1 Introduction
Nature-based Solutions (NbS) have emerged as critical strategies for addressing global climate change, biodiversity loss, and environmental degradation
[1]. By leveraging ecosystem services, NbS address these challenges through the protection, restoration, and sustainable management of natural or modified ecosystems
[2–
3]. This approach offers high cost-effectiveness, significant co-benefits, and broad social acceptance
[4–
5]. However, effective NbS implementation faces a critical bottleneck: the need for high-precision 3D structural data to support spatial and environmental modelling
[6–
7]. Traditional data collection methods (e.g., field surveys, 2D remote sensing) exhibit limited precision and efficiency in complex terrains, dense vegetation, and urban environments, making it difficult to satisfy the requisite demand
[8–
10]. Particularly in critical applications such as stormwater management, urban heat island (UHI) mitigation, and ecological restoration, precise data on topography, vegetation structure, and surface characteristics are indispensable for supporting decision-making in site selection, scale control, and performance evaluation
[11–
13].
Light Detection and Ranging (LiDAR) technology offers a transformative solution to aforementioned data acquisition challenge. As an active remote sensing technology, LiDAR can penetrate vegetation canopies to acquire high-precision 3D structural information, including topography, canopy architecture, and surface features
[14]. LiDAR operates by emitting laser pulses and measuring return signals to calculate the distance between the sensor and the target, generating high-fidelity 3D point cloud data
[15]. This capability enables effective surface information acquisition in complex environments, including dense forests and urban areas
[16]. Furthermore, LiDAR technology has evolved into multiple operational platforms—including satellite (SLS), airborne (ALS), terrestrial (TLS), and mobile (MLS) laser scanning
[17]—providing comprehensive data support for quantitative NbS assessment and targeted design. In recent years, LiDAR applications in NbS have expanded across critical domains, including ecosystem restoration, stormwater management, and urban resilience enhancement
[18–
19]. However, the absence of a systematic review of LiDAR applications in NbS obscures the overall research trajectory, evolutionary patterns, and technical bottlenecks that constrain broader adoption.
To address these knowledge gaps, this study employed bibliometric analysis to systematically review LiDAR applications in NbS from 2000 to 2024. The specific objectives are to: 1) identify the core application domains of LiDAR in NbS and their technical characteristics; 2) analyze the primary bottlenecks and challenges in technical integration; and 3) propose technical development pathways oriented towards design practice. The findings would provide a comprehensive reference for promoting the integration of LiDAR technology with landscape planning and design, facilitating the digital and intelligent transformation of NbS practices.
2 Data and Methods
2.1 Data Acquisition
Data collection was conducted using the Web of Science Core Collection, employing a search strategy that combined LiDAR technology and NbS application scenario terms. Drawing upon established research frameworks
[20–
24], a comprehensive search system was constructed, encompassing four core dimensions (Table 1): 1) ecosystem restoration and biodiversity conservation, 2) hydrological regulation and flood control, 3) urban resilience and blue–green infrastructure, and 4) climate regulation and ecosystem service assessment.
The inclusion criteria were established as follows: 1) time span ranged between January 1, 2000 and December 31, 2024; 2) document types restricted to articles and review articles; and 3) language limited to English. Data retrieval was completed on October 6, 2025. By pooling the four search sets, a unified dataset was generated. The dataset was deduplicated based on unique Digital Object Identifiers (DOIs) and further validated through a preliminary manual review of titles and abstracts to ensure topical relevance. Ultimately, a total of 4,275 valid publications were confirmed and included to construct the dataset for macro-trend analysis.
2.2 Bibliometric Analysis Methods
This study employed CiteSpace and the Bibliometric R package to conduct a comprehensive bibliometric analysis
[25]. The analytical framework comprised three key procedures: 1) topological optimization of the co-occurrence network via the Pathfinder algorithm, which prunes redundant and non-significant links while preserving global network connectivity to delineate core research clusters; 2) mapping the technological trajectory through temporal citation analysis; and 3) systematically synthesizing the core functions, developmental trends, and challenges of LiDAR within the NbS domain. To mitigate the inherent limitations of quantitative metrics, this study incorporated secondary validation through supplementary literature review and expert consultation. This qualitative assessment focuses on technical application characteristics, workflow integration bottlenecks, and implementation pathways for planning and design practice, ensuring both scientific rigor and targeted relevance of the research conclusions.
