Vegetation Mapping Through Multiscale Remote Sensing

1. Introduction

The accurate mapping of vegetation types and their spatial extent, together with the long-term monitoring of vegetation dynamics, has become a major research priority because of its critical role in assessing ecosystem condition, biodiversity conservation, land degradation, and climate change impacts [1,2,3]. Accordingly, vegetation monitoring has gained increasing relevance within the framework of the United Nations Sustainable Development Goals (SDGs), particularly SDG 13 (Climate Action) and SDG 15 (Life on Land), as reliable information on vegetation change is essential to evaluating ecosystem resilience, biodiversity trends, and the effectiveness of conservation and restoration policies [4,5].

To address these challenges, Earth Observation (EO) has become indispensable to vegetation mapping and long-term monitoring, providing consistent, spatially explicit, and repeated observations from local to global scales. Recent advances in satellite remote sensing have substantially enhanced the ability to characterize vegetation structure, composition, and temporal dynamics, enabling more accurate assessments of ecosystem change and supporting evidence-based environmental decision making [6,7]. As a result, EO has become a key component of international frameworks for assessing progress towards the Sustainable Development Goals, providing standardized indicators of ecosystem condition, biodiversity conservation, land degradation, and ecosystem restoration [8].

The rapid evolution of EO capabilities over the past decade has revolutionized vegetation mapping and monitoring. The continuous deployment of new satellite missions, together with the expansion of commercial constellations, long-term observation programs, and airborne and unmanned aerial vehicle (UAV) platforms, has dramatically improved the spatial, spectral, temporal, and radiometric characteristics of remotely sensed data. In parallel, open-data policies adopted by major space agencies have democratized access to long-term EO archives, fostering unprecedented opportunities for analyzing vegetation patterns and dynamics across multiple spatial and temporal scales [9]. These technological advances have enabled vegetation to be characterized consistently across hierarchical ecological levels, from individual plants and local communities to regional, continental, and global ecosystems, providing new insights into ecosystem structure, functioning, resilience, and responses to climate variability, land-use change, and natural disturbances [10,11]. Furthermore, the complementary capabilities of optical, hyperspectral, thermal, LiDAR, and synthetic aperture radar (SAR) sensors have considerably expanded the range of ecological variables that can be retrieved, promoting the development of multiscale and multisensor approaches to vegetation mapping and long-term ecosystem monitoring [6,12].

Despite these advances, significant scientific and methodological challenges remain in fully exploiting multiscale EO for vegetation mapping and monitoring. Vegetation is inherently hierarchical, with ecological patterns and processes operating across nested spatial and temporal scales [13,14]. Individual plants form communities, communities shape landscapes, and landscapes evolve into ecosystems under the influence of climate, disturbance regimes, and human activities. As a result, observations acquired at a single spatial, spectral, or temporal resolution are often insufficient to capture the complexity of vegetation structure, composition, functioning, and dynamics. Understanding how ecological information changes across scales therefore remains a central challenge in vegetation remote sensing and is essential to developing reliable and transferable mapping approaches [9,15,16].

Within this context, one of the major research priorities is the accurate characterization of vegetation composition, spatial patterns, and temporal dynamics across increasingly heterogeneous landscapes. Although recent advances in EO have substantially improved land-cover mapping, monitoring vegetation change remains challenging because of the complex interactions among climate variability, land-use change, ecological succession, and natural disturbances. Furthermore, the increasing frequency and severity of biotic and abiotic stressors—including drought, pest outbreaks, storms, and wildfires—require monitoring systems capable of detecting both gradual ecosystem transitions and abrupt disturbances across multiple spatial and temporal scales. Such information is essential to understanding ecosystem resilience, supporting adaptive management strategies, and evaluating post-disturbance recovery [5,10,17,18].

Another major research direction concerns the retrieval of Essential Vegetation Variables (EVVs) and other biophysical parameters describing vegetation structure and functioning, including aboveground biomass, leaf area index (LAI), canopy height, canopy water content, primary productivity, and carbon dynamics. These variables provide fundamental information for ecosystem modeling, carbon accounting, habitat characterization, biodiversity assessment, and climate change mitigation. Recent advances in remote sensing have considerably improved the capability to retrieve these parameters consistently across multiple spatial and temporal scales by combining structural and spectral observations, thereby enabling more comprehensive assessments of ecosystem functioning, biodiversity, and habitat quality [6,10].

