Spatiotemporal Variability of Anthropogenic Film Pollution in Avacha Gulf near the Kamchatka Peninsula Based on Synthetic-Aperture Radar Imagery

Abstract

The paper addresses the spatiotemporal variability of anthropogenic film pollution (AFP) in Avacha Gulf near the Kamchatka Peninsula based on satellite synthetic-aperture radar (SAR) imagery. Coastal waters of the study area are subject to significant anthropogenic impacts associated with intensive marine traffic, as well as the flow of household and industrial wastewater from factories located on the coast. A quantitative approach to the registration and quantitative analysis of spatiotemporal AFP distributions was applied. This approach is based on the processing of long-term time series of SAR imagery, taking into account inhomogeneous observation coverage and changing hydrometeorological conditions of different regions of water areas in various time periods. In total, 318 cases of AFP were detected in 2014–2023 in Avacha Gulf, covering 332 km² of the total area (~3% of the water area) based on the 1134 processed radar Sentinel-1A/B scenes. The average value of AFP exposure, e, was about 93 ppm, evidencing the high level of AFP in the studied water area (comparable to areas of the Black Sea with intensive marine traffic, for which e was previously determined to be between ~90 and ~130 ppm). An interannual positive trend was revealed, indicating that over the 10-year period under study, the exposure of the waters of Avacha Bay (the most polluted part of Avacha Gulf) to AFP increased ~3-fold. An analysis of AFP spatial distributions and marine traffic maps indicates that this type of activity is a significant source of anthropogenic film pollution in Avacha Gulf (including Avacha Bay). It was shown that the generated quantitative information products using the introduced AFP exposure concept can be interpreted and used, for example, for making management decisions.

1. Introduction

One of the indicators of negative impacts on coastal water areas is anthropogenic film pollution (AFP). The appearance of AFP on the sea surface is caused, first of all, by ship operational or accidental spills (including oil spills) [1,2], surfactants due to wastewater discharges [2,3], polluted river runoff, including pollutants, released from landfills [4], etc. Despite the ability of the marine environment to undergo natural purification and restoration, negative consequences of anthropogenic impact on its ecological condition (including the intake of substances that cause film pollution on the surface) manifest themselves in the systematic violation of water quality standards and lead to severe eutrophication of the marine environment, deterioration of its gas composition, etc. [2,5].

Among the most developed areas of AFP monitoring, satellite, including radar, imagery of marine surfaces are used [3,6], as well as various sea truth sensors [5] for AFP detection and the use of geoinformation systems (GISs) for the comprehensive analysis of pollution [7,8,9]. Also, modeling methods are used to study these anthropogenic impacts on marine water areas [3,10].

The interpretation and identification of AFPs in satellite radar imagery is a difficult task. It should be solved taking into account the physics of the studied phenomena and the used methods of sea surface remote sensing in the radio range of electromagnetic waves (see, e.g., [3,11,12]). A large number of studies dedicated to the detection of film pollution and the determination of its types and sources (including [1,13,14]) indicate the relevance of such studies. However, at present, there is another problem consisting of the insufficient development of methods for obtaining quantitative assessments characterizing the level (intensity) of impact of sea waters by film pollution based on the results of long-term monitoring. This paper mainly addresses this problem.

Such studies often (see, e.g., [7,15,16]) come down to generalized maps of detected AFP and its comparison with a GIS-based situation map. Such an approach allows us to analyze the characteristics and probable causes of AFP generation but does not provide the opportunity for reasonable classification and ranking of the studied water areas according to the pollution intensity. Determining the number of cases [17] or areas [18] of the detected pollution events within the selected regions of the studied water area may not show the real situation of the exposure of water areas to pollution, since hydrometeorological conditions and satellite imagery coverage have significant spatial and temporal variability. This results in spatial and temporal variability in the chances of detecting the pollutants under study. Conducting a quantitative comparison of data obtained for different water areas is particularly difficult.

Hydrometeorological conditions, including near-surface wind speed, current fields, and internal waves, etc., greatly affect successful AFP detection using remote sensing data [19,20,21]. It should be noted that the coverage of satellite images varies due to the peculiarities of the orbital structure and the survey programs of the satellite constellation used. For example, in [22], it was shown that in the coastal areas of the Black Sea, the spatial distribution of the total number of surveys carried out in 2019 by Sentinel-1A/B at surface wind speeds from 2 m/s to 9 m/s (the acceptable range for detecting AFPs [20]) and by Sentinel-2A/B and Landsat-8 satellites in the absence of dense clouds has a complex structure and varies from ~75 to ~450 surveys, depending on the location of the observed pixel. It is obvious that in such conditions a certain turn of events may occur when an intensively polluted region may be observed infrequently, while another slightly polluted region may be observed frequently. In these cases, the number of pollution detections, as well as their areas, will not be adequate criteria for assessing the exposure of these regions to pollution.

