European Journal of Remote Sensing ISSN: (Print) (Online) Journal homepage: www.tandfonline.com/journals/tejr20 Advances in vegetation mapping through remote sensing and machine learning techniques: a scientometric review Charles Matyukira & Paidamwoyo Mhangara To cite this article: Charles Matyukira & Paidamwoyo Mhangara (2024) Advances in vegetation mapping through remote sensing and machine learning techniques: a scientometric review, European Journal of Remote Sensing, 57:1, 2422330, DOI: 10.1080/22797254.2024.2422330 To link to this article: https://doi.org/10.1080/22797254.2024.2422330 © 2024 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. Published online: 30 Oct 2024. Submit your article to this journal Article views: 626 View related articles View Crossmark data Full Terms & Conditions of access and use can be found at https://www.tandfonline.com/action/journalInformation?journalCode=tejr20 https://www.tandfonline.com/journals/tejr20?src=pdf https://www.tandfonline.com/action/showCitFormats?doi=10.1080/22797254.2024.2422330 https://doi.org/10.1080/22797254.2024.2422330 https://www.tandfonline.com/action/authorSubmission?journalCode=tejr20&show=instructions&src=pdf https://www.tandfonline.com/action/authorSubmission?journalCode=tejr20&show=instructions&src=pdf https://www.tandfonline.com/doi/mlt/10.1080/22797254.2024.2422330?src=pdf https://www.tandfonline.com/doi/mlt/10.1080/22797254.2024.2422330?src=pdf http://crossmark.crossref.org/dialog/?doi=10.1080/22797254.2024.2422330&domain=pdf&date_stamp=30%20Oct%202024 http://crossmark.crossref.org/dialog/?doi=10.1080/22797254.2024.2422330&domain=pdf&date_stamp=30%20Oct%202024 https://www.tandfonline.com/action/journalInformation?journalCode=tejr20 Advances in vegetation mapping through remote sensing and machine learning techniques: a scientometric review Charles Matyukira and Paidamwoyo Mhangara School of Geography, Archaeological & Environmental Studies, Faculty of Science, University of the Witwatersrand, Johannesburg, South Africa ABSTRACT This study explores the rapid growth in remote-sensing technologies for vegetation mapping, driven by the integration of advanced machine learning techniques. An analysis of publication trends from Scopus indicates significant expansion from 2019 to 2023, reflecting technological advancements and improved accessibility. Incorporating algorithms like random forest, sup- port vector machines, neural networks, and XGBRFClassifier has enhanced the monitoring and analysis of vegetation dynamics at various scales. This progress supports addressing global environmental challenges such as climate change by providing timely data for conservation strategies. China leads in research output, followed by the United States and India, under- scoring the field’s global significance. Key journals, including “Remote Sensing,” and confer- ences like IGARSS, play pivotal roles in disseminating findings. The majority of publications are research articles, emphasizing the reliance on original research and empirical data. The field’s multidisciplinary nature is evident, with contributions spanning Earth Sciences, Agriculture, Environmental Science, and Computer Science. Visualisations using VOSviewer reveal inter- connected themes, highlighting topics like land use, climate change, and aboveground bio- mass. The findings emphasise the importance of continued research and international collaboration to develop innovative solutions for environmental sustainability. ARTICLE HISTORY Received 5 July 2024 Revised 22 September 2024 Accepted 17 October 2024 KEYWORDS Vegetation mapping; remote sensing; machine learning; climate change; environmental monitoring Introduction The ability of remote-sensing technologies to create detailed vegetation maps has made them vital for environmental management and monitoring. These technologies enable the assessment of vegetation cov- ering, condition, and temporal changes, offering cru- cial insights into the dynamics of ecosystems. Advanced sensors, satellite platforms, and unmanned aerial vehicles (UAVs) have recently improved the precision and breadth of environmental data gather- ing. These advancements make it easier to monitor and evaluate a wide range of environmental data, and they also allow for multiscale studies (Cui et al., 2023; Marques et al., 2024; Szpakowski & Jensen, 2019; Timilsina et al., 2020; Vidican et al., 2023). Remote sensing facilitates efficient decision-making in con- servation, land use planning, and resource manage- ment by offering continuous observation and thorough analysis (Rocchini, 2014; Szpakowski & Jensen, 2019). Advances in satellite imaging technol- ogy, such as the increased spectral and geographic resolution of sensors like Landsat, Sentinel, and MODIS, have significantly enhanced the precision and reliability of vegetation maps (Hansen et al., 2013; Morell-Monzó et al., 2020; Pastick et al., 2018, 2020). Machine learning in the field of vegetation map- ping has made significant advancements, greatly enhancing the accuracy, efficiency, and detail of vege- tation classification and monitoring. Integrating advanced machine learning algorithms with remote sensing data has further improved the accuracy and depth of vegetation analysis. Algorithms such as Random Forest (RF), Support Vector Machines (SVM), neural networks, and gradient boosting classi- fiers have facilitated the accurate grouping of different vegetation types and the detection of changes over time (Jozdani et al. 2019). These algorithms can pro- cess large volumes of multi-source data, including spectral, spatial, and temporal information, to differ- entiate between vegetation types, assess health condi- tions, and monitor changes over time. Deep learning, particularly through Convolutional Neural Networks (CNNs), has enabled sophisticated image analysis, such as the automated extraction of vegetation fea- tures and high-resolution mapping of plant commu- nities (Corbane et al., 2021; Maxwell et al., 2021; Timilsina et al., 2020). Integration with advanced sen- sors and cloud-based platforms like Google Earth Engine allows for large-scale, real-time analysis, mak- ing machine learning an indispensable tool in CONTACT Paidamwoyo Mhangara paida.mhangara@wits.ac.za School of Geography, Archaeological & Environmental Studies, Faculty of Science, University of the Witwatersrand, Johannesburg 2000, South Africa EUROPEAN JOURNAL OF REMOTE SENSING 2024, VOL. 57, NO. 1, 2422330 https://doi.org/10.1080/22797254.2024.2422330 © 2024 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The terms on which this article has been published allow the posting of the Accepted Manuscript in a repository by the author(s) or with their consent. http://www.tandfonline.com https://crossmark.crossref.org/dialog/?doi=10.1080/22797254.2024.2422330&domain=pdf&date_stamp=2024-11-07 biodiversity conservation, land management, and eco- logical research (Google Earth Engine Team, 2024). However, machine learning in the domain of vege- tation mapping faces several challenges. One of the primary issues is the need for high-quality labelled training data. Accurate vegetation mapping often requires large, well-labelled datasets, which are time- consuming and costly to collect, particularly for diverse or remote regions (Kaiser et al., 2017; Liang et al., 2020; Stanimirova et al., 2023). Additionally, model performance can be hindered by the quality and resolution of remote sensing data. Factors such as cloud cover, atmospheric conditions, and sensor limitations can affect data quality, leading to potential inaccuracies in vegetation classification. Spectral and spatial resolutions are also critical; lower spectral reso- lution can limit the differentiation of vegetation types, while lower spatial resolution can obscure fine-scale patterns in heterogeneous landscapes (Di & Yu, 2023; Dou et al., 2024; Lin et al., 2022; Wilson et al., 2016). Sensors like Landsat, Sentinel, and MODIS offer vary- ing levels of spectral and spatial resolution, which may not always meet the requirements for detailed vegeta- tion analysis. Recent advances, such as hyperspectral imaging and the use of finer spatial resolution sensors like WorldView-3 and PlanetScope, have begun to address these limitations by providing richer spectral information and finer spatial details (Liu et al., 2020; Satellite Imaging Corporation, 2024; Whig et al., 2024; Xie et al., 2008). These advances allow machine learn- ing models to capture more subtle differences in vege- tation characteristics, improving classification accuracy and the detection of small-scale changes. However, integrating such high-resolution data still poses challenges in terms of data processing and sto- rage, requiring further advancements in computa- tional methods (Ayhan et al., 2020; Lu et al., 2022). Furthermore, despite the effectiveness of machine learning models, they often function as “black boxes”, making it difficult to interpret their decisions and understand the ecological processes they detect. This lack of interpretability can be a barrier to their accep- tance in ecological research and management practices (Shams et al., 2024; Welchowski et al., 2022). Additionally, the generalization of models across dif- ferent regions and conditions is another challenge, as models trained in one area may not perform well in other regions with different vegetation types, climate conditions, or spectral characteristics. This necessi- tates ongoing model tuning and adaptation to ensure accuracy across diverse landscapes (Ayhan et al., 2020; Turner et al., 2019; Yang et al., 2022). Despite these challenges, advancements in machine learning, com- bined with evolving remote sensing technologies, con- tinue to push the boundaries of vegetation mapping, offering increasingly precise and scalable tools for ecological analysis. The aim of this research is twofold: first, to conduct a comprehensive literature review exploring the appli- cations of remote sensing in monitoring vegetation growth patterns, detecting land cover changes, asses- sing land degradation and fragmentation, evaluating the impact of topography on vegetation health, and estimating biomass and evapotranspiration. This research seeks to contribute to the field by highlighting the advancements in combining remote sensing tech- nologies with machine learning algorithms, which offer significant improvements in the accuracy, depth, and scale of vegetation analysis. These methods can better track environmental changes, support bio- diversity conservation, and enhance land management practices (Bauer et al., 2024; Matyukira & Mhangara, 2023b; Mullissa et al., 2024; Pettorelli et al., 2005). Additionally, this paper aims to offer insights into the evolving techniques that allow more precise assess- ment of ecosystem health and changes in vegetation over time, thereby contributing to the body of knowl- edge essential for sustainable environmental monitor- ing and management. By systematically reviewing the literature, this research identifies critical themes and advancements that highlight the growing potential of remote sensing and machine learning in tackling environmental challenges. Approaches to vegetation mapping and analysis using remote-sensing technologies Advancements in machine learning algorithms for vegetation classification Remote sensing and machine learning have evolved significantly over the past few decades, each contribut- ing to vegetation mapping and classification advance- ments. The use of remote sensing to map vegetation began in the 1960s with the advent of aerial photo- graphy and early satellite imaging, marking the start of widespread remote sensing applications. During this initial phase, these technologies facilitated the visual evaluation of vegetation patterns across vast terri- tories, initiating the first phase of remote sensing applications in environmental monitoring (NASA Earth Observatory, 2023, Geological Survey, 2023). The launch of the Landsat program in 1972 was a pivotal event, producing the first multi-spectral satellite images capable of efficiently studying vegeta- tion. Early studies utilised