Monitoring recent changes of vegetation in Fildes Peninsula (King George Island, Antarctica) through satellite imagery guided by UAV surveys

Monitoring of Antarctica's Fragile Vegetation Using Drone-Based Remote Sensing, Multispectral Imagery and AI

Vegetation in East Antarctica, such as moss and lichen, vulnerable to the effects of climate change and ozone depletion, requires robust non-invasive methods to monitor its health condition. Despite the increasing use of unmanned aerial vehicles (UAVs) to acquire high-resolution data for vegetation analysis in Antarctic regions through artificial intelligence (AI) techniques, the use of multispectral imagery and deep learning (DL) is quite limited. This study addresses this gap with two pivotal contributions: (1) it underscores the potential of deep learning (DL) in a field with notably limited implementations for these datasets; and (2) it introduces an innovative workflow that compares the performance between two supervised machine learning (ML) classifiers: Extreme Gradient Boosting (XGBoost) and U-Net. The proposed workflow is validated by detecting and mapping moss and lichen using data collected in the highly biodiverse Antarctic Specially Protected Area (ASPA) 135, situated near Casey Station, between January and February 2023. The implemented ML models were trained against five classes: Healthy Moss, Stressed Moss, Moribund Moss, Lichen, and Non-vegetated. In the development of the U-Net model, two methods were applied: Method (1) which utilised the original labelled data as those used for XGBoost; and Method (2) which incorporated XGBoost predictions as additional input to that version of U-Net. Results indicate that XGBoost demonstrated robust performance, exceeding 85% in key metrics such as precision, recall, and F1-score. The workflow suggested enhanced accuracy in the classification outputs for U-Net, as Method 2 demonstrated a substantial increase in precision, recall and F1-score compared to Method 1, with notable improvements such as precision for Healthy Moss (Method 2: 94% vs. Method 1: 74%) and recall for Stressed Moss (Method 2: 86% vs. Method 1: 69%). These findings contribute to advancing non-invasive monitoring techniques for the delicate Antarctic ecosystems, showcasing the potential of UAVs, high-resolution multispectral imagery, and ML models in remote sensing applications.

The use of sentinel-2 imagery to generate vegetations maps for the Northern Antarctic peninsula and offshore islands

We used Sentinel-2 imagery time series to generate a vegetation map for the Northern part of the Antarctica Peninsula and offshore islands, including the South Shetlands. The vegetation cover was identified in the NDVI maximum value composite image. The NDVI values were associated with the occurrence of algae (0.15 - 0.20), lichens (0.20 - 0.50), and mosses (0.50 - 0.80). The vegetation cover distribution map was validated using the literature information. Generating a vegetation map distribution on an annual basis was not possible due to high cloud cover in the Antarctic region, especially in coastal áreas, so optical images from 2016 to 2021 were necessary to map the vegetation distribution in the entire study área. The final map analyzed in association with the weather data shows the occurrence of a microenvironment over the western islands of the Antarctic Peninsula that provided vegetation growth conditions. The Sentinel-2 images with 10m spatial resolution allow the assembly of accurate vegetation distribution maps for the Antarctica Peninsula and Islands, the Google Earth Engine cloud computing being essential to process a large amount of the satellite images necessary for processing these maps.

Accuracy of Vegetation Indices in Assessing Different Grades of Grassland Desertification from UAV

Grassland desertification has become one of the most serious environmental problems in the world. Grasslands are the focus of desertification research because of their ecological vulnerability. Their application on different grassland desertification grades remains limited. Therefore, in this study, 19 vegetation indices were calculated for 30 unmanned aerial vehicle (UAV) visible light images at five grades of grassland desertification in the Mu Us Sandy. Fractional Vegetation Coverage (FVC) with high accuracy was obtained through Support Vector Machine (SVM) classification, and the results were used as the reference values. Based on the FVC, the grassland desertification grades were divided into five grades: severe (FVC < 5%), high (FVC: 5-20%), moderate (FVC: 21-50%), slight (FVC: 51-70%), and non-desertification (FVC: 71-100%). The accuracy of the vegetation indices was assessed by the overall accuracy (OA), the kappa coefficient (k), and the relative error (RE). Our result showed that the accuracy of SVM-supervised classification was high in assessing each grassland desertification grade. Excess Green Red Blue Difference Index (EGRBDI), Visible Band Modified Soil Adjusted Vegetation Index (V-MSAVI), Green Leaf Index (GLI), Color Index of Vegetation Vegetative (CIVE), Red Green Blue Vegetation Index (RGBVI), and Excess Green (EXG) accurately assessed grassland desertification at severe, high, moderate, and slight grades. In addition, the Red Green Ratio Index (RGRI) and Combined 2 (COM₂) were accurate in assessing severe desertification. The assessment of the 19 indices of the non-desertification grade had low accuracy. Moreover, our result showed that the accuracy of SVM-supervised classification was high in assessing each grassland desertification grade. This study emphasizes that the applicability of the vegetation indices varies with the degree of grassland desertification and hopes to provide scientific guidance for a more accurate grassland desertification assessment.