Addressing these challenges requires continuous methodological innovation. Vegetation patterns are strongly scale-dependent, and observations acquired at different spatial, spectral, radiometric, and temporal resolutions often provide complementary rather than redundant information. Consequently, increasing attention is being devoted to multiscale modeling, multisensor and multi-temporal data fusion, artificial intelligence, cloud computing, and upscaling/downscaling strategies capable of integrating observations acquired from satellite, airborne, and UAV platforms. These approaches are progressively improving the accuracy, transferability, and operational applicability of vegetation mapping while facilitating consistent ecosystem characterization across observation scales and environmental conditions [6,9,12].

Against this background, the Special Issue “Vegetation Mapping through Multiscale Remote Sensing” aims to provide a comprehensive overview of recent advances in multiscale vegetation monitoring and mapping, bringing together methodological developments and application-oriented studies across a broad range of ecosystems and environmental conditions. The contributions included in this Special Issue address key challenges related to vegetation classification, ecosystem monitoring, disturbance assessment, wildfire analysis, retrieval of biophysical parameters, biodiversity conservation, multiscale data fusion, and innovative processing strategies. Collectively, these studies demonstrate the growing potential of integrating observations acquired at different spatial, spectral, and temporal scales to improve our understanding of vegetation patterns and processes while supporting more effective ecosystem management and environmental decision making.

2. Overview of the Published Contributions in This Special Issue

To address the challenge of vegetation’s hierarchical nature and the insufficiency of single-resolution observations, the research in this Special Issue proposes the synergistic integration of multiple platforms and sensors to capture both three-dimensional structure and spectral dynamics across various scales. A recurrent approach is the combination of active and passive sensors; for instance, Fernández-Alonso et al. [19] demonstrate that adding LiDAR metrics to Sentinel-2 multispectral imagery significantly improves post-fire aboveground biomass modeling, allowing for an evaluation of the vertical structure that optical remote sensing alone cannot fully capture. Similarly, Tamudo et al. [20] utilize UAVs equipped with LiDAR and multispectral sensors to descend to the individual tree scale, enabling near-real-time detection of forest die-off that medium-resolution satellite observations often average out or ignore. This multiscale approach is further enhanced by incorporating geospatial auxiliary data, such as topographic variables and linear transformations (PCA), which, according to Aghababaei et al. [21], are essential to reducing spectral confusion in heterogeneous landscapes and improving mapping accuracy for plant ecological units (PEUs).

Furthermore, the contributions included in this Special Issue address the problem of continuity and information transferability between scales through advanced harmonization and data fusion techniques. Jargalsaikhan et al. [22] present a key innovation by adapting a high-resolution biomass model (PlanetScope, 3 m) to daily low-resolution imagery (VIIRS, 375 m) using spectral band adjustment factors (SBAFs), bridging the gap between spatial detail and the temporal frequency required for regional monitoring. The role of spatial resolution is also critically evaluated by Omoniyi et al. [23], who find that while 10 m resolution offers slight advantages over 20 m for estimating forest volume, the integration of auxiliary data is the primary driver for improving estimation efficiency. Likewise, Muxiye and Yonezawa [24] emphasize the importance of combining medium-resolution historical archives (Landsat) with high-resolution data (WorldView) to achieve a holistic perspective of long-term pasture changes, compensating for the limitations of individual sensors. Finally, Sun et al. [25] validate the temporal consistency between Landsat-8 and Landsat-9, ensuring the continuity and reliability of NDVI time series for long-term climate monitoring despite intersensor differences.

