Multi-temporal Sentinel-2 Data Research Articles

Detection of Coral Reef Bleaching by Multitemporal Sentinel-2 Data Using the PU-Bagging Algorithm: A Feasibility Study at Lizard Island

Coral reef bleaching events have become more frequent all over the world and pose a serious threat to coral reef ecosystems. Therefore, there is an urgent need for better detection of coral reef bleaching in a time- and cost-saving manner. In recent years, remote sensing technology has often been utilized and gained recognition for coral reef bleaching detection. However, bleaching corals in the water always have weak spectral change signals, causing difficulties in using remote sensing data. Additionally, uneven change samples make it challenging to adequately capture the details of coral reef bleaching detection and produce thematic maps. To resolve these problems, a novel method named coral reef bleaching detection by positive-unlabeled bagging (CBD-PUB) is proposed in this paper. To test the capacity of the method, a series of multi-temporal Sentinel-2 remote sensing images are utilized, and Lizard Island in Australia is taken as a case study area. The pseudo-invariant feature atmospheric correction (PIF) algorithm is adopted to improve coral reef bleaching spectral signals. After that, CBD-PUB is employed to effectively explore coral reef bleaching variation and its corresponding influence relations. The experimental results show that the overall accuracy of bleaching detection by the proposed algorithm reaches 92.1% and outperforms the traditional method. It fully demonstrates the feasibility of the model for the field of coral reef bleaching detection and provides assistance in the monitoring and protection of coral environments.

Parcel level temporal variance of remotely sensed spectral reflectance predicts plant diversity

Over the last two decades, considerable research has built on remote sensing of spectral diversity to assess plant diversity. The spectral variation hypothesis (SVH) proposes that spatial variation in reflectance data of an area is positively associated with plant diversity. While the SVH has exhibited validity in dense forests, it performs poorly in highly fragmented and temporally dynamic agricultural landscapes covered mainly by grasslands. Such underperformance can be attributed to the mosaic-like spatial structure of human-dominated landscapes with fields in varying phenological and management stages. Therefore, we argued for re-evaluating SVH’s flawed window-based spatial analysis and underutilized temporal component. In particular, we captured the spatial and temporal variation in reflectance and assessed the relationships between spatial and temporal components of spectral diversity and plant diversity at the parcel level as a unit that relates to management patterns. Our investigation spanned three grasslands on two continents covering a wide spectrum of agricultural usage intensities. To calculate different components of spectral diversity, we used multi-temporal spaceborne Sentinel-2 data. We showed that plant diversity was negatively associated with the temporal component of spectral diversity across all sites. In contrast, the spatial component of spectral diversity was related to plant diversity in sites with larger parcels. Our findings highlighted that in agricultural landscapes, the temporal component of spectral diversity drives the spectral diversity-plant diversity associations. Consequently, our results offer a novel perspective for remote sensing of plant diversity globally.

Remote Quantification of Soil Organic Carbon: Role of Topography in the Intra-Field Distribution

