Search

Bush encroachment threatens rangelands' biodiversity and productivity, impacting savannah ecosystems based on location, management practices, and factors like erratic rainfall, climate change, and environmental variability. Considering these challenges, this study therefore seeks to evaluate bush encroachment changes over-time (1992-2022) in the Southern part of Kruger National Park and surrounding communities of South Africa. The study estimated the proportion and extent of encroacher plants in relation to native bush species. To achieve this objective, bioclimatic variables, and a digital elevation model in conjunction with the Random Forest model were utilized. Classified satellite imageries achieved an overall accuracy of 72 and 93%, respectively. A gradual increase in bush encroachment was observed from 41,947 hectares (ha) (3.4%) in 1992 to 61,225 ha (10%) in 2022. Additionally, this study observed a decline in the spatial extent of native plant species by 178,163.4 ha, while invasive species have expanded by 44,022.17 ha from 1992 to 2022 wet season. There is a significant increase in the invasive woody species across the years; because the p-value (p<0.000) is <5% (p<0.05), the significance level is <5%. The findings of this study offer a significant understanding of the historical patterns of bush encroachment over the past three decades. This information serves as a baseline for formulating strategic and sustainable management approaches to combat bush encroachment and promote effective rangeland preservation in predominantly savannah ecosystems. [ABSTRACT FROM AUTHOR]

This article addresses the class imbalance problem in urban gain modelling (UGM) of Tabriz and Isfahan megacities in Iran by proposing novel cost-sensitive machine learning models, namely costsensitive support vector machine (CSVM), random forest (CRF) and artificial neural network (CANN). Random sampling, a frequently utilized method, fails to effectively tackle this issue by biasing models towards no change samples, which outnumber change samples. The results showed that CRF exhibited the highest accuracy (AUC = 0.560), followed by CANN (AUC = 0.557) and CSVM (AUC = 0.448) in Isfahan. In Tabriz, CRF (AUC = 0.809) and CANN (AUC = 0.818) excelled, outperforming balanced sampling models constructed with ANN, RF and SVM with the AUROC of ANN and RF boosted by 15% and 2% in validation. By emphasizing the significance of addressing class imbalance appropriately, this research highlights the improvement in modelling outcomes achievable through the cost-sensitive models especially in Tabriz case. [ABSTRACT FROM AUTHOR]

Assessing the impact of floods on crop cultivation is critical for ensuring food security and resilience in flood-prone regions. This study relied on cloud-free imagery from Landsat to construct six vegetation indices. These indices were used to conduct a comparative analysis of the flood disaster of 2022 in Pakistan. A random forest classifier was employed on pre-flood images to identify agricultural zones. A model was developed using two indices, Brightness (DBSI) and Wetness (TCW), along with the vegetative index of crop vigor (MSAVI), during the pre-flood months of 2021. The model's accuracy was validated and exhibited an RMSE of 0.03 in predicting MSAVI values. The model was used to forecast MSAVI during post-flood periods. Spatial analysis identified approximately 30-47% of the study area as 'Suitable' and 46-64% as 'Highly suitable' for early recultivation of Rabi Crops. The approach helps in planning for fast agricultural recovery in the aftermath of floods. [ABSTRACT FROM AUTHOR]

This study presents a novel Spectral Triad feature selection (STfs) technique based on music theory and compares it to the entire Sentinel-2 feature space and Random Forest-Recursive Feature Elimination (RF-RFE). The optimal subsets were evaluated with Random Forest for retrieving Leaf Area Index (LAI), Leaf Chlorophyll Content (LCab), and Canopy Chlorophyll Content (CCC) in a semi-arid agricultural landscape. The results indicated that the proposed STfs algorithm obtained equivalent or better (i.e. by 1 - 3%) retrieval accuracies for LAI (R2 cv of 66%, root mean squared error of cross-validation [RMSEcv] of 0.53 m2 m-2), LCab (R2 cv: 74%, RMSEcv: 7.09 mg cm-2) and CCC (R2 cv: 77%, RMSEcv: 33.69 mg cm-2), using only 5, 7 and 7 variables, respectively, when compared to RF-RFE and entire Sentinel-2 feature space. Overall, the proposed STfs algorithm has great potential to optimize the spectral feature space of quasi-hyperspectral sensors for rapid crop biophysical and biochemical parameter retrieval. [ABSTRACT FROM AUTHOR]

