Selection of test areas

In order to verify and validate the proposed methodology, a selection of test areas was first made through visual interpretation and initial revision of satellite images from the Copernicus program. Due to the frequent and dense cloud cover, the selected images shall be of different days to maximise the size of the test area. The test sites should include various objects in different states, which should be located both on the ground and in the sea. In this way, certain territories can be better compared and the indices and classifications made would be more precise. For each image, the so-called area of interest (AOI) has been selected precisely. Each of these areas should be the same for optical and radar images for matching images in subsequent processing or reuse of the information. The selection of the individual studied objects was performed during the field campaign on 12/12/2022 and 24/12/2022, during which Ground Control points of each individual object were generated.

Ground Control points (GCP)

For the purpose of the study, the following Ground Control Point (GCP) locations were generated and applied: 25 GCP, generated on 12/12/2022, and 6 GCP – on 4/12/2022. The points were generated via OruxMap in offline mode and published for Open access on a platform (catalogue) such as open data at www.Destination-earth-antarctica.eu. The generated point locations are freely available on a catalogue page, aiming freely publishing data from the Antarctic polar studies in line with the DCAT-AP data standard³⁴.

Spectrometer data

The field spectral surveys were performed using a spectrometer in the visible range (Sekonic Spectrometer C-800), a full-spectrum color meter that can accurately measure any type of light³⁵. The data from the spectrometer is with the following parameters³⁶: Color temperature, CCT (Correlated Color Temperature), Δuv (Deviation) between the correlated color temperature and the place of emission of the black body, Hue (in degrees), Sat – Saturation is the index to express the intensity or chroma. Using these data, a study of the color characteristics of the objects was conducted. For each spectral reflectance characteristic (SRC), the color characteristics (color coordinates (x,y) and dominant wavelength) in the spectral range 380-780 nm were calculated relative to CIE 1931 and a standard source of electromagnetic radiation³⁶.

Satellite data

The satellite data used in the study are from the Sentinel-1-A and Sentinel-2-B satellites of the European Space Agency (ESA). Sentinel-1-A is Synthetic Aperture Radar (SAR)³⁷ sensor, and Sentinel-2B is a Multi-Spectral Instrument (MSI)^32,33 recording data in the optical range with different spectral resolutions. Data by dates and satellites used are represented in Table 1.

Table 1.

Satellite images acquisition dates

*

Ground Sample Distance (GSD)

**

Pixel Spacing Resolution (rg x az)

Automatic recording weather station (AWG) data

Another type of data, which at this stage is part of the catalogue, but still is not involved in the analysis, was derived from the Automatic recording weather station (AWG), presented in detail in Table 2.The data were generated from field surveys on Livingstone Island in December 2022 and provided by Spasova and Avetisyan, Department of Aerospace Information, Space Research and Technology Institute, Bulgarian Academy of Sciences and by Popov and Shirov, OXYMET³⁸.

Table 2.

Raw Data from Automatic recording weather station (AWG), an Еcological Magnesium-Air battery (11/12/2022)

Drone data

The last data source is the DJI Mavic 2 drone. The drone photography was used to correctly verify the environment, in which the experiment was made and for the synchronized remote sensing monitoring approach. The Drone images are from the days of field measurements (12/12/2022 and 12/24/2022) (Figure 3, 4).

Figure 3

Hanna Point – 12/12/2022, Drone data; Personal archive, Autor: Temenuzhka Spasova

Figure 4

Hanna Point – 24/12/2022, Drone data; Personal archive, Autor: Temenuzhka Spasova

Description of the synchronized experiment

During the experiment, field measurements of the spectral reflectance characteristics (SRC) of the studied objects and drone photography were performed. The timing of the experiment was consistent with the passage of Sentinel-1 and Sentinel-2 satellites over the study area. Due to the distance of Hannah Point from the Bulgarian Antarctic Base (BAB) and availability of logistical support, the field surveys often do not coincide with the optimal weather conditions for satellite remote sensing with optical data.The generation of SRC of the studied objects and the drone photography were performed on 12/12/2022 and 24/12/2022. Unfortunately, the weather conditions were not suitable for satellite remote sensing through optical imagery and the Sentinel-2 images from these dates are with clouds on them. Therefore, we used the Sentinel-2 image acquired on the closest date before the day with ground-based research (7/12/2022). We assume that there weren’t significant changes in the onshore sites during these 4-5 days. The reference date for the verification of the dynamics was 16/01/2023.The surveys were carried out within the same time range, which defines them as quasi-synchronous for the optical data. As regards the Synthetic Aperture Radar (SAR) images from the Sentinel-1 sensor, there is sufficient representative data and dates close to the dates with field surveys, such as 11/12/2022, 23/12/2023, 29/11/202022, and 16/01/2023.