3 Application Progress and Core Domains of LiDAR in NbS
3.1 Research Trends and Hotspot Evolution
3.1.1 Publication Trends
The temporal distribution of publications reveals three distinct developmental stages in LiDAR applications within the NbS domain (Fig. 1): 1) the technology accumulation phase (2000–2006), characterized by limited publication volume but the establishment of foundational technologies; 2) the rapid growth phase (2007–2015), marked by a significant surge in annual output that expanded the disciplinary influence and broadened cross-domain applications; and 3) the high-quality development phase (2016–present), where annual publication volume has consistently exceeded 200 papers and maintained a continuous upward trend, indicating the progressive maturation of the field. Geographically, North American and European nations maintain a dominant position in publication output. The USA ranks first with 1,306 publications and exhibits the highest betweenness centrality in the international collaboration network. This metric underscores the pivotal hub role of the USA in connecting disparate research clusters and integrating global scientific resources. China ranks second with 758 publications, with its volume rapidly approaching that of traditional research powerhouses, signifying the swift emergence of Asia-Pacific nations. European nations such as England, Germany, and Spain demonstrate high centrality despite trailing the USA and China in total volume, highlighting their critical bridging function within the global collaborative network.
3.1.2 Research Hotspot Evolution
The temporal visualization of keywords reveals a distinct development trajectory in LiDAR application within the NbS domain, characterized by a "technology–application–intelligence" progress (Fig. 2). Corresponding to the initial technology accumulation phase, research primarily focused on digital elevation model (DEM) generation and topographic mapping (cluster #2), aiming to resolve fundamental technical issues related to the precise acquisition and filtering of surface 3D information and laying the data foundation for subsequent applications. As the field entered the rapid growth phase, research hotspots extended toward ecological parameter retrieval, forming application clusters represented by forest inventory (cluster #0) and sediment transport (cluster #1). This phase marked a transition from single structural parameter extraction to complex biomass estimation and habitat assessment. In the current high-quality development phase, keyword clusters such as deep learning (cluster #3) and UHI (cluster #7) have become dominant in the high-frequency co-occurrence network. This shift indicates that current research is accelerating toward intelligent and refined directions, including semantic segmentation, multi-source data fusion, and urban resilience modeling.
3.2 Identification of Core Application Domains
3.2.1 Cluster Analysis and Visualization
The dataset was analyzed using annual time slicing, with cited references and keywords serving as network nodes. The g-index threshold was set at
k = 25
[25]. As an enhanced algorithm over the h-index, the g-index accounts for the cumulative contribution of highly cited papers, assigning greater weight to high-impact nodes. This approach filtered out low-impact information while preserving core knowledge units within each time slice. Based on this parameter setting, the initial co-citation network was constructed and optimized via the Pathfinder algorithm. The resulting network comprises 1,196 nodes and 1,789 links, from which 19 sub-clusters with a size greater than 30 were extracted. The network modularity (
Q = 0.9045) and weighted mean silhouette score (
S = 0.9442) indicate that the clustering structure is highly reliable and robust (Fig. 3).
3.2.2 Core Application Domains
Based on Latent Semantic Indexing (LSI) labels and Log-Likelihood Ratio (LLR) optimization, combined with an analysis of highly cited references and burst years, the sub-clusters were synthesized into four core application domains (Table 2). 1) Point cloud analysis and ecological structural foundations: this domain focuses on critical processes such as semantic segmentation, point cloud segmentation, and terrain reconstruction using airborne, terrestrial, and mobile LiDAR data, providing structured data support for high-precision 3D modeling of watershed and urban ecosystems. 2) Vegetation structure and ecological assessment: this domain encompasses aboveground biomass retrieval, tree species classification, canopy structure, and biodiversity assessment, achieving the quantitative characterization of vegetation from morphological features to ecological functions, and represents a core direction for evaluating the ecosystem service benefits of NbS. 3) River restoration and stormwater management: underpinned by LiDAR-supported topographic mapping and hydrological modeling, this domain focuses on floodplain evolution, flood risk analysis, and watershed ecological restoration, emphasizing disaster prevention and mitigation strategies driven by natural processes. 4) Urban resilience and blue–green infrastructure (BGI) optimization: this domain centers on UHI effects and climate regulation mechanisms, utilizing LiDAR data to quantify urban green volume, ventilation corridors, and microclimate characteristics, and provides a scientific basis for optimizing BGI layout and climate-adaptive design.
3.3 Technical Analysis of Representative Application Domains
3.3.1 Point Cloud Analysis and Ecological Structural Foundations
(1) Point cloud processing and terrain modeling
LiDAR terrain modeling typically adheres to a technical workflow of "data preprocessing–feature extraction–3D modeling," transforming raw point cloud data into digital models suitable for NbS applications (Fig. 4).