Soil organic carbon (SOC) measurements are an indicator of soil health and an important parameter for the study of land-atmosphere carbon fluxes. Field sampling provides precise measurements at the sample location but entails high costs and cannot provide detailed maps unless the sampling density is very high. Remote sensing offers the possibility to quantify SOC over large areas in a cost-effective way. As a result, numerous studies have sought to quantify SOC using Earth observation data with a focus on inter-field or regional distributions. This study took a different angle and aimed to map the spatial distribution of SOC at the intra-field scale, since this distribution provides important insights into the biophysiochemical processes involved in the retention of SOC. Instead of solely using spectral measurements to quantify SOC, topographic and spectral features act as predictor variables. The necessary data on study fields in South-East England was acquired through a detailed SOC sampling campaign, including a LiDAR survey flight. Multi-spectral Sentinel-2 data of the study fields were acquired for the exact day of the sampling campaign, and for an interval of 18 months before and after this date. Random Forest (RF) and Support Vector Regression (SVR) models were trained and tested on the spectral and topographical data of the fields to predict the observed SOC values. Five different sets of model predictors were assessed, by using independently and in combination, single and multidate spectral data, and topographical features for the SOC sampling points. Both, RF and SVR models performed best when trained on multi-temporal Sentinel-2 data together with topographic features, achieving validation root-mean-square errors (RMSEs) of 0.29% and 0.23% SOC, respectively. These RMSEs are competitive when compared with those found in the literature for similar models. The topographic wetness index (TWI) exhibited the highest permutation importance for virtually all models. Given that farming practices within each field are the same, this result suggests an important role of soil moisture in SOC retention. Contrary to findings in dryer climates or in studies encompassing larger areas, TWI was negatively related to SOC levels in the study fields, suggesting a different role of soil wetness in the SOC storage in climates characterized by excess rainfall and poorly drained soils.

Combining Multitemporal Optical and Radar Satellite Data for Mapping the Tatra Mountains Non-Forest Plant Communities

Climate change is significantly affecting mountain plant communities, causing dynamic alterations in species composition as well as spatial distribution. This raises the need for constant monitoring. The Tatra Mountains are the highest range of the Carpathians which are considered biodiversity hotspots in Central Europe. For this purpose, microwave Sentinel-1 and optical multi-temporal Sentinel-2 data, topographic derivatives, and iterative machine learning methods incorporating classifiers random forest (RF), support vector machines (SVMs), and XGBoost (XGB) were used for the identification of thirteen non-forest plant communities (various types of alpine grasslands, shrublands, herbaceous heaths, mountain hay meadows, rocks, and scree communities). Different scenarios were tested to identify the most important variables, retrieval periods, and spectral bands. The overall accuracy results for the individual algorithms reached RF (0.83–0.96), SVM (0.87–0.93), and lower results for XGBoost (0.69–0.82). The best combination, which included a fusion of Sentinel-1, Sentinel-2, and topographic data, achieved F1-scores for classes in the range of 0.73–0.97 (RF) and 0.66–0.95 (SVM). The inclusion of topographic variables resulted in an improvement in F1-scores for Sentinel-2 data by one–four percent points and Sentinel-1 data by 1%–9%. For spectral bands, the Sentinel-2 10 m resolution bands B4, B3, and B2 showed the highest mean decrease accuracy. The final result is the first comprehensive map of non-forest vegetation for the Tatra Mountains area.

Forest type identification by combining multi-temporal GF-6 WFV and Sentinel-2 data

Forest type identification by combining multi-temporal GF-6 WFV and Sentinel-2 data

Analysis of the spatio-temporal dynamics of Buxus hyrcana Pojark defoliation using spaceborne satellite data

Abstract This study aims to assess the spatio-temporal defoliation dynamics of box tree, one of the few evergreen species of the Hyrcanian Forests. For this, we integrated multi-temporal leaf-off optical Sentinel-2 and radar Sentinel-1 data from 2017 to 2021 with elevation data. A state-of-the-art sample migration approach was used to generate annual reference samples of two categories (defoliated and healthy box tree) for a set of target years 2017–2020. This approach is based on field samples of the reference year 2021 and two similarity measures, the Euclidean distance and the spectral angle distance. The analysis of spectral and radar profiles showed that the migrated samples were well representative of both defoliated and healthy box trees categories. The migrated samples were then used for spatially mapping the two classes using support vector machine classification. The results of support vector machine classification indicated a large extent of box tree mortality. The most significant changes from healthy box trees to defoliated ones, or vice versa, occurred during the years 2017 and 2018. In the consecutive years of 2019, 2020, and 2021, no significant changes in the distribution of healthy or defoliated box trees were observed. The statistical assessment also revealed that mortality of evergreen understory tree species can be mapped with practically sufficient overall accuracies reaching from 84% (in 2017) to 91%–92% (in 2020 and 2021) using spaceborne remote sensing data. This information using freely accessible satellite data can benefit forest managers responsible for monitoring landscapes affected by the box moth and facilitates the identification of optimal control programs.