To overcome critical issues in landslide susceptibility modeling, a multifactor landslide susceptibility prediction model based on deep belief networks (DBN) with optimized learning rate control (LRC-DBN) was introduced in this study. The LRC-DBN model was applied to predict landslide susceptibility in Zhidan County, Shaanxi and its performance was compared against random forest, support vector machine and logistic regression models. The results show that the LRC-DBN model achieves a maximum AUC value of 0.941, which demonstrating higher predictive performance. The interpretability of LRC-DBN reveals that the development of loess landslides is more likely in areas with elevations ranging from 894.7 m to 998.5 m, slopes ranging from 49.6° to 64°, terrain undulations spanning 31.7 m to 91.0 m, rock type T3y and terrain humidity ranging from 2.4 to 3.9. The result provides invaluable insights for local landslide prevention endeavors and can assist decision-making in urban planning. [ABSTRACT FROM AUTHOR]

Multispectral (MS) images offer essential spectral information for monitoring paddy crops' Aboveground-biomass (AGB), but efficiency decreases due to background materials and high canopy biomass. Texture reveals canopy structure and can be employed in vegetation-indices (VIs) to enhance monitoring accuracy. This study focuses to estimate AGB of paddy crop by exploring the combined potential of spectral and textural features of unmanned aerial vehicle (UAV)-MS images using linear regression (LR), multilinear regression (MLR), and random forest (RF) models. Results demonstrate that near infrared (NIR)-based VIs outperform Colour-Indices. Normalised difference texture indices (NDTIs) composed of NIR, red-edge (RE) and blue (B) bands outperform all-evaluated VIs and grey-level co-occurrence matrix (GLCM)-textures for different growth stages. Combining VIs and NDTIs, RF performs best compared to other models. The outcomes suggest that the combined spectral and texture information can significantly improve estimation of AGB in paddy crops compared to using either of them alone. [ABSTRACT FROM AUTHOR]

Traditional landslide susceptibility assessment methods often rely on single models, which can be biased and less accurate. In this article, we introduce a two-tiered strategy to enhance landslide susceptibility predictions. Initially, we employ an ensemble stacking technique that combines the strengths of three machine learning classifiers. This combination leverages the support vector classifier (SVC) as the key meta-classifier to optimize and refine predictions. Subsequently, we integrate the extreme gradient boosting (XGB), random forest (RF) and gradient boosting decision tree (GBDT) models with SVC to create hybrid approaches. In this study, we evaluate and compare the effectiveness of six machine learning algorithms for predicting landslide susceptibility in Anhua County, Hunan Province, China. The results demonstrated that the stacking ensemble model outperforms traditional models. The XBGþSVC model achieves the highest AUC value (0.9468), which is followed by the GBDTþSVC (0.9316), RFþSVC (0.9162), XGB (0.9393), GBDT (0.9009), and RF (0.8693). These findings indicate that stacking machine learning approaches hold promise for landslide susceptibility mapping. [ABSTRACT FROM AUTHOR]