Unmanned aerial vehicles (UAVs) as a tool for hazard assessment: The 2021 eruption of Cumbre Vieja volcano, La Palma Island (Spain)

Monitoring for assessment of natural disasters, such as volcanic eruptions, presents a methodological challenge for the scientific community. Here, we present Unmanned Aerial Vehicles (UAVs) as a feasible, precise, rapid and safe tool for real time monitoring of the impacts of a volcanic event during the Cumbre Vieja eruption on La Palma Island, Spain (2021). UAV surveys with optical RGB (Red-Green-Blue), thermal and multispectral sensors, and a water sampling device, were carried out in different areas affected by the lava flow, including the upper volcanic edifice and the lava delta formed on the coastal fringe of the island. Our results have provided useful information for the monitoring of the advance of the lava flow and its environmental consequences during the volcanic emergency. Our data shows how La Palma island's growth, with the formation of a new lava delta of 28 ha and a total volume of lava injected into the sea of 5,138,852 m³. Moreover, our Digital Elevation Model (DEM) simulated, with a 70 % accuracy, the probabilistic simulation of the possible path followed by the lava flow in the vicinity of the fissure from which the magma emanates. In addition, significant changes of seawater physical-chemical parameters were registered in coastal surface waters by the in situ seawater samples collected with the automatic water sampling device of our UAV. The first meters of the water column, due to the instant evaporation of the seawater in contact with the hot lava, produce an increase of temperature and salinity of up to 4-5 °C and up to 5 units, respectively.

Sustainable Oil Palm Resource Assessment Based on an Enhanced Deep Learning Method

Knowledge of the number and distribution of oil palm trees during the crop cycle is vital for sustainable management and predicting yields. The accuracy of the conventional image processing method is limited for the hand-crafted feature extraction method and the overfitting problem occurs due to the insufficient dataset. We propose a modification of the Faster Region-based Convolutional Neural Network (FRCNN) for palm tree detection to reduce the overfitting problem and improve the detection accuracy. The enhanced FRCNN (EFRCNN) leads to improved performance for detecting objects (in the same image) when they are of multiple sizes by using a feature concatenation method. Transfer learning based on a ResNet50 model is used to extract the features of the input image. High-resolution images of oil palm trees from a drone are used to form the data set, containing mature, young, and mixed oil palm tree regions. We train and test the EFRCNN, the FRCNN, a CNN used recently for oil palm image detection, and two standard methods, namely, the support vector machine (SVM) and template matching (TM). The results reveal an overall accuracy of ≥96.8% for the EFRCNN on the three test sets. The accuracy is higher than the CNN and FRCNN and substantially higher than SVM and TM. For large-scale plantations, the accuracy improvement is significant. This research provides a method for automatically counting the oil palm trees in large-scale plantations.

Antarctic biological soil crusts surface reflectance patterns from landsat and sentinel-2 images

The remote sensing techniques must be used to obtain long-term information in remote areas, like the Antarctic continent, to monitor the environmental productivity and its changes. The aim of this work was to analyze the surface reflectance profile patterns for the Antarctic biological soil crusts (algae, lichens, and mosses) in an area of Nelson Island (South Shetland Islands, maritime Antarctic), calculated from Landsat and Sentinel-2 images to identify its similarities and differences due to targets, sensors and acquired date. The surface reflectance values for Antarctic biological soil crusts are similar for those observed for biological soil crusts in other Earth extreme environments, like deserts. In Landsat images, the differences among biological soil crusts surface reflectance were identified at visible and near-infrared wavelengths and for Sentinel-2 images, the differences occur at visible, red-edge and shortwave infrared wavelengths, showing the feasibility of using surface reflectance products to identify these different crusts, despite its inherent pixel spectral mixture. Long-term biophysical parameters from such crusts as retrieved from orbital data is not possible due to very low cloud-free images over the Antarctic, which prevents building a consistent surface reflectance time-series which covers all biological soil crusts growth season.