Multi-Temporal Sentinel-1 and Sentinel-2 Data for Orchards Discrimination in Khairpur District, Pakistan Using Spectral Separability Analysis and Machine Learning Classification

Generating orchards spatial distribution maps within a heterogeneous landscape is challenging and requires fine spatial and temporal resolution images. This study examines the effectiveness of Sentinel-1 (S1) and Sentinel-2 (S2) satellite data of relatively high spatial and temporal resolutions for discriminating major orchards in the Khairpur district of the Sindh province, Pakistan using machine learning methods such as random forest (RF) and a support vector machine. A Multicollinearity test (MCT) was performed among the multi-temporal S1 and S2 variables to remove those with high correlations. Six different feature combination schemes were tested, with the fusion of multi-temporal S1 and S2 (scheme-6) outperforming all other combination schemes. The spectral separability between orchards pairs was assessed using Jeffries-Matusita (JM) distance, revealing that orchard pairs were completely separable in the multi-temporal fusion of both sensors, especially the indistinguishable pair of dates-mango. The performance difference between RF and SVM was not significant, SVM showed a slightly higher accuracy, except for scheme-4 where RF performed better. This study concludes that multi-temporal fusion of S1 and S2 data, coupled with robust ML methods, offers a reliable approach for orchard classification. Prospectively, these findings will be helpful for orchard monitoring, improvement of yield estimation and precision based agricultural practices.

Flood susceptibility mapping using a novel integration of multi-temporal sentinel-1 data and eXtreme deep learning model

Flash floods (FFs) are amongst the most devastating hazards in arid regions in response to climate change and can cause the loss of agricultural land, human lives and infrastructure. One of the major challenges is the high-intensity rainfall events affecting low-lying areas that are vulnerable to FF. Several works in this field have been conducted using ensemble machine learning models and geohydrological models. However, the current advancement of eXtreme deep learning, which is named eXtreme deep factorisation machine (xDeepFM), for FF susceptibility mapping (FSM) is lacking in the literature. The current study introduces a new model and employs a previously unapplied approach to enhance FSM for capturing the severity of floods. The proposed approach has three main objectives: (i) During- and after-flood effects are assessed through flood detection techniques using Sentinel-1 data. (ii) Flood inventory is updated using remote sensing-based methods. The derived flood effects are implemented in the next step. (iii) An FSM map is generated using an xDeepFM model. Therefore, this study aims to apply xDeepFM to estimate susceptible areas using 13 factors in the emirates of Fujairah, UAE. The performance metrics show a recall of 0.9488), an F1-score of 0.9107), precision of (0.8756) and an overall accuracy of 90.41%. The accuracy of the applied xDeepFM model is compared with that of traditional machine learning models, specifically the deep neural network (78%), support vector machine (85.4%) and random forest (88.75%). Random forest achieves high accuracy, which is due to its strong performance that depends on factors contribution, dataset size and quality, and available computational resources. Comparatively, the xDeepFM model works efficiently for complicated prediction problems having high non-collinearity and huge datasets. The obtained map denotes that the narrow basins, lowland coastal areas and riverbank areas up to 5 km (Fujairah) are highly prone to FF, whilst the alluvial plains in Al Dhaid and hilly regions in Fujairah show low probability. The coastal city areas are bounded by high-rise steep hills and the Gulf of Oman, which can elevate the water levels during heavy rainfall. Four major synchronised influencing factors, namely, rainfall, elevation, drainage density, distance from drainage and geomorphology, account for nearly 50% of the total factors contributing to a very high flood susceptibility. This study offers a platform for planners and decision makers to take timely actions on potential areas in mitigating the effects of FF.