The Karakoram Highway (KKH) is a pivotal gateway within the framework of the China-Pakistan Economic Corridor. Nevertheless, its distinct and intricate geographical characteristics make it susceptible to recurrent landslides. As an essential tool, landslide susceptibility mapping (LSM) is significant in managing and alleviating landslides near the KKH. Moreover, landslide conditioning factors (LCFs) are crucial determinants influencing the outcomes of LSM. However, existing methods primarily rely on machine learning algorithms, which do not adequately account for the intricate spatial characteristics and patterns between LCFs and landslide occurrences. In response, this study introduces the feature fusion transformer (FFTR) module, constructed based on the foundations of the transformer framework, to fuse the spatial information features of all LCFs. Subsequently, the abstract high-level spatial features obtained are fed into diverse machine learning classifiers, including random forest (RF), extreme gradient boosting (XGBoost), gradient boosting decision trees (GBDT), categorical boosting (CatBoost), and extremely randomized trees (ET), to generate landslide susceptibility maps. Displacement velocity calculated by Persistent Scatterer Interferometric Synthetic Aperture Radar (PSInSAR) is incorporated in the LSM. The results demonstrate that FFTR-RF achieves premium performance in the area under the curve (AUC) (94%), accuracy (87.31%), precision (87.21%), recall (88.02%), and F1-score (87.61%). Incorporating displacement velocity into LSM results predicted by models enhances the comprehensiveness of LSM. These methods will furnish early warning systems for landslide disasters along the KKH, thus aiding recommendations for mitigating landslides' social and economic losses. [ABSTRACT FROM AUTHOR]

Fire was a serious natural disturbance to forest ecosystems. In this study, a national-scale forest fire risk model (RF_NTL) was proposed in mainland China by incorporating a new human activity factor (NTL) from nighttime light data. For objective verification of model's performance, the logistic regression model (LR) and random forest model (RF) were also built in this paper. Results showed that the RF_NTL model fitted with an AUC value up to 0.95, which was superior to the conventional LR model and RF model. The RF_NTL model also had good generalization ability and stability on annual and monthly scales, helping to accurately forecast forest fire risk. Moreover, the spatial distribution of forest fire risk based on RF_NTL model was detailed and reasonable on a national scale. Forest fire high-risk areas in China were mainly concentrated in the northeast and southwest regions. This study could provide references for future relevant research. [ABSTRACT FROM AUTHOR]

This study makes a significant contribution to the field of groundwater potential mapping (GWPM) by exploring the application of ensemble learning models (ELMs), specifically boosting ensemble models (BEMs), which have not been fully utilized in GWPM. By employing six ELMs (random forest, AdaBoost, XGBoost, CatBoost, GBDT and LightGBM), along with Tree-structured Parzen Estimator in Luoning County, China, this study identifies key indicators (topographic position index, distance to rivers and topographic wetness index) and demonstrates the superior model performance of XGBoost compared to other ELMs. Additionally, correlation analysis confirms the accuracy of XGBoost in predicting relationships between important indicators and groundwater potentials. Finally, the findings provide valuable insights for sustainable groundwater management strategies in Luoning County and emphasize the need for further exploration of ELMs, development of comprehensive performance evaluation and indicator systems, reduction of the inconsistencies between indicators and predication results and practical research to support future sustainable groundwater management. [ABSTRACT FROM AUTHOR]

Groundwater contamination is a serious health problem affected by land use/land cover (LULC) changes. Therefore, this study aims to investigate the impacts of LULC on groundwater levels and quality in the Emirate of Abu Dhabi over the past two decades. Two sets of Landsat images were used to classify LULC using the random forest (RF) classifier, while an image-different tool was used to monitor changes in LULC from 2000 to 2020. After that, a spatial analysis was performed by comparing LULC maps against the hydrological maps. The results show a rapid increase in all LULC classes, whereas the groundwater level has depleted by 40 m. The results also show that the area of high-water quality across the study area was reduced by 479 km2 (0.79%). The results reveal significant hydrological changes in response to rapid urbanization. [ABSTRACT FROM AUTHOR]

Population growth and environmental shifts have elevated the pressure on land use and cover (LULC), necessitating vital management and adaptive strategies to preserve the balance between ecosystem services and human well-being in watersheds. It's pivotal to understand the implications of human-induced shifts from natural to human-dominated surroundings This study utilized Google Earth Engine (GEE) to analyze 31 years of LULC changes in Letaba watershed. Using GEE's random forest, LULC classes were mapped with 93% to 99% accuracy across four timeframes (1990, 2000, 2010, and 2021). Trends revealed declining water bodies (-4%), bare surfaces (-2%), natural forest (-3%), and grassland (-3%), while shrublands, plantations, and built-up areas increased at annual rates of 41%, 24%, and 47% respectively. This transformation reflects population-driven shifts, necessitating adaptive strategies. Given the importance of plantations for income, embracing climate-smart agriculture could ensure long-term food and environmental security, thus addressing the evolving dynamics in Letaba watershed. [ABSTRACT FROM AUTHOR]

