Figure - available from: Remote Sensing
This content is subject to copyright.
Spectral signatures of selected two training data polygons; (a) spectral signature of training polygon 331—Cropland and (b) spectral signature of training polygon 488—Grassland.

Spectral signatures of selected two training data polygons; (a) spectral signature of training polygon 331—Cropland and (b) spectral signature of training polygon 488—Grassland.

Source publication
Article
Full-text available
Land use, land-use change and forestry (LULUCF) is a greenhouse gas inventory sector that evaluates greenhouse gas changes in the atmosphere from land use and land-use change. This study focuses on the development of a Sentinel-2 data classification according to the LULUCF requirements on the cloud-based platform Google Earth Engine (GEE). The meth...

Similar publications

Article
Full-text available
Hypsometric or area-altitude analysis unveils the geomorphic evolution of a river basin and its interactions with tectonics, lithology, and climate. The results of hypsometric analysis i.e., hypsometric curve (HC) along with the hypsometric parameters (HP) namely, hypsometric integral (HI), skewness (SK), kurtosis (KU), density skewness (DSK), and...
Preprint
Full-text available
Al Saud (2023) worked on detailed cartography for all drainage systems of Saudi Arabia using SRTM DEM and topographic maps. However, a number of watercourses were noted with anomalous morphometric behavior. This includes abrupt termination of stream networks, traces for dry channels and presence of linear wet and saline soil horizons. This was attr...
Article
Full-text available
Quantification of drainage networks in relation to geomorphology, lithology, climate, and tectonics gives important evidence of an area’s hydro-geomorphic, denudation characteristics, and drainage development. The current research used the Shuttle Radar Topographic Mission (SRTM) DEM to investigate drainage morphometry and its impact upon the envir...
Article
Full-text available
The mutation of spaces observed in the Katangese Copper Belt (KCB) causes significant topographical changes. Some colonial geodetic markers are easily noticeable on many of the hills making up the KCB. These hills are subject to mining which ruins the completeness of the network of triangulations: geometric and trigonometric Katangese. In order to...
Article
Full-text available
The Puruogangri Ice Field (PIF), classified as an ultra-continental glacier, is considered extremely stable. However, several glaciers in this area have recently experienced surge events with significant instability and information on surge-type glaciers (STGs) in this region remains scarce. In this study, we identified six STGs and reported the ob...

Citations

... Land cover classification using satellite imagery requires the selection of an appropriate algorithm to produce accurate maps. One of the most widely used algorithms in image classification is Random Forest (RF) [6]. This algorithm is a machine learning method based on an ensemble of decision trees, known for its ability to handle complex and heterogeneous data [7]. ...
... SAGA GIS 9.6 software was used in this study to analyze and classify land cover using the Random Forest algorithm. Random Forest Classification is a method used for land cover classification by leveraging a robust and efficient machine learning algorithm [6]. This algorithm works by constructing multiple decision trees generated from random sampling of the training dataset, where each tree provides predictions based on features extracted from satellite imagery [6]. ...
... Random Forest Classification is a method used for land cover classification by leveraging a robust and efficient machine learning algorithm [6]. This algorithm works by constructing multiple decision trees generated from random sampling of the training dataset, where each tree provides predictions based on features extracted from satellite imagery [6]. The strength of Random Forest lies in its ability to reduce the risk of overfitting and improve classification accuracy through the majority voting technique across all constructed trees [12]. ...
Article
Full-text available
Rapid population growth and increased human activities have caused significant changes in land cover in this region. These changes can impact the environment, including a decline in environmental quality. This study aims to classify land cover in Ternate City using Sentinel-2A satellite imagery with the Random Forest method in SAGA GIS 9.6 software. The classification results show that out of a total area of 10,163.41 hectares, built-up land accounts for 2,242.60 hectares (22.07%), while forests dominate with an area of 5,854.76 hectares (57.61%). This research highlights the impact of urbanization and population growth on land cover changes, as well as the importance of managing and protecting natural resources to maintain ecosystem balance. By utilizing remote sensing technology and machine learning algorithms, this study is expected to contribute significantly to understanding land cover dynamics and supporting decision-making in spatial planning and environmental conservation in tropical regions.
