Deep learning models for monitoring landscape changes in a UNESCO Global Geopark.

By identifying Earth heritage sites, UNESCO Global Geoparks (UGGps) have promoted geo-tourism and regional economic prosperity. However, commercial and tourism development has altered the natural contexts of these geoparks, diminishing their initial value. Before implementing land use policies, spatial landscape parameters should be monitored in multiple dimensions and in real time. This study aims to develop Bilateral Segmentation Network (BiSeNet) models employing an upgraded U-structured neural network in order to monitor land use/cover changes and landscape indicators in a Vietnamese UGGp. This network has proven effective at preserving input image data and restricting the loss of spatial information in decoding data. To demonstrate the utility of deep learning, eight trained BiSeNet models were evaluated against Maximum Likelihood, Support Vector Machine, and Random Forest. The trained BSN-Nadam model (128x128), with a precision of 94% and an information loss of 0.1, can become a valuable instrument for analyzing and monitoring monthly changes in land uses/covers once tourism activities have been rapidly expanded. Three tourist routes and 41 locations in the Dak Nong UGGp were monitored for 30 years using three landscape indices: Disjunct Core Area Density (DCAD), Total Edge Contrast Index (TECI), Shannon's Diversity Index (SHDI), based on the results of the model. As a result, 18 identified geo-sites in the Daknong Geopark have been influenced significantly by agricultural and tourist activities since 2010, making these sites less uniform and unsustainable management. It promptly alerts UNESCO management to the deterioration of geological sites caused by urbanization and tourist development.

Coastal landscape classification using convolutional neural network and remote sensing data in Vietnam.

The length of global coastline is about 356 thousand kilometers with various dynamic natural and anthropogenic. Although the number of studies on coastal landscape categorization has been increasing, it is still difficult to distinguish precisely them because the used methods commonly are traditional qualitative ones. With the leverage of remote sensing data and GIS tools, it helps categorize and identify a variety of features on land and water based on multi-source data. The aim of study is using different natural - social profile data obtained from ALOS, NOAA, and multi-temporal Landsat satellite images as input data of the convolutional-neural-network (CvNet) models for coastal landscape classification. Studies used 900 cut-line samples which represent coastal landscapes in Vietnam for training and optimizing CvNet models. As a result, nine coastal landscapes were identified including: deltas, alluvial, mature and young sand dunes, cliff, lagoon, tectonic, karst, and transitional landscapes. Three CvNet models using three different optimizer types classified the landscapes of other 1150 cut-lines in Vietnam with the accuracies about 98% and low loss function value. Excepting dalmatian, karst and delta coastal landscapes, five others distribute heterogeneous along the coasts in Vietnam. Therefore, the evaluation of additional natural components is necessary and CvNet model have ability to update new landscape types in variety of tropical nation as a step toward coastal landscape classification at both national and global scales.

Application of deep learning models to detect coastlines and shorelines.

Identifying and monitoring coastlines and shorelines play an important role in coastal erosion assessment around the world. The application of deep learning models was used in this study to detect coastlines and shorelines in Vietnam using high-resolution satellite images and different object segmentation methods. The aims are to (1) propose indicators to identify coastlines and shorelines; (2) build deep learning (DL) models to automatically interpret coastlines and shorelines from high-resolution remote sensing images; and (3) apply DL-trained models to monitor coastal erosion in Vietnam. Eight DL models were trained based on four artificial-intelligent-network structures, including U-Net, U2-Net, U-Net3+, and DexiNed. The high-resolution images collected from Google Earth Pro software were used as input data for training all models. As a result, the U-Net using an input-image size of 512 × 512 provides the highest performance of 98% with a loss function of 0.16. The interpretation results of this model were used effectively for the coastline and shoreline identification in assessing coastal erosion in Vietnam due to sea-level rise in storm events over 20 years. The outcomes proved that while the shoreline is ideal for observing seasonal tidal changes or the immediate motions of current waves, the coastline is suitable to assess coastal erosion caused by the influence of sea-level rise during storms. This paper has provided a broad scope of how the U-Net model can be used to predict the coastal changes over vietnam and the world.

Integrated methods and scenarios for assessment of sand dunes ecosystem services.