The aim of this paper is to provide a quantitative assessment of the spatiotemporal dynamics of marine anthropogenic film pollution in Avacha Gulf (Kamchatka), which has an area of about ~10 thousand km² and is exposed to quite intense pollution. In this research, our approach to the registration and quantitative analysis of spatiotemporal AFP dynamics is based on the processing of long-term satellite data time series, taking into account spatiotemporal variability of observation density.

Coastal waters not only of the study area but also of the entire Far Eastern region of Russia are subject to significant anthropogenic impacts associated with intensive shipping, as well as the flow of household and industrial wastewater from factories located on the coast [23,24,25]. The combination of these anthropogenic impacts causes the risks of deterioration and destruction of marine ecosystems in the region [26,27,28,29].

2. Materials and Methods

2.1. Studied Water Area and Data Used

The Avacha Gulf is located on the southeastern coast of the Kamchatka Peninsula, where this study was carried out. It consists of two parts: open (the gulf itself) and closed (Avacha Bay). The shores of Avacha Gulf are formed by spurs of mountain ranges, between which river valleys that reach the sea are located. The Avacha and Paratunka rivers fall into Avacha Bay, and the Nalycheva River falls into Avacha Gulf. The depths of the Avacha Gulf vary. After elevations of 150–200 m, there is a sharp slope, turning into the valleys of the Pacific Ocean. The climate of Avacha Gulf is extremely unstable; the gulf is open to winds and sea waves. The Kamchatka Current passes along the eastern coast of the Kamchatka Peninsula, which includes Avacha Gulf. The water temperature in winter varies from 1 °C to 3 °C, and in summer months the water warms up to an average of 15 °C [30].

In the coastal waters off the Kamchatka Peninsula, especially in the Avacha Gulf area, where the bay of the same name (Avacha) is located (see Figure 1), various researchers have recorded significant anthropogenic impacts [2,31,32,33]. This bay is being intensively developed and is used in various sectors of human activity [28]. Shipping routes run through Avacha Gulf to the bay (and also directly within the bay); along the coast, there are ports of various purposes and oil storage facilities, as well as ship repair, fish processing, and other plants. In addition, for a long time, household and industrial wastewater from the Petropavlovsk–Yelizovo–Vilyuchinsk city agglomeration, where three quarters of the population of Kamchatka Krai (up to ~300 thousand people) live, has been discharged into Avacha Bay [31].

Therefore, in order to quantify the extent of anthropogenic activities that negatively affect the marine ecosystems in this region, the monitoring of Avacha Gulf waters—including Avacha Bay—is becoming more and more relevant [2,28,31].

In this study, Sentinel-1A/B SAR images obtained for the period 2014-2023 were used to identify and subsequently analyze the spatiotemporal dynamics of the AFP in the waters of Avacha Gulf. The processing level of original radar scenes is SAR GRD [34]. Such vertically co-polarized (VV) C-band data of a 10 m pixel size are currently widely used in tasks related to remote monitoring of various types of sea surface pollution (see, for example [13,18,35]). The NCEP Climate Forecast System Version 2 archive was selected as a source of hydrometeorological data [36]. Using this archive, raster data on surface wind speed were selected, consistent in dates and times of day with the radar surveys. Marine Traffic service was used as a shipping density map for the Avacha Gulf source [37].

2.2. Method

2.2.1. A Brief Review of Approaches to a Quantitative Analysis of Marine Film Pollution Detection Results Based on Remote Sensing Data

Without dwelling on the problem of AFP detection and identification using remote sensing data, which are important and deserve attention but are not included in the scope of this study, we present a brief review of approaches to quantitatively analyze the results obtained from the detection of pollution during long-term monitoring. Here, special attention is paid to studies in which the results of pollution detection were standardized based on the number of observations performed.

There is a study [8] dedicated to the long-term monitoring of oil pollution in European seas in the period from 1999 to 2004. That study suggested an approach to quantitatively assess the density of such pollution based on a relation between the number of such pollutants and the area of processed radar satellite images. Notably, this study takes into account spatial variability in the number of surveys. In a study [38] devoted to the analysis of the spatial distribution of oil spills in the Bohai Sea for the period from 2009 to 2013, the concept of a normalized oil pollution index is proposed, which is calculated as the ratio of the number of oil pollution records in a regular grid cell to the number of satellite images used.

