Performance comparison of deep learning and machine learning methods in determining wetland water areas using EuroSAT dataset

Wetland classification based on depth-adaptive convolutional neural networks using leaf-off SAR imagery

The recent development of deep learning (DL) techniques has created opportunities for classifying wetlands from remote sensing data (mainly optical data). However, the methods for accurately and efficiently classifying large-scale wetlands using DL and radar data that can be more effective than optical data still needs evaluation. In this study, we developed an end-to-end depth-adaptive convolutional neural network (CNN) for mapping wetlands using leaf-off time-series Sentinel-1 Synthetic Aperture Radar (SAR) imagery along with ancillary data. We examined the inclusion of multi-land cover proximity information and a CNN-based self-supervised SAR denoising procedure for enhancing wetland classification accuracy. The depth-adaptive CNN based on U-Net architecture was designed to classify wetland classes (emergent wetland, scrub-shrub wetland, forested wetland, and open water) in Delaware, U.S. while achieving optimization between model complexity (network depths) and accuracy. Results show that our proposed DL method (OA = 0.93, MIoU = 0.60) not only produced a higher classification accuracy than the traditional RF method (OA = 0.89, MIoU = 0.18) but also had a significantly reduced computational cost compared to established state-of-the-art CNNs (e.g., DeepLabV3+ and DANet) without loss of accuracy. The inclusion of multi-land cover proximity information (especially distances to forest and water) and the CNN-based self-supervised SAR denoising procedure can both enhance wetland classification accuracy, especially for forested wetland using traditional RF methods. These results demonstrated the novelty and efficiency of our proposed DL method for classifying wetlands by combing denoised SAR imagery and ancillary information, which provides insights on integration of DL approach and radar data for supporting operational wetland mapping at large spatial scales.

Exploring the forensic effectiveness and population genetic differentiation in Guizhou Miao and Bouyei group by the self-constructed panel of X chromosomal multi-insertion/deletions

In this research, a self-developed panel comprising 22 X chromosomal multi-InDels and one X-STR was used to explore the genetic polymorphisms and forensic characteristics of these loci in Guizhou Miao and Guizhou Bouyei populations. Besides, genetic affiliations among Guizhou Miao, Guizhou Bouyei and Guizhou Han populations were investigated using principal component analysis, STRUCTURE and machine learning methods. The findings indicated that these loci in the male and female samples had comprehensive discrimination powers greater than 0.999999999. Meanwhile, the cumulative mean exclusion chance of these 23 loci for trio and duo cases were also greater than 0.9999 in Guizhou Miao and Guizhou Bouyei populations. Population genetic analyses of three Guizhou populations revealed that there were relatively low genetic divergences among these populations based on the self-constructed panel. In conclusion, this system could be utilized as the valuable tool for forensic personal identification and parentage testing in Guizhou Miao and Guizhou Bouyei populations.

Multi-Source Remote Sensing Data for Wetland Information Extraction: A Case Study of the Nanweng River National Wetland Reserve

Wetlands play a vital role in regulating the global carbon cycle, providing biodiversity, and reducing flood risks. These functions maintain ecological balance and ensure human well-being. Timely, accurate monitoring of wetlands is essential, not only for conservation efforts, but also for achieving Sustainable Development Goals (SDGs). In this study, we combined Sentinel-1/2 images, terrain data, and field observation data collected in 2020 to better understand wetland distribution. A total of 22 feature variables were extracted from multi-source data, including spectral bands, spectral indices (especially red edge indices), terrain features, and radar features. To avoid high correlations between variables and reduce data redundancy, we selected a subset of features based on recursive feature elimination (RFE) and Pearson correlation analysis methods. We adopted the random forest (RF) method to construct six wetland delineation schemes and incorporated multiple types of characteristic variables. These variables were based on remote sensing image pixels and objects. Combining red-edge features, terrain data, and radar data significantly improved the accuracy of land cover information extracted in low-mountain and hilly areas. Moreover, the accuracy of object-oriented schemes surpassed that of pixel-level methods when applied to wetland classification. Among the three pixel-based schemes, the addition of terrain and radar data increased the overall classification accuracy by 7.26%. In the object-based schemes, the inclusion of radar and terrain data improved classification accuracy by 4.34%. The object-based classification method achieved the best results for swamps, water bodies, and built-up land, with relative accuracies of 96.00%, 90.91%, and 96.67%, respectively. Even higher accuracies were observed in the pixel-based schemes for marshes, forests, and bare land, with relative accuracies of 98.67%, 97.53%, and 80.00%, respectively. This study's methodology can provide valuable reference information for wetland data extraction research and can be applied to a wide range of future research studies.