This study, which aims to make predictions using a previously unused deep hybrid Convolutional Neural Network-Support Vector Regression approach for hydropower generation, was carried out using a unique meteorological data set as input data. The unique dataset was collected from Kaman Meteorology Directorate and Hirfanlı HEPP in Turkey, from 2007 to 2021, to estimate the HEPP's Net Head and Hydroelectric Power Generation on a daily basis. The performances of the prediction models were benchmarked in addition to the used CNN-SVR model with Machine Learning (ML) models (Boosting Random Forest Regression (BRFR) and Weighted K-Nearest Neighbor Regression (WKNNR)) and Deep Learning (DL) models (Long-Short Term Memory (LSTM) and Deep Belief Network (DBN)). The comparison results of the used CNN-SVR model with other alternative models showed that CNN-SVR performed effectively with the highest correlation coefficient of 0.971 for NH and 0.968 for PP. [ABSTRACT FROM AUTHOR]

As urbanization accelerates, the degree of human impact on land use is increasing. land use land cover change (LULC) is acknowledged as crucial factor in environmental change. The best way to understand historical land use patterns, changes, drivers, and developments is through a rigorous assessment of LULC changes. In this study, we aim to identify LULC changes from 2015 to 2022, and predict changes for 2030. Sentinel-2 images were employed to analyze LULC change patterns and predict future trends. The Random Forest algorithm was used to classify the various LULC classes with high accuracy and reliability. Multilayer Perceptron and Markov Chain Analysis (MLP-MCA) based Hybrid-Approach was employed to predict the future dynamics of LULC change for 2030. The study revealed that built-up area expanded 90.64 km2 from 2015 to 2022 due to natural resource substitution. Predictions indicate that 58.84% of the study area will be transform into built-up by 2030. [ABSTRACT FROM AUTHOR]

In this paper, we examined the degree to which inherent spatial structure in soil properties influences the outcomes of machine learning (ML) approaches to predicting soil spatial variability. We compared the performances of four ML algorithms (support vector machine, artificial neural network, random forest, and random forest for spatial data) against two non-ML algorithms (ordinary least squares regression and spatial filtering regression). None of the ML algorithms produced residuals that had lower mean values or were less autocorrelated over space compared with the non-ML approaches. We recommend the use of random forest when a soil variable of interest is weakly autocorrelated (Moran's I < 0.1) and spatial filtering regression when it is relatively strongly autocorrelated (Moran's I > 0.4). Overall, this work opens the door to a more consistent selection of model algorithms through the establishment of threshold criteria for spatial autocorrelation of input variables. [ABSTRACT FROM AUTHOR]

Urban land use/land cover (LULC) classification has long been a hotspot for remote sensing applications. With high spatio-temporal resolution and multispectral, the recently launched GF-6 satellite provides ideal open imagery for LULC mapping. In this study, we utilized multitemporal GF-6 images to generate six types of land features, including spectral bands, texture features, built-up, waterbody, vegetation, and red-edge indices. The minimum Redundancy Maximum Relevance (mRMR) algorithm was employed to optimize feature selection. Subsequently, Random Forest (RF) and Extreme Gradient Boosting (XGBT) were assessed using different feature selections. Besides, various feature configurations were designed for LULC classification and comparison. The results indicate that the mRMR-based RF method achieved the highest overall accuracy of 91.37%. The temporal red-edge indices were important features for urban LULC classification and contributed mainly to grassland and cropland. These results supplement existing classification methods and assist in improving LULC mapping in urban areas with complex landscapes. [ABSTRACT FROM AUTHOR]