... In this regard, several studies have used Landsat satellite images for land cover mapping in different ecosystems around the world (Rodriguez-Galiano et al., 2012;Taati et al., 2015;Nguyen et al., 2018;Jamali, 2021;Tikuye et al., 2023;Zafar et al., 2024). However, by starting Sentinel-2 satellite mission in 2015, many researchers have applied Sentinel-2A and Sentinel-2B for mapping the land uses, vegetation cover, and forest cover in different areas around the world (Puletti et al., 2017;Hawryło et al., 2018;Abdi, 2019;Junior et al., 2020;Svoboda et al., 2022;Zhou and Feng, 2023;Billah et al., 2023;Cecili et al., 2023;Lasko et al., 2024;Zhang et al., 2024). The results of these studies have demonstrated the good efficiency of Sentinel-2 satellite images to map the land uses, vegetation cover, and forest cover in different natural ecosystems. ...
... Regarding the efficiency of Sentinel-2A satellite images with good spatial resolution (10 m) and temporal resolutions (every 5 days) for forest cover and land use mapping (Puletti et al., 2017;Hawryło et al., 2018;Abdi, 2019;Junior et al., 2020;Svoboda et al., 2022;Zhou and Feng, 2023;Billah et al., 2023;Cecili et al., 2023;Lasko et al., 2024;Zhang et al., 2024), these satellite images along with ground truth data would be very useful for land cover management and forest monitoring in semi-arid areas of Iran. In addition, usage of new classification methods for land use mapping and also vegetation indices for forest density mapping in Zagros forest area of Iran will lead to get the accurate results of recent situation of these forests and land uses inside these forests. ...
... The efficiency of random forest algorithm for classifying the vegetation cover has already been confirmed in previous studies (Karlson et al., 2015;Jin et al., 2018;Abdi, 2019;Sellami and Rhinane, 2022;Zhou and Feng, 2023;Mahmoud et al., 2023;Lasko et al., 2024). In addition, the good accuracy of this algorithm for land cover mapping has been confirmed in some studies (Nguyen et al., 2018;Svoboda et al., 2022;Tikuye et al., 2023;Billah et al., 2023;Zafar et al., 2024;Sun et al., 2024) which is according to the results of this research. Random forest is a powerful, simple and flexible machine learning algorithm which is very efficient for organizing many data and analyses across space (Donges, 2024;Tech and Tales, 2023). ...
... Automated pattern recognition and classification in LULC mapping have been possible because of the combination of ML and GIS [5]. The complexity and variability of LULC data have been successfully handled by machine learning, in particular, supervised learning algorithms like Random Forest [6][7][8], Support Vector Machines [9][10][11][12], and Neural Networks [13,14], which provide high classification accuracy (> 85%) even in heterogeneous landscapes [15,16]. ...
Article
Full-text available
Mapping land use and land cover (LULC) is essential for comprehending changes in the environment and promoting sustainable planning. To achieve accurate and effective LULC mapping, this work investigates the integration of Geographic Information Systems (GIS) with Machine Learning (ML) methodology. Different types of land covers in the Lucknow district were classified using the Random Forest (RF) algorithm and Landsat satellite images. Since the research area consists of a variety of landforms, there are issues with classification accuracy. These challenges are met by combining supplementary data into the GIS framework and adjusting algorithm parameters like selection of cloud free images and homogeneous training samples. The result demonstrates a net increase of 484.59 km 2 in built-up areas. A net decrement of 75.44 km 2 was observed in forest areas. A drastic net decrease of 674.52 km 2 was observed for wetlands. Most of the wastelands have been converted into urban areas and agricultural land based on their suitability with settlements or crops. The classifications achieved an overall accuracy near 90%. This strategy provides a reliable way to track changes in land cover, supporting resource management, urban planning, and environmental preservation. The results highlight how sophisticated computational methods can enhance the accuracy of LULC evaluations.
... For both algorithms, to achieve higher classification accuracy, we also used the NDVI and NDBI spectral indices as input, similar to the study of Svoboda et al. (2022). These indices were used to train the models to better distinguish between vegetation and built-up areas. ...