these images to classify various types of vegetation and monitor changes in land cover (Cohen & Goward, 2004; Tucker, 1979) demonstrating the potential of satellite imagery for comprehensive environmental monitoring. In parallel, the field of machine learning began to take shape. In the mid-20th century, early machine learning efforts focused on pattern recognition and statistical learning, laying the groundwork for more 2 C. MATYUKIRA AND P. MHANGARA sophisticated algorithms. The development of decision trees, neural networks, and support vector machines (SVM) in the 1980s and 1990s marked significant milestones (Breiman, 2001; Cortes & Vapnik, 1995). These advancements allowed for better handling of complex datasets and improved predictive accuracy. As machine learning techniques became more advanced, they began to be applied to remote sensing data, enhancing vegetation classification accuracy (Chaves et al., 2020; Maxwell et al., 2018). Technical advancements in digital image processing throughout the 1990s significantly improved the analysis of data obtained from remote sensing. The precision of vege- tation maps has increased due to the development of algorithms for image classification, such as the Maximum Likelihood Classifier (MLC), providing environmental scientists with more accurate tools. Additionally, the development of vegetation indices, such as the Enhanced Vegetation Index (EVI) and Normalized Difference Vegetation Index (NDVI), enabled quantitative analysis of vegetation health and productivity (Arifeen et al., 2021; Lillesand et al., 2004; Rahman et al., 2017; Rani et al., 2023). These indices became widespread for analysing environmental impacts and monitoring vegetation growth and are directly related to machine learning as they serve as key input features for various models. By incorporat- ing vegetation indices into machine learning algo- rithms like decision trees, neural networks, and SVM, researchers can enhance the accuracy and effi- ciency of vegetation classification and monitoring tasks. Machine learning models leverage these indices to identify patterns, distinguish between different vegetation types, assess vegetation health, and predict changes over time (Pan et al., 2023; Turhal, 2022; Xie et al., 2019). Furthermore, the availability of higher- resolution satellite images from sensors like SPOT and the continuous enhancements of Landsat sensors sig- nificantly improved spatial resolution and spectral capabilities, facilitating more detailed and accurate vegetation analysis (Hansen et al., 2013; Morell- Monzó et al., 2020; Qarallah et al., 2023; Radoux et al., 2016; Townshend et al., 2000; Geological Survey, 2023). Since its first use, vegetation mapping has advanced significantly with the integration of remote sensing data into Geographic Information Systems (GIS) in the 2000s. This combination allowed for comprehen- sive geographical analysis, enabling the mapping of vegetation patterns in relation to other environmental factors such as topography, soil types, and climate (Brown et al., 2016; Foody, 2002; Lv et al., 2019). GIS technology facilitated the management and analysis of large datasets, enhancing the ability to monitor changes in land cover and assess their impact on ecosystems (Arifeen et al., 2021; Brown et al., 2016; Foody, 2002; Rouse et al., 1974). Platforms like Google Earth Engine further revolutionised the field with cloud-based processing capabilities and access to a vast library of satellite images, making large-scale environmental assessments more accessible and effi- cient (Gorelick et al., 2017). In recent years, the integration of machine learning techniques with remote-sensing data has become increasingly important to tackle the complexities and scale of environmental data. Advanced algorithms, particularly deep learning and neural networks, have greatly enhanced image classification tasks. Convolutional Neural Networks (CNNs) have demon- strated superior performance by automatically learn- ing hierarchical features from raw satellite imagery, significantly improving the accuracy and comprehen- siveness of vegetation mapping (Ma et al., 2019; Zhu et al., 2017). Additionally, machine learning algo- rithms such as Random Forest (RF) and Support Vector Machine (SVM) have been widely used for classification tasks. RF employs an ensemble of deci- sion trees to manage high-dimensional data and iden- tify key features, aiding in differentiating vegetation types based on their spectral signatures while mini- mizing overfitting (Berhane et al., 2018; Dobrinić et al., 2021; Wang et al., 2018). SVM, on the other hand, finds an optimal hyperplane in a high- dimensional space to classify vegetation using spectral and spatial features, making it particularly useful in scenarios with limited training data (Fu et al., 2017; Hasan et al., 2019; Xie et al., 2019). Unsupervised learning methods like K-means clustering also play a valuable role in grouping similar spectral signatures, helping to identify various vegetation zones without labelled training samples (Cohn & Holm, 2021; Tavallali et al., 2021). The integration of remote sensing with machine learning is thus crucial for facilitating more precise, detailed, and scalable ecological research. For example, a study by Kluczek et al. (2024) demonstrates this integration by combining multitemporal optical and radar satellite data to enhance vegetation mapping accuracy, account for seasonal variability, and provide a robust methodology for biodiversity conservation, land management, and ecological monitoring in mountainous regions. Similarly, Bartold and Kluczek (2024) show that augmenting deep neural networks with metadata significantly improves wild animal clas- sification accuracy, offering valuable tools for biodi- versity conservation, wildlife management, and ecological research. Furthermore, Sharma (2022) highlights the practical applications and successes of integrating machine learning with multispectral data, achieving detailed, high-resolution (10-meter) coun- trywide mapping of plant ecological communities using CNNs, thereby enhancing land management, biodiversity conservation, and ecological research. By integrating these advanced machine learning EUROPEAN JOURNAL OF REMOTE SENSING 3 algorithms with remote sensing platforms such as Google Earth Engine, researchers can process and analyse satellite imagery in real-time, leading to further advancements in the field of environmental monitoring and ecosystem assessment (Gorelick et al., 2017). Another significant advancement in the application of machine learning to remote sensing is the introduc- tion of extreme gradient-boosting algorithms, such as XGBoost and its variants (Chen & Guestrin, 2016). These algorithms have complemented traditional machine learning techniques by offering superior per- formance in handling large datasets and high- dimensional features, which are common in remote sensing data (Chen & Guestrin, 2016; Matyukira & Mhangara, 2023b). Extreme gradient boosting com- bines the strengths of decision trees and gradient boosting, providing robust and efficient classification results. Its ability to handle missing values and its scalability have made it a popular choice for vegetation classification tasks, further enhancing the accuracy and efficiency of remote sensing analyses (Pham et al., 2020; Zhang et al., 2020). Vegetation growth dynamics using remote-sensing technologies Remote sensing technologies have become indispen- sable for environmental monitoring and management due to their ability to map vegetation comprehen- sively. These technologies enable the evaluation of vegetation coverage, condition, and temporal fluctua- tions, providing crucial insights into ecosystem dynamics (Knauer et al., 2014). The continuous advancements in satellite remote sensing, such as the development of high-resolution sensors and sophisti- cated data processing algorithms, have significantly enhanced researchers’ ability to monitor vegetation health, productivity, and changes over time (Li et al. 2023c; Pradhan et al., 2024). Time-series analysis of remote sensing data is a crucial technique for monitoring temporal changes in vegetation phenology and growth patterns (Kooistra et al., 2024). Common methods for assessing vegetation health and productivity include using NDVI, EVI, and other vegetation indices that provide accurate measurements of plant greenness, which are closely correlated with photosynthetic activity and biomass (Li et al. 2023b; Pradhan et al., 2024). These indices are instrumental in detecting seasonal varia- tions, identifying periods of maximum growth and evaluating long-term trends in vegetation dynamics. By analysing temporal changes through remote sen- sing data, researchers can better understand vegeta- tion’s response to climate fluctuations, human activities, and natural disturbances (Nyamjav et al., 2024; Obuchowicz et al., 2024; Pettorelli et al., 2005). Furthermore, phenological metrics derived from remote sensing data, such as the start of the season (SOS), end of the season (EOS), and length of the growing season (LOS), provide essential information about plant growth over time (White et al., 2009). These measures help detect changes in phenological events due to climate change and other environmental variables. For instance, higher temperatures often result in an earlier onset of SOS and a longer duration of LOS. By observing these changes, scientists can more accurately predict the effects of climate change on ecosystems, adjust agricultural practices, and con- serve biodiversity (Chen & Zhang, 2023; Fang et al., 2023). In addition to vegetation indices, several advanced methods are employed for assessing vegetation health and analysing time-series data. One of the most com- mon approaches is the use of radiative transfer models such as PROSPECT + SAIL models, which simulate light interaction with vegetation to estimate biophysi- cal parameters like leaf area index (LAI) and chloro- phyll content (Berger et al., 2018; Gupta & Pandey, 2022; Jacquemoud et al., 2009). A recent study utilised PROSAIL model inversion on Google Earth Engine to estimate aboveground biomass continuously over time on the Tibetan Plateau (Xie et al., 2022). This method provided detailed insights into vegetation structure and function, highlighting the benefits of combining biophysical models with remote sensing data for com- prehensive vegetation analysis. Additionally, LiDAR and radar technologies are increasingly being used to measure vegetation height and biomass, offering information that cannot be captured by optical sensors alone (Edson & Wing, 2011; Maesano et al., 2020). However, despite these advancements, several chal- lenges remain in analysing vegetation growth dynamics using remote-sensing data. One of the major challenges is the data quality and resolution. While temporal resolution has improved, there are still limitations in capturing rapid vegetation changes, especially in areas with high variability due to seasonal or anthropogenic factors (Fan et al., 2022; Li et al., 2023a). In regions with frequent cloud cover, optical data collection can be interrupted, necessitating the use of radar or LiDAR technologies. Additionally, lower spectral resolution may hinder the differentia- tion between various vegetation types in heteroge- neous landscapes, making it difficult to capture subtle vegetation characteristics (Ougahi et al., 2022; Wegehenkel, 2009). Moreover, vegetation indices such as NDVI, while widely used, have limitations in areas of dense vegeta- tion where they tend to saturate, reducing sensitivity to biomass changes (Fu et al., 2022; Hassan et al., 2023). To overcome these challenges, alternative approaches such as data fusion techniques – which combine optical, radar, and LiDAR data – are being 4 C. MATYUKIRA AND P. MHANGARA developed to offer a more comprehensive assessment of vegetation health (Illarionova et al., 2024; Tian et al., 2024). Furthermore, the analysis of vegetation dynamics is complicated by the variability in environ- mental factors, such as climate variability and human- induced