1) Data preprocessing phase. Based on raw point clouds, this phase constructs the digital terrain model (DTM), encompassing noise removal and ground point separation
[16]. Noise removal eliminates outliers and anomalies caused by sensor errors, environmental interference, or data acquisition constraints, thereby enhancing overall data quality. Ground point separation utilizes algorithms to distinguish ground points from non-ground points (e.g., vegetation, buildings, vehicles), preserving the authentic undulations of the terrain surface. Common algorithms include Cloth Simulation Filtering (CSF)
[26–
27], Progressive Morphological Filter (PMF), and Simple Morphological Filter (SMRF)
[28]. However, manual intervention and editing remain necessary in areas with abrupt terrain changes, complex feature types, or severe noise interference.
2) Feature extraction phase. This phase transforms the geometric and reflectance attributes of point clouds into structured information. Traditional methods typically rely on geometric features such as normal vectors, curvature, and height thresholds for preliminary separation
[29–
30]. Data-driven supervised learning algorithms, e.g., Random Forest (RF), Support Vector Machine (SVM), k-Nearest Neighbors (kNN), Decision Trees, can integrate multidimensional features, including intensity, return number, and waveform width, to accomplish finer-grained classification and semantic segmentation
[31–
34]. Moreover, deep learning approaches can directly process unstructured point sets, improving robustness in complex scenarios such as urban environments and under-canopy areas
[35].
3) 3D modeling phase. DEMs are generated via Delaunay triangulation or voxelization
[27,
36]. Parametric variables such as canopy height models (CHM), roughness, canopy porosity, and slope can be derived using DEMs, providing a 3D data foundation for hydrological analysis and ecological assessment
[14,
37–
39]. In contrast to traditional 2D imagery, LiDAR's active ranging and high-density sampling maintain geometric continuity within complex environments, significantly enhancing data quality. At the macro scale, airborne LiDAR-derived DEMs are extensively applied in regional ecological restoration
[19,
40–
41], natural disaster assessment, and flood simulation coupled with hydrodynamic models
[42–
44]. At the micro scale, high-precision DTMs/DEMs provide sub-meter or even millimeter-level spatial references for infrastructure layout, vertical design, earthwork calculation, and construction stakeout
[16,
45].
(2) Characterization of land cover features
The refined characterization of land cover and ecological features serves as a pivotal link bridging 3D structure and ecological function, typically achieved through multi-source data fusion. The integration of LiDAR and hyperspectral imaging (HSI) consolidates geometric and spectral information. By leveraging the spectral dimension for species discrimination and surface moisture detection, this approach enhances identification accuracy and model generalization in areas with complex vegetation and mixed surfaces
[46–
47], achieving robust land cover recognition across multiple scales
[48–
49].
LiDAR-based land cover analysis technologies can effectively address the practical demands of NbS governance. At the macro scale, this technology enables the precise classification of complex heterogeneous surfaces, e.g., urban–rural transitional zones
[50]. Combined with multi-temporal data to monitor dynamic ecological processes such as secondary forest succession
[51], it provides high-precision baseline data and continuous monitoring evidence for simulating urban expansion boundaries and delineating ecological protection red lines
[52–
53]. At the micro scale, the synergistic integration of multi-source remote sensing data enables the accurate retrieval of soil moisture and erosion dynamics
[10,
32,
39]. This facilitates the identification of key site-scale units for soil and water conservation, providing a scientific basis for the layout and optimization of low impact development (LID) facilities, e.g., ecological swales and rain gardens
[11].
3.3.2 Vegetation Structure and Ecological Assessment
(1) Urban vegetation structure monitoring
Within the NbS framework, LiDAR technology has emerged as a fundamental pillar for the inventory and dynamic management of urban vegetation structures. Unlike traditional 2D spectral remote sensing, LiDAR's active ranging capability enables effective penetration of canopy layers to acquire comprehensive 3D structural information
[20,
54]. This technological advantage facilitates consistent output of multi-scale structural data ranging from individual tree to community, providing a reliable data foundation for subsequent ecosystem service assessments
[55–
57].
At the individual tree scale, algorithms such as density-based clustering and region growing effectively facilitate single-tree segmentation and crown reconstruction
[58–
59]. Recent advances in deep learning, particularly Graph Convolutional Networks (GCNs), have significantly improved the accuracy of tree species identification
[60]. The incorporation of multi-source imagery and novel segmentation frameworks (e.g., Segment Anything Model) has further enhanced the system adaptability to complex urban scenarios
[61]. Additionally, multi-view scanning techniques combined with crowdsourced annotation methods have expanded survey coverage of urban green spaces while significantly improving model robustness and cross-domain transferability
[62–
63].
Regarding structural parameters and biomass retrieval, LiDAR technology enables direct and stable extraction of core structural parameters, including tree height, diameter at breast height (DBH), leaf area index (LAI), and CHM
[64–
65]. Through DBH prediction models and LAI quantification methods based on canopy profiles
[66–
69], high-precision structural data layers can be constructed
[70–
72], providing a unified data basis for biomass estimation at both individual tree and stand levels. Besides, the integration of multi-platform data (e.g., MLS–ALS) significantly reduces uncertainties caused by occlusion, scale discrepancies, and sampling angle variations compared with single-platform observation. The synergistic fusion of LiDAR and hyperspectral features demonstrates superior explanatory power and robustness in species/functional group discrimination and structure–physiology coupling analysis, reflecting a technological trend toward "geometry–spectrum–scale" integration
[19,
73].
