Accurate building height information at the individual footprint scale is essential for material stock accounting and post-disaster damage assessments yet remains difficult to obtain at city scale in the Global South where airborne LiDAR coverage is rare and commercial very high-resolution imagery is cost-prohibitive or unavailable. While recent works have demonstrated building height estimation using freely available Sentinel imagery, the resolution ceiling of resulting products is still coarse for material stock analysis. This study incorporates products derived from data freely accessible under scientific research licenses, TerraSAR-X StripMap and PlanetScope, alongside Sentinel-1 to predict building heights in a large city in Brazil. To account for the spatial autocorrelation in the training set, features from all sources are integrated in a geographically weighted random forest model, returning an RMSE of 5.34 m and R2 of 0.756 against a LiDAR reference dataset. Local feature importance showed predictor dominance to vary consistently across intra-urban contexts, with footprint geometry dominating for low-rise buildings, shadow-derived height for taller and more isolated structures, and spectral reflectance for the tallest buildings in the set. Sentinel-1 backscatter and InSAR occupy complementary spatial niches, with no single sensor uniformly preferable across the set. Results provide optioneering guidance and insight over satellite-derived products predictive relevance in distinct contexts, which global machine learning or neural network models cannot offer.
The near-worldwide coverage of spaceborne LiDAR data from the GEDI offers unprecedented opportunities for mapping canopy height (CH). Notwithstanding the sensitivity of the GEDI to forests’ vertical structure, it provides sparse sampling measurements, which hinder gap-free mapping. Several machine and deep learning models that resort to optical, radar, and GEDI have been tested to produce gap-free CH maps. Not all factors affecting the accuracy and consistency of these GEDI-fused products have been explored. Specifically, the sampling fraction coverage and the geolocation correction of footprints on-orbit positions have not been deeply studied. In this article, a collection of data from 15 study areas, characterised by Mediterranean landscapes, had their CH mapped resorting to Sentinel-1/2, ALOS-2, ancillary data, and a locally fitted extreme gradient boosting regressor. The produced maps had an average %RMSE of 32.41%, outperforming the other two global products in Mediterranean regions. Additionally, the geolocation correction of the GEDI footprints was limited to 1.23 percentage points. This experiment was able to demonstrate three key points: (1) the importance of including InSAR in the optical/SAR synergy; (2) the reduced impact of collocating GEDI footprints at the track level; and (3) the benefits of increasing GEDI-footprint coverage using multi-year data are hindered by temporal variation effects.
João E. Pereira-Pires, J. Guerra-Hernández, Adrián Pascual et al.· International Journal of Dig...· 0 citations
Accurate estimation of forest height from satellite imagery is essential for applications such as carbon accounting, biodiversity monitoring, and ecosystem management. While recent deep learning approaches provide accurate predictions, they typically do not quantify predictive uncertainty. This limitation is particularly relevant in geospatial settings characterized by sparse supervision and geographic distribution shift. In this work, we investigate Deep Evidential Regression (DER) for forest height estimation on the TreeUQ benchmark, a large-scale dataset designed for the joint estimation of tree count and average tree height at 10 m resolution, based on Sentinel-1/-2 data as well as tree inventory data over the federal state of Bavaria. To account for the extreme label sparsity of the tree inventory data, we introduce a masked evidential loss for dense geospatial prediction. Using a U-Net architecture with multimodal Sentinel-1 and Sentinel-2 inputs, the proposed approach jointly predicts tree height and associated uncertainty estimates in a single forward pass. Experimental results show that DER achieves predictive performance comparable to a deterministic U-Net while additionally providing well-calibrated uncertainty estimates. These findings demonstrate the potential of evidential learning as an efficient framework for uncertainty-aware forest structure estimation from Earth observation data.
Laura Bader, Muhammad Ammar Ahmed, Xiao Xiang Zhu et al.· 0 citations
Digital elevation models (DEMs) are essential for infrastructure design and flood modeling, yet publicly available DEMs typically exhibit vertical accuracies of one to three meters, insufficient for these applications. This study develops and systematically evaluates a constraint-based DEM enhancement approach that integrates sparse, high-accuracy light detection and ranging (LiDAR) reference data collected at bridge locations to improve regional DEM quality. The methodology employs nearest-neighbor spatial interpolation to generate elevation correction surfaces from constraint points (points of known accurate elevations), followed by Gaussian smoothing to preserve topographic continuity. The approach was evaluated using LiDAR data from 15 bridge locations in the Austin metropolitan area in Central Texas. Four spatial interpolation methods (nearest neighbor, inverse distance weighting, natural neighbor, and linear interpolation) were compared, with nearest neighbor achieving optimal performance with a 28.79% improvement in mean Root Mean Square Error (RMSE) within the study area. Gaussian filtering with an optimized smoothing parameter (
σ
=
0.3
m
) further enhanced accuracy, achieving a mean RMSE to 0.105 m. Spatial configuration analysis across six-, nine-, and 15-bridge configurations revealed critical dependencies: distributed constraint arrangements consistently achieved optimal accuracy with six to eight constraint locations, while clustered configurations (constraint points clustered inside the area of interest) with peripheral constraints exhibited performance degradation despite increasing constraint count. This performance resulted from nearest-neighbor interpolation’s reliance on geometric proximity without terrain similarity consideration. The findings provide practical guidance for transportation agencies and flood management programs seeking to leverage existing infrastructure-derived LiDAR surveys for regional DEM enhancement, demonstrating that strategic constraint placement throughout the area of interest is essential for maximizing performance.