Transformer-based land use and land cover classification with explainability using satellite imagery

Transformer-based models have greatly improved Land Use and Land Cover (LULC) applications. Their revolutionary ability to analyze and extract key information has greatly advanced the field. However, the high computational cost of these models presents a considerable obstacle to their practical implementation. Therefore, this study aims to strike a balance between computational cost and accuracy when employing transformer-based models for LULC analysis. We exploit transfer learning and fine-tuning strategies to optimize the resource utilization of transformer-based models. Furthermore, transparency is the core principle of our methodology to promote fairness and trust in applying LULC models across various domains, including forestry, environmental studies, and urban or rural planning. To ensure transparency, we have employed Captum, which enables us to uncover and mitigate potential biases and interpret AI-driven decisions. Our results indicate that transfer learning can potentially improve transformer-based models in satellite image classification, and strategic fine-tuning can maintain efficiency with minimal accuracy trade-offs. This research highlights the potential of Explainable AI (XAI) in Transformer-based models for achieving more efficient and transparent LULC analysis, thereby encouraging continued innovation in the field.

Generation and classification of patch-based land use and land cover dataset in diverse Indian landscapes: a comparative study of machine learning and deep learning models

In the context of environmental and social applications, the analysis of land use and land cover (LULC) holds immense significance. The growing accessibility of remote sensing (RS) data has led to the development of LULC benchmark datasets, especially pivotal for intricate image classification tasks. This study addresses the scarcity of such benchmark datasets across diverse settings, with a particular focus on the distinctive landscape of India. The study entails the creation of patch-based datasets, consisting of 4000 labelled images spanning four distinct LULC classes derived from Sentinel-2 satellite imagery. For the subsequent classification task, three traditional machine learning (ML) models and three convolutional neural networks (CNNs) were employed. Despite facing several challenges throughout the process of dataset generation and subsequent classification, the CNN models consistently attained an overall accuracy of 90% or more. Notably, one of the ML models stood out with 96% accuracy, surpassing CNNs in this specific context. The study also conducts a comparative analysis of ML models on existing benchmark datasets, revealing higher prediction accuracy when dealing with fewer LULC classes. Thus, the selection of an appropriate model hinges on the given task, available resources, and the necessary trade-offs between performance and efficiency, particularly crucial in resource-constrained settings. The standardized benchmark dataset contributes valuable insights into the relative performance of deep CNN and ML models in LULC classification, providing a comprehensive understanding of their strengths and weaknesses.

Remote sensing and monitoring of water resources: A comparative study of different indices and thresholding methods