Article
Full-text available
In recent decades, global climate change and rapid urbanization have aggravated the urban heat island (UHI) effect, affecting the well-being of urban citizens. Although this significant phenomenon is more pronounced in larger metropolitan areas due to extensive impervious surfaces, small- and medium-sized cities also experience UHI effects, yet research on UHI in these cities is rare, emphasizing the importance of land surface temperature (LST) as a key parameter for studying UHI dynamics. Therefore, this paper focuses on the evaluation of LST and land cover (LC) changes in the city of Prešov, Slovakia, a typical medium-sized European city that has recently undergone significant LC changes. In this study, we use the relationship between Landsat-8/Landsat-9-derived LST and spectral indices Normalized Difference Built-Up Index (NDBI), Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI) derived from Landsat-8/Landsat-9 and Sentinel-2 to downscale LST to 10 m. Two machine learning (ML) algorithms, support vector machine (SVM) and random forest (RF), are used to assess image classification and identify how different types and LC changes in selected years 2017, 2019, and 2023 affect the pattern of LST. The results show that several decisions made during the last decade, such as the construction of new urban fabrics and roads, caused the increase in LST. The LC change evaluation, based on the RF classification algorithm, achieved overall accuracies of 93.2% in 2017, 89.6% in 2019, and 91.5% in 2023, outperforming SVM by 0.8% in 2017 and 4.3% in 2023. This approach identifies UHI-prone areas with higher spatial resolution, helping urban planning mitigate the negative effects of increasing urban LSTs.
... During training, Random Forest creates several decision trees as a method of collaborative learning and outputs the mode of classes for each tree. This method successfully reduces the overfitting issue that distinct decision trees can have, improving the model's predictability (Gislason et al. 2006;Svoboda et al. 2022;Tikuye et al. 2023). In this study, the Random Forest algorithm was particularly adept at handling the complexity and heterogeneity inherent in land use data. ...
Article
Urban and economic growth around the world have rapidly degraded the ecosystem and environmental quality. Chattogram, a renowned city in Bangladesh known for its scenic environment, has experienced significant urban growth, which has degraded the environmental and ecological quality to the greatest extent. To investigate this issue, we assessed the Chattogram Metropolitan Area (CMA)’s ecological quality using the remote sensing-based ecological index (RSEI), incorporating four ecological indices such as heat index (LST), wetness (WET), dryness (NDBSI), and vegetation index (NDVI) through principal component analysis (PCA) and random forest (RF) model to analyze land use transitions (LUTs) using an advanced google earth engine (GEE). The PCA1 accounted for 80% of the dataset, delving into the highly significant features. Land use and land cover classes (LULCs) showed a gradual expansion of urban areas over the last two decades, indicating socio-economic progress. The urban area expanded from 34.08 km² in 2003 to 101.99 km² in 2023, an increase of 67.91 km² over 20 years, with an annual expansion rate of 2.34 km²/year. The RSEI values remained consistently > 0.6, indicating good ecological levels, primarily due to sustainable urban planning. RSEI negatively correlated with LST and NDBSI, suggesting a detrimental influence, particularly in urban and economic zones. Conversely, the NDVI and WET indices showed a positive correlation, indicating a constructive impact on the local ecological landscape. This study will assist policymakers in formulating strategies for ecological conservation to achieve sustainable development goals (SDGs) and implement the Chattogram Metropolitan Master Plan (2020–2041).
... Remote sensing technologies, particularly satellite-based earth observation missions such as Landsat and Sentinel, have revolutionized global-scale monitoring of CLCLU dynamics [19], [20], [21], [22], [23]. These missions provide continuous, open-access imagery, yielding valuable datasets for environmental analysis, including derived products like the Cropland Data Layer (CDL) and various vegetation indices [24], [25], [26]. Recent advancements in artificial intelligence (AI), specifically machine learning (ML) and deep learning (DL), have significantly enhanced the analysis of remotely sensed data for CLCLU classification. ...
Article
Full-text available
This study introduces two novel explainable AI (XAI) frameworks, Interclass-Grad-CAM and Spectral-Grad-CAM, designed to enhance the interpretability of semantic segmentation models for Crop and Land Cover Land Use (CLCLU) mapping. Interclass-Grad-CAM provides insights into interactions between land cover classes, revealing complex spatial arrangements, while Spectral-Grad-CAM quantifies the contributions of individual spectral bands to model predictions, optimizing spectral data use. These XAI methods significantly advance understanding of model behavior, particularly in heterogeneous landscapes, and ensure enhanced transparency in CLCLU mapping. To demonstrate the effectiveness of these innovations, we developed a framework that addresses data asymmetry between the United States and Mexico in the transboundary Middle Rio Grande region. Our approach integrates pixel-level multi-sensor fusion, combining dual-month moderate-resolution optical imagery (July and December 2023), Synthetic Aperture Radar (SAR), and Digital Elevation Model (DEM) data, processed using a Multi-Attention Network (MANet) with a modified Mix Vision Transformer (MiT) encoder to process multiple spectral inputs. Results indicate a uniform improvement in class-specific Intersection over Union (IoU) by approximately 1% with multi-sensor integration compared to optical imagery alone. Optical bands proved most effective for crop classification, while SAR and DEM data enhanced predictions for non-agricultural types. This framework not only improves CLCLU mapping accuracy but also offers a robust tool for broader environmental monitoring and resource management applications.