changes. Phenological shifts, for instance, may vary significantly across ecosystems, making it difficult to create generalisable models for predicting vegetation responses (Hassan et al., 2023; Roberts et al., 2015) Finally, integrating environmental variables like temperature, precipitation, and soil moisture into remote sensing data remains a complex task. Although studies have demonstrated correlations between these factors and vegetation indices (Tsai & Der Yang, 2016; Zhong et al., 2010), predicting vegeta- tion dynamics accurately requires sophisticated mod- els capable of capturing these intricate interactions. Additionally, the increasing occurrence of extreme weather events adds further complexity, making it essential to develop robust models that integrate remote sensing data with in-situ measurements. These advancements are crucial for improving predic- tions of vegetation responses and guiding ecosystem management practices (Fang et al., 2023). Analysis of land cover and land use change utilising remote-sensing technologies Remote sensing technologies are crucial in identifying and examining changes in land cover and land use. These changes are essential for comprehending envir- onmental shifts and guiding sustainable land manage- ment strategies. Monitoring these changes over time allows for valuable observations of the effects of human activities and natural processes on landscapes. Researchers use various remote sensing methods to obtain precise information on the evolution of land cover and land use. This knowledge is invaluable for developing effective conservation strategies and informing policy decisions (Fang et al., 2023; Knauer et al., 2014; Pradhan et al., 2024). Change detection methods, such as post-classification comparison, image differencing, and principal component analysis (PCA), are often used to detect changes in land cover using multi-temporal remote sensing data. Post- classification comparison is the separate categorisa- tion of images from various periods, followed by com- paring the outcomes to identify any alterations (Liu & Zhou, 2004). Image differencing is a process where the pixel values of one date are subtracted from the pixel values of another date, which helps to identify places where there have been substantial changes (Panuju et al. 2020). Principal Component Analysis (PCA) decreases the number of dimensions in the data, improving the ratio of useful signal to unwanted noise and increasing the efficiency of detecting changes (Li et al., 2024b; Wu et al., 2015). These tools have shown efficacy in monitoring several land cover changes, such as urbanisation, deforestation, and agricultural development (Asokan & Anitha, 2019; Cheng et al., 2023; Marukatat, 2023; Yu et al., 2023). Remote sensing technology has been instrumental in tracking urbanisation and deforestation, key drivers of environmental transformation. High-resolution satellite imagery facilitates detailed mapping of land cover changes, revealing the extent and patterns of urban growth and forest clearance (Panuju et al. 2020). By analysing remote sensing data, researchers can observe the impact of urban development on natural environments, such as the proliferation of impervious surfaces and built-up areas. Similarly, deforestation activities can be scrutinised to ascertain the pace and spatial pattern of forest depletion. These insights are vital for understanding the consequences of land-use changes and formulating interventions to mitigate their adverse effects (Joseph & Jaya Surya, 2019; Kumar, 2011). Building on this understanding, evaluating the effects of changes in land cover and land use on ecosystems and biodiversity becomes a critical next step for implementing sustainable management prac- tices (D. Kumar, 2011). Remote sensing data provide significant information for estimating the impacts of these changes on habitat fragmentation, species dis- tribution and ecosystem services (Matyukira & Mhangara, 2023b). One such impact is habitat frag- mentation, often a result of urbanisation or deforesta- tion. To understand its effects on biodiversity, researchers can analyse landscape metrics derived from remote sensing data (Musetsho et al., 2021; Rong & Fu, 2023). In addition, remote sensing may be used to observe changes in land use patterns, such as the transformation of forests into agricultural areas, which has substantial consequences for carbon seques- tration, water cycles, and soil quality (Assede et al., 2023; Cheng et al., 2023). Researchers may create complete models to forecast future land use scenarios and their possible effects by combining remote sensing with ecological and socio-economic data. This integra- tion helps formulate policies for sustainable develop- ment and conservation (Jiang et al., 2021; Vitale & Salvo, 2024). While remote sensing technologies have greatly enhanced the examination of land cover and land use changes, specific limitations intrinsic to these tools must be acknowledged. A significant difficulty is the accessibility and quality of remote sensing data. High-resolution satellite imagery, essential for com- prehensive monitoring, may be inaccessible due to expense, cloud cover interference, or limited satellite revisit time, especially in tropical areas susceptible to persistent cloud cover (Breunig et al., 2023). EUROPEAN JOURNAL OF REMOTE SENSING 5 Furthermore, the validation of remote sensing data necessitates access to accurate ground truth data, which can be costly and logistically difficult to acquire, particularly in distant or politically dangerous areas. The absence of dependable validation data may under- mine the accuracy of land cover classifications, result- ing in inaccuracies in identifying land use changes (Olofsson et al., 2014; Stehman & Foody, 2019). The classification algorithms employed in remote sensing are frequently constrained by their sensitivity to spectral similarities among various land cover types. For instance, differentiating between specific vegeta- tion types and urban areas might be challenging when their spectral signatures overlap, potentially resulting in misclassification (Gündüz & Orman, 2024; Mehmood et al., 2022). The intricacy of data proces- sing and the substantial computational expenses linked to managing extensive multi-temporal remote sensing data may be prohibitive for certain researchers or organizations. These challenges, including the time and resources required for data preprocessing and analysis, must be balanced against the powerful insights that remote sensing offers in understanding land cover and land use dynamics (Southworth & Muir, 2021, Di and Yu 2023; Yao et al., 2020). Consequently, although remote sensing is an invalu- able instrument, these constraints must be recognized to guarantee precise monitoring and informed deci- sion-making in land management. Utilising remote sensing technologies to monitor land degradation and fragmentation Remote sensing technologies are efficient tools for monitoring land degradation and fragmentation, which are significant environmental concerns impact- ing ecosystem health and species diversity. These tech- nologies provide detailed, multi-temporal data that are essential for the ongoing evaluation of land degrada- tion indicators such as soil erosion, loss of plant cover, and desertification (Giri Tejaswi Rome, 2007; Reddy et al., 2018; Shange, 2020; Symeonakis, 2022). Spectral indices like the NDVI and the Soil Adjusted Vegetation Index (SAVI) are instrumental in measur- ing the extent and condition of vegetation, offering valuable insights into the health of the land (Reddy et al., 2018). By aiding in detecting regions undergoing deterioration and assessing the magnitude of these changes over time, NDVI and SAVI enhance our ability to monitor and address the critical issue of land degradation. The integration of remote sensing data with spectral indices allows for a more compre- hensive understanding and management of environ- mental changes, ensuring the preservation of ecosystems and biodiversity (Giri Tejaswi Rome, 2007; Lu et al., 2015; Mambo & Archer, 2007; Yengoh et al., 2016). Various remote sensing methods are key for effec- tively monitoring land cover changes, which in turn can indicate land degradation. Supervised and unsu- pervised classification techniques are commonly used for categorizing land cover types. Supervised classifi- cation utilises predefined training datasets, enabling the algorithm to classify images according to known land cover categories. In contrast, unsupervised clas- sification relies on the spectral properties of pixels to group them into clusters without prior knowledge of land types (Parashar, 2023). Object-Based Image Analysis (OBIA) is another important method that segments high-resolution satellite images into objects and classifies them based on both spectral and spatial properties such as texture and shape, improving clas- sification accuracy in complex landscapes (Dronova, 2015; Kumar et al., 2020). Additionally, change detection techniques such as post-classification comparison, image differencing, and NDVI differencing play a vital role in identifying land cover transitions over time. These techniques allow for detecting changes like deforestation, urbani- zation, and agricultural expansion, which often con- tribute to land degradation (Asokan & Anitha, 2019; Cheng et al., 2024; Parelius, 2023). Incorporating these methods into land monitoring systems enhances the detection of subtle changes in land cover and provides better insights into ongoing degradation processes. Similarly, fragmentation analysis is an essential application of remote sensing in environmental mon- itoring. It addresses habitat fragmentation – dividing continuous habitats into smaller, isolated patches. Metrics such as patch size, edge density, and connect- edness, derived from remote sensing data, provide insights into habitats’ spatial arrangement and integ- rity (Singh et al., 2014). Analysing these metrics helps researchers understand fragmentation’s extent and impact on ecosystems. This knowledge is vital for conservation efforts and land-use planning to mitigate negative effects on biodiversity and ecosystem services (Dupin et al., 2013; Mambo & Archer, 2007; Singh et al., 2014; Zlinszky et al., 2015). In parallel, monitoring desertification is crucial for evaluating land deterioration, especially in dry and semi-dry regions. Remote sensing approaches, includ- ing the Land Degradation Neutrality (LDN) frame- works, leverage satellite data to track changes in land cover and soil characteristics over time (Kust et al., 2023). Such monitoring is key to identifying areas vulnerable to desertification. The data obtained from these tools are essential for mapping and observing desertification trends, which in turn supports the development of policies aimed at combating land degradation. By integrating remote sensing data with ground-based observations and climate factors, researchers can create comprehensive models to pre- dict and manage desertification, thereby helping to 6 C. MATYUKIRA AND P. MHANGARA maintain land productivity and ecological balance in susceptible areas (Dharumarajan et al., 2022; El Hassan, 2004; Hill & Helldén, 2005; Xu, 2023). Assessing the impact of terrain on plant health using remote sensing technologies Topography is crucial in influencing vegetation distri- bution and vitality because variations in elevation, slope, and aspect create a diversity of microhabitats (Xun et al., 2023). Remote sensing technologies, including Digital Elevation Models (DEMs), facilitate a comprehensive analysis of topographic factors on vegetation. DEMs derived from remote sensing data offer detailed topographical information, allowing researchers to explore the effects of elevation gradients on vegetation patterns (Guth et al., 2021). Remote sensing data such as those derived from radar (e.g. SRTM – Shuttle Radar Topography Mission, or TanDEM-X) and LiDAR (Light Detection and Ranging) can be used to directly obtain ground eleva- tion information through the generation of Digital Elevation Models (DEMs). These datasets