Water resources are essential for the ecological system and the development of civilization. Water is imperative factor for health preservation and sustaining various human activities, including industrial production, agriculture, and daily life. Remote sensing provides a cost-effective and practical means to detect and monitor water bodies, offers valuable insights into the impact of climatic events on water structures, especially in coastal lake regions. The research primarily utilizes Landsat-9 OLI-2 satellite images to evaluate the effectiveness of various water indices (WRI, NWI, MNDWI, NDWI) in combination with global automatic thresholding methods (K-Means, Zhenzhou's, Adaptive, Intermodes, Prewitt and Mendelsohn's Minimum, Maximum Entropy, Median, Concavity, Percentile, Intermeans, Kittler and Illingworth's Minimum Error, Tsai's Moments, Otsu's, Huang's fuzzy, Triangle, Mean, IsoData, Li's). The study was carried out on Lake Nazik, Lake Iznik, and Lake Beyşehir, which have unique geographical characteristics, and examined the adaptability and robustness of the selected indices and thresholding methods. MNDWI consistently stands out as a robust index for water extraction, delivering accurate results across different thresholding methods in regions all three lakes. As a result of quite extensive analysis, it is obtained that MNDWI and NDWI are reliable choices for water feature extraction in various lake environments, but the specific index should consider the thresholding method and unique lake characteristics. The Minimum thresholding method stands out as the most effective thresholding technique, demonstrating impressive results across different lakes. Specifically, it achieved an average Peak Signal-to-Noise Ratio (PSNR) of 78.97 and Structural Similarity Index (SSIM) of 99.37 for Lake Nazik, 74.08 PSNR and 98.34 SSIM for Lake Iznik, and 63.96 PSNR and 93.61 SSIM for Lake Beyşehir.

Spatiotemporal analysis of transition probabilities of wet and dry days under SSPs scenarios in the semi-arid Susurluk Basin, Türkiye

Precipitation, especially in regions dominated by the Mediterranean climate, is one of the most critical parameters of the hydrological cycle and the environment affected by climate change. One the one hand, the transition probabilities of wet and dry days in precipitation occurrence are a relatively new topic, on the other hand these are essential in defining the regional climate. For the first time, spatiotemporal variations of transition probabilities of wet and dry days in the Susurluk Basin, northwestern Türkiye, dominated by a semi-arid Mediterranean climate and also having a mountain climate, were analyzed based on the observation (1979-2014) and future terms (2030-2059 as short and 2070-2099 as long), under four Shared Socioeconomic Pathways (SSPs) scenarios. To do this, statistical downscaling was performed for 14 general circulation models (GCMs) from the CMIP6. By applying an ensemble of four high-performing GCMs, four indices for transition probabilities of wet and dry, i.e., a dry day following a dry day (FDD), a wet day following a dry day (FDW), a dry day following a wet day (FWD), and a wet day following a wet day (FWW), were calculated, and their changes were determined statistically. Monotonic and partial trends of the indices were also analyzed. According to the results, the FDD will increase in water year and wet period and autumn in the future, especially for the long term, in the basin dominated by the FDD (75 % in water year). The risks are higher in the western part of the basin, where human activities are intense, as the FDD is higher in this part than other parts especially in summer (90-100 %) in SSP3-7.0 and SSP5-8.5 scenarios for the long term. So, the length of consecutive dry days in the wet period and water year will increase in the basin, thus increasing the likelihood of droughts. As for the intra-term trends, the FDD increases and the FWW decreases in the water year and seasons in SSP3-7.0 and SSP5-8.5, contrary to the observation term.

Analysis of surface water area dynamics and driving forces in the Bosten Lake basin based on GEE and SEM for the period 2000 to 2021

As an inland dryland lake basin, the rivers and lakes within the Lake Bosten basin provide scarce but valuable water resources for a fragile environment and play a vital role in the development and sustainability of the local societies. Based on the Google Earth Engine (GEE) platform, combined with the geographic information system (GIS) and remote sensing (RS) technology, we used the index WI2019 to extract and analyze the water body area changes of the Bosten Lake basin from 2000 to 2021 when the threshold value is -0.25 and the slope mask is 8°. The driving factors of water body area changes were also analyzed using the partial least squares-structural equation model (PLS-SEM). The result shows that in the last 20 years, the area of water bodies in the Bosten Lake basin generally fluctuated during the dry, wet, and permanent seasons, with a decreasing trend from 2000 to 2015 and an increasing trend between 2015 and 2019 followed by a steadily decreasing trend afterward. The main driver of the change in wet season water bodies in the Bosten Lake basin is the climatic factors, with anthropogenic factors having a greater influence on the water body area of dry season and permanent season than that of wet season. Our study achieved an accurate and convenient extraction of water body area and drivers, providing up-to-date information to fully understand the spatial and temporal variation of surface water body area and its drivers in the basin, which can be used to effectively manage water resources.