UAVs for Science in Antarctica

Remote sensing is a very powerful tool that has been used to identify, map and monitor Antarctic features and processes for nearly one century. Satellite remote sensing plays the main role for about the last five decades, as it is the only way to provide multitemporal views at continental scale. But the emergence of small consumer-grade unoccupied aerial vehicles (UAVs) over the past two decades has paved the way for data in unprecedented detail. This has been also verified by an increasing noticeable interest in Antarctica by the incorporation of UAVs in the field activities in diversified research topics. This paper presents a comprehensive review about the use of UAVs in scientific activities in Antarctica. It is based on the analysis of 190 scientific publications published in peer-reviewed journals and proceedings of conferences which are organised into six main application topics: Terrestrial, Ice and Snow, Fauna, Technology, Atmosphere and Others. The analysis encompasses a detailed overview of the activities, identifying advantages and difficulties, also evaluating future possibilities and challenges for expanding the use of UAV in the field activities. The relevance of using UAVs to support numerous and diverse scientific activities in Antarctica becomes very clear after analysing this set of scientific publications, as it is revolutionising the remote acquisition of new data with much higher detail, from inaccessible or difficult to access regions, in faster and cheaper ways. Many of the advances can be seen in the terrestrial areas (detailed 3D mapping; vegetation mapping, discrimination and health assessment; periglacial forms characterisation), ice and snow (more detailed topography, depth and features of ice-sheets, glaciers and sea-ice), fauna (counting penguins, seals and flying birds and detailed morphometrics) and in atmosphere studies (more detailed meteorological measurements and air-surface couplings). This review has also shown that despite the low environmental impact of UAV-based surveys, the increasing number of applications and use, may lead to impacts in the most sensitive Antarctic ecosystems. Hence, we call for an internationally coordinated effort to for planning and sharing UAV data in Antarctica, which would reduce environmental impacts, while extending research outcomes.

Applications of unmanned aerial vehicles in Antarctic environmental research

Antarctica plays a fundamental role in the Earth's climate, oceanic circulation and global ecosystem. It is a priority and a scientific challenge to understand its functioning and responses under different scenarios of global warming. However, extreme environmental conditions, seasonality and isolation hampers the efforts to achieve a comprehensive understanding of the physical, biological, chemical and geological processes taking place in Antarctica. Here we present unmanned aerial vehicles (UAVs) as feasible, rapid and accurate tools for environmental and wildlife research in Antarctica. UAV surveys were carried out on Deception Island (South Shetland Islands) using visible, multispectral and thermal sensors, and a water sampling device to develop precise thematic ecological maps, detect anomalous thermal zones, identify and census wildlife, build 3D images of geometrically complex geological formations, and sample dissolved chemicals (< 0.22 µm) waters from inaccessible or protected areas.

Trends in Satellite Earth Observation for Permafrost Related Analyses—A Review

Climate change and associated Arctic amplification cause a degradation of permafrost which in turn has major implications for the environment. The potential turnover of frozen ground from a carbon sink to a carbon source, eroding coastlines, landslides, amplified surface deformation and endangerment of human infrastructure are some of the consequences connected with thawing permafrost. Satellite remote sensing is hereby a powerful tool to identify and monitor these features and processes on a spatially explicit, cheap, operational, long-term basis and up to circum-Arctic scale. By filtering after a selection of relevant keywords, a total of 325 articles from 30 international journals published during the last two decades were analyzed based on study location, spatio-temporal resolution of applied remote sensing data, platform, sensor combination and studied environmental focus for a comprehensive overview of past achievements, current efforts, together with future challenges and opportunities. The temporal development of publication frequency, utilized platforms/sensors and the addressed environmental topic is thereby highlighted. The total number of publications more than doubled since 2015. Distinct geographical study hot spots were revealed, while at the same time large portions of the continuous permafrost zone are still only sparsely covered by satellite remote sensing investigations. Moreover, studies related to Arctic greenhouse gas emissions in the context of permafrost degradation appear heavily underrepresented. New tools (e.g., Google Earth Engine (GEE)), methodologies (e.g., deep learning or data fusion etc.) and satellite data (e.g., the Methane Remote Sensing LiDAR Mission (Merlin) and the Sentinel-fleet) will thereby enable future studies to further investigate the distribution of permafrost, its thermal state and its implications on the environment such as thermokarst features and greenhouse gas emission rates on increasingly larger spatial and temporal scales.

Molecular and isotopic biogeochemistry on recently-formed soils on King George Island (Maritime Antarctica) after glacier retreat upon warming climate