provide detailed topographical information at various resolu- tions. SRTM, for example, provides global elevation data at a 30-meter resolution, while LiDAR offers even more precise elevation data, typically down to meter or sub-meter accuracy (Guth et al., 2021, 2024; Wei & Bartels, 2012). These remote sensing methods invert elevation data directly from satellite or aerial observa- tions, allowing researchers to analyse terrain features like slope, aspect, and elevation, which are critical for understanding vegetation patterns. For instance, ele- vation can significantly influence plant growth and health by affecting temperature and moisture avail- ability, which are vital factors. By examining these elevation-related factors, scientists can gain a deeper understanding of the spatial distribution of various plant species and their responses to topographic changes (Brocard et al., 2023; Khalaf et al., 2021; Matsuura & Suzuki, 2013; Odgaard et al., 2014). In tandem with the analysis provided by DEMs, remote sensing is also instrumental in evaluating vege- tation vigour through topographic correction. Since remote sensing data frequently require modifications to account for topographic impacts such as slope and aspect on reflectance measurements, topographic cor- rection techniques are employed to alter the reflec- tance data (Zhang, 2011). This process ensures that vegetation indices like the NDVI are more accurate. Precise vegetation indices are critical for assessing vegetation health and productivity because they pro- vide reliable assessments of plant biomass and photo- synthetic activity. Therefore, topographic correction not only complements the insights gained from DEMs but also enhances the accuracy and relevance of vegetation vigour evaluations by removing the impact of the underlying topography (Gao & Zhang, 2009; Ma et al., 2021; Riaño et al., 2003; Yao et al., 2022). Furthermore, topography-induced microclimates significantly impact vegetation vigour in the Southern Hemisphere, as differences in terrain may generate localised climatic conditions that affect plant develop- ment. North-facing slopes in the Southern Hemisphere receive more sunshine, resulting in higher temperatures and less moisture, while south-facing slopes are colder and have more moisture (Barry & Blanken, 2016). These microclimatic differences impact the species makeup, growth rates, and general health of plants. By combining remote sensing data with topography infor- mation, researchers can examine microclimates’ influ- ence on plants’ distribution and health (Jucker et al., 2018). Such assessments are crucial for forecasting the impact of climate change on vegetation patterns since variations in temperature and precipitation patterns may modify the microclimatic conditions that plants rely on for their survival and development. Understanding these relationships enables scientists to formulate more efficient conservation and land man- agement techniques (Bramer et al., 2018; Chen et al., 2013; De Frenne et al., 2013; Jung et al., 2016). In addition to topography, vegetation growth and health are also influenced by both environmental fac- tors and human activities, which can be monitored using remote-sensing technologies. Environmental factors such as temperature, precipitation, and soil moisture can be captured using satellite data from missions like MODIS (Moderate Resolution Imaging Spectroradiometer) and Sentinel-2, which track cli- mate variables over time (Dhillon et al., 2023; Lange et al., 2017). Similarly, human activities such as land use changes, deforestation, and urbanization can be monitored through remote sensing, using methods such as land cover classification and change detection. These factors significantly impact vegetation by alter- ing local climates and soil conditions, affecting plant growth and health (Abebe et al., 2022; Nedd et al., 2021). For example, a study using the CCDC algo- rithm with optical and SAR images investigated marsh vegetation phenology in the Honghe National Nature Reserve, China, and found that hydro- meteorological factors significantly drive phenological changes (Fu et al., 2022). Additionally, another study in the same reserve employed the LandTrendr algo- rithm and Google Earth Engine to analyse the dynamic relationship between marsh vegetation and hydrological changes over a 35-year period, further demonstrating how remote sensing can be used to monitor complex environmental interactions over time (Fu et al. 2022). Integrating data on climate and human-induced changes, alongside topographic infor- mation, provides a more holistic understanding of the drivers affecting vegetation patterns EUROPEAN JOURNAL OF REMOTE SENSING 7 Remote sensing technologies for estimating biomass and evapotranspiration Estimating biomass and quantifying evapotranspira- tion are crucial for comprehending ecosystem produc- tion and water use, which are fundamental environmental and resource management aspects (Khan et al., 2019; Zhou et al., 2008). Remote sensing technologies are used to estimate biomass using many sophisticated approaches such as vegetation indices, radar, LiDAR, and UAVs. Vegetation indices, such as NDVI, are commonly used for biomass estimation by correlating the reflectance of vegetation to biomass levels. Radar data, particularly from Synthetic Aperture Radar (SAR), provide structural information by penetrating vegetation canopies, enabling biomass estimation. LiDAR offers a precise 3D measurement of plant structures, including height, which is essential for biomass calculations (Borsah et al., 2023; Kumar et al., 2015; Sinha et al., 2015). The use of UAVs equipped with advanced RGB and multispectral sensors has significantly enhanced the precision and spatial resolution of biomass estimation. These drones, as highlighted by Tait et al. (2019) and Bazrafkan et al. (2023), are capable of capturing high- resolution images across various spectral bands, enabling precise spatial and spectral vegetation analysis. RGB sensors offer insights into the visible spectrum for visual assessments, while multispectral sensors provide critical data at specific wavelengths like the near- infrared for in-depth vegetation studies. The detailed imagery from UAVs is particularly beneficial for monitoring minute changes in plant structure, proving invaluable in hard-to-reach areas or when high-detail information is required (Lussem et al., 2019; Nex & Remondino, 2014). However, despite their high resolu- tion and flexibility, UAVs face challenges in covering large spatial scales due to their limited flight range, data volume constraints, and regulatory restrictions. For larger areas, satellite-based methods such as synthetic aperture radar (SAR) and LiDAR remain more effective (Nedd et al., 2021; Saadatseresht et al., 2015) Advancements in remote sensing techniques for measuring evapotranspiration have been marked by the sophisticated use of thermal infrared data and energy balance models. Thermal infrared sensors, which Holmes (2019) elaborates on, are instrumental in gauging vegetation surface temperatures to estimate rates of water transpiration and soil evaporation. Energy balance models like SEBAL and METRIC, dis- cussed by (Allen et al., 2007) and Elhag et al. (2011), leverage these thermal data alongside meteorological and environmental factors to precisely calculate eva- potranspiration rates. These models are particularly suited for large-scale applications where satellite- based remote sensing can efficiently capture vast areas, a task where UAVs often struggle due to their operational limitations. Such large-scale, accurate eva- potranspiration data are crucial for managing water resources effectively, particularly in arid and semi-arid regions, where efficient water use is critical for the sustainability of agriculture and ecosystem health (Gxokwe et al., 2020; Nedd et al., 2021; Saadatseresht et al., 2015). Furthermore, the combination of biomass estima- tion and evapotranspiration data with carbon and water flow measurements enriches our understanding of ecosystem dynamics. Remote sensing is pivotal in providing spatially accurate data for modelling carbon and water movements on various scales, assisting in assessing the impact of different plant types on the carbon cycle, as noted by Allen et al. (2007), Srinet et al. (2022) and Huang et al. (2024). UAVs, as in Bendig et al. (2014), suggest augmenting these cap- abilities with their flexible and immediate data collec- tion, adapting them to specific research needs and schedules. These models are essential for appraising climate change effects on ecosystem services and for- mulating mitigation strategies. Integrating remote sensing data with terrestrial observations and other environmental variables offers researchers a deeper understanding of carbon-water cycle interactions, leading to the development of comprehensive models that enhance environmental management practices, as highlighted by Initiative (2016) and Niu et al. (2021) and Niu et al. (2021). Materials and method Employing Scopus and Excel for targeted literature review in vegetation mapping Scopus is a comprehensive, multidisciplinary biblio- graphic database that indexes various peer-reviewed literature, including journals, conference proceedings, and patents, across diverse science, technology, med- icine, social sciences, and arts and humanities fields (Burnham, 2006). Its extensive coverage and robust search capabilities make it an invaluable resource for conducting thorough literature reviews (Kadam et al., 2020). Scopus provides advanced search functional- ities that allow for the precise construction of queries using logical operators and field-specific filters, which are essential for refining searches and retrieving highly relevant studies (Burnham, 2006; Kadam et al., 2020). Additionally, its analytical tools enable detailed exam- ination of citation patterns, authorship networks, and research trends, offering insights into the impact and progression of specific fields. These features make Scopus particularly well-suited for our research on vegetation mapping through remote sensing, as it ensures access to high-quality, peer-reviewed articles and facilitates the identification of key contributions and emerging trends in the intersection of remote 8 C. MATYUKIRA AND P. MHANGARA sensing technologies and advanced machine learning techniques (Burnham, 2006; Kadam et al., 2020; Matyukira & Mhangara, 2023a). Leveraging the extensive coverage and advanced search capabilities of Scopus, our research on vegeta- tion mapping through remote sensing was able to tap into a rich repository of high-quality, peer-reviewed articles. By utilising Scopus’s precise query construc- tion using logical operators and field-specific filters, we were able to define the research scope clearly and identify key terminologies as detailed in Table 1. These included terms related to both vegetation mapping and remote sensing, as well as advanced machine learning techniques such as “random forest”, “support vector machine”, “neural networks”, “deep learning”, and “XGBRFClassifier”. The additional search terms related to specific applications further honed our exploration, ensuring a targeted yet comprehensive review of the subject matter. The construction of the Scopus query was meticu- lously executed, incorporating both primary and sec- ondary keywords. Logical operators such as AND, OR, and NOT were employed to interconnect relevant concepts, as demonstrated in Table 1. This systematic strategy expedited the retrieval of studies pertinent to vegetation mapping and remote sensing. The search was further refined by focusing on specific research disciplines listed in Table 1, including earth and pla- netary sciences, environmental science, agricultural and biological sciences, computer science, engineer- ing, geography, planning, and development. This refinement was instrumental in aligning the search results with the study’s objectives by concentrating on the intersection of vegetation mapping, remote sensing, sophisticated computational methods, and interdisciplinary environmental