Anthropogenic impacts on urban blue space and its reciprocal effect on human and socio-ecological health

Quantifying anthropogenic impacts on blue space (BS) and its effect on human and socio-ecological health was least explored. The present study aimed to do this in reference to the urban BS transformation scenario of Eastern India. To measure BS transformation, Landsat image-based water indices were run from 1990 to 2021. Anthropogenic impact score (AIS) and 7 components scores of 78 selected BS on 70 parameters related data driven from the field. Total 345 respondents were taken for human and socio-ecological health assessment. For this, depression (DEP), anxiety (ANX), stress (STR), physical activities (PA), social capital (SC), therapeutic landscape (TL) and environment building (EB) parameters were taken. The result exhibited that BS was reduced. About 50% of urban core BS was reported highly impacted. Human and socio-ecological health was identified as good in proximity to BS, but it was observed better in the cases of larger peripheral BS. AIS on BS was found to be positively associated with mental health (0.47-0.63) and negatively associated with PA, SC, TL and EB (-0.50 to -0.90). Standard residual in ordinary least square was reported low (-1.5 to 1.5) in 95% BS. Therefore, BS health restoration and management is crucial for sustaining the living environment.

Multi-spatial-scale land/use land cover influences on seasonally dominant water quality along Middle Ganga Basin

Studying spatiotemporal water quality characteristics and their correlation with land use/land cover (LULC) patterns is essential for discerning the origins of various pollution sources and for informing strategic land use planning, which, in turn, requires a comprehensive analysis of spatiotemporal water quality data to comprehend how surface water quality evolves across different time and space dimensions. In this study, we compared catchment, riparian, and reach scale models to assess the effect of LULC on WQ. Using various multivariate techniques, a 14-year dataset of 20 WQ variables from 20 monitoring stations (67,200 observations) is studied along the Middle Ganga Basin (MGB). Based on the similarity and dissimilarity of WQPs, the K-means clustering algorithm classified the 20 monitoring stations into four clusters. Seasonally, the three PCs chosen explained 75.69% and 75% of the variance in the data. With PCs > 0.70, the variables EC, pH, Temp, TDS, NO2 + NO3, P-Tot, BOD, COD, and DO have been identified as dominant pollution sources. The applied RDA analysis revealed that LULC has a moderate to strong contribution to WQPs during the wet season but not during the dry season. Furthermore, dense vegetation is critical for keeping water clean, whereas agriculture, barren land, and built-up area degrade WQ. Besides that, the findings suggest that the relationship between WQPs and LULC differs at different scales. The stacked ensemble regression (SER) model is applied to understand the model's predictive power across different clusters and scales. Overall, the results indicate that the riparian scale is more predictive than the watershed and reach scales.

Mapping small inland wetlands in the South-Kivu province by integrating optical and SAR data with statistical models for accurate distribution assessment

There are several techniques for mapping wetlands. In this study, we examined four statistical models to assess the potential distribution of wetlands in the South-Kivu province by combining optical and SAR images. The approach involved integrating topographic, hydrological, and vegetation indices into the four most used classifiers, namely Artificial Neural Network (ANN), Random Forest (RF), Boosted Regression Tree (BRT), and Maximum Entropy (MaxEnt). A wetland distribution map was generated and classified into 'wetland' and 'non-wetland.' The results showed variations in predictions among the different models. RF exhibited the most accurate predictions, achieving an overall classification accuracy of 95.67% and AUC and TSS values of 82.4%. Integrating SAR data improved accuracy and precision, particularly for mapping small inland wetlands. Our estimations indicate that wetlands cover approximately 13.5% (898,690 ha) of the entire province. BRT estimated wetland areas to be ~ 16% (1,106,080 ha), while ANN estimated ~ 14% (967,820 ha), MaxEnt ~ 15% (1,036,950 ha), and RF approximately ~ 10% (691,300 ha). The distribution of these areas varied across different territories, with higher values observed in Mwenga, Shabunda, and Fizi. Many of these areas are permanently flooded, while others experience seasonal inundation. Through digitization, the delineation process revealed variations in wetland areas, ranging from tens to thousands of hectares. The geographical distribution of wetlands generated in this study will serve as an essential reference for future investigations and pave the way for further research on characterizing and categorizing these areas.