Vegetation hinders the acquisition of bedrock spectra and makes it difficult to classify bedrock with remote sensing data. Previous studies indicated bedrock can control vegetation growth through soluble nutrients and water-holding capacity. However, the potential of using vegetation spectral features to map bedrock has been rarely explored. This study first derived reflectance and vegetation indices from time series Sentinel-2 products, then did a spatiotemporal analysis of vegetation spectral features on different bedrock, and finally combined vegetation features and random forest classifiers to map bedrock. The results demonstrated that (a) the close relationship between vegetation growth and bedrock can be captured by Sentinel-2 images; (b) both VIs' combination and reflectance derived from the growing season can achieve reasonably classified maps, with classification accuracies of 70.06% and 73.46%, respectively; (c) NDVI was more sensitive to the bedrock than other VIs. Overall, vegetation spectral features showed great potential to map bedrock underneath vegetation. [ABSTRACT FROM AUTHOR]

Maize is a staple food, but productivity has stagnated due to limited access to advanced farming methods and knowledge. To promote sustainable agriculture, understanding the factors affecting maize yield at the farm level is crucial. This study used panel data on maize yield and agronomic practices in Northern Ghana and Malawi from 2014 to 2020. Satellite-based environmental variables were extracted at household locations, and Random Forest modeling was used to identify factors influencing maize yield variability. The models performance was sub-par with low R2 values (∼0.1 and ∼0.24 for Northern Ghana and Malawi). Fertilizer and precipitation were the most important factors explaining maize yield variability. Spatial maps showed that Malawi's maize yield can increase with more fertilizer, but rainfall is essential. In Northern Ghana, relying solely on fertilizer may not be enough to boost maize production. Survey data on maize is limited in making accurate yield predictions. Fertilizer use can increase maize yield in both Northern Ghana and Malawi. Fertilizer use intervention strategies should be region-specific. The efficiency of fertilizer use is dependent on adequate rainfall availability. [ABSTRACT FROM AUTHOR]

Transboundary forests in the Himalayas are one of the world's biodiversity hotspots, but exploring the patterns and influencing factors related to forest disturbance-recovery is still limited. In this study, we used a random forests model, a LandTrendr temporal segmentation algorithm and geographical detector to investigate the spatial and temporal evolution characteristics of forest disturbance and recovery and their influencing factors in the central Himalayas from 1995 to 2018. The results showed that forest disturbance was dominant (∼2.94 × 104 km2) and greater than recovery (∼0.56 × 104 km2). Both disturbance and recovery showed a significant decreasing trend (p < 0.01); disturbance occurred mainly at low elevations and the Gandaki basin, while recovery was in eastern Nepal. Disturbance rates were highest in Nepal (1.14%) and recovery rates in China (0.29%). The main influencing factors on forest dynamics were elevation, temperature and population. The interaction of all factors was synergistically enhanced. [ABSTRACT FROM AUTHOR]

In 2014, the State of Telangana in southern India began repairing and restoring more than 46,000 irrigation water tanks (artificial reservoirs) under the Mission Kakatiya project with an investment in excess of USD 2 billion. In this study, we attempted to map the temporal changes that have occurred in cropland areas and water bodies as a result of the project, using remote sensing imagery and applying land use/land cover (LULC) mapping algorithms. We used 16-day time series data from Landsat 8 to study the spatial distribution of changes in water bodies and cropland areas over the 2013–18 period. Ground survey information was used to assess the pixel-based accuracy of the Landsat-derived data. The areas served by these tanks were identified on the basis of training data and Random Forest algorithms using Google Earth Engine. Our spatial analysis revealed a substantial increase in cropped area under irrigation and expansion of water bodies over the study period. We observed a 20% increase in total tank area in 2017–18 and total cropland and irrigated area expansion of the order of 0.6 M ha and 0.2 M ha, respectively. A comparison of ground survey data and four LULC classes derived from Landsat temporal imagery showed an overall accuracy of 87%, significantly correlated with national agriculture statistics. Periodic monitoring based on remote sensing has proved to be an effective method of capturing LULC changes resulting from the Mission Kakatiya interventions. Higher-resolution satellite data can further improve the accuracy of estimates. [ABSTRACT FROM AUTHOR]