... Its usefulness is reflected in the different applications of these images for the detection of invasive species [26], the identification and classification of tree species [27], or the use of multitemporal data for the characterization of regenerating forest stands [28]. To enhance the discrimination of forest species, various studies have demonstrated the effectiveness of combining S1 and S2 data [21,24], as well as their integration with topographic information [29], in Cloud Computing platforms such as Google Earth Engine (GEE) [30][31][32]. On the other hand, there is currently spectral information with better spatial resolution available through the Plan-etScope platform, acquiring images with 3 m/pixel spatial resolution and capturing data in the red (R), green (G), blue (B), and near-infrared (NIR) channels. ...
... Combining remote sensing data with image processing and classification methods is useful to analyze and identify forest entities [39]. Several works present the feasibility of the evaluation of forest ecosystems through Machine Learning (ML) algorithms such as Random Forest (RF) [22,30,35,36,38,[40][41][42], Support Vector Machines [22,35,36,43], Gradient Boosting [44], Naive Bayes [45], being RF one of the most used, especially in the discrimination and identification of mountain forests and shrublands [22]. Additionally, some researchers used hyperspectral images obtained from Unmanned Aerial Vehicles (UAVs) for the identification of Polylepis [10] and other forest species [46][47][48]. ...
Article
Full-text available
Globally, there is a significant trend in the loss of native forests, including those of the Polylepis genus, which are essential for soil conservation across the Andes Mountain range. These forests play a critical role in regulating water flow, promoting soil regeneration, and retaining essential nutrients and sediments, thereby contributing to the soil conservation of the region. In Ecuador, these forests are often fragmented and isolated in areas of high cloud cover, making it difficult to use remote sensing and spectral vegetation indices to detect this forest species. This study developed twelve scenarios using medium- and high-resolution satellite data, integrating datasets such as Sentinel-2 and PlanetScope (optical), Sentinel-1 (radar), and the Sigtierras project topographic data. The scenarios were categorized into two groups: SC1–SC6, combining 5 m resolution data, and SC7–SC12, combining 10 m resolution data. Additionally, each scenario was tested with two target types: multiclass (distinguishing Polylepis stands, native forest, Pine, Shrub vegetation, and other classes) and binary (distinguishing Polylepis from non-Polylepis). The Recursive Feature Elimination technique was employed to identify the most effective variables for each scenario. This process reduced the number of variables by selecting those with high importance according to a Random Forest model, using accuracy and Kappa values as criteria. Finally, the scenario that presented the highest reliability was SC10 (Sentinel-2 and Topography) with a pixel size of 10 m in a multiclass target, achieving an accuracy of 0.91 and a Kappa coefficient of 0.80. For the Polylepis class, the User Accuracy and Producer Accuracy were 0.90 and 0.89, respectively. The findings confirm that, despite the limited area of the Polylepis stands, integrating topographic and spectral variables at a 10 m pixel resolution improves detection accuracy.
... Every decision tree within the RF is trained using randomly selected features and a random subset of data. At the end, all the individual predictions were summed up to form a final prediction (Rodriguez-Galiano et al., 2012;Adam et al., 2014;Svoboda et al., 2022). To train the RF model, the training data were labeled with ground control points (GCPs) and their corresponding land use/land cover classes. ...