Sensitivity Analysis of Parameters Affecting Wetland Water Levels: A Study of Flood Detention Basin, Colombo, Sri Lanka

Wetlands play a vital role in ecosystems. They help in flood accumulation, water purification, groundwater recharge, shoreline stabilization, provision of habitats for flora and fauna, and facilitation of recreation activities. Although wetlands are hot spots of biodiversity, they are one of the most endangered ecosystems on the Earth. This is not only due to anthropogenic activities but also due to changing climate. Many studies can be found in the literature to understand the water levels of wetlands with respect to the climate; however, there is a lack of identification of the major meteorological parameters affecting the water levels, which are much localized. Therefore, this study, for the first time in Sri Lanka, was carried out to understand the most important parameters affecting the water depth of the Colombo flood detention basin. The temporal behavior of water level fluctuations was tested among various combinations of hydro-meteorological parameters with the help of Artificial Neural Networks (ANN). As expected, rainfall was found to be the most impacting parameter; however, apart from that, some interesting combinations of meteorological parameters were found as the second layer of impacting parameters. The rainfall-nighttime relative humidity, rainfall-evaporation, daytime relative humidity-evaporation, and rainfall-nighttime relative humidity-evaporation combinations were highly impactful toward the water level fluctuations. The findings of this study help to sustainably manage the available wetlands in Colombo, Sri Lanka. In addition, the study emphasizes the importance of high-resolution on-site data availability for higher prediction accuracy.

Wetland transformation and its impact on the livelihood of the fishing community in a flood plain river basin of India

Discretely wetland transformations and livelihood vulnerability related works are profoundly found worldwide, but their linkage is not investigated often. The present study aimed to explore the after damming transformation of wetland's eco-hydrological status and water quality and assessed its effects on livelihood vulnerability state of the fishermen community in the lower part of the Tangon river basin. Based on 15 field and satellite image-driven indicators of transformation, multiple machine learning (ML) algorithms were used to model the eco-hydrological state (EHS) of the wetland. Livelihood Vulnerability Index (LVI) of 45 fishing-dominated villages was computed using a balanced weighted LVI score. The result revealed that 60.55 % wetland area was obliterated between the pre- dam and post-dam periods, and the existing wetland area (21.06 km²) witnessed noticeable eco-hydrological and water quality degradation. Correlation and kernel density estimation (KDE) plot clearly revealed that rate of EHS degradation and water quality changes was negatively associated (at ≤0.01 level of significance) and both controlled LVI. So, such changes not only pose pressure on the aquatic species like fishes but also hampered the well-being of the fishermen communities evolving. The findings of the work would be useful in this transition while deciding the alternative strategies to build a resilient community. Moreover, since the eco-hydrological state were explored this would be effective for wetland restoration planning.