For carbon sequestration and temporal monitoring, the combined application of LiDAR and high-resolution remote sensing effectively maps the spatial distribution patterns of carbon density
[74]. Sub-pixel modeling techniques combined with biomass expansion factors (BEF) significantly mitigate the impact of mixed pixel effects
[75]. Multi-temporal point clouds shift carbon cycle assessment from static analysis to dynamic time-series analysis, offering deep insights into the response mechanisms between urban forest carbon storage and climatic disturbances
[76]. Additionally, emerging 3D carbon models have achieved a preliminary "observation–simulation–validation" cycle through parameter transfer and uncertainty optimization, providing critical data support for climate-adaptive planning and urban carbon neutrality pathway analysis
[77].
(2) Ecosystem service assessment
Building on refined vegetation structure characterization, current research increasingly quantifies LiDAR-derived 3D structural variables as ecosystem service indicators. These indicators delineate key dimensions (e.g., risk regulation, habitat provision, health exposure, social equity) and are integrated into decision-making workflows to underpin site prioritization, scale control, and performance evaluation for NbS
[63,
78–
79]. LiDAR serves not merely to visualize vegetation structure but to establish a traceable "structure–function–governance" linkage, enabling the spatialized and quantitative integration of ecosystem services into urban and regional governance practices
[78].
For risk prevention and resilience management, LiDAR-derived 3D forest models play a pivotal role in urban forest planning and disaster management
[57]. Leaf-off point cloud data significantly enhances vertical structure analysis precision by reducing parameter uncertainty
[80]. The dynamic mapping of coarse woody debris (CWD) using bi-temporal point clouds provides temporal evidence for targeted renewal and patrol strategies, improving the resilience and timeliness of urban forest governance
[81–
82]. By integrating spatial metrics such as canopy cover and patch connectivity, fire risk models for the wildland–urban interface can spatially visualize ecological safety assessments
[83].
Regarding habitat provision and biodiversity, LiDAR's paramount value lies in revealing vertical canopy stratification, gap patterns, and spatial heterogeneity across various heights. These structural features are significantly correlated with species richness, habitat quality, and habitat preferences of terrestrial fauna, particularly birds
[84–
85]. Within urban parks, corridors, and fringe habitats, LiDAR-derived metrics (e.g., canopy height distribution, layer thickness, horizontal heterogeneity) offer superior explanatory power for bird species richness and distribution patterns compared with traditional 2D coverage indices, effectively identifying green patches that structurally approximate high-quality habitats
[58,
86–
87]. Specifically, stands exhibiting distinct vertical stratification alongside horizontal heterogeneity provide multifunctional niches for shelter, foraging, and nesting, thereby enhancing overall biodiversity
[84–
85,
87].
For health benefits and environmental equity, LiDAR concurrently characterizes environmental stressors and spatial distribution of ecosystem services, directly linking public health and social vulnerability with vegetation structure
[78–
79]. Regarding thermal exposure, LiDAR characterizes 3D features (e.g., tree height, canopy thickness, continuity) at the block scale, enabling quantitative comparison with land surface temperature (LST) and near-surface thermal loads
[88]. Neighborhoods with taller, contiguous canopies exhibit more pronounced cooling effects than those explained by 2D green coverage alone. Additionally, LiDAR-derived individual tree parameters (e.g., height, crown width, lean angle) can be integrated with species traits and local meteorological data to identify potentially unstable trees and high-risk green infrastructure under extreme weather, providing a spatial basis for precision maintenance and renewal prioritization
[89–
90]. Urban-scale analysis reveals that low-income or historically marginalized communities often feature lower tree heights, fragmented canopy continuity, and diminished shading capacity, placing them at a systemic disadvantage regarding heat exposure, environmental stress, and accessibility to buffering services
[91]. LiDAR's significance lies in both unveiling and quantifying these disparities and localizing them to specific units, providing a targeted spatial decision-making foundation for NbS interventions.