Aiming to provide a scientific basis for the prevention and control of soil erosion in the Huaihe region, this study uses the InVEST-SDR model to estimate the temporal and spatial changes of soil erosion and soil conservation in the Huaihe region in 2010, 2015 and 2020, and the driving factors are evaluated and quantified using Random Forest and Geodetector, respectively. The major findings are as follows: (1) From 2010 to 2020, soil loss and soil retention in the Huaihe region gradually intensified, while the spatial distribution trend was consistent from year to year. (2) The results from Random Forest and Geodetector show that changes in soil loss and soil retention are mainly influenced by slope and rainfall. (3) The steepness and length of slopes are the most influencing factors for soil erosion. Soil loss and soil retention vary widely between different land uses. Forests play an important role in soil conservation. [ABSTRACT FROM AUTHOR]

It is a fundamental task to monitor the topography and understand the changes of intertidal zone for rational utilization and sustainable development. A new method is proposed for identifying the terrain of the intertidal zone, using ICESat-2 data to replace a large amount of on-site observation data, thereby reducing costs and improving efficiency. Based on pre-experiments and correlation analysis, time phase index, water index, water transparency index and suspended sediment concentration index are added as features for the random forest (RF). Compared with using only the original band as the model input, the RMSE is reduced by 0.08 m. The results show that the inverted terrain has an RMSE of 0.45 m compared with handheld RTK data, and the RMSE at the mudflat from UAV data is 0.20 m. Based on the analysis of terrain changes over the four-year period, the trend towards sedimentation closer to land becomes more pronounced. [ABSTRACT FROM AUTHOR]

We used the Cellular Automata Markov (CA-Markov) integrated technique to study land use and land cover (LULC) changes in the Cholistan and Thal deserts in Punjab, Pakistan. We plotted the distribution of the LULC throughout the desert terrain for the years 1990, 2006 and 2022. The Random Forest methodology was utilized to classify the data obtained from Landsat 5 (TM), Landsat 7 (ETM+) and Landsat 8 (OLI/TIRS), as well as ancillary data. The LULC maps generated using this method have an overall accuracy of more than 87%. CA-Markov was utilized to forecast changes in land usage in 2022, and changes were projected for 2038 by extending the patterns seen in 2022. A CA-Markov-Chain was developed for simulating long-term landscape changes at 16-year time steps from 2022 to 2038. Analysis of urban sprawl was carried out by using the Random Forest (RF). Through the CA-Markov Chain analysis, we can expect that high density and low-density residential areas will grow from 8.12 to 12.26 km2 and from 18.10 to 28.45 km2 in 2022 and 2038, as inferred from the changes occurred from 1990 to 2022. The LULC projected for 2038 showed that there would be increased urbanization of the terrain, with probable development in the croplands westward and northward, as well as growth in residential centers. The findings can potentially assist management operations geared towards the conservation of wildlife and the eco-system in the region. This study can also be a reference for other studies that try to project changes in arid are as undergoing land-use changes comparable to those in this study. [ABSTRACT FROM AUTHOR]

Vietnam is the world's second-largest coffee exporter. Information on coffee-growing areas is thus important for production estimation. This research developed an approach for coffee mapping and change detection using Landsat and DEM data in Central Highlands, Vietnam. The data were processed for 1995, 2005, and 2020 using random forests (RF). The mapping results, verified with reference data, showed overall accuracy and Kappa coefficient higher than 86.9% and 0.74, respectively. These findings were reaffirmed by a close relationship between the satellite-based coffee area and official statistics, with the root mean squared percentage error (RMSPE) and mean absolute percentage error (MAPE) greater than 9.9% and 7.6%, respectively. From 1995 to 2020, the newly planted coffee area increased by 135,520 ha, due to the conversion of forests to coffee plantations. Such distributions of coffee areas and decadal changes in farming activities are critically useful for policymakers to formulate successful crop planning strategies. [ABSTRACT FROM AUTHOR]