applications. To enhance the precision of our search out- comes, we strategically excluded generic or unre- lated terms that could yield irrelevant or overly broad information for our inquiry. As indicated in Table 1, terms such as “antennas”, “artificial intelligence” (in a broader context), “China”, “crops”, “engines”, “forestry”, “image analysis”, “image enhancement”, “image processing”, “India”, “procedures”, “satellite data”, “soils”, and “United States” were omitted. This exclusionary tactic ensured a focused aggregation of literature pertinent to vegetation mapping via remote sensing and machine learning techniques. By judiciously selecting and excluding keywords while honing in on essential subject areas, our methodology not only streamlined the literature collection process but also guaranteed the incorporation of high- calibre research directly relevant to our study goals. Incorporating visual tools such as Scopus’s export function alongside Excel’s conditional for- matting and filtering features enabled us to effec- tively identify and eliminate duplicate document titles. After deactivating these duplicates, we har- nessed Scopus’s analytical tools to scrutinise the data based on various parameters like country of origin and authorship. This analysis facilitated the distillation of key findings into succinct summary tables and graphical representations, including bar charts, line charts, pie charts, and global maps. The creation of an array of tables and figures was pivo- tal in providing a structured synthesis of the litera- ture review findings. Visualising the geographical distribution of research efforts highlighted the glo- bal scope of scholarly activities in this domain. Author contribution analysis was employed to pin- point leading researchers and their collaborative networks within this specialised field. The trend analysis of publications over time underscored the growing scholarly interest in common machine- learning approaches applied to vegetation mapping through remote sensing. Bibliometric visualisation and analysing of data in VOSviewer VOSviewer is a powerful software tool designed for constructing and visualising bibliometric networks, offering an intuitive interface for analysing Table 1. Scopus search engine and queries used for the scope of this study. Search Engine Website Query Scopus scopus.com TITLE-ABS-KEY (“vegetation mapping” AND “remote sensing” AND (“machine learning” OR “Random Forest” OR “Support Vector Machine” OR “neural networks” OR “deep learning” OR “XGBRFClassifier”) AND (“vegetation classification” OR “vegetation growth dynamics” OR “land cover change” OR “land degradation” OR “fragmentation” OR “topographic influences” OR “vegetation vigor” OR “biomass estimation” OR “evapotranspiration”)) AND (LIMIT-TO (SUBJAREA, “EART”) OR LIMIT-TO (SUBJAREA, “ENVI”) OR LIMIT-TO (SUBJAREA, “AGRI”) OR LIMIT-TO (SUBJAREA, “COMP”) OR LIMIT-TO (SUBJAREA, “ENGI”) OR LIMIT-TO (SUBJAREA, “GEOG”)) AND (EXCLUDE (DOCTYPE, “le”) OR EXCLUDE (DOCTYPE, “re”)) AND (LIMIT-TO (LANGUAGE, “English”)) AND (EXCLUDE (EXACTKEYWORD, “China”) OR EXCLUDE (EXACTKEYWORD, “Antennas”) OR EXCLUDE (EXACTKEYWORD, “Crops”) OR EXCLUDE (EXACTKEYWORD, “Times Series”) OR EXCLUDE (EXACTKEYWORD, “Soils”) OR EXCLUDE (EXACTKEYWORD, “Time Series”) OR EXCLUDE (EXACTKEYWORD, “Article”) OR EXCLUDE (EXACTKEYWORD, “India”) OR EXCLUDE (EXACTKEYWORD, “United States”) OR EXCLUDE (EXACTKEYWORD, “Engines”) OR EXCLUDE (EXACTKEYWORD, “Image Enhancement”) OR EXCLUDE (EXACTKEYWORD, “Image Processing”) OR EXCLUDE (EXACTKEYWORD, “United Kingdom”) OR EXCLUDE (EXACTKEYWORD, “Study Areas”)) EUROPEAN JOURNAL OF REMOTE SENSING 9 relationships within large datasets derived from aca- demic publications (Jan van Eck & Waltman, 2023). Its primary advantages include handling extensive bibliographic data and generating detailed visual maps that illustrate complex networks of authorship, co-citation, and keyword co-occurrence (Centre for Science and Technology Centre for Science and Technology & Studies, 2023). VOSviewer supports the use of thesaurus files to manage synonyms and term variations, ensuring consistency in data analysis (Centre for Science and Technology Centre for Science and Technology & Studies, 2023). Its visuali- sation capabilities are particularly robust, allowing researchers to create both network and density visua- lisations that reveal underlying patterns and trends in the data (Jan van Eck & Waltman, 2023). These fea- tures make VOSviewer an indispensable tool for our research, as it enables the effective synthesis and inter- pretation of vast amounts of bibliometric data, facil- itating a deeper understanding of the research landscape in vegetation mapping through remote sen- sing and advanced machine learning techniques (Matyukira & Mhangara, 2023a). Building on VOSviewer’s robust capabilities for constructing and visualising bibliometric networks, our research methodology involved seamless integra- tion of this tool with our Scopus-derived bibliometric data. By importing data in CSV format, we captured essential details such as authors, titles, sources, and citation counts, which laid the foundation for our subsequent analysis. The creation of a thesaurus file was a critical step in managing synonyms and term variations, ensuring consistency across our dataset. With VOSviewer’s intuitive interface, we generated a map using these bibliographic data, applying the thesaurus file during import to standardise terms. Both computational efficiency and expert judgment guided our systematic approach to determining cri- teria for item inclusion. This allowed us to focus on cluster formation that was relevant to our research topic, creating a clear and informative visualisation that highlighted key patterns and relationships within the field of vegetation mapping through remote sen- sing and advanced machine learning techniques. This meticulous process resulted in a network visualisation that effectively displayed nodes and edges representing elements such as authors or key- words and their connections. Transitioning to density visualisation mode enabled us to analyse item distri- bution and identify areas of high concentration, dis- cern clusters, and detect trends that emphasised significant themes or authors. After conducting a systematic study, we exported our visualisations as image files for presentation purposes and optionally extracted the raw data for further detailed analysis. This methodical approach enhanced our understand- ing of the research landscape and provided us with valuable insights into the progression of vegetation mapping through remote sensing. Results and analysis Publication trends in remote sensing and vegetation mapping The analysis of the publication trend from Scopus, as illustrated in Figure 1(a) and Table 2, highlights the dynamic growth and increasing interest in applying remote sensing for vegetation mapping, particularly with integrating advanced machine learning techni- ques. The number of articles published yearly has notably increased, particularly recently. For instance, starting with a modest count of 1 article per year from 2002 to 2004 and sporadic increases thereafter, the trend significantly accelerated from 2018 onwards. The number of publications rose from 6 articles in 2018 and 2019 to a remarkable 48 articles in 2023. This upward trend underscores the critical importance and necessity of continuing research in this field, reflecting the growing recognition of the value of combining remote sensing technologies with advanced analytical methods to enhance vegetation mapping efforts. The total number of articles reaching 162 by 2024 further emphasises this research area’s expanding scope and relevance. Remote sensing technologies, coupled with machine learning algorithms such as random forest, support vector machines, neural networks, and XGBRFClassifier, have significantly enhanced our ability to monitor, analyse, and understand vegetation dynamics across various spatial and temporal scales (Belgiu & Drăgu, 2016; Hansen et al., 2013; Maxwell et al., 2018; Song et al., 2016). The sharp rise in pub- lications from 2019 to 2023 reflects the technological advancements and increased accessibility of these tools, making conducting more precise and compre- hensive vegetation studies possible (Maxwell et al., 2018). Continuing research in this area is essential for several reasons. Firstly, ongoing climate change and environmental degradation pose significant challenges to global ecosystems. Advanced remote sensing and machine learning techniques are crucial for monitor- ing these changes, providing timely data that can inform conservation and management strategies (Joshi et al., 2016; Pham et al., 2019; Wulder et al., 2004). Secondly, integrating these technologies facil- itates the detection of subtle changes in vegetation health and distribution, which are vital for early warn- ing systems and disaster management, such as in the case of forest fires, pest infestations, and drought con- ditions (Hansen et al., 2013). Lastly, continuous improvement and adaptation of these methods will drive innovation in environmental research, allowing 10 C. MATYUKIRA AND P. MHANGARA Figure 1. Examination of research publications in remote sensing and vegetation mapping. Table 2. Worldwide articles per year from the Scopus database from 2000 to 2024. Year 2002 2004 2006 2007 2008 2009 2010 2011 2012 2013 2014 No. Articles 1 1 1 3 1 1 3 2 2 1 4 Year 2015 2016 2018 2019 2020 2021 2022 2023 2024 Total No. Articles 2 2 6 6 6 11 35 48 26 162 EUROPEAN JOURNAL OF REMOTE SENSING 11 for more accurate predictions and effective interven- tions. As global environmental challenges become more complex, the need for advanced tools to monitor and mitigate their impacts becomes increasingly urgent, making sustained research in remote sensing and vegetation mapping not just necessary but imperative for future sustainability (Bhandari et al., 2012; Jiang et al., 2007). Geographical breakdown of research on remote sensing and vegetation mapping China has the highest number of research publications as shown in Figure 1(b), in the field of remote sensing and vegetation mapping, indicating its leadership in this area. The United States closely follows China, also making significant contributions and showing leader- ship in this research field. India’s ranking as the third- largest contributor underscores its increasing empha- sis on using remote sensing technology for the pur- pose of environmental monitoring and vegetation analysis. Germany, the United Kingdom, France, Italy, Spain, Belgium, and Brazil, among other European nations, demonstrate significant contribu- tions, indicating a broad international interest in this topic. Many geographical locations emphasise the worldwide significance and cooperative character of research in remote sensing and vegetation mapping. The reason for China’s significant role in remote sensing research is its huge investment in satellite technology and environmental monitoring programs. The United States’ robust position results from its enduring leadership in technology innovation and environmental research. Both nations have developed comprehensive remote-sensing infrastructure and research institutes, which have facilitated progress in this domain (Chen & Chen, 2018; Wulder et al., 2018). India’s rising production is probably fueled by its focus on sustainable agriculture and forest manage- ment, bolstered by national efforts such as the Indian Space Research Organisation’s remote-sensing pro- jects (Roy et al., 2015). The European nations have made noteworthy advances in comprehending and tackling environmental difficulties by using improved remote sensing methods, as shown by the studies conducted by Anjos et al. (2015), Dalouman et al. (2023), and de Souza Soler & Verburg (2010). The worldwide dissemination of research emphasises the significance of international cooperation and informa- tion sharing in promoting remote sensing for vegeta- tion mapping and environmental sustainability. Distribution of publications in remote sensing and vegetation mapping by source