of disturbance was lower. In some cases, we could see patterns formed by ground fire, such as burned circles around trees or tree stools surrounded by unburned areas. Post-fire regeneration is very fast, and even after one year, vegetation growth can be detected using LIDAR and photogrammetric point clouds. Derived information on fire severity, detailed 3D stand structure and health status are to be used as a proxy of the fire disturbance impact on biodiversity and patterns of regeneration.

This study evaluates the accuracy of various preprocessing methods applied to scanned archival aerial photographs for reconstructing historical terrain in the Czech Republic. Seven workflows were tested on identical imagery and control points, varying parameters such as resolution unification, brightness normalization, focal length calibration, and AI-based denoising. Accuracy was assessed using GNSS checkpoints and high-resolution LiDAR data. Results show that basic brightness correction reduced the vertical RMSE by 59% (to 5.69 m). In contrast, standalone AI preprocessing was associated with increased geometric instability (RMSE 16.48 m) due to over-smoothing and the loss of essential micro-texture. However, the evaluated “Fusion AI” workflow—combining AI enhancement with strict focal length constraints—successfully mitigated this degradation. By restricting the internal orientation, it stabilized the vertical accuracy at 6.48 m, closely matching the best traditional approaches. Statistical analysis revealed strong spatial autocorrelation and non-normal error distributions, highlighting the need for robust validation. Ultimately, this study confirms that AI can be effectively utilized to enhance visual clarity in data-scarce historical reconstruction without sacrificing spatial reliability, provided it is strictly geometrically constrained. This offers an optimal compromise and a tested, reproducible workflow that supports heritage preservation and long-term environmental analysis.