Accurate land cover mapping is challenging in Southeast Asia where cloud coverage is prevalent and landscape is heterogenous. Object-based mapping, multi-temporal images and combined use of optical and microwave data, provide abundant features in spectral, spatial, temporal, geometric and polarimetric dimensions. And random forest is usually employed due to robustness and efficiency in handling high-dimensional and noisy data. This study assesses whether feature selection and ensemble analysis, which are rarely adopted, yield improved result. Recursive feature elimination decreases original 568 features into a subset of 53 features, achieving the optimal combination of features. Ensemble analysis of random forest, support vector machine, and K-nearest neighbors leads to an overall accuracy of 0.816. Comparison experiments demonstrated the merits of the multi-temporal, multi-source approach, feature elimination and ensemble analysis. [ABSTRACT FROM AUTHOR]

This study proposed a comprehensive approach that utilized PlanetScope imagery for classifying tropical-marine benthic habitats after retrieving bathymetry from ICESat-2 data and water-column correction for areas around Lyson Islands, Vietnam. Exact bathymetry derivation and water column correction were applied to the PlanetScope imagery, making it an effective method for mapping marine benthic habitats. Water column correction was achieved by applying Depth Invariant Index (DII) and Bottom Reflectance Index (BRI). Moreover, two conventional machine learning algorithms, including Random Forest and Support Vector Machine, and a current deep Convolutional Neural Network (CNN) was employed to classify the benthic features. The overall accuracy of these classifiers are 80.74%, 84.19%, and 89.80% with the BRI, 80.17%, 82.75%, and 87.85% with the DII compared to 37.64%, 42.5%, and 47.2% of without corrected water columns respectively. The CNN model demonstrated that the approach significantly maximizes the improvement in benthic classification results in coastal region. [ABSTRACT FROM AUTHOR]

Seagrass ecosystems are crucial to the carbon cycle, biodiversity, fisheries, and erosion prevention. However, there are gaps in mapping seagrass in tropical coastal waters due to turbidity and constantly changing tidal dynamics. To tackle such issues, tidal heights and reflectance of 33 Sentinel-2 images were used to calculate a multitemporal composite of Sentinel-2 images, which was classified using the Random Forest classifier. UAV images classified with Object-based image analysis provided ground truth for training and validation. The effectiveness of water column correction and using the multitemporal composite was evaluated by comparing overall classification accuracy. As a result, applying water column correction improved classification accuracy from 80.4% to 80.6%, while using the multi-temporal composite improved accuracy to 88.6%, among the highest accuracy achieved for seagrass classification in turbid waters. The combined protocol could lead to advances in quantifying seagrass distribution in the tropical coastal waters and their carbon sequestration capacity, resolving significant uncertainties in blue carbon estimation. [ABSTRACT FROM AUTHOR]

Estimation of carbon stock in seagrass meadows is in challenges of paucity of assessment and low accuracy of the estimates. In this study, we used a fusion of the synthetic aperture radar (SAR) Sentinel-1 (S-1), the multi-spectral Sentinel-2 (S-2), and coupled this with advanced machine learning (ML) models and meta-heuristic optimization to improve the estimation of total organic carbon (TOC) stock in the Zostera muelleri meadows in Tauranga Harbour, New Zealand. Five scenarios containing combinations of data, ML models (Random Forest, Extreme Gradient Boost, Rotation Forest, CatBoost) and optimization were developed and evaluated for TOC retrieval. Results indicate a fusion of S1, S2 images, a novel ML model CatBoost and the grey wolf optimization algorithm (the CB-GWO model) yielded the best prediction of seagrass TOC (R2, RMSE were 0.738 and 10.64 Mg C ha−1). Our results provide novel ideas of deriving a low-cost, scalable and reliable estimates of seagrass TOC globally. [ABSTRACT FROM AUTHOR]