Evaluating the potential of multi-temporal Sentinel-1 and Sentinel-2 data for regional mapping of olive trees

ABSTRACT Olives are a crucial economic crop in Mediterranean countries. Detailed spatial information on the distribution and condition of crops at regional and national scales is essential to ensure the continuity of crop quality and yield efficiency. However, most earlier studies on olive tree mapping focused mainly on small parcels using single-sensor, very high resolution (VHR) data, which is time-consuming, expensive and cannot feasibly be scaled up to a larger area. Therefore, we evaluated the performance of Sentinel-1 and Sentinel-2 data fusion for the regional mapping of olive trees for the first time, using the Izmir Province of Türkiye, an ancient olive-growing region, as a case study. Three different monthly composite images reflecting the different phenological stages of olive trees were selected to separate olive trees from other land cover types. Seven land-cover classes, including olives, were mapped separately using a random forest classifier for each year between 2017 and 2021. The results were assessed using the k-fold cross-validation method, and the final olive tree map of Izmir was produced by combining the olive tree distribution over two consecutive years. District-level areas covered by olive trees were calculated and validated using official statistics from the Turkish Statistical Institute (TUIK). The K-fold cross-validation accuracy varied from 94% to 95% between 2017 and 2021, and the final olive map achieved 98% overall accuracy with 93% producer accuracy for the olive class. The district-level olive area was strongly related to the TUIK statistics (R2 = 0.60, NRMSE = 0.64). This study used Sentinel data and Google Earth Engine (GEE) to produce a regional-scale olive distribution map that can be scaled up to the entire country and replicated elsewhere. This map can, therefore, be used as a foundation for other scientific studies on olive trees, particularly for the development of effective management practices.

Examining the potential for early detection of spruce bark beetle attacks using multi-temporal Sentinel-2 and harvester data

Forests are invaluable terrestrial ecosystems with considerable economic, ecological, and environmental benefits. Bark beetles have been recognized as one of the major causes of forest disturbance, and climate change can exacerbate their impact, leading to more tree mortality. Early detection of bark beetle attacks is vital to reduce forest loss and devastating consequences. This study examines the potential for early detection of European spruce bark beetle (Ips typographus L.) attacks in southeastern Sweden using comprehensive harvester data and time series of Sentinel-2 images, 2015–2021. Specifically, it aims at 1) determining the most pronounced wavelength bands and vegetation indices (VIs) of Sentinel-2 for early detection, 2) determining the number of attacked trees in a Sentinel-2 pixel required to enable detection, 3) testing three change detection approaches, Detecting Breakpoints and Estimating Segments in Trend (DBEST), Mean-Level-Shift (MLS), and Cumulative Sum (CUSUM) to investigate the potential for early detection of bark beetle attacks. The greatest separation in reflectance between healthy and attacked pixels, from first swarming peak (May 2018) till harvesting (April 2019), was observed in the SWIR1 (0.018) and SWIR2 (0.011) bands followed by red-edge (0.008), red (0.007), NIR (0.005), and the green band (0.004). The blue band showed the least separation (0.003). All VIs showed a change in their base level after the swarming and this was more prominent for NDRS with an increase of 0.14, followed by NDWI (-0.13), CCI (-0.11) and NDVI (-0.09), all with decreasing values. The observed responses of VIs in relation to the number of attacked trees in a Sentinel-2 pixel increased gradually for pixels having one to ten infested trees, with the strongest response observed for pixels with 9 to 14 attacked trees. Pixels including more than 14 attacked trees did not show any further substantial change in VIs. DBEST, on average, indicated that the infestation impact on VIs is detectable one month after the swarming peak with a 15–31 days precision. MLS and CUSUM with up to two months' accuracies were ranked next. NDRS and CCI showed superior detection performance compared to NDVI and NDWI. The detection was based on smoothed time series of Sentinel-2 data to reduce the influence of noise and missing data and cannot be directly applied to a near real-time detection method.