The analysis of remote sensing and vegetation map- ping research conducted between 2002 and 2024 reveals a substantial rise in publications, particularly from 2018 onwards, as depicted in Figure 1(c) and Table 3, reaching a prominent peak in 2023. The journal “Remote Sensing” has the largest number of articles, especially from 2019 onwards, highlighting its significant role in sharing research on this subject. This increase in publications may be credited to the rapid progress in remote sensing technology and the expanding use of machine learning methods in vege- tation research. Publications such as “Remote Sensing of Environment” and “International Journal of Remote Sensing”, as well as conferences like the “International Geoscience and Remote Sensing Symposium (IGARSS)”, also make significant contri- butions to the field of remote sensing in environmen- tal monitoring and management. The journal “Remote Sensing” showed a notable increase in articles, espe- cially from 2019 onwards, with 11 articles in 2021 and 9 in 2022, Table 3. Similarly, “Remote Sensing of Environment” maintained a consistent presence, con- tributing significantly to the field. The “International Journal of Remote Sensing” and “IGARSS” confer- ences also played pivotal roles, with a notable increase in their contributions in the latter years of the study period sources reflect the diverse interests and wide range of applications of remote sensing (Hansen et al., 2013; Maxwell et al., 2018). The significant rise in publications in the journal “Remote Sensing” corresponds to the wider pattern of incorporating sophisticated analytical techniques, such as deep learning and neural networks, in Table 3. Top 5 articles per journal from the Scopus database from 2000 to 2024. Year 2002 2004 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Remote Sensing 1 1 Remote Sensing of Environment 1 1 1 1 1 1 Forests International Geoscience and Remote Sensing Symposium IGARSS 1 International Journal of Remote Sensing Year 2016 2018 2019 2020 2021 2022 2023 2024 Remote Sensing 2 11 9 6 Remote Sensing of Environment 1 1 1 2 1 Forests 1 1 2 3 2 International Geoscience and Remote Sensing Symposium IGARSS 1 5 International Journal of Remote Sensing 2 1 2 2 12 C. MATYUKIRA AND P. MHANGARA vegetation mapping research. This tendency empha- sises the crucial significance of the journal in advan- cing groundbreaking approaches and applications in the discipline. The ongoing contributions from “Remote Sensing of Environment” and the “International Journal of Remote Sensing” underscore the well-established venues these publications provide for impactful research (Belgiu & Drăgu, 2016; Wulder et al., 2018). Moreover, the inclusion of IGARSS in the publishing landscape indicates the significance of con- ference proceedings in spreading advanced research and promoting cooperation among scientists and practitioners. The increase in publications from these sources highlights the growing research community and the urgent need for more investigation and advancement in remote sensing technologies to tackle intricate environmental issues (Joshi et al., 2016; Matyukira & Mhangara, 2023a; Pham et al., 2019; Roy et al., 2015). Types of documents in remote sensing and vegetation mapping research The distribution of various remote sensing and vege- tation mapping publications, Figure 1(d), demon- strates that a substantial majority, 85.8%, of the papers are research articles, highlighting a strong emphasis on original research and experimental dis- coveries in this discipline. Conference papers account for 12.3% of the publications, underscoring the sig- nificance of academic conferences in sharing recent progress and promoting dialogue among scholars. The other categories, such as book chapters, letters, and reviews, account for just 0.6% of the total publications. This indicates that these publication platforms are less often used in this field of study. The prevalence of research publications highlights the field’s dependence on comprehensive experi- mental investigations and empirical data to progress knowledge and innovate new technologies. Research publications include in-depth analysis of methodol- ogy, findings, and debates, which are crucial for the scientific community to expand on previous research and foster innovation (Hansen et al., 2013; Maxwell et al., 2018). The significant percentage of conference papers demonstrates the ever-changing nature of this field of research, where quick techno- logical progress and the implementation of novel methods require frequent sharing of knowledge and collaboration, which is facilitated by conferences like the International Geoscience and Remote Sensing Symposium (IGARSS) (Joshi et al., 2016; Roy et al., 2015). The limited number of reviews and book chapters suggests that although these for- mats are useful for consolidating existing knowledge and offering a broader perspective, they are not the main means of presenting innovative research find- ings in remote sensing and vegetation mapping (Pham et al., 2019). Subject Area Distribution of Research in Remote Sensing and Vegetation Mapping The extensive array of fields involved in remote sen- sing and vegetation mapping indicates that the major- ity of papers, Figure 1(e), namely 33.3%, pertain to Earth and Planetary Sciences, highlighting the essen- tial relevance of geospatial science in this particular study domain. The prevalence of this dominance underscores the need to comprehend the Earth’s sur- face and atmosphere since they are pivotal for precise vegetation mapping and environmental monitoring (Chen & Chen, 2018; Wulder et al., 2018). Agricultural and Biological Sciences represent 17.5% of the published works, emphasising the practical use of remote sensing in agriculture. This includes activ- ities such as monitoring crops, estimating yields, and implementing sustainable land management practices (Roy et al., 2015; Shoko et al., 2015, 2016; Wulder et al., 2018). The substantial contributions from Environmental Science (11.6%) and Computer Science (13.7%) emphasise the multidisciplinary char- acter of this subject, combining environmental mon- itoring with sophisticated computational tools and data analysis methodologies (Maxwell et al., 2018, 2021; Pham et al., 2019). The inclusion of other disciplines, Figure 1(e), such as Engineering (7.4%), Physics and Astronomy (7.0%), and Social Sciences (3.2%), underscores the diverse range of applications and research interests in remote sensing. Engineering plays a crucial role in advancing the development of novel remote sensing devices and platforms, including unmanned aerial vehicles (UAVs) and satellites. These technologies are vital for collecting data (Hansen et al., 2013; Matyukira & Mhangara, 2023a). Physics and Astronomy are essen- tial in remote sensing technologies since they include physical concepts to comprehend how the electromag- netic spectrum interacts with plants. Incorporating Social Sciences indicates a desire to explore the social consequences and policy dimensions of environmen- tal monitoring and land use management. The wide range of subject areas covered in remote sensing and vegetation mapping research demonstrates this field’s extensive practicality and interdisciplinary character. This emphasises the importance of ongoing collabora- tion among different scientific disciplines to tackle complex environmental issues effectively (Dalouman et al., 2023; Joshi et al., 2016; Roy et al., 2015; Chen & Chen, 2018). EUROPEAN JOURNAL OF REMOTE SENSING 13 Analysis of research affiliations in remote sensing and vegetation mapping According to the prominent research institutes in remote sensing and vegetation mapping, the Chinese Academy of Sciences (CAS) stands out as the institu- tion with the most publications, Figure 1(f), surpass- ing other affiliations by a substantial margin. China’s dominant position in scientific research and environ- mental monitoring may be linked to its significant investment and strategic focus on remote sensing technology (Chen & Chen, 2018). Chinese universi- ties, such as the University of Chinese Academy of Sciences and the Aerospace Information Research Institute, part of the CAS network, play a significant role in promoting remote sensing research. These institutions are well known for their thorough research programs and large cooperation networks, which enable them to produce influential research results (Liu & Zhang, 2024; Rakhmankulova et al., 2024; Xu & Bai, 2024). Additional significant associations include the Ministry of Education and the Institute of Geographic Sciences and Natural Resources Research, both of the People’s Republic of China and the Ministry of Natural Resources. These organisations represent the Chinese government’s comprehensive scientific research strat- egy to enhance environmental management and policy development (Wulder et al., 2018). Notable interna- tional collaborators include Texas A&M University and the Indian Space Research Organisation (ISRO). Texas A&M University is renowned for its strong research programs in geospatial science and engineer- ing, while ISRO’s proficiency in satellite technology and remote sensing applications showcases India’s increas- ing focus on sustainable agriculture and natural resource management (Roy et al., 2015). The wide range of these associations demonstrates the worldwide cooperation and multidisciplinary character of remote sensing and vegetation mapping research, highlighting the significance of multinational alliances in tackling intricate environmental issues (Anjos et al., 2015; Bhandari et al., 2012; Chen & Chen, 2018; Dalouman et al., 2023; Jiang et al., 2007; Roy et al., 2015; Wulder et al., 2018). Summary of the significance and need of remote sensing and vegetation mapping research using the Scopus database The analysis of the Scopus database reveals a substantial increase in publications related to remote sensing and vegetation mapping. This development may be attributed to technological improvements and machine learning methods like Random Forest and neural networks. The data shows a steady increase in the number of articles from 2002 to 2016, foll owed by a significant surge from 2018 to 2023, with the number of articles peaking at 48 in 2023. This substantial growth highlights the urgent need for con- tinuous research to tackle climate change and envir- onmental difficulties. China and the United States are at the forefront regarding publication volume, indicating significant expenditures in satellite technology and environmental monitoring. Well-known publications such as “Remote Sensing” and “Remote Sensing of Environment” are crucial in spreading new approaches, as evidenced by their consistent and increasing contributions to the field. For instance, “Remote Sensing” saw a notable rise in articles, particularly from 2019 onwards, while “Remote Sensing of Environment” has maintained a consistent presence. The significant increase in publications, parti- cularly between 2019 and 2023, underscores the field’s dependence on factual information and cooperation across many disciplines to enhance understanding and encourage creativity. This trend underscores the critical importance of continuing research and innovation in remote sensing and vegetation mapping to address pressing environmental challenges. Interconnected themes in remote sensing and vegetation mapping research Examination of VOSviewer network visualization The VOSviewer network visualisation Figure 2(a) and Table 4 showcase the interconnections between important issues in vegetation mapping research using remote sensing. The primary focal points consist of “vegetation mapping”, “remote sensing”, and “machine learning”, signifying their pivotal positions within this domain, as indicated by their high total link strengths and occurrences. Specifically, “vegeta- tion mapping” has a total link strength of 159 and 59 occurrences, “remote sensing” has a total link strength of 157 and 58 occurrences, and “machine learning” has a total link strength of 141 and 50 occurrences. The links between these nodes and other