The main shortcomings of these works are that the variability of hydrometeorological conditions was not taken into account, as well as the fact that the number of pollution cases was what was addressed and not their area.

In studies [39,40] focused on the satellite monitoring of oil pollution of the sea surface in the south-eastern part of the Baltic Sea for the period from 2004 to 2015, compilation and quantitative analysis were carried out, taking into account the correction (standardization) for the annual number of surveys. In these works, the results of long-term remote monitoring of sea surface pollution were compiled using geoinformation technologies and statistical methods.

The analysis of the spatiotemporal distribution of film pollution presented in another study [41] involved integrating the vector layers of pollution detected from 2009 to 2015 into a generalized map with the calculated statistics, including the total number and area of detected pollution cases, the area of the largest pollution, etc. A similar approach, which involves generalizing the outlines of anthropogenic pollution and their subsequent analysis, taking into account a priori data on possible sources, was used in the work [9].

In another work [42], which contains the results of the analysis of oil pollution in the Mediterranean Sea for the period from 1970 to 2018, a methodology is proposed for quantitative assessment of the spatiotemporal dynamics of such pollution based on the combination of information from several databases. The methodology involved studying the density of pollution cases for ~20 separate areas of the Mediterranean Sea, corresponding to the special economic zones of different countries. In this case, the number of pollution cases and the area of each zone were considered compared to the total number of pollution cases detected and to the area of the entire Mediterranean Sea, respectively. This made it possible to determine the relative density of pollution cases for each zone and express it as a percentage. In addition, in [42], the entire studied water area was divided into equal sections (cells of a regular spatial grid), and the number of pollution cases in each of them was calculated. However, like most of the cited studies, this work did not take into account the spatial and temporal variability of the number of observations and hydrometeorological conditions.

As follows from the cited works addressing the problem under consideration, when summarizing and analyzing the results of long-term satellite monitoring of sea surface pollution, researchers sought to take into account the variability of the number of observations (affecting the reliability of the resulting final pollution assessments), as well as to relate the resulting statistical assessments with the areas of the studied sections of the water region. In addition, a promising approach seems to be the division of large water regions into equal-area cells based on a regular spatial grid, and obtaining estimates for a set of such cells. As a result, it became possible to classify the examined portions of the water area based on the level of pollution and to trace the dynamics of pollution in each cell.

Taking into account the above, in this work we used an approach that involves measuring the total area of registered pollution of a given type (in this case, film pollution of anthropogenic origin) in cells of a regular spatial grid within the studied bay. In this case, in each individual cell, the area of pollution was related to the area of satellite observations performed under acceptable hydrometeorological conditions (i.e., conditions under which detection of pollution was theoretically possible).

2.2.2. The Approach to Analyze the AFP Spatiotemporal Dynamics

The approach to analyzing AFP spatiotemporal dynamics is based on the calculation and further analysis of annual spatial distributions of a certain value, e_AFP, provisionally named the exposure to AFP. It is determined for each regular spatial grid cell as the ratio of the total area of registered AFP to the total area of performed SAR observations in this cell under acceptable hydrometeorological conditions. The value of e_AFP, determined for each cell, is essentially equivalent to the value of the probability [43] of registering AFP. This is due to the fact that this value is the ratio of the number of registered outcomes of the event of interest (in our case, the fact of the presence of pollution) to the possible number of outcomes of the event (in our case, to the total number of observations in which the registration of the fact of the presence of pollution was physically feasible).

Let us consider a time series of N satellite observations of the same minimum possible area of the sea surface (a pixel with fixed coordinates) over a certain period of time and determine the portion of observations that results in the detection of anthropogenic film pollution.

f = N_AFP/N

(1)

where N_AFP is the number of pollution cases.