To address the challenge of characterizing vegetation composition and temporal dynamics across heterogeneous landscapes, several studies in this Special Issue illustrate the benefits of integrating three-dimensional structural information and multiscale spectral observations. Ruscalleda-Alvarez et al. [26] demonstrate that high-density aerial LiDAR is capable of identifying vegetation structural types in arid rangelands that traditional optical remote sensing fails to distinguish, revealing a spatial heterogeneity that is critical to managing degraded ecosystems. This precision in pattern delineation is complemented by the work by Li et al. [27], who use multisource thematic data together with morphological algorithms to automatically delineate the distribution of key species like Populus euphratica, overcoming single-sensor limitations in environments with high variability. Furthermore, the inclusion of auxiliary topographic variables, such as elevation and slope, alongside Landsat-8 data has proven, according to Aghababaei et al. [21], to be essential to reducing spectral confusion and improving the separability of Plant Ecological Units (PEUs) in complex landscapes. Abrupt disturbances and gradual ecosystem transitions can also be effectively monitored using high-temporal- and high-spatial-resolution platforms that enable the assessment of ecosystem resilience. In response to stressors such as drought, Tamudo et al. [20] propose a drone-based workflow that integrates LiDAR and multispectral sensors to track canopy vigor at the individual tree scale, facilitating the detection of early warning signals for forest die-off. Wildfire applications are represented by studies addressing both pre-fire conditions and post-fire ecosystem recovery. Lee et al. [28] link vegetation functional traits (such as the NDVI) and structural traits (such as tree density) with fire severity, while Fernández-Alonso et al. [19] model post-fire aboveground biomass recovery by combining Sentinel-2 and LiDAR observations. Finally, to ensure that these monitoring systems remain consistent over time, Sun et al. [25] validate data continuity between Landsat-8 and Landsat-9 missions, while Jargalsaikhan et al. [22] introduce spectral harmonization techniques to scale high-resolution biomass models to sensors with daily global coverage, allowing for continuous and robust regional monitoring.

To address the challenge of retrieving EVVs and other biophysical parameters in heterogeneous landscapes, the research presented in this Special Issue proposes the synergistic integration of structural and spectral observations across multiple platforms. A predominant approach is the use of LiDAR (aerial or UAV-based) combined with multispectral satellite imagery to accurately model aboveground biomass (AGB) and canopy height, variables that are critical to carbon accounting and fire risk management. For instance, Fernández-Alonso et al. [19] demonstrate that adding LiDAR metrics to Sentinel-2 spectral indices significantly improves the estimation of vertical structure and biomass in burned shrublands, enabling a detailed quantification of post-fire recovery that optical remote sensing alone cannot capture. In arid rangeland environments, Ruscalleda-Alvarez et al. [26] employ high-density aerial LiDAR to identify eight distinct vegetation structural classes, revealing a spatial heterogeneity that is fundamental for habitat characterization and ecosystem degradation assessment. Likewise, the use of drones allows for a descent to the individual tree scale; Tamudo et al. [20] integrate LiDAR and multispectral sensors on UAVs to monitor canopy vigor using greenness indices (NDVI and EVI) and height, detecting early signs of drought-induced forest die-off that are essential to modeling climate resilience.

Additionally, the research presented in this Special Issue resolves the problem of continuity and consistency of these parameters across different temporal and spatial scales through advanced harmonization and upscaling techniques. Jargalsaikhan et al. [22] introduce a key innovation by adapting high-resolution biomass models (PlanetScope) to daily global coverage imagery (VIIRS) using spectral band adjustment factors (SBAFs), allowing for the robust and frequent regional monitoring of primary productivity in grasslands. In the forestry sector, Omoniyi et al. [23] integrate national inventory data with Sentinel-2 variables to improve accuracy in estimating forest volume and its changes, demonstrating that the use of multi-temporal auxiliary data significantly reduces uncertainty in population-level biomass assessments. This long-term consistency is reinforced by Sun et al. [25], who validate NDVI continuity between the Landsat-8 and Landsat-9 missions, ensuring stable time series for monitoring vegetation dynamics in the face of climate change. Finally, the combination of functional traits (such as NDVI greenness) and structural traits (such as tree density and canopy cover) allows, according to Lee et al. [28], for prediction of the severity of abrupt disturbances like wildfires, thus integrating vegetation structure and functioning into an operational environmental management framework.