Spectrometric measurements and data interpretation

The research aimed to create a methodology for a complex synchronous approach and interoperability in remote sensing in polar latitudes. Through spectrometric measurements of various objects on Hannah Point, Livingston Island, Antarctica, a database of SRC (spectral library) was created. This database was used and is available for the calibration of satellite sensors for remote sensing of various natural objects, for environmental protection, and for the needs of comparative analysis of objects in polar conditions and latitudes. The database will be upgraded and enriched with additional data after future studies.

Colour is a major factor in Remote sensing and a leading characteristic for each of the objects studied.The analysis of the SRC includes a study of the color characteristics of the objects. For each SRC of a given object, color characteristics (color coordinates (x, y) and dominant wavelength in the spectral range 380-780 nm relative to CIE 1931 and a standard source of electromagnetic radiation D₆₅) were obtained³⁶. The color coordinates of rocks, soils, and vegetation can be successfully used for object recognition by spectral data. According to the methodology proposed, only the selected points from the study area were characterized. The final selection of the objects was facilitated by drone photography. In-situ (field) spectrometric measurements of the SRC of the selected objects in their natural conditions were performed. A GPS control point was taken for each of the profiles using OruxMaps. This product works in offline conditions, but it generates enough GPS data and statistics successfully.

The next step of the methodology was the selection of algorithms for the interpretation of the measured spectral data for the recognition of the type of studied objects. A comparative analysis of the obtained library of SRC with reference spectral libraries for the studied objects and those obtained through the Sentinel 2B MSI sensor was made.

Remote sensing uses a wide range of the electromagnetic spectrum - from ultraviolet to radio waves, but the visible, infrared, and microwave regions of the Electromagnetic spectrum (EMS) have the most applications and importance³⁹.

Using the spectrometer, measurements in the visible and a narrow part of the near-infrared range (0.38–0.78) were made. The average height (h) of the spectral profiles taken is ~1m. In the same spectral range, there is also a measurement from the Sentinel 2 sensor (from band 1-0.44 to band 7-0.783). The in-situ measurements were used for validation of the satellite-based optical data.

In the visible range, primary colors with the following wavelengths are distinguished: Violet 0.380–0.440; Blue 0.440–0.490; Green 0.490–0.565; Yellow 0.565–0.595; Orange 0.595–0.620; Red 0.620–0.760 ^16,39.

Another EMS range that has been used in remote sensing for polar latitudes is near- and mid-infrared and in this case 0.7–2.19 μm. In the present study, we analysed the SRC of the objects from 0.70 to 0.78 μm using both the data obtained by a spectrometer and by Sentinel-2 satellite sensor, which is a prerequisite for data comparison and validation.

Each of the spectral ranges has its advantages and disadvantages, which determine the specifics of the measurement technology and the type of tasks to be solved ³⁹. Due to a number of advantages, the most widely used spectral ranges for polar studies of objects on the Earth’s surface are the visible and near-infrared spectral ranges between 0.4 and 1.3 μm. Of course, there are also disadvantages to using satellite remote sensing for the polar latitudes and they are related to the lighting conditions and the state of the atmosphere. Therefore, in our experiment, we used only satellite data acquired during the summer season.

Classification of satellite optical and radar data

The measured reflectance of satellite sensors depends on the local characteristics of the Earth’s surface, which need to be detected to retrieve information from image data. Theoretically, a single spectral band of a remote sensing image must be sufficient to classify objects, but the classification of complex multispectral data produces much better results². Using imaging software, satellite images from different periods, over the same territory of interest, and for each of the profiles studied were processed, classified, and analysed. To study the changes occurring in the individual objects, numerous indices were used. They were based on optical and radar data. The indices for classification optical data are presented with the formulas for their calculation in Table 3.