Eucalyptus Plantation Area Extraction Based on SLPSO-RFE Feature Selection and Multi-Temporal Sentinel-1/2 Data

An accurate and efficient estimation of eucalyptus plantation areas is of paramount significance for forestry resource management and ecological environment monitoring. Currently, combining multidimensional optical and SAR images with machine learning has become an important method for eucalyptus plantation classification, but there are still some challenges in feature selection. This study proposes a feature selection method that combines multi-temporal Sentinel-1 and Sentinel-2 data with SLPSO (social learning particle swarm optimization) and RFE (Recursive Feature Elimination), which reduces the impact of information redundancy and improves classification accuracy. Specifically, this paper first fuses multi-temporal Sentinel-1 and Sentinel-2 data, and then carries out feature selection by combining SLPSO and RFE to mitigate the effects of information redundancy. Next, based on features such as the spectrum, red-edge indices, texture characteristics, vegetation indices, and backscatter coefficients, the study employs the Simple Non-Iterative Clustering (SNIC) object-oriented method and three different types of machine-learning models: Random Forest (RF), Classification and Regression Trees (CART), and Support Vector Machines (SVM) for the extraction of eucalyptus plantation areas. Each model uses a supervised-learning method, with labeled training data guiding the classification of eucalyptus plantation regions. Lastly, to validate the efficacy of selecting multi-temporal data and the performance of the SLPSO–RFE model in classification, a comparative analysis is undertaken against the classification results derived from single-temporal data and the ReliefF–RFE feature selection scheme. The findings reveal that employing SLPSO–RFE for feature selection significantly elevates the classification precision of eucalyptus plantations across all three classifiers. The overall accuracy rates were noted at 95.48% for SVM, 96% for CART, and 97.97% for RF. When contrasted with classification outcomes from multi-temporal data and ReliefF–RFE, the overall accuracy for the trio of models saw an increase of 10%, 8%, and 8.54%, respectively. The accuracy enhancement was even more pronounced when juxtaposed with results from single-temporal data and ReliefF-RFE, at increments of 15.25%, 13.58%, and 14.54% respectively. The insights from this research carry profound theoretical implications and practical applications, particularly in identifying and extracting eucalyptus plantations leveraging multi-temporal data and feature selection.

Smallholder Maize Farm Mapping With Multi-Temporal Sentinel-1 Data in Support of the Sustainable Development Goals

Smallholder Maize Farm Mapping With Multi-Temporal Sentinel-1 Data in Support of the Sustainable Development Goals

Land-Use Mapping with Multi-Temporal Sentinel Images Based on Google Earth Engine in Southern Xinjiang Uygur Autonomous Region, China

Land-use maps are thematic materials reflecting the current situation, geographical diversity, and classification of land use and are an important scientific foundation that can assist decision-makers in adjusting land-use structures, agricultural zoning, regional planning, and territorial improvement according to local conditions. Spectral reflectance and radar signatures of time series are important in distinguishing land-use types. However, their impact on the accuracy of land-use mapping and decision making remains unclear. Also, the many spatial and temporal heterogeneous landscapes in southern Xinjiang limit the accuracy of existing land-use classification products. Therefore, our objective herein is to develop reliable land-use products for the highly heterogeneous environment of the southern Xinjiang Uygur Autonomous Region using the freely available public Sentinel image datasets. Specifically, to determine the effect of temporal features on classification, several classification scenarios with different temporal features were developed using multi-temporal Sentinel-1, Sentinel-2, and terrain data in order to assess the importance, contribution, and impact of different temporal features (spectral and radar) on land-use classification models and determine the optimal time for land-use classification. Furthermore, to determine the optimal method and parameters suitable for local land-use classification research, we evaluated and compared the performance of three decision-tree-related classifiers (classification and regression tree, random forest, and gradient tree boost) with respect to classifying land use. Yielding the highest average overall accuracy (95%), kappa (95%), and F1 score (98%), we determined that the gradient tree boost model was the most suitable for land-use classification. Of the four individual periods, the image features in autumn (25 September to 5 November) were the most accurate for all three classifiers in relation to identifying land-use classes. The results also show that the inclusion of multi-temporal image features consistently improves the classification of land-use products, with pre-summer (28 May–20 June) images providing the most significant improvement (the average OA, kappa, and F1 score of all the classifiers were improved by 6%, 7%, and 3%, respectively) and fall images the least (the average OA, kappa, and F1 score of all the classifiers were improved by 2%, 3%, and 2%, respectively). Overall, these analyses of how classifiers and image features affect land-use maps provide a reference for similar land-use classifications in highly heterogeneous areas. Moreover, these products are designed to describe the highly heterogeneous environments in the study area, for example, identifying pear trees that affect local economic development, and allow for the accurate mapping of alpine wetlands in the northwest.