3.3.3 River Restoration and Stormwater Management
(1) River geomorphological restoration
Floodplain systems are characterized by the inextricable "geometry–connectivity–habitat" coupling, where LiDAR technology serves as a foundational tool for river ecological restoration. LiDAR's active ranging capability and high-density 3D sampling enable the stable reconstruction of baseline topography in river reaches obscured by dense canopy cover or complex gravel micro-topography. This provides a consistent data foundation for identifying connectivity bottlenecks and estimating retention potential
[92–
93]. Specifically, applying hydrologic enforcement to multi-platform DTMs effectively eliminates spurious sinks and restores continuous flow paths. This process rectifies connectivity misinterpretations caused by artificial barriers (e.g., roads, levees) being incorrectly identified as impediments to flow
[92]. On this basis, key metrics, including sinuosity, width-to-depth ratio, bank breaklines, and roughness, can be extracted to precisely locate lateral depressions, paleo-channels, and floodways requiring priority reconnection
[93–
94]. Furthermore, by leveraging canopy height and gap distribution to characterize the vertical structure of riparian zones, the "morphology–habitat" coupling degree can be quantified using the probability of connectivity and integral indices of connectivity, effectively identifying severe fragmentation points
[94].
Regarding the spatial scaling of restoration, LiDAR-constrained relationships between threshold water levels and inundation, along with that between area/volume and water level, enable direct estimation of available flood storage volumes and peak reduction potential under varying design water levels. These metrics facilitate reverse calculation of geometric control parameters such as setback width, excavation depth, and bench elevation, providing a scientific basis for site selection and scaling of interventions including levee setbacks, lateral expansion, compound wetlands, reconnected channels, and ecological embankments
[93,
95–
96]. Compared with traditional 2D imagery or sparse cross-section extrapolation, this LiDAR-based quantitative workflow significantly shortens the inference chain from morphology to function, reduces reliance on empirical roughness coefficients and coarse-resolution DEMs, and enhances the scientific rigor and operability of restoration schemes.
For performance verification and adaptive management, multi-temporal LiDAR facilitates the comparison of sediment erosion and deposition volumes and bankline retreat at both event-based and inter-annual scales. Furthermore, point cloud-based statistical tests (e.g., M3C2-EP) are employed to assess the statistical significance of topographic changes, distinguishing authentic morphological evolution from measurement errors
[95]. The integration of topo-bathymetric LiDAR and full-waveform ALS for inundation extent identification enables evidence-based reconstruction of flood processes and the verification of connectivity pathways
[96–
97], supporting the long-term monitoring and adaptive management of river ecological restoration.
(2) Urban stormwater management
NbS advocates a stormwater management philosophy anchored in source reduction, on-site infiltration, hierarchical retention, and peak attenuation
[98]. LiDAR technology serves as a foundational tool for urban flood simulation and regulation by translating micro-topography, underlying surface characteristics, roughness, and canopy structure into physical parameter layers that can be seamlessly integrated into hydrodynamic models. This significantly mitigates the reliance on empirical parameterization and low-resolution terrain data
[97,
99]. Hydrologically enforced LiDAR-DEMs enable the identification of curbs, stormwater inlets, micro-depressions, and drainage channels at the centimeter-to-decimeter level. When integrated with CHM and land cover information to derive key parameters such as Manning's roughness, imperviousness, and canopy interception, this approach significantly enhances the simulation of flow paths, inundation duration, and retention nodes
[11,
99].
For model integration, LiDAR-derived parameters can be coupled with hydrologic models (e.g., SWMM, HEC-HMS) and 1D/2D hydrodynamic models (e.g., HEC-RAS, CityCAT) to establish a comparative assessment workflow
[11,
97,
99]. Hydrodynamic models driven by LiDAR-constrained DTMs and digital surface models (DSMs) significantly outperform traditional remote sensing topographic basemaps in reconstructing inundation extents, peak flow processes, and block-scale flow paths
[10,
99]. Supplementary local elevation data from LiDAR can further elevate precision and interpretability of ultra-high-resolution scenario assessments
[97,
99]. Simultaneously, the direct parameterization of roughness and sub-grid topographic variability from LiDAR geometry incorporates the coupling of "morphology–roughness–hydraulic resistance" into the computational framework, improving stability and reliability of 2D hydrodynamic solutions
[11,
100].
In terms of LID and sustainable drainage systems (SuDS) configuration and optimization, LiDAR-driven parameters can drive scenario analysis at both street and catchment scales. This facilitates comparative analysis of performance improvements—such as peak reduction, lag time increase, and node overflow frequency—offered by interventions like rain gardens, sunken green belts, permeable pavements, and retention ponds. Utilizing a multi-objective Pareto framework, this process outputs a definitive list of critical vulnerability and priority intervention units
[99]. Overall, by leveraging technical advantages of parameter traceability, indicator quantifiability, and scalar interoperability, LiDAR propels urban NbS stormwater governance from empirical practice toward an structure-framed and evidence-based engineering-oriented framework.