On a scale of one pixel, the value, f, corresponds to the concept of exposure to pollution introduced above. Taking this into account, we proceed to an expression that allows us to calculate the exposure to pollution for an area of arbitrary size. In [22], it was shown that when monitoring using several satellite constellations, the exposure e_AFP for an area of arbitrary size can be calculated as follows:

$$e_{AFP}={\displaystyle \frac{\sum _{j=1}^Y\left(s_{pixelSCj}·{\sum }_1^X{\left(n_{afpSCj}\right)}_i\right)}{\sum _{j=1}^Y\left(s_{pixelSCj}·{\sum }_1^X{\left(n_{hmSCj}\right)}_i\right)}}$$

(2)

where Y is the number of remote sensing satellite constellations (SCs) involved (SC means a series of same-type satellites, e.g., Sentinel-1A and Sentinel-1B);

• j is the SC counting number (j = 1, …, Y);
• $s_{pixelSCj}$ is the area of a pixel for jth SC;
• X is the number of surveys for the selected SC for the processed time period;
• i = 1, …, X is the counting number of a survey;
• ${\left(n_{afpSCj}\right)}_i$ is the number of pixels corresponding to AFP registered as a result of interpreting the results of ith survey performed by jth SC;
• ${\left(n_{hmSCj}\right)}_i$ is the number of pixels of the water surface (including polluted ones) for which hydrometeorological conditions during the ith survey allow for the detection of pollution of a given type.

The numerator in Equation (2) corresponds to the total area of pollution observed in a given section of marine waters using all involved SCs.

The denominator in Equation (2) shows the total number of those resolution elements of all analyzed images, expressed in units of area, in which it is theoretically possible to detect pollution of a given type.

This effort involves one satellite constellation (Sentinel-1A/B) and 10 m fixed pixel size imagery (100 m² was the area), allowing for this:

$$j\equiv 1,$$

(3)

$$s_{pixelSCj}\equiv 100 {\mathrm{m}}^2$$

(4)

Considering the persistency of $s_{pixelSCj}$, these terms in the numerator and denominator of Equation (2) can be reduced. Thus, in this study, the $e_{{AFP}_{SENTINEL1}}$ can be calculated as a special case of (2), as follows:

$$e_{{AFP}_{SENTINEL1}}={\displaystyle \frac{{\sum }_1^X{\left(n_{afp}\right)}_i}{{\sum }_1^X{\left(n_{hm}\right)}_i}}$$

(5)

The essence of the proposed approach consists of obtaining the spatiotemporal distribution of $e_{{AFP}_{SENTINEL1}}$, calculated using Equation (5), by sequentially processing satellite SAR images in the cells of a spatial grid in the selected time intervals, followed by aggregation of the results into a systematized spatiotemporal array.

2.3. Processing Algorithm

An integrated flowchart for the data processing algorithm according to the proposed approach is given in Figure 2. This figure illustrates the main stages of SAR and other data acquisition and processing in the interests of studying AFP spatiotemporal dynamics.

The parallelograms in Figure 2 represent data, and the rectangles represent actions. The blue-shaded blocks show the stages of data collection and processing; the yellow-shaded blocks illustrate the stages of systematization of data processing results and calculation of output information products; and the green-shaded blocks correspond to output information products (see Figure 2).

Steps of satellite data collection and processing were performed using the Google Earth Engine (GEE) cloud platform [44] through the activation of program scripts developed during the study.

Data collection and processing steps for each year included the following (see blue blocks in Figure 2):

-

Selecting an array of SAR images (Sentinel-1A/B) for the current year to be processed;

-

Selecting an array of wind data for each SAR image (NCEP);

-

Finding SAR image areas for which the wind speed met the conditions 2 < V < 9 (m/s), under which radar detection of film pollution is possible [1,20], by combining SAR images with NCEP data, as a result of which the corresponding SAR image masks were generated;

-

Interactive AFP detection by the operator–interpreter in each radar image in areas where the condition 2 < V < 9 (m/s) was met, taking into account the interpretation features of AFPs [1,6,12];

-

Calculating AFP areas and perimeters (metadata).

Let us briefly mention the interactive AFP detection process. During the detection, a team of operators–interpreters (3 people) identified film formations of an anthropogenic origin. The operators’ task included detecting slicks on the sea surface, determining the physical mechanism of slick formation (film/non-film), and discarding non-film and biogenic film slicks. An observed object was recognized as AFP if at least two of the operators agreed it was. The exact set of AFP subclasses was not recorded in this effort, since only in some cases could the true genesis of AFP be unambiguously determined (for example, in the case of accidental ship spills). It was assumed that the main share of AFP in the studied water area consists of operational and accidental ship spills and also the consequences of river runoffs and municipal discharges (Section 3.5 of this paper shows the significant role of shipping in the AFP generation in this area). Despite the difficulties, visual interpretation of SAR images, performed by skilled specialists, in combination with the capabilities of the so-called geoinformation approach [7,22], allows for obtaining generally adequate results.