Natura 2000 Grassland Habitats Mapping Based on Spectro-Temporal Dimension of Sentinel-2 Images with Machine Learning

Habitat mapping is essential for the management and monitoring of Natura 2000 sites. Time-consuming field surveys are still the most frequently used solution for the implementation of the European Habitats Directive, but the use of remote sensing tools for this is becoming more common. The high temporal resolution of Sentinel-2 data, registering the visible, near, and shortwave infrared ranges of the electromagnetic spectrum, makes them valuable material in this context. In this study, we aimed to use multitemporal Sentinel-2 data for mapping three grassland Natura 2000 habitats in Poland. We performed the classification based on spectro-temporal features extracted from data collected from eight different terms within the year 2017 using Convolutional Neural Networks (CNNs), and we also tested other widely used machine learning algorithms for comparison, such as Random Forests (RFs) and Support Vector Machines (SVMs). Based on ground truth data, we randomly selected training and validation polygons and then performed the evaluation iteratively (100 times). The best resulting median F1 accuracies that we obtained for habitats were as follows: 6210, 0.85; 6410, 0.80; and 6510, 0.84 (with SVM). Finally, we concluded that the accuracy of the results was comparable, but we obtained the best results using SVM (median OA = 88%, with 86% for RF and 84% for CNNs). In this work, we confirmed the usefulness of the spectral dimension of Sentinel-2 time series data for mapping grassland habitats, and researchers of future work can further develop the use of CNNs for this purpose.

USING MACHINE LEARNING CLASSIFIERS AND REGRESSION MODELS FOR ESTIMATING THE STAND AGES OF FALCATA PLANTATIONS FROM SENTINEL DATA

Abstract. Falcata is a widely planted Industrial Tree Plantation (ITP) species in the Caraga Region, Mindanao, Philippines, significantly contributing to more than 80% of the country's Falcata log production in recent years. Currently, Falcata plantations face several challenges, especially regarding plantation monitoring and management. The information on stand age is essential for the efficient monitoring and sustainable management of ITPs. With advances in technology and freely available satellite data, primarily from the Sentinel satellites, remote sensing has provided an alternative approach to determining stand age information. In this context, this study used multivariate regression models and machine learning classifiers, namely Support Vector Machine (SVM) and Random Forest (RF) to estimate Falcata stand ages from single-date and multitemporal Sentinel-1 and Sentinel-2 images and vegetation indices (VIs). The eleven multivariate regression models have R2 values ranging from 0.23 and 0.81 and with estimation errors of 1.72 to 3.58 years. The best multivariate regression model developed is an exponential model that relates Falcata stand age with the year 2021 Sentinel-2 surface reflectance bands and year 2021 Sentinel-1 VV and VH polarization bands. The model has training and validation data R2 of 0.61 and 0.55, which are the most consistent among the 11 regression models developed. When used for stand age estimation, this model may underestimate, or overestimate stand age by 1.72 years. When a machine learning approach is to be employed, the RF classifier performed better than SVM, particularly when estimating stand ages using multitemporal Sentinel-1 and Sentinel-2 data and VIs. The classifier has an overall accuracy of 84.69%, the highest among the eight classification results generated by the study.