3.3.4 Urban Resilience and BGI Optimization
(1) UHI effect and climate regulation
For NbS-oriented urban climate optimization, the critical objective is to translate 3D geometric structural information directly into interpretable thermal environment control variables, not merely to increase green coverage. LiDAR technology provides essential technical support for mitigating the UHI effect and regulating local climates. DSMs, DTMs, and CHM derived from airborne or terrestrial LiDAR facilitate the stable derivation of key parameters, including sky view factor (SVF), canopy height/thickness, street canyon aspect ratio, and roughness. When integrated with thermal infrared (TIR) data and building vector data, these parameters enable mechanistic interpretation and correction of LST and near-surface air temperature
[101–
104]. Specifically, utilizing LiDAR-derived SVF to refine the SVF-based urban effective emissivity (UEM-SVF) and LST retrieval accuracy significantly mitigates biases induced by the radiation trapping effect within street canyons
[101]. Furthermore, continuous SVF fields computed by LiDAR demonstrate high consistency with hemispherical photography, maintaining spatial continuity and methodological reproducibility in complex, mixed vegetation–building environments
[105]. When coupled with airborne thermal imagery, LiDAR also enables error control in single- or multi-band temperature retrieval
[104].
For vegetation cooling pathways, vertical canopy structure can be directly measured by LiDAR. A case study in Tampa, Florida, revealed that LiDAR-extracted vegetation height possesses greater explanatory power for LST than 2D coverage, indicating that the vertical height of trees is more determinant of cooling magnitude than horizontal canopy extent
[88]. For building morphology, the joint analysis of LiDAR and airborne TIR data demonstrates a significant correlation between 3D morphological metrics (e.g., building height, volume, fractal dimension) and LST. This correlation exhibits heterogeneity across different functional zones, providing a quantitative basis for zoning and regulation
[102].
Overall, the technical advantages of LiDAR in mitigating UHI and regulating climate are manifested in three aspects: 1) direct vertical structure observation, where SVF, shading, and canopy layering are derived from a unified measurement system, minimizing cross-source co-registration errors
[101,
103–
104]; 2) shortened parameter–performance linkage, which allows design variables (e.g., aspect ratio, tree height, canopy gaps) to be directly mapped to temperature and thermal exposure indicators
[88,
102]; and 3) cross-scale scalability, enabling a unified parametric framework to be extrapolated from the block scale to district and city scales, facilitating identification of hotspots and priority intervention units
[88,
102,
104].
(2) Ventilation corridors and ecosystem services
Ventilation corridors are conceptualized as critical spatial networks connecting source areas (with strong cooling and purification capacities) to high-heat, high-exposure neighborhoods, aiming to enhance the dilution and transport efficiency of heat and pollutants within high-density built environments
[106–
107]. LiDAR's core contribution lies in providing a homologous data foundation for 3D morphology and vegetation structure. This enables the construction of roughness/resistance fields and SVF/openness layers, reducing reliance on empirical parameterization and low-resolution data for corridor identification
[108–
109]. A case study in Hong Kong, China, utilized airborne LiDAR to calculate the frontal area index (FAI) of buildings and trees, combined with least cost path (LCP) analysis, identified optimal ventilation corridors under dominant wind directions. Existing comparisons with computational fluid dynamics (CFD) results demonstrated consistency in corridor location and ventilation accessibility, validating the reliability and operability of the LiDAR-based method
[109].
Regarding data and assessment, LiDAR-derived building height fields and canopy thickness can be directly parameterized into ventilation resistance grids, while SVF supplements radiative openness information, jointly constructing ventilation potential maps and resistance maps
[110–
111]. Furthermore, novel metrics such as the tree canopy view factor (TC-VF) and potential impact intensity grade (PIIG), calculated from airborne LiDAR 3D point clouds, have shown significant correlations with thermal risk and energy consumption
[112]. These metrics measure the localized interplay of shading, heat dissipation, and ventilation, thereby facilitating the identification of high-resistance bottlenecks, ventilation interruptions, and candidate corridors with modification potential. Consequently, this provides a spatial basis for removing aerodynamic obstacles, connecting flow paths, and expanding ventilation inlets in cold source areas (e.g., waterfronts, open green spaces)
[112].
In terms of monitoring and verification, Doppler LiDAR can reconstruct urban vertical wind profiles, enabling site-scale independent verification and model calibration for corridor accessibility and cooling efficacy, significantly enhancing the evidence base for corridor planning
[113]. Furthermore, LiDAR-based 3D footprints of buildings and tree canopies facilitate the rapid construction of roughness and openness indices required for urban climate maps (UCM). These indices support corridor alignment, prioritization, and phased implementation between waterfront cold sources, open green spaces, and high-heat exposure blocks
[114,
105].