Remote sensing and machine learning based framework for the assessment of spatio-temporal water quality in the Middle Ganga Basin

Understanding the dynamics of water quality in any water body is vital for the sustainability of our water resources. Thus, investigating spatio-temporal changes of dominant water quality parameters (WQPs) in any study is indeed critical for proposing the appropriate treatment for the water bodies. Traditionally, concentrations of WQPs have been measured through intensive fieldwork. Additionally, many studies have attempted to retrieve concentrations of WQPs from satellite images using regression-based methods. However, the relationship between WQPs and satellite data is complex to be modeled accurately by using simple regression-based methods. Our study attempts to develop a machine learning model for mapping the concentrations of dominant optical and non-optical WQPs such as electrical conductivity (EC), pH, temperature (Temp), total dissolved solids (TDS), silicon dioxide (SiO₂), and dissolved oxygen (DO). In this context, a remote sensing framework based on the extreme gradient boosting (XGBoost) and multi-layer perceptron (MLP) regressor with optimized hyper parameters (HPs) to quantify concentrations of different WQPs from the Landsat-8 satellite imagery is developed. We evaluated six years of satellite data stretching spatially from upstream to downstream Ankinghat to Chopan (20 stations under Central Water Commission (CWC), Middle Ganga Basin) for characterizing the trends of dominant physico-chemical WQPs across the four clusters identified in our previous study. Through the developed XGBoost and MLP regression models between measured WQPs and the reflectance of the pixels corresponding to the sampling stations, a significant coefficient of determination (R²) in the range of 0.88-0.98 for XGBoost and 0.72-0.97 for MLP were generated, with bands B1-B4 and their ratios more consistent. Indeed, these findings indicate that from a small number of in-situ measurements, we can develop reliable models to estimate the spatio-temporal variations of physico-chemical and biological WQPs. Therefore, models generated from Landsat-8 could facilitate the environmental, economic, and social management of any waterbody.

Convolutional neural networks for approximating electrical and thermal conductivities of Cu-CNT composites

This article explores the deep learning approach towards approximating the effective electrical and thermal conductivities of copper (Cu)-carbon nanotube (CNT) composites with CNTs aligned to the field direction. Convolutional neural networks (CNN) are trained to map the two-dimensional images of stochastic Cu-CNT networks to corresponding conductivities. The CNN model learns to estimate the Cu-CNT composite conductivities for various CNT volume fractions, interfacial electrical resistances, R_c = 20 Ω–20 kΩ, and interfacial thermal resistances, R^″_t,c = 10⁻¹⁰–10⁻⁷ m²K/W. For training the CNNs, the hyperparameters such as learning rate, minibatch size, and hidden layer neurons are optimized. Without iteratively solving the physical governing equations, the trained CNN model approximates the electrical and thermal conductivities within a second with the coefficient of determination (R²) greater than 98%, which may take longer than 100 min for a convectional numerical simulation. This work demonstrates the potential of the deep learning surrogate model for the complex transport processes in composite materials.

Identifying Urban Wetlands through Remote Sensing Scene Classification Using Deep Learning: A Case Study of Shenzhen, China

Urban wetlands provide cities with unique and valuable ecosystem services but are under great degradation pressure. Correctly identifying urban wetlands from remote sensing images is fundamental for developing appropriate management and protection plans. To overcome the semantic limitations of traditional pixel-level urban wetland classification techniques, we proposed an urban wetland identification framework based on an advanced scene-level classification scheme. First, the Sentinel-2 high-resolution multispectral image of Shenzhen was segmented into 320 m × 320 m square patches to generate sample datasets for classification. Next, twelve typical convolutional neural network (CNN) models were transformed for the comparison experiments. Finally, the model with the best performance was used to classify the wetland scenes in Shenzhen, and pattern and composition analyses were also implemented in the classification results. We found that the DenseNet121 model performed best in classifying urban wetland scenes, with overall accuracy (OA) and kappa values reaching 0.89 and 0.86, respectively. The analysis results revealed that the wetland scene in Shenzhen is generally balanced in the east–west direction. Among the wetland scenes, coastal open waters accounted for a relatively high proportion and showed an obvious southward pattern. The remaining swamp, marsh, tidal flat, and pond areas were scattered, accounting for only 4.64% of the total area of Shenzhen. For scattered and dynamic urban wetlands, we are the first to achieve scene-level classification with satisfactory results, thus providing a clearer and easier-to-understand reference for management and protection, which is of great significance for promoting harmony between humanity and ecosystems in cities.