Anthropogenic and natural ecosystems in coastal dunes provide considerable benefits to human well-being. However, to date, we still lack a good understanding of how ecosystem services (ES) supply varies from young dunes (e.g., embryo and fore dunes) to mature dunes (e.g., brown and red dunes). This study proposed a novel modelling methodology by integrating an expert-based matrix, a Bayesian Belief Network (BBN), a structural equation model, and a scenario development method. It aims at evaluating dune ecosystem services for the sustainable development of coastal areas. The model was tested using data collected from dunes in Vietnam. An expert-based matrix to assess the supply capacity of 18 ES in different types of dunes was generated with the participation of 21 interdisciplinary scientists. It was found that red dune ecosystems could supply the most regulation and cultural ecosystem services, while gray dunes provided the least amount. Results from a scenario analysis recommended that decision-making is able to optimize multiple ES by: (i) keeping embryo/fore dunes in their natural state instead of using them for mineral mining and urbanization; (ii) enlarging certified and protected forests areas in gray and yellow dunes; and (iii) optimizing cultural ES supply in red dunes.

Ecosystem service value assessment of a natural reserve region for strengthening protection and conservation.

Ecosystem Services (ESs) refer to the direct and indirect contributions of ecosystems to human well-being and subsistence. Ecosystem valuation is an approach to assign monetary values to an ecosystem and its key ecosystem goods and services, generally referred to as Ecosystem Service Value (ESV). We have measured spatiotemporal ESV of 17 key ESs of Sundarbans Biosphere Reserve (SBR) in India using temporal remote sensing (RS) data (for years 1973, 1988, 2003, 2013, and 2018). These mangrove ecosystems are crucial for providing valuable supporting, regulatory, provisioning, and cultural ecosystem services. We have adopted supervised machine learning algorithms for classifying the region into different ecosystem units. Among the used machine learning models, Support Vector Machine (SVM) and Random Forest (RF) algorithms performed the most accurate and produced the best classification estimates with maximum kappa and an overall accuracy value. The maximum ESV (derived from both adjusted and non-adjusted units, million US$ year-1) is produced by mangrove forest, followed by the coastal estuary, cropland, inland wetland, mixed vegetation, and finally urban land. Out of all the ESs, the waste treatment (WT) service is the dominant ecosystem service of SBR. Additionally, the mangrove ecosystem was found to be the most sensitive to land use and land cover changes. The synergy and trade-offs between the ESs are closely associated with the spatial extent. Therefore, accurate estimates of ES valuation and mapping can be a robust tool for assessing the effects of poor decision making and overexploitation of natural resources on ESs.

Comparison between U-shaped structural deep learning models to detect landslide traces.

Landslides endanger lives and public infrastructure in mountainous areas. Monitoring landslide traces in real-time is difficult for scientists, sometimes costly and risky because of the harsh terrain and instability. Nowadays, modern technology may be able to identify landslide-prone locations and inform locals for hours or days when the weather worsens. This study aims to propose indicators to detect landslide traces on the fields and remote sensing images; build deep learning (DL) models to identify landslides from Sentinel-2 images automatically; and apply DL-trained models to detect this natural hazard in some particular areas of Vietnam. Nine DL models were trained based on three U-shaped architectures, including U-Net, U2-Net, and U-Net3+, and three options of input sizes. The multi-temporal Sentinel-2 images were chosen as input data for training all models. As a result, the U-Net, using an input image size of 32 × 32 and a performance of 97 % with a loss function of 0.01, can detect typical landslide traces in Vietnam. Meanwhile, the U-Net (64 × 64) can detect more considerable landslide traces. Based on multi-temporal remote sensing data, a different case study in Vietnam was chosen to see landslide traces over time based on the trained U-Net (32 × 32) model. The trained model allows mountain managers to track landslide occurrences during wet seasons. Thus, landslide incidents distant from residential areas may be discovered early to warn of flash floods.

Monitoring the effects of urbanization and flood hazards on sandy ecosystem services.

Urbanization, storms, and floods have compromised the benefits derived from various types of sand dune landscapes, particularly in developing countries located in humid monsoon tropical regions. One pertinent question is which driving forces have had a dominant impact on the contributions of sand dune ecosystems to human well-being. Has the decline in sand dune ecosystem services (ES) been primarily due to urbanization or flooding hazards? This study aims to address these issues by developing a Bayesian Belief Network (BBN) to analyze six different sand dune landscapes worldwide. The study uses various data types, including multi-temporal and -sensor remote sensing (SAR and optical data), expert knowledge, statistics, and GIS to analyze the trends in sand dune ecosystems. A support tool based on probabilistic approaches was developed to assess changes in ES over time due to the effects of urbanization and flooding. The developed BBN has the potential to assess the ES values of sand dunes during both rainy and dry seasons. The study calculated and tested the ES values in detail over six years (from 2016 to 2021) in Quang Nam province, Vietnam. The results showed that urbanization has led to an increase in the total ES values since 2016, while floods only had a minimal impact on dune ES values during the rainy season. The fluctuations of ES values were found to be more significant due to urbanization than floods. The study's approach can be useful in future research on coastal ecosystems.

A new deep learning approach based on bilateral semantic segmentation models for sustainable estuarine wetland ecosystem management.