Table 3

Calculation formulas and description of the optical indices used in the study

In present study, we introduced the Disturbance Index (DI) for monitoring objects with SRC in the environment of Livingston Island. The index is primarily applied for assessment vegetation disturbances and it has shown to be more accurate, compared to the standard spectral indices²⁹. The DI is based on the linear orthogonal transformation of multispectral satellite images - Tasseled Cap Transformation (TCT)^48,49 and uses unique transformation matrices for each sensor. In present study, we used TCT matrix proposed by⁵⁰ involving all spectral bands of the Sentinel-2 sensor. Utilization of all spectral information increases significantly the degree of identification of the three main components of the Earth’s surface - soil, vegetation and moisture⁵¹. The outputs from TCT are multi-band images containing three layers – Wetness (TCW), Brightness (TCB), and Greenness (TCG). After their generation, decomposition on each of them is being applied. The next step is the calculation of the normalized values of the Tasseled cap components, the values of mean and standard deviations. After the normalization, the three component indices were combined linearly to acquire the DI²⁸.

Another index based on the TCT is the NDGI. Before its final calculation, it is needed the variables of the index to be calculated. The formula for calculation of the variables is as follows:

,where GR_n(t₁) and GR_n(t₂) are the normalized values of the Greenness component at time points t₁ and t₂; |GR_n(t₁)| and |GR_n(t₂)| are the absolute values of the same component; E{GR(t)} is the mean of GR(t).

And last but not least, optical indices for snow and ice were used in the study. To study the snow cover dynamics, the Normalized Difference Snow Index is calculated and classified, which is extremely accurate in areas with a constant presence of snow and ice (Table 3).

For calculation of the radar coefficients in the complex synchronous approach, the SR between two SAR images from different time points was used. The coefficient is a dimensionless quantity, and the formula is as follows:

,where σ is the microwave reflectance coefficient for time instant t1 (satellite image of earlier date) and t₂ for time instant (satellite image of later date), i, j are the row and column numbers of pixels in a given SAR image, and r is the relative soil moisture content reported for the corresponding period between two images of one polarization. This mathematical approach was used between four images over a period of about a month, showing promising results in identifying areas of moderate and high change in vegetation, wet snow, ice, and water¹³. The values above 2 register serious changes in the reflectance of a given object in selected areas, and low values below 1.0 indicate little or no change in an object’s reflectance. A threshold value of 2.0 was used to discriminate large changes in reflectance or the presence of a new object¹⁷.

Automatic recording weather station (AWG)

An ecological Air-Magnesium battery was tested for the needs of meteorological independent information in the area. It was used in the Automatic recording weather station (AWG), which is an innovative instrument for measuring and recording ambient temperature, illumination, UV radiation, GPS coordinates, barometric pressure, and altitude.

The battery is activated with seawater and can operate for several weeks, with the measured data recorded on a SD memory card. The device is built on the Arduino hardware platform. The equipment is intended for operation at the Bulgarian Antarctic station on Livingston Island, but due to the similar environmental conditions, the data obtained are representative of the entire island ¹¹(Figure 5).

Figure 5

Innovative Automatic recording weather station (AWG)³⁸, Livingston Island. Autor: Temenuzhka Spasova, 2022

Dynamics and territorial distribution of the optical indices values

The spectral indices showed a weaker ability to differentiate the individual objects on the Earth’s surface in early December compared to mid-January. The only exception is DI, where the opposite trend was observed (Figure 12d). This is due to the similar SRH of the individual objects, which results in similar values of the spectral indices applied. As the summer progresses and the state change of the studied objects intensifies, their SRCs become more distinct, which also affects the indices values. This is particularly evident in the transition of ice and snow to water and the growth of areas occupied by different types of vegetation. DI and NDWI showed to be superior, amongst the vegetation indices, in the assessment of ice and snow covers dynamics (Figure 12d,12e). DI has not been used so far to evaluate disturbances in snowy/icy areas. From the results obtained, it can be seen that it is distinguished by higher values at the beginning of summer and stabilizes as the season progresses (Figure 12d). MSAVI2 showed to be the most suitable among the vegetation indices applied in the present study to assess the dynamics of terrestrial vegetation. The maximum MSAVI2 values increased by 0.31 in the period between 7/12/2022 and 16/01/2023, and the number of pre-defined indices classes increased from four to seven (Figure12b). In comparison, the NDVI maximum values increased by 0.26, and the pre-defined classes by two numbers (from three to five) (Figure 12a). MCARI2 has accounted for both terrestrial and aquatic vegetation dynamics (Figure 12c).