Article
Full-text available
Water hyacinth (Pontederia crassipes) is an invasive weed that covers a significant portion of Lake Tana. The infestation has an impact on the lake’s ecological and socioeconomic systems. Early detection of the spread of water hyacinth using geospatial techniques is crucial for its effective management and control. The main objective of this study was to examine the spatiotemporal distribution of water hyacinth from 2016 to 2022 using a random forest machine learning model. The study used 16 variables obtained from Sentinel-2A, Sentinel-1 SAR, and SRTM DEM, and a random forest supervised classification model was applied. Seven spectral indices, five spectral bands, two Sentinel-1 SAR bands, and two topographic variables were used in combination to model the spatial distribution of water hyacinth. The model was evaluated using the overall accuracy and kappa coefficient. The findings demonstrated that the overall accuracy ranged from 0.91 to 0.94 and kappa coefficient from 0.88 to 0.92 in the wet season and 0.93 to 0.95 and 0.90 to 0.93 in the dry season, respectively. B11 and B5 (2022), VH, soil adjusted vegetation index (SAVI), and normalized difference water index (NDWI) (2020), B5 and B12 (2018), and VH and slope (2016) are the highly important variables in the classification. The study found that the spatial coverage of water hyacinth was 686.5 and 650.4 ha (2016), 1,851 and 1,259 ha (2018), 1,396.7 and 1,305.7 ha (2020), and 1,436.5 and 1,216.5 ha (2022) in the wet and dry seasons, respectively. The research findings indicate that variables derived from optical (Sentinel-2A and SRTM) and non-optical (Sentinel-1 SAR) satellite imagery effectively identify water hyacinth and display its spatiotemporal spread using the random forest machine learning algorithm.
... Semantička segmentacija pokrivenosti zemljišta, zasnovana na podacima sa Sentinel-2 satelitskih slika zasniva se na klasifikaciji svakog pojedinačnog piksela nezavisno, upotrebom nekog algoritma mašinskog učenja [1]. Ovaj rad predstavlja primenu ovog pristupa na problem detekcije različitih slojeva koji pokrivaju oblast Gornjeg Podunavlja. ...
Article
U okviru ovog rada opisana je seman¬tička segmentacija pokrivenosti zemljišta u Gornjem Po¬dunavlju sa Sentinel-2 satelitskih slika. Segmentacija je izvršena u 6 klasa: krošnje drveća, vodene površine, trav¬nate površine, tlo, poljoprivredno zemljište i vegetacija na vodenim površinama. Podaci predstavljaju multispektral¬ne slike Sentinel-2 satelitskih slika koje obuhvataju pod¬ručje Gornjeg Podunavlja. Za obuku su korišćeni XG-Boost klasifikator i Random Forest klasifikator. Dronske slike ovog područja služile su za kreiranje činjeničlnog stanja. Kao dodatna obeležja prilikom klasifikacije korišćeni su vegetacioni indeksi. Maksimalna postignuta tačnost iznosila je 93%.
... An overview of the parameters and their respective values used to train the SVM model is presented in Table 3. RF is implemented using the tool "ee.Classi er.smileRandomForest()" and uses 6 input parameters (Table 3). For both algorithms, to achieve higher classi cation accuracy, we also used the NDVI and NDBI spectral indices as input, similarly to the study of Svoboda et al. (2022). These indices were used to train the models to better distinguish between vegetation and built-up areas. ...
Preprint
Full-text available
In recent decades, global climate change and rapid urbanisation have aggravated the urban heat island (UHI) effect, affecting the well-being of urban citizens. Although this significant phenomenon is more pronounced in larger metropolitan areas due to extensive impervious surfaces, small and medium-sized cities also experience UHI effects, yet research on UHI in these cities is rare, emphasizing the importance of Land Surface Temperature (LST) as a key parameter for studying UHI dynamics. Therefore, this paper focusses on the evaluation of LST and land cover (LC) changes in the city of Prešov, Slovakia, a typical medium-sized European city that has recently undergone significant LC changes. In this study, we use the relationship between Landsat-8/-9 derived LST and spectral indices NDBI, NDVI, NDWI derived from Landsat-8/-9 and Sentinel-2 to downscale LST to 10 m. Two machine learning (ML) algorithms, Support Vector Machine (SVM) and Random Forest (RF), are used to assess image classification and identify how different types and LC changes in selected years 2017, 2019, 2023 affect the pattern of LST. The results show that several decisions made during the last decade, such as the construction of new urban fabrics and roads, caused the increase of LST. The LC change evaluation, based on the RF classification algorithm, achieved overall accuracies of 93.2% in 2017, 89.6% in 2019, and 91.5% in 2023, outperforming SVM by 0.8% in 2017 and 4.3% in 2023. This approach identifies UHI-prone areas with higher spatial resolution, helping urban planning mitigate the negative effects of increasing urban LSTs.