4 Challenges and Prospects of LiDAR in NbS Applications
4.1 Technical Challenges and Constraints
4.1.1 Data Quality and Precision Limitations
The foremost challenge confronting LiDAR technology in NbS applications pertains to data quality and precision. In complex terrains, point cloud preprocessing procedures (e.g., filtering, resampling, surface fitting) often result in the loss of critical geometric information. Specifically, filtering algorithms, while removing noise, may inadvertently eliminate micro-topographic features sensitive to hydrological processes, including flow paths, micro-depressions, and terraced structures. This "over-smoothing" phenomenon directly compromises the accuracy of hydrological models, introducing systematic biases when translating geometric morphology into ecological functions
[10]. Furthermore, data acquisition under dense vegetation cover presents inherent limitations. Restricted laser penetration within dense forests or multi-layer canopies constrains extraction of parameters (e.g., individual tree identification, crown delineation, DBH, tree height), which rely on penetration rates, point density, and waveform characteristics. Even with advanced deep learning segmentation models, complete elimination of omission and commission errors remains unachievable
[8]. This data incompleteness restricts the efficacy of LiDAR in ecological assessments.
4.1.2 Multi-source Data Fusion Barriers
Multi-source remote sensing data fusion is essential for enhancing the efficacy of LiDAR applications; however, it faces significant technical challenges. The synergistic application of LiDAR with hyperspectral or multispectral, TIR, and synthetic aperture radar (SAR) data necessitates resolving fundamental issues such as geometric registration, scale matching, and feature space mapping
[115]. Discrepancies in data formats, spatial resolutions, and temporal baselines across different sensors underscores a lack of unified fusion standards and uncertainty assessment frameworks. More critically, there is a deficit in standardized methodologies for validating and accurately assessing of fused data products. Consequently, quantifying data uncertainty and establishing robust confidence propagation mechanisms have become technical bottlenecks constraining the effective application of multi-source data in NbS.
4.1.3 Semantic Understanding and Automated Processing Bottlenecks
The semantic understanding of point cloud data remains a core challenge in LiDAR applications. Because raw point clouds lack inherent semantic information, feature recognition and classification require complex feature engineering and extensive manual criteria
[116]. Although deep learning technologies have significantly improved automation in point cloud processing, model training still relies heavily on massive, high-quality annotated datasets, and cross-scene generalization capability remain limited
[30,
117–
118]. In NbS applications, this semantic gap hinders the direct translation of point cloud data into design parameters. Throughout the "point cloud–semantics–design indicators" chain, information loss and uncertainty accumulation at each stage comprises the reliability and efficiency of data-driven design.
4.2 Workflow Integration and Engineering Challenges
4.2.1 Fragmented Toolchains and Data Silos
Current LiDAR applications are hindered by highly fragmented toolchains. Point cloud data processing and analysis rely heavily on specialized remote sensing software (e.g., CloudCompare, MeshLab), whereas planning and design workflows predominantly operate within CAD/BIM environments
[119–
120]. This fragmentation often leads to format incompatibility, precision degradation, and semantic information loss during data migration across platforms. Although standard frameworks such as CityGML, Industry Foundation Classes (IFC), and GeoBIM are advancing, technical difficulties persist regarding geometric consistency, topological integrity, and semantic mapping
[119–
120]. The absence of unified data exchange standards renders LiDAR-derived parameters difficult to reuse directly within design software, thereby impeding data-driven design iteration and optimization.
4.2.2 Cost-Benefit and Accessibility Challenges
Cost remains a pivotal factor constraining the widespread adoption of LiDAR. The acquisition costs for high-precision LiDAR systems range from tens of thousands to hundreds of thousands of yuan. When compounded with the expenses of specialized personnel training and computational resources for data processing, these costs become prohibitive for small-to-medium-scale projects
[121]. The data processing phase necessitates high-performance computing infrastructure and professional software; and the entire workflow may span weeks to months, making it challenging to meet the rapid-response demands of planning and design projects. Furthermore, the high professional threshold of LiDAR technology necessitates operators with multidisciplinary backgrounds in remote sensing, geographic information science, and computer vision. This skill requirement restricts the broader application of the technology, particularly within traditional planning and design domains.
4.3 Future Prospects and Development Pathways
Building on the preceding analysis and practical insights, this section outlines the future trajectory of LiDAR technology in the NbS domain, focusing on data platform construction, intelligent algorithm research, and interoperability enhancement. Recommendations for professional capacity building and cross-disciplinary collaboration are also proposed.
4.3.1 Open Data and Artificial Intelligence Empowerment
Looking forward, constructing open data sharing platforms, utilizing artificial intelligence (AI) to empower point cloud processing, and unifying data exchange standards are critical for promoting the widespread application of LiDAR technology in the NbS domain (Fig. 5).