The systematization of data processing results and calculation of output information products were performed using ENVI Classic v.5.2 + IDL v.8.4 software and scripts and included (see yellow blocks in Figure 2) the following:

-

Integrated annual raster map of AFP cases created by summing single-raster AFP masks;

-

Integrated annual raster map of the number of surveys under acceptable wind conditions created by summing raster image masks of SAR areas in which the wind speed met the conditions 2 < V < 9 (m/s);

-

Calculation of AFP exposure spatial distribution using Equation (5) for each processed spatial grid cell (10 m pixel) using raster algebra tools;

-

Generalization of the spatial resolution of AFP exposure distribution by calculating an average value of all 10 m size pixels, included in new pixels with a side size of 3 km, representing cells of a 3 km step regular spatial grid (the grid step was selected upon the need to obtain a visual map illustrating the final information product, taking into account the experience of previous efforts [22,38]).

The results of the generalization made it possible to form a color-coded map of the distribution of the AFP exposure value ($e_{{AFP}_{SENTINEL1}}$) for the current year, convenient for visual perception and further analysis.

The useful output data of the algorithm during the processing cycle of the annual data array were as follows:

-

AFP contours and metadata;

-

Map of the number of SAR surveys (under acceptable wind conditions);

-

AFP exposure ($e_{{AFP}_{\mathrm{S}\mathrm{E}\mathrm{N}\mathrm{T}\mathrm{I}\mathrm{N}\mathrm{E}\mathrm{L}1}}$) map.

Next, data for another year were processed, etc. (see Figure 2). In this way, data for the entire study period (2014–2023) were sequentially processed, resulting in $e_{{AFP}_{\mathrm{S}ENTINEL1}}$ maps for each year.

After processing all data (from 2014 to 2023), we proceeded to the processing and integration of data for the whole monitoring period, including the calculated all-period AFP exposure map. This was accomplished by using not annual but multi-year raster maps of the number of AFP registration cases and the number of surveys under acceptable wind conditions.

After that, the resulting set of maps of the AFP exposure index, $e_{{AFP}_{SENTINEL1}}$, were analyzed using GIS together with maps of marine traffic and data on the location of possible coastal sources of anthropogenic impacts.

In addition, time series of annual area-averaged $e_{{AFP}_{SENTINEL1}}$ values obtained separately for the water area of Avacha Bay and separately for the water area of Avacha Gulf (excluding the bay), as well as for the combined water area (Avacha Gulf, including Avacha Bay) were generated and studied. For better visualization of small values, the e values were converted to parts per million (ppm) by multiplying by 10⁶ (e_ppm = e × 10⁶).

3. Results and Discussion

3.1. Distributions of the Number of SAR Surveys, Examples of AFP Detections, and a Preliminary Integrated AFP Map

Figure 3 shows information products illustrating the distribution of the number of SAR surveys, examples of AFP detections, and an integrated map of all AFPs detected in the period from 2014 to 2023. Figure 3a shows a time series of annual color-coded maps of the number of SAR surveys of the studied water area, carried out at acceptable wind speeds, with the numbers of detected AFPs in Avacha Gulf and Avacha Bay indicated in black and red fonts, respectively.

An analysis of Figure 3a shows that the survey coverage of different regions of the studied water area varies widely in different years (approximately from 10⁰ to 10²). This fact confirms the feasibility of recording the spatial and temporal variability of the survey coverage, implemented in the proposed approach. It is obvious that even with quite frequent pollution of a section of water area that is probed only once a year, the chances of detecting pollution are small. At the same time, no detected pollution on the integrated AFP map in such a region may be mistakenly interpreted as if this region is not exposed to pollution. And, vice versa, several AFPs detected in a region of water area observed, for example, 10² times a year, may lead to the misconception that this section is the one most exposed to anthropogenic impacts. Figure 3c shows an integrated map of AFPs detected over the entire monitoring period (2014–2023).

Figure 3b shows some typical examples of AFPs detected in SAR images in Avacha Gulf and in Avacha Bay (the blue-framed images are mainly ship spills, and the brown-framed images are AFPs mainly related to coastal anthropogenic impacts).

Integration of the obtained data collection and processing results shows that, based on the results of processing 1134 SAR images (Sentinel-1A/B), 318 AFPs were identified within the test waters in 2014–2023 (see the number of identified AFPs over years, shown in Figure 3a in black and red for Avacha Bay and Avacha Gulf, respectively). The total area of AFPs registered over the entire observation period, calculated using GIS QGIS v.3.28, was 332 km², which exceeds the area of Avacha Bay by more than 1.5-fold.

3.2. Spatial Distribution of the AFPs over the Entire Monitoring Period

Figure 4a,b show the