Figure 12

Spatial distribution of the optical indices values throughout the study are

For the snow optical index, there was an increasing trend of the areas occupied by snow and wet snow, which is due to the transition period in which its values vary from 0 to 1 (one). Values above 0.80 could be considered when investigating older snow cover and ice. There is no ice presence in the considered study area because the values are below 0.75 (Figure 12f).

Optical indices values in the locations of sample point objects

The analysis of the Sentinel-2 –based spectral indices calculated for the individual sample point objects showed the highest values for P61 (Moss) for the most of indices (MCARI2, MSAVI2, NDVI, NDWI, MSI) (Table 4). DI had the highest value in P45, which is also a moss object (Table 4). Only NDGI, which accounts for the change in the greenness between 07/12/2022 and 16/01/2023 showed a greater diversity of the objects with the highest NDGI value = 1. The objects with NDGI=1 are P5 (rock), P67 (Wet sand), P2 (Mud), P6 (Soil on scree), P52 (Wet snow), and P50 (Muddy water) (Table 5, 6). These results show that new vegetation has developed in all these places.

Table 4

Optical indices values for mosses, lichens, and herbaceous vegetation

Table 5

Optical indices values for water, wet snow, and glacier

Table 6

Optical indices values for sand, soil, scree, and rock

Dynamics and territorial distribution of the radar coefficients throughout the study area

Composite radar images from one, two, or more dates were used in pseudocolors to detect wet snow, ice, water, vegetation, and rocks. The results clearly showed that locations of water, wet snow, ice and its melting can be mapped using SAR data. The same applies to vegetation such as mosses and lichens, but with a more detailed classification of radar indices. Based on the spatial distributions of the SR, it can be assumed that the data are suitable for tracking the dynamics of objects even in polar latitudes. SAR images in dB (decibels) are suitable for detecting water, wet snow and ice, as precise levels of their values can be determined. It is more complicated to use them for snow with less water content, but such objects have been successfully observed by optical indices developed for snow and ice monitoring such as NDSI. Promising results for snow and ice monitoring showed also the DI and the NDWI. However, more experiments are needed to confirm their potential in monitoring such objects.

After processing and creating composite images, the pixel size at SAR is 12.05/16.35 m and is preserved. This is a good enough resolution to follow the general dynamics of objects, and by transforming the SAR images into decibels, an even greater logarithmic accuracy was achieved where there are already proven threshold values of water, snow, and ice. The values in the raneg 22-24 indicates wet snow, those in the range 25-27 indicates snow, and above 27 are the values representative for ice (Figure 13)^3,4,.

Figure 13

SAR images of study area (HH polarization) in decibels (dB) values

Within the framework of the conducted research, a trend is clearly observed in the change of the land cover state and in the areas occupied by ice and snow, which is also characteristic of the seasonal dynamics.

SR values showing the change between two images close to 2.0 and above 2.0 (Figure 14) indicate that a change has occurred through the appearance or absence of a new object. Figure 14 presents the changes that occurred in the period between 29/11/2022 and 23/12/2022 (the latter date is from the synchronous experiment). Values are high in the coastal area and for water bodies, which is typical for the season and ice soup movement. For the second period of observation (11/12/2022 – 16/01/2023), a change is observed, which, according to the values of 1.7, indicates moderate changes.

Figure 14

Spatial distribution of SAR index values throughout the study area

Automatic recording weather station (AWG)

Test data of the eco-friendly battery has shown that it works successfully in polar conditions. As an ecological source of energy, it works from 11/12/2022 until the base closes at the end of the season for the 31st Antarctic Expedition in early March. It is yet to be reinstalled and analyzed, but at this first stage, it has shown that it can run on salt water, which is a renewable energy source. The battery is an absolutely harmless source of energy, as it contains Magnesium. The data from the AWG established the operability of the ecological magnesia-air electrochemical system for real-world operation in Antarctica.

Data generated by AWG are still being processed and analysed at this stage. The results will be published in future articles. However, energy was produced in three battery cells, which shows that this system could be successfully applied under polar conditions or in the marine environment. The energy is generated in W (Watt). In one of the cells, the voltage throughout the test reaches values above 3. The geographical coordinates and temperatures in F (Fahrenheit) table were correctly reported (Table 2).