Mountain Tree Species Mapping Using Sentinel-2, PlanetScope, and Airborne HySpex Hyperspectral Imagery

Europe’s mountain forests, which are naturally valuable areas due to their high biodiversity and well-preserved natural characteristics, are experiencing major alterations, so an important component of monitoring is obtaining up-to-date information concerning species composition, extent, and location. An important aspect of mapping tree stands is the selection of remote sensing data that vary in temporal, spectral, and spatial resolution, as well as in open and commercial access. For the Tatra Mountains area, which is a unique alpine ecosystem in central Europe, we classified 13 woody species by iterative machine learning methods using random forest (RF) and support vector machine (SVM) algorithms of more than 1000 polygons collected in the field. For this task, we used free Sentinel-2 multitemporal satellite data (10 m pixel size, 12 spectral bands, and 21 acquisition dates), commercial PlanetScope data (3 m pixel size, 8 spectral bands, and 3 acquisitions dates), and airborne HySpex hyperspectral data (2 m pixel size, 430 spectral bands, and a single acquisition) with fusion of the data of topographic derivatives based on Shuttle Radar Topography Mission (SRTM) and airborne laser scanning (ALS) data. The iterative classification method achieved the highest F1-score with HySpex (0.95 RF; 0.92 SVM) imagery, but the multitemporal Sentinel-2 data cube, which consisted of 21 scenes, offered comparable results (0.93 RF; 0.89 SVM). The three images of the high-resolution PlanetScope produced slightly less accurate results (0.89 RF; 0.87 SVM).

Evaluating capabilities of machine learning algorithms for aquatic vegetation classification in temperate wetlands using multi-temporal Sentinel-2 data

Different perspectives use of machine learning (ML) algorithms have proven their performance depends on the quality of reference data. This is particularly true when targets are complex environments, such as wetlands, on which a vast majority of studies are site-specific and based on a single date. With this work, an extensive reference dataset of about 400,000 samples was collected, covering nine different sites and multiple seasons, to be considered representative of temperate wetland vegetation communities at continental scale. Starting from this dataset, the performance of selected ML classifiers was compared for detailed wetland vegetation type mapping, using spectral indices (SI) derived from multi-temporal composites of Sentinel-2 as input. Global and per-class accuracy metrics were computed based on four independent training and testing subsets, extracted from the overall dataset, and the impacts of input features variation in number and sites covered were assessed. Our results show a generally higher predictive power for ensemble methods, such as Random Forest (RF) and eXtreme Gradient Boosting (XGBoost), compared to standalone ones, with the notable exception of Support Vector Machine (SVM); the latter in fact, in the algorithm that scored the highest overall accuracy (0.977 ± 0.001) and F-score for all the target classes. Decreasing the number of input features generally resulted in classification accuracy losses, less marked for RF than for SVM, while site-specific algorithms training showed more stability of SVM, thus indicating SVM stronger transferability than RF and XGBoost.

Machine learning based combinatorial analysis for land use and land cover assessment in Kyiv City (Ukraine)