Xinke Huang, R. S. Wilkho, N. Gharaibeh· Journal of Surveying Enginee...· 0 citations
Abstract. Reliable urban risk assessment requires accurate and up-to-date information on building characteristics, particularly construction year and height, which are often incomplete or unavailable in existing databases. This study presents a cloud-based methodology for the automatic estimation of these parameters using multispectral and very high-resolution Earth Observation (EO) data. The proposed approach integrates temporal analysis of multispectral satellite imagery (Sentinel-2 and Landsat) with photogrammetric processing of very high-resolution stereo imagery (Pléiades). Building construction year is estimated by detecting temporal changes in spectral indices using spline-based modeling and discrete-difference analysis, achieving an accuracy of better than ±3 years. Building height is derived from digital surface models generated from satellite stereo imagery, with a mean accuracy of less than 2 m relative to LiDAR reference data (~1.40 m). The methodology was implemented in a cloud computing environment (Google Earth Engine and Google Colab) and tested in the City of Zagreb, Croatia. Validation results show robust performance, with an F1-score of 0.819 for construction year estimation and strong agreement between EO-derived and LiDAR-based height values. The results demonstrate the potential of EO-based methods for scalable, reliable extraction of building information, thereby supporting improved urban risk assessment and decision-making.
M. Gašparović, Filip Radić, I. Gasparovic et al.· The International Archives o...· 0 citations
Abstract. Accurate building footprint extraction is critical for applications ranging from population estimation to disaster management. Although optical imagery provides detailed spectral information, it often struggles with shadows, occlusions, and background clutter in dense urban environments. Lidar data, by contrast, offer precise elevation and structural attributes but face challenges such as variable point density and noise. This study integrates multispectral imagery from the U.S. Department of Agriculture (USDA) National Agriculture Imagery Program (NAIP) with lidar-derived feature height and intensity from the U.S. Geological Survey (USGS) 3D Elevation Program (3DEP) to improve footprint extraction using a U-Net–based deep learning model. A six-band input stack (RGB, near-infrared, height, intensity) was developed, normalized, and tiled for training and evaluation against Microsoft Global Building Footprints (GBF). Results from the Houston, TX test site show that the six-band model achieved a precision of 0.86, recall of 0.88, F1 score of 0.87, and Intersection-over-Union (IoU) of 0.76, consistently outperforming four-band baselines by reducing false positives while maintaining sensitivity. Predictions on withheld Houston tiles confirmed strong within-region generalization, yielded a precision of 0.78, recall of 0.81, F1 score of 0.79, and IoU of 0.66. Qualitative analysis further revealed limitations stemming from both training label quality and vegetation–building confusion. These findings demonstrate the complementary value of integrating spectral and structural information for robust building footprint extraction and how domain adaptation strategies can be used to enhance cross-regional transferability.
Jung-Kuan Liu, Rongjun Qin, S. Arundel et al.· ISPRS Annals of the Photogra...· 0 citations
Abstract. High-resolution Digital Elevation Models (DEMs) are essential for urban flood modelling, where small elevation differences govern drainage and inundation extent. However, DEMs frequently contain hydrological inconsistencies: bridges, tunnels and culverts appear as artificial barriers disrupting flow continuity, while flood defence structures may be poorly represented at the available resolution. This paper presents an automated open-source Python pipeline for generating hydrologically conditioned DEMs by integrating classified airborne LiDAR data with OpenStreetMap (OSM) infrastructure features. The workflow is tested on a 16 km² area over Copenhagen city centre using a 2023 national LiDAR acquisition (13.5 pts/m²). A 0.5 m resolution DSM is generated from LiDAR ground and building classes via Inverse Distance Weighting (k=12, power=2, max radius 5 m), with Nearest Neighbour gap-filling. Hydrological conditioning applies four sequential operations: bridge burning at 108 footprints, tunnel enforcing at 53 shallow underpasses, culvert enforcing at 8 subsurface passages, and barrier rasterization raising 199 flood defence structures to their LiDAR-measured top-of-wall elevations. In total, 198,733 pixels were lowered (median 0.69 m) and 27,022 raised (median 3.00 m above ground), reducing the trapped depression volume in the DSM by 718,496 m³ (−5.0%). Vertical accuracy is assessed against the Danish national terrain model DHM/Terræn (NMAD = 0.066 m, LE90 = 0.265 m). The conditioned DEM feeds the CLEAR-EO urban flood simulation chain at the Danish Meteorological Institute; full hydraulic validation is outside the scope of this work. The pipeline is modular and transferable to other urban contexts with pre-classified LiDAR and OSM data.
Tommaso Destefanis, E. Durando, M. Oliveti et al.· The International Archives o...· 0 citations
A new machine-learning framework aims to improve the success rate of computational protein design while moving away from results that reproduce sequences found in nature.
MIT News · Artificial Intelligence· news.mit.eduAug 24, 2026