terms like “land use”, “land cover”, “climate change”, “evapo- transpiration”, “aboveground biomass”, and “accuracy assessment” suggest that researchers often investigate these subjects together. For instance, “land use” and “land cover” both have a total link strength of 38 and 10 occurrences each, highlighting their connection to the primary focal points. The term “accuracy assess- ment” is also prominent, with a total link strength of 49 and 14 occurrences, reflecting its importance in validating the research methods used. The interlinking of several disciplines highlights the diverse character of the study, as it combines sophisticated machine- learning methods to tackle different areas of environ- mental monitoring and vegetation analysis. The sig- nificant correlations among these topics indicate a resilient and expanding collection of research that 14 C. MATYUKIRA AND P. MHANGARA Figure 2. Visualisation of key research themes in vegetation mapping using remote sensing, (a) network visualisation, (b) density visualisation. Table 4. Network visualisation map structural information. Label cluster weight Weight Weight score Vegetation mapping 1 8 159 59 22 Remote sensing 1 8 157 58 22 Machine learning 1 8 141 50 23 Land use 3 7 38 10 16 Land cover 2 7 38 10 18 Evapotranspiration 3 7 19 5 3 climate change 1 8 33 8 11 Accuracy assessment 2 7 49 14 26 Aboveground biomass 1 4 16 5 4 EUROPEAN JOURNAL OF REMOTE SENSING 15 uses remote sensing and machine learning to enhance our understanding of vegetation dynamics and asso- ciated environmental processes. Examination of VOSviewer density visualisation The VOSviewer density visualisation Figure 2(b) showcases the significance and dominance of major research topics in vegetation mapping via remote sen- sing. The nodes that stand out the most are “vegeta- tion mapping”, “remote sensing”, and “machine learning”, indicating their significant contributions to the literature. The research area encompasses sev- eral interconnected subjects such as “land use”, “land cover”, “climate change”, “evapotranspiration”, “aboveground biomass”, and “accuracy assessment”, highlighting the complex nature of this field of study. The high density at these nodes indicates significant research activity and strong linkages, highlighting the need to use advanced machine learning algorithms to improve the precision and effectiveness of remote sensing methods for vegetation monitoring. Research topics such as vegetation mapping, remote sensing, and machine learning are fundamental while studying land use and land cover change is crucial for compre- hending environmental effects. It is important to con- sider climate change when researching how vegetation changes over time. We use metrics like evapotran- spiration, which is the combined process of water evaporation from the land and plant transpiration, and aboveground biomass, which is the total weight of plant material above the ground, as important indi- cators of the overall health of vegetation. The valida- tion of remote sensing data remains dependent on the accuracy evaluation. The Cradle Nature Reserve in the COHWHS offers a varied karst topography that is ideal for sophisticated approaches to monitoring and assessing vegetation efficiently. This setting empha- sises the potential for effective study in vegetation dynamics and environmental monitoring by utilising the interrelated themes found in the visualisation processes. Findings from the scientometric review (1) The analysis highlights a substantial increase in publications related to remote sensing and vegetation mapping from 2002 to 2024, with a marked acceleration from 2018 onwards. The number of articles rose significantly from 6 in 2018 and 2019 to 48 in 2023, reach- ing a total of 162 articles by 2024. This upward trend underscores the growing recognition of the importance of combining remote sensing technologies with advanced analytical meth- ods for vegetation mapping. (2) There has been a significant rise in publica- tions reflecting advancements in remote sensing technology and the expanding use of machine learning algorithms such as random forests, support vector machines, neural net- works, and XGBRFClassifier. These technolo- gies have enhanced the ability to monitor, analyse, and understand vegetation dynamics across various spatial and temporal scales. (3) The study reveals that China leads in the number of research publications, followed clo- sely by the United States. India ranks as the third-largest contributor, highlighting its increasing emphasis on remote sensing tech- nology for environmental monitoring and vegetation analysis. Significant contributions also come from European nations, emphasis- ing the global and cooperative nature of research in this field. (4) Journals like “Remote Sensing”, “Remote Sensing of Environment”, and the “International Journal of Remote Sensing”, along with conferences such as the “International Geoscience and Remote Sensing Symposium (IGARSS)”, play crucial roles in disseminating research. “Remote Sensing” showed a notable increase in articles, particularly from 2019 onwards, reflecting its significant role in advancing the field. (5) The VOSviewer network visualisation and Table 4 identify “vegetation mapping”, “remote sensing”, and “machine learning” as the primary focal points, with high total link strengths and occurrences. “Vegetation map- ping” has a total link strength of 159 with 59 occurrences, “remote sensing” has 157 with 58 occurrences, and “machine learning” has 141 with 50 occurrences. (6) The analysis reveals significant interconnec- tions between primary themes and other important topics such as “land use”, “land cover”, “climate change”, “evapotranspira- tion”, “aboveground biomass”, and “accuracy assessment”. For instance, “land use” and “land cover” both have a total link strength of 38 and 10 occurrences each, highlighting their relevance in vegetation mapping research. (7) Research Gaps and Emerging Areas: ● Despite its importance, “evapotranspira- tion” has a relatively low occurrence (5) and total link strength (19), as shown in Table 4, indicating a gap in comprehensive studies. ● Aboveground Biomass estimation also has low occurrences (5) and total link strength (16), suggesting the need for more research in biomass estimation using remote sensing technologies. 16 C. MATYUKIRA AND P. MHANGARA ● As shown in Table 4, climate change is identified as a crucial but less integrated area, presenting opportunities for further investigation. Climate Change has (8) occurrences and a total link strength of 33. (8) Most publications (85.8%) are research arti- cles emphasising original research and experi- mental discoveries. Conference papers account for 12.3%, highlighting the impor- tance of academic conferences in sharing recent progress. Other categories, such as book chapters, letters, and reviews, account for just 0.6%, indicating that these formats are less commonly used in this field. (9) The subject area distribution reveals an exten- sive range of fields involved, including Earth and Planetary Sciences (33.3%), Agricultural and Biological Sciences (17.5%), Environmental Science (11.6%), and Computer Science (13.7%). Other fields such as Engineering, Physics and Astronomy, and Social Sciences also contribute, underscoring the multidisciplinary nature of remote sensing and vegetation mapping research. (10) The Chinese Academy of Sciences (CAS) is the leading institution in terms of publica- tions, supported by significant national invest- ments in remote sensing technology. Notable international collaborators include Texas A&M University and the Indian Space Research Organisation (ISRO), demonstrating the global and collaborative nature of the research. These findings highlight the evolving landscape of remote sensing and vegetation mapping research, emphasising both established areas and potential gaps for future investigation. Future research directions, emerging technologies, and potential challenges The future of remote sensing and vegetation mapping research depends on the integration of rising technol- ogy and the resolution of existing issues. There is a distinct necessity to enhance research in domains such as evapotranspiration and aboveground biomass estimation, where substantial gaps persist. Future stu- dies should focus on leveraging multi-source data fusion, combining optical, radar, and LiDAR data to enhance accuracy. Additionally, climate change remains underexplored in this context, and future research should aim to better understand its impact on vegetation dynamics by utilizing long-term satellite data and machine learning models for predictive ana- lysis (Balestra et al., 2024; Jin & Mountrakis, 2022; Li et al., 2024a; Zhang, 2010). Deep learning techniques, including convolutional neural networks (CNNs), provide a means to augment the extraction of vegeta- tive characteristics and refine temporal analysis. Integrating UAVs with satellite data to extend their application beyond small-scale investigations could address existing limits concerning flight range and regulatory restrictions (Nex & Remondino, 2014). Emerging technologies such as hyperspectral ima- ging, advanced LiDAR, and synthetic aperture radar (SAR) hold great promise for enhancing the accuracy of vegetation mapping. These technologies, in con- junction with edge computing and AI, can enable real- time monitoring and analysis; nevertheless, their inte- gration necessitates addressing data processing and storage difficulties (Khonina et al., 2024; Kuras et al., 2021; Li et al., 2024b). Furthermore, the advancement of quantum computing may provide a remedy for managing extensive datasets, facilitating swifter and more efficient analysis. Ensuring global access to these technologies is essential for equitable advance- ment, as numerous locations, especially in poor nations, lack the resources for efficient implementa- tion of these innovations (Khonina et al., 2024; McKinsey & Company, 2024; World Economic Forum, 2022). Fostering international collaborations will be key to addressing global environmental chal- lenges, ensuring that research benefits both local eco- systems and the global scientific community (Avilés Irahola et al., 2022; Mariani et al., 2022). Conclusion This study highlights significant advancements in combining remote sensing technologies with machine learning algorithms for vegetation mapping. A comprehensive literature review using the Scopus database provided an in-depth understanding of the current research landscape. Employing a systematic search strategy focused on key terminologies related to vegetation mapping and advanced machine learn- ing techniques, we gathered and analysed high-quality studies, ensuring a thorough exploration of the field. Integrating VOSviewer into our methodology enabled effective visualisation and analysis of bibliometric data, revealing key patterns and relationships within the research field. Network and density visualisations highlighted significant clusters and trends, enhancing our understanding of the research progression and focus areas on vegetation mapping through remote sensing. The systematic review via Scopus revealed a robust and growing body of research emphasising the effectiveness of remote sensing technologies combined with machine learning algorithms. Techniques such as random forest, support vector machines, and neural networks have significantly enhanced remote sensing data analysis. VOSviewer EUROPEAN JOURNAL OF REMOTE SENSING 17 proved valuable in visualising complex relationships and trends and identifying influential research areas, key authors, and collaborative networks. Our research underscores the critical importance of advanced remote sensing and machine learning tech- niques in addressing global environmental chal- lenges. These technologies show significant potential in tackling issues like land cover change, vegetation health assessment, and biomass estima- tion. However, several research gaps remain. There is a need for more comprehensive studies on evapo- transpiration and aboveground biomass, which have relatively