The main goal of this study is to evaluate different models for further improvement of the accuracy of land use and land cover (LULC) classification on Google Earth Engine using random forest (RF) and support vector machine (SVM) learning algorithms. Ten indices, namely normalized difference vegetation index, normalized difference soil index, index-based built-up index, biophysical composition index, built-up area extraction index (BAEI), urban index, new built-up index, band ratio for built-up area, bare soil index, and normalized built up area index, were used as input parameters for the machine learning algorithms to improve classification accuracy. The combinatorial analysis of the Sentinel-2 bands and the aforementioned indices allowed us to create four combinations based on surface reflectance characteristics. The study includes data from April 2020 to September 2021 and April 2022 to June 2022. The multitemporal Sentinel-2 data with spatial resolutions of 10 m were used to determine the LULC classification. The major land use classes such as water, forest, grassland, urban areas, and other lands were obtained. Generally, the RF algorithm showed higher accuracy than the SVM. The overall accuracy for RF and SVM was 86.56% and 84.48%, respectively, and the mean Kappa was 0.82 and 0.79, respectively. Using the combination 2 with the RF algorithm and combination 4 with the SVM algorithm for LULC classification was more accurate. The additional use of vegetation indices allowed to increase in the accuracy of LULC classification and separate classes with similar reflection spectra.

A generalized framework for drought monitoring across Central European grassland gradients with Sentinel-2 time series

Fractional cover time series of photosynthetic vegetation (PV), non-photosynthetic vegetation (NPV), and soil from remote sensing provide essential detail to understand how grasslands are affected by recent and future drought periods in the 21st century. In this regard, Sentinel-2A/B offer frequent large-area observations, which have not yet been fully exploited for a spatially continuous drought monitoring of highly dynamic Central European grasslands. In this study, we developed a generalized drought monitoring framework for Central European grasslands linking Sentinel-2 data, field survey information, and spectral unmixing. We first implemented a consistent and repeatable strategy to obtain a grassland spectral library supported by the Europe-wide Land Use/Cover Area frame statistical Survey (LUCAS) and multitemporal Sentinel-2 data. Our library captured the spectral variability of PV, NPV, and soil cover from 12 grassland areas distributed along typical environmental and land use gradients of Central Europe. We trained a generalized regression-based unmixing model with synthetic data generated from the spectral library and compared fractional cover estimates to a multitemporal reference dataset. PV, NPV, and soil were estimated with good accuracy, achieving MAEs of 6.54%, 13.7%, and 12.2%, respectively. Local unmixing models trained on area-specific library subsets were overall outperformed by the generalized model highlighting the value of a comprehensive grassland library for generalized spectral unmixing. Based on fractional cover time series from 2017 to 2021, we calculated time series of the grassland-specific Normalized Difference Fraction Index (NDFI) capturing proportions of NPV and soil relative to PV. Comparison of annual growing season drought metrics derived from the NDFI to annual meteorological drought statistics from the Standardized Precipitation Evapotranspiration Index (SPEI) as well as the Soil Moisture Index (SMI) revealed widespread drought impacts on grasslands during the persistent drought period in Central Europe from 2018 to 2020. While impacts on grasslands overall closely followed meteorological and soil drought conditions, regionally varying drought metrics underline that local to regional environmental and hydrological conditions shaped the drought response of Central European grasslands. Our study emphasizes the value of combining Sentinel-2 data, field survey information, and spectral unmixing to enable drought monitoring across grassland gradients of Central Europe with Sentinel-2 time series.

Regional-Scale Analysis of Vegetation Dynamics Using Satellite Data and Machine Learning Algorithms: A Multi-Factorial Approach

Abstract Accurate vegetation analysis is crucial amid accelerating global changes and human activities. Achieving precise characterization with multi-temporal Sentinel-2 data is challenging. In this article, we present a comprehensive analysis of 2021's seasonal vegetation cover in Greater Sydney using Google Earth Engine (GEE) to process Sentinel-2 data. Using the random forest (RF) method, we performed image classification for vegetation patterns. Supplementary factors such as topographic elements, texture information, and vegetation indices enhanced the process and overcome limited input variables. Our model outperformed existing methods, offering superior insights into season-based vegetation dynamics. Multi-temporal Sentinel-2 data, topographic elements, vegetation indices, and textural factors proved to be critical for accurate analysis. Leveraging GEE and rich Sentinel-2 data, our study would benefit decision-makers involved in vegetation monitoring.