Maritime Antarctica is a climate-sensitive region that has experienced a continuous increase of temperature over the last 50 years. This phenomenon accelerates glacier retreat and promotes the exposure of ice-covered surfaces, triggering physico-chemical alteration of the ground and subsequent soil formation. Here, we studied the biogeochemical composition and evolution extent of soil on three recently exposed peninsulas (Fildes, Barton and Potter) on Southwest (SW) King George Island (KGI). Nine soil samples were analyzed for their lipid biomarkers, stable isotope composition, bulk geochemistry and mineralogy. Their biomarkers profiles were compared to those of local fresh biomass of microbial mats (n = 3) and vegetation (1 moss, 1 grass, and 3 lichens) to assess their contribution to the soil organic matter (SOM). The molecular and isotopic distribution of lipids in the soil samples revealed contributions to the SOM dominated by biogenic sources, mostly vegetal (i.e. odd HMW n-alkanes distributions and generally depleted δ¹³C ratios). Microbial sources were also present to a lesser extent (i.e. even LMW n-alkanes and n-alkanoic acids, heptadecane, 1-alkenes, 9-octadecenoic acid, or iso/anteiso 15: 0 and 17:0 alkanoic acids). Additional contribution from petrogenic sources (bedrock erosion-derived hydrocarbons) was also considered although found to be minor. Results from mineralogy (relative abundance of plagioclases and virtual absence of clay minerals) and bulk geochemistry (low chemical weathering indexes) suggested little chemical alteration of the original geology. This together with the low content of total nitrogen and organic carbon, as well as moderate microbial activity in the soils, confirmed little edaphological development on the recently-exposed KGI surfaces. This study provides molecular and isotopic fingerprints of SOM composition in young Antarctic soils, and contributes to the understanding of soil formation and biogeochemistry in this unexplored region which is currently being affected by thermal destabilization.

Vegetation Abundance and Health Mapping Over Southwestern Antarctica Based on WorldView-2 Data and a Modified Spectral Mixture Analysis

In polar regions, vegetation is especially sensitive to climate dynamics and thus can be used as an indicator of the global and regional environmental change. However, in Antarctica, there is very little information on vegetation distribution and growth status. To fill this gap, we evaluated the ability of both linear and nonlinear spectral mixture analysis (SMA) models, including a group of newly developed modified Nascimento’s models for Antarctic vegetated areas (MNM-AVs), in estimating the abundance of major Antarctic vegetation types, i.e., mosses and lichens. The study was conducted using WorldView-2 satellite data and field measurements over the Fildes Peninsula and its surroundings, which are representative vegetated areas in Antarctica. In MNM-AVs, we introduced secondary scattering components for vegetation and its background to account for the sparsity of vegetation cover and reassigned their coefficients. The new models achieved improved performances, among which MNM-AV3 achieved the lowest error for mosses (lichens) abundance estimation with RMSE = 0.202 (0.213). Compared with MNM-AVs, the linear model performed particularly poor for lichens (RMSE = 0.322), which is in contrast to the case of mosses (RMSE = 0.212), demonstrating that spectral signals of lichens are more prone to mix with their backgrounds. Abundance maps of mosses and lichens, as well as a map of moss health status for the entire study area, were then obtained based on MNM-AV3 with around 80% overall accuracy. Moss areas account for 0.7695 km2 in Fildes and 0.3259 km2 in Ardley Island; unhealthy mosses amounted to 40% (49%) of the area in the summer of 2018 (2019), indicating considerable environmental stress.

A feature-supervised generative adversarial network for environmental monitoring during hazy days

The adverse haze weather condition has brought considerable difficulties in vision-based environmental applications. While, until now, most of the existing environmental monitoring studies are under ordinary conditions, and the studies of complex haze weather conditions have been ignored. Thence, this paper proposes a feature-supervised learning network based on generative adversarial networks (GAN) for environmental monitoring during hazy days. Its main idea is to train the model under the supervision of feature maps from the ground truth. Four key technical contributions are made in the paper. First, pairs of hazy and clean images are used as inputs to supervise the encoding process and obtain high-quality feature maps. Second, the basic GAN formulation is modified by introducing perception loss, style loss, and feature regularization loss to generate better results. Third, multi-scale images are applied as the input to enhance the performance of discriminator. Finally, a hazy remote sensing dataset is created for testing our dehazing method and environmental detection. Extensive experimental results show that the proposed method has achieved better performance than current state-of-the-art methods on both synthetic datasets and real-world remote sensing images.

Detection and Delineation of Sorted Stone Circles in Antarctica

Sorted stone circles are natural surface patterns formed in periglacial environments. Their relation to permafrost conditions make them very helpful for better understanding the past climates where they were formed and have evolved and also for monitoring current underlying processes in case circles are active. These metric scale patterns that occur in clusters of tens to thousands of circular elements, can be more comprehensively characterized if automated methods are used. This paper addresses their identification and delineation through the development and testing of a set of automated approaches, namely, template matching, sliding band filter, and dynamic programming. All of these methods take advantage of the 3D shape of the structures conveyed by digital elevation models (DEM), built from ultra-high resolution imagery captured by unmanned aerial vehicles (UAV) surveys developed in Barton Peninsula, King George Island, Antarctica (62°S). The best detection results achieve scores above 85%, while the delineations are performed with errors as low as 7%.