Nowadays, estuarial areas have been strongly affected by the construction of electrical power dams from upstream, downstream urbanization and many types of hazards along the coastal regions. It has resulted in significant changes in estuarine wetland ecosystems between rainy and dry seasons. To avoid estuary vulnerability, monitoring and evaluation of the estuarine ecosystems are very critical tasks. The main goal of this research is to propose and implement a novel deep learning method in monitoring various ecosystems in estuarine regions. The processing speed and accuracy of common neural networks is improved more than ten times through spatial and context paths integrated into a novel Bilateral Segmentation Network (BiSeNet). The multi-sensor and multi-temporal satellite images (including Sentinel-2, ALOS-DEM, and NOAA-DEM images) served as input data. As a result, four BiSeNet models out of 20 trained models achieved a greater than 90% accuracy, especially for interpreting estuarine waters, intertidal forested wetlands, and aquacultural lands in subtidal regions. These models outperformed Random Forest and Support Vector Machine approaches. The best one was used to map estuarine ecosystems from 12 satellite images over a five-year period in the largest estuary in northern Vietnam. The ecosystem changes between dry and rainy seasons were analyzed in detail to assess the ecological succession in estuaries. Furthermore, this model can potentially update new estuarine ecosystem types in other estuarine areas across the world, making possible real-time monitoring and assessing estuarine ecological conditions for sustainable management of wetland ecosystem.

A novel intelligence approach based active and ensemble learning for agricultural soil organic carbon prediction using multispectral and SAR data fusion.

Monitoring agricultural soil organic carbon (SOC) has played an essential role in sustainable agricultural management. Precise and robust prediction of SOC greatly contributes to carbon neutrality in the agricultural industry. To create more knowledge regarding the ability of remote sensing to monitor carbon soil, this research devises a state-of-the-art low cost machine learning model for quantifying agricultural soil carbon using active and ensemble-based decision tree learning combined with multi-sensor data fusion at a national and world scale. This work explores the use of Sentinel-1 (S1) C-band dual polarimetric synthetic aperture radar (SAR), Sentinel-2 (S2) multispectral data, and an innovative machine learning (ML) approach using an integration of active learning for land-use mapping and advanced Extreme Gradient Boosting (XGBoost) for robustness of the SOC estimates. The collected soil samples from a field survey in Western Australia were used for the model validation. The indicators including the coefficient of determination (R2) and root - mean - square - error (RMSE) were applied to evaluate the model's performance. A numerous features computed from optical and SAR data fusion were employed to build and test the proposed model performance. The effectiveness of the proposed machine learning model was assessed by comparing with the two well-known algorithms such as Random Forests (RF) and Support Vector Machine (SVM) to predict agricultural SOC. Results suggest that a combination of S1 and S2 sensors could effectively estimate SOC in farming areas by using ML techniques. Satisfactory accuracy of the proposed XGBoost with optimal features was achieved the highest performance (R2 = 0.870; RMSE = 1.818 tonC/ha) which outperformed RF and SVM. Thus, multi-sensor data fusion combined with the XGBoost lead to the best prediction results for agricultural SOC at 10 m spatial resolution. In short, this new approach could significantly contribute to various agricultural SOC retrieval studies globally.

New approach of water quantity vulnerability assessment using satellite images and GIS-based model: An application to a case study in Vietnam.

Water deficiency due to climate change and the world's population growth increases the demand for the water industry to carry out vulnerability assessments. Although many studies have been done on climate change vulnerability assessment, a specific framework with sufficient indicators for water vulnerability assessment is still lacking. This highlights the urgent need to devise an effective model framework in order to provide water managers and authorities with the level of water exposure, sensitivity, adaptive capacity and water vulnerability to formulate their responses in implementing water management strategies. The present study proposes a new approach for water quantity vulnerability assessment based on remote sensing satellite data and GIS ModelBuilder. The developed approach has three layers: (1) data acquisition mainly from remote sensing datasets and statistical sources; (2) calculation layer based on the integration of GIS-based model and the Intergovernmental Panel on Climate Change's vulnerability assessment framework; and (3) output layer including the indices of exposure, sensitivity, adaptive capacity and water vulnerability and spatial distribution of remote sensing indicators and these indices in provincial and regional scale. In total 27 indicators were incorporated for the case study in Vietnam based on their availability and reliability. Results show that the most water vulnerable is the South Central Coast of the country, followed by the Northwest area. The novel approach is based on reliable and updated spatial-temporal datasets (soil water stress, aridity index, water use efficiency, rain use efficiency and leaf area index), and the incorporation of the GIS-based model. This framework can then be applied effectively for water vulnerability assessment of other regions and countries.