To address scale dependence and the need for methodological innovation, contributions in this Special Issue propose the synergistic integration of multiple platforms and sensors, demonstrating that structural and spectral observations are complementary in robust ecosystem characterization. A recurrent approach is the fusion of active and passive sensors; for instance, Fernández-Alonso et al. [19] integrate LiDAR metrics with Sentinel-2 spectral indices to model post-fire biomass recovery, capturing the vertical structure that optical remote sensing alone cannot define. On a finer scale, Tamudo et al. [20] employ UAVs equipped with LiDAR and multispectral cameras to descend to the individual tree resolution, facilitating near-real-time detection of forest die-off that medium-resolution satellite observations often average or ignore. This multi-platform synergy is reinforced by the work by Ruscalleda-Alvarez et al. [26], who use high-density aerial LiDAR to identify structural types in arid rangelands, revealing spatial heterogeneity invisible to traditional optical sensors. Likewise, Lee et al. [28] link LiDAR-derived structural traits with functional data from Landsat-8 to predict fire severity, illustrating how multiscale information improves the accuracy of disturbance response models.

In parallel, the studies integrated into this Special Issue explore challenges regarding transferability and operational applicability through the use of artificial intelligence and advanced upscaling strategies. Jargalsaikhan et al. [22] present a key innovation by adapting high-resolution biomass models (PlanetScope, 3 m) to daily low-resolution imagery (VIIRS, 375 m) via spectral harmonization techniques, enabling frequent and consistent regional monitoring. In terms of operational efficiency, Omoniyi et al. [23] apply Random Forest algorithms to integrate Sentinel-2 data with national forest inventories, achieving up to a fivefold reduction in variance compared with traditional estimation methods. Mapping automation is enhanced by the technical framework by Li et al. [27], who fuse multi-source thematic data and morphological algorithms to automatically delineate species distribution outlines, overcoming the subjectivity of manual photointerpretation. Finally, consistency in long-term monitoring is guaranteed by validating data continuity between missions, as shown by Sun et al. [25] in confirming the interchangeability of Landsat-8 and Landsat-9 data and by Muxiye and Yonezawa [24] in combining historical archives with modern sensors to holistically capture successional landscape changes.

3. Conclusions

In conclusion, the contributions integrated into this Special Issue, “Vegetation Mapping through Multiscale Remote Sensing”, offer a comprehensive and cutting-edge vision of how multiscale remote sensing is transforming our ability to monitor and manage terrestrial ecosystems. The studies presented demonstrate that the synergistic integration of active and passive sensors—such as combining LiDAR metrics with multispectral data from Sentinel-2 or Landsat—allows us to overcome the limitations of single-resolution observations, capturing both the three-dimensional structure of vegetation and its functional dynamics with unprecedented precision. From the characterization of structural types in arid rangelands using high-density LiDAR to the detection of early forest die-off signals at the individual tree scale with drones, these investigations underscore that spatial detail and temporal continuity are fundamental pillars for understanding the hierarchy of ecological processes.

Furthermore, this Special Issue highlights the crucial role of methodological innovation and data democratization in achieving the Sustainable Development Goals (SDGs) related to “Climate Action” and “Life on Land”. The application of advanced artificial intelligence and Random Forest algorithms, along with novel spectral harmonization strategies for data upscaling, has proven highly effective for regional biomass estimation, forest volume mapping, and the automatic delineation of key species. Additionally, the validation of data continuity between consecutive satellite missions, such as Landsat-8 and Landsat-9, ensures the robustness of the time series required for global change monitoring.

Ultimately, the results compiled here not only generate high-fidelity mapping products but also establish transferable and operational workflows for forest managers and decision makers. The transition toward the use of digital twins and near-real-time monitoring, facilitated by the increasing availability of 3D data and open access to multi-temporal information, opens new frontiers for biodiversity conservation and adaptive management in the face of natural disturbances such as wildfires and droughts. This Special Issue reaffirms that multiscale remote sensing is an indispensable tool for navigating emerging environmental challenges and ensuring ecosystem resilience on a constantly transforming planet.