Open data sharing is a vital pathway to reducing barriers to entry for LiDAR technology, necessitating concerted efforts from governments, academic institutions, and industry stakeholders
[122]. Currently, platforms such as OpenTopography provide open-access LiDAR data covering diverse landforms alongside online processing services
[123]. Additionally, city-scale open point clouds like DublinCity offer benchmark data for multi-scale analysis
[124]; while high-quality benchmark datasets such as WHU-TLS have advanced algorithm evaluation and model training
[125]. Future efforts should focus on constructing a more comprehensive open data ecosystem, encompassing data standardization, quality control, and metadata management.
AI technologies are key to address the challenges in LiDAR application. Deep learning models can automatically recognize landscape elements such as vegetation, buildings, and pavements, significantly reducing reliance on specialized skills
[18]. For instance, large-scale annotated benchmarks like Semantic3D and Paris-Lille-3D have facilitated standardized evaluation for segmentation or classification in complex urban scene
[18]. Moreover, few-shot learning and transfer learning techniques can enhance the cross-scene generalization capabilities of point cloud segmentation and classification models
[30]. End-to-end point cloud networks, such as PointNet/PointNet++, have reduced dependency on traditional feature engineering
[126]. Future research should prioritize weakly supervised and self-supervised learning methods to minimize the need for extensive manual annotation, thereby enhancing model practicality and scalability.
Establishing unified data standards and interface specifications is fundamental to promoting LiDAR technology adoption. Lightweight semantic mapping based on CityGML/CityJSON and IFC can advance GeoBIM workflows, enabling the seamless integration of LiDAR-derived parameters into design software
[119–
120]. This process effectively links manipulable spatial control variables with quantifiable performance indicators, allowing design adjustments to be tracked, compared, and validated. The next critical step lies in unifying cross-platform data standards and APIs to ensure interoperability among LiDAR data, environmental assessment models, and design software, thereby reducing format conversion costs and enhancing decision-making transparency.
4.3.2 Professional Capacity Building and Cross-Disciplinary Collaboration
To implement NbS effectively, the landscape architecture profession must cultivate talents who combine ecological and computational thinking with digital proficiency. Designers should master spatial data analysis and integrate LiDAR results into NbS design workflow. Higher education curricula should systematically incorporate modules on geospatial data analysis, digital tools, ecological process modeling, and parametric design. Through interdisciplinary collaboration, these programs should empower designers to use LiDAR data for NbS site selection, scale regulation, and performance assessment.
Cross-disciplinary collaboration in NbS must evolve from ad-hoc, project-based approaches to systematic, institutionalized mechanisms. Implementing NbS requires the deep integration of ecology, hydrology, landscape architecture, and remote sensing. This integration requires clear protocols for data acquisition, iterative design charrettes, and interdisciplinary workshops. These mechanisms facilitate the translation of ecological process insights into actionable design interventions. Future work should establish standardized cross-disciplinary collaboration workflows for NbS, encompassing ecological data quality control, ecosystem service assessment validation, and knowledge dissemination mechanisms. Fostering long-term partnerships and shared platforms will promote deep integration across relevant disciplines, ultimately elevating the precision and ecological benefits of LiDAR-based NbS designs. Such collaborative efforts provide scientific evidence-based support for the quantitative assessment of ecological benefits and the construction of climate-adaptive landscapes
[127].
5 Conclusions
Using a bibliometric approach, this study systematically analyzed LiDAR applications in NbS from 2000 to 2024, identifying four core application domains: point cloud analysis and ecological structural foundations, vegetation structure and ecological assessment, river restoration and stormwater management, and urban resilience and BGI optimization. The findings reveal that leveraging its high precision, 3D capability, and multi-platform adaptability, LiDAR consistently translates geometric and structural information into computable ecological metrics. It significantly outperforms traditional 2D remote sensing in observability, spatial continuity, and cross-scale consistency within complex environments. LiDAR thus provides a robust data support for NbS site selection, spatial scaling, and performance evaluation. However, the widespread adoption of LiDAR in NbS still faces challenges including cumbersome data processing, barriers in multi-source data fusion, fragmented toolchains, and cost-benefit imbalances.
To address these challenges, this study proposes a development pathway centered on open data sharing, AI-driven automation, and standardized interoperability, aiming to facilitate the integration of LiDAR technology with NbS practices. The role of LiDAR in NbS should evolve from being merely a high-precision measurement tool to serving as a public infrastructure for evidence-based design and governance. By establishing a complete workflow from data acquisition to design application, LiDAR can support precise implementation of NbS goals, including disaster risk reduction, climate adaptation, biodiversity conservation, and equitable distribution of ecosystem services. With rapid advancement of AI, cloud computing, and the Internet of Things (IoT), LiDAR-empowered NbS practices will drive the landscape planning and design toward a more digital, intelligent, and sustainable development.
PDF
(4734KB)