low occurrences and link strengths. Additionally, integrating climate change data with remote sensing technologies is underexplored, pre- senting opportunities for further investigation. Advances in satellite and UAV technologies, com- bined with sophisticated data processing algorithms, enable detailed and large-scale environmental moni- toring, supporting more accurate and timely decision- making in environmental management. Continuous research and innovation are essential for developing effective strategies for environmental management, conservation, and sustainable development. Highlights ● Remote sensing and vegetation mapping research has grown, notably using random forest, support vector machines, neural networks, and XGBRFClassifier. ● Chinese researchers publish the most, followed by the USA and India, suggesting worldwide interest and investment in remote sensing and vegetation mapping. ● More research papers are published in key jour- nals like “Remote Sensing” and “Remote Sensing of Environment” and at significant conferences like IGARSS. ● Vegetation mapping, remote sensing, and machine learning are strongly linked in the VOSviewer network and density visualisations Disclosure statement No potential conflict of interest was reported by the author(s). Funding This research was funded through a bursary by the Lee Burger Foundation. Author contributions Conceptualisation, P.M. and C.M.; methodology, P.M. and C.M.; software, C.M.; validation, C.M.; formal analysis, C. M.; investigation, C.M.; writing – original draft preparation, C.M and P.M.; writing – review and editing, C.M and P.M.; visualisation, C.M.; supervision, P.M. All authors have read and agreed to the published version of the manuscript. Data availability statement The corresponding author can provide the datasets used in the current study upon reasonable request. References Abebe, G., Getachew, D., & Ewunetu, A. (2022). Analysing land use/land cover changes and its dynamics using remote sensing and GIS in Gubalafito district, Northeastern Ethiopia. SN Applied Sciences, 4(1). https://doi.org/10.1007/s42452-021-04915-8 Allen, R. G., Tasumi, M., & Trezza, R. (2007). Satellite-based energy balance for mapping evapotranspiration with internalized calibration (metric)—model. Journal of Irrigation & Drainage Engineering, 133(4), 380–394. https://doi.org/10.1061/(asce)0733-9437(2007)133:4(380) Anjos, D., Lu, D., Dutra, L., & Sant’anna, S. (2015). Change detection techniques using multisensor data. Remotely Sensed Data Characterization, Classification, and Accuracies, 1, 375–395. Boca Raton: CRC Press. https:// doi.org/10.1201/b19355 Arifeen, H. M., Phoungthong, K., Mostafaeipour, A., Yuangyai, N., Yuangyai, C., Techato, K., & Jutidamrongphan, W. (2021). Determine the land-use land-cover changes, urban expansion and their driving factors for sustainable development in gazipur Bangladesh. Atmosphere, 12(10), 1353. https://doi.org/ 10.3390/atmos12101353 Asokan, A., & Anitha, J. (2019). Change detection techni- ques for remote sensing applications: A survey. Earth Science Informatics, 12(2), 143–160. https://doi.org/10. 1007/s12145-019-00380-5 Assede, E. S. P., Orou, H., Biaou, S. S. H., Geldenhuys, C. J., Ahononga, F. C., & Chirwa, P. W. (2023). Understanding drivers of land use and land cover change in Africa: A review. Current Landscape Ecology Reports, 8(2), 62–72. https://doi.org/10.1007/s40823-023-00087-w Avilés Irahola, D., Mora-Motta, A., Barbosa Pereira, A., Bharati, L., Biber-Freudenberger, L., Petersheim, C., Quispe-Zuniga, M. R., Schmitt, C. B., & Youkhana, E. (2022). Integrating scientific and local knowledge to address environmental conflicts: The role of academia. Human Ecology, 50(5), 911–923. https://doi.org/10.1007/ s10745-022-00344-2 Ayhan, B., Kwan, C., Budavari, B., Kwan, L., Lu, Y., Perez, D., Li, J., Skarlatos, D., & Vlachos, M. (2020). Vegetation detection using deep learning and conven- tional methods. Remote Sensing, 12(15), 2502. https:// doi.org/10.3390/RS12152502 Balestra, M., Marselis, S., Sankey, T. T., Cabo, C., Liang, X., Mokroš, M., Peng, X., Singh, A., Stereńczak, K., Vega, C., Vincent, G., & Hollaus, M. (2024). LiDAR data fusion to improve forest attribute estimates: A review. Current Forestry Reports, 10(4), 281–297. https://doi.org/10. 1007/s40725-024-00223-7 Barry, R. G., & Blanken, P. D. (Eds.). (2016). Topoclimatic effects on microclimate. In Microclimate and local climate (pp. 261–274). Cambridge University Press. https://doi. org/10.1017/CBO9781316535981.014 18 C. MATYUKIRA AND P. MHANGARA https://doi.org/10.1007/s42452-021-04915-8 https://doi.org/10.1007/s42452-021-04915-8 https://doi.org/10.1061/(asce)0733-9437(2007)133:4(380) https://doi.org/10.1061/(asce)0733-9437(2007)133:4(380) https://doi.org/10.1201/b19355 https://doi.org/10.1201/b19355 https://doi.org/10.3390/atmos12101353 https://doi.org/10.3390/atmos12101353 https://doi.org/10.1007/s12145-019-00380-5 https://doi.org/10.1007/s12145-019-00380-5 https://doi.org/10.1007/s40823-023-00087-w https://doi.org/10.1007/s40823-023-00087-w https://doi.org/10.1007/s10745-022-00344-2 https://doi.org/10.1007/s10745-022-00344-2 https://doi.org/10.3390/RS12152502 https://doi.org/10.3390/RS12152502 https://doi.org/10.1007/s40725-024-00223-7 https://doi.org/10.1007/s40725-024-00223-7 https://doi.org/10.1017/CBO9781316535981.014 https://doi.org/10.1017/CBO9781316535981.014 Bartold, M., & Kluczek, M. (2024). Estimating of chlorophyll fluorescence parameter Fv/Fm for plant stress detection at peatlands under Ramsar convention with sentinel-2 satellite imagery. Ecological Informatics, 81, 81. https:// doi.org/10.1016/j.ecoinf.2024.102603 Bauer, L., Huth, A., Bogdanowski, A., Müller, M., & Fischer, R. (2024). Edge effects in Amazon forests: Integrating remote sensing and modelling to assess changes in biomass and productivity. Remote Sensing, 16(3), 501. https://doi.org/10.3390/rs16030501 Bazrafkan, A., Delavarpour, N., Oduor, P. G., Bandillo, N., & Flores, P. (2023). An overview of using unmanned aerial system mounted sensors to measure plant above-ground biomass. Remote Sensing, 15(14). Multidisciplinary Digital Publishing Institute (MDPI). https://doi.org/10.3390/rs15143543 Belgiu, M., & Drăgu, L. (2016). Random forest in remote sensing: A review of applications and future directions. In ISPRS journal of photogrammetry and remote sensing (Vol. 114, pp. 24–31). Elsevier B.V. https://doi.org/10. 1016/j.isprsjprs.2016.01.011 Bendig, J., Bolten, A., Bennertz, S., Broscheit, J., Eichfuss, S., & Bareth, G. (2014). Estimating biomass of barley using crop surface models (CSMs) derived from uav-based RGB imaging. Remote Sensing, 6(11), 10395–10412. https://doi.org/10.3390/rs61110395 Berger, K., Atzberger, C., Danner, M., D’Urso, G., Mauser, W., Vuolo, F., & Hank, T. (2018). Evaluation of the PROSAIL model capabilities for future hyperspectral model environments: A review study. Remote Sensing, 10 (1). MDPI AG. 85. https://doi.org/10.3390/rs10010085 Berhane, T. M., Lane, C. R., Wu, Q., Autrey, B. C., Anenkhonov, O. A., Chepinoga, V. V., & Liu, H. (2018). Decision-tree, rule-based, and random forest classifica- tion of high-resolution multispectral imagery for wetland mapping and inventory. Remote Sensing, 10(4), 580. https://doi.org/10.3390/rs10040580 Bhandari, A. K., Kumar, A., & Singh, G. K. (2012). Feature extraction using normalized difference vegetation index (NDVI): A case study of Jabalpur City. Procedia Technology, 6, 612–621. https://doi.org/10.1016/j.protcy. 2012.10.074 Borsah, A. A., Nazeer, M., & Wong, M. S. (2023). LIDAR- Based forest biomass remote sensing: A review of metrics, methods, and assessment criteria for the selection of allometric equations. Forests, 14(10). Multidisciplinary Digital Publishing Institute (MDPI). 2095. https://doi. org/10.3390/f14102095 Bramer, I., Anderson, B. J., Bennie, J., Bladon, A. J., De Frenne, P., Hemming, D., Hill, R. A., Kearney, M. R., Körner, C., Korstjens, A. H., Lenoir, J., Maclean, I. M. D., Marsh, C. D., Morecroft, M. D., Ohlemüller, R., Slater, H. D., Suggitt, A. J., Zellweger, F., & Gillingham, P. K. (2018). Chapter three - advances in monitoring and modelling climate at ecologically relevant scales. In D.A. Bohan, A.J. Dumbrell, G. Woodward, & M. Jackson (Eds.), Advances in ecological research (Vol. 58, pp. 101–161). Academic Press. https://doi.org/10. 1016/bs.aecr.2017.12.005 Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32. https://doi.org/10.1023/A:1010933404324 Breunig, M., Kuper, P., Reitze, F., Landgraf, S., Al-Doori, M., Stefanakis, E., Abdulmuttalib, H., & Kugler, Z. (2023). Improving data quality and management for remote sen- sing analysis: Use-cases and emerging research questions. ISPRS Annals of the Photogrammetry, Remote Sensing & Spatial Information Sciences, 10(1), 41–49. https://doi. org/10.5194/isprs-annals-X-1-W1-2023-41-2023 Brocard, G., Willebring, J. K., Scatena, F. N., & Van Stan, J. T. (2023, March). Shaping of topography by topographically-controlled vegetation in tropical mon- tane rainforest. PLoS One, 18(3), e0281835. https://doi. org/10.1371/journal.pone.0281835 Brown, D. R. N., Jorgenson, M. T., Kielland, K., Verbyla, D. L., Prakash, A., & Koch, J. C. (2016). Landscape effects of wildfire on permafrost distribution in interior Alaska derived from remote sensing. Remote Sensing, 8(8), 654. https://doi.org/10.3390/rs8080654 Burnham, J. F. (2006). Scopus database: A review. Biomedical Digital Libraries, 3(1), 1–8. https://doi.org/ 10.1186/1742-5581-3-1 Chaves, M. E. D., Picoli, M. C. A., & Sanches, I. D. (2020). Recent applications of landsat 8/OLI and Sentinel-2/MSI for land use and land cover mapping: A systematic review. Remote Sensing, 12(18). MDPI AG. 3062. https://doi.org/10.3390/rs12183062 Chen, J., & Chen, J. (2018). GlobeLand30: Operational glo- bal land cover mapping and big-data analysis. Science China Earth Sciences, 61(10), 1533–1534. https://doi. org/10.1007/s11430-018-9255-3 Chen, J. M., Chen, X., & Ju, W. (2013). Effects of vegetation heterogeneity and surface topography on spatial scaling of net primary productivity. Biogeosciences, 10(7), 4879–4896. https://doi.org/10.5194/bg-10-4879-2013 Chen, T., & Guestrin, C. (2016). Xgboost: A scalable tree boosting system. In Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, San Francisco California, USA, 13-17- August, 785–794. https://doi.org/10.1145/2939672. 2939785 . Chen, X., & Zhang, Y. (2023). The impact of vegetation phenology changes on the relationship between climate and net primary productivity in Yunnan, China, under global warming. Frontiers in Plant Science, 14, 14. https:// doi.org/10.3389/fpls.2023.1248482 Cheng, G., Huang, Y., Li, X., Lyu, S., Xu, Z., Zhao, H., Zhao, Q., & Xiang, S. (2024). Change detection methods for remote sensing in the Last Decade: A comprehensive review. Remote Sensing, 16(13), 2355. https://doi.org/10. 3390/rs16132355 Cheng, G., Huang, Y., Li, X., Lyu, S., Xu, Z., Zhao, Q., & Xiang, S. (2023). Change detection methods for remote sensing in the Last Decade: A comprehensive review. Current Landscape Ecology Reports, 9(2). http://arxiv. org/abs/2305.05813 Cohen, W. B., & Goward, S. N. (2004). Landsat’s role in ecological applications of remote sensing. BioScience, 54 (6), 535–545. https://doi.org/10.1641/0006-3568(2004) 054[0535]2.0.CO;2 Cohn, R., & Holm, E. (2021). Unsupervised machine learn- ing via transfer learning and k-means clustering to clas- sify materials image data. Integrating Materials and Manufacturing Innovation, 10(2), 231–244. https://doi. org/10.1007/s40192-021-00205-8 Corbane, C., Syrris, V., Sabo, F., Politis, P., Melchiorri, M., Pesaresi, M., Soille, P., & Kemper, T. (2021). Convolutional neural networks for global human settle- ments mapping from sentinel-2 satellite imagery. Neural Computing & Applications, 33(12), 6697–6720. https:// doi.org/10.1007/s00521-020-05449-7 Cortes, C., & Vapnik, V. (1995). Support-vector networks. Machine Learning, 20(3), 273–297. https://doi.org/10. 1007/BF00994018 EUROPEAN JOURNAL OF REMOTE SENSING 19 https://doi.org/10.1016/j.ecoinf.2024.102603 https://doi.org/10.1016/j.ecoinf.2024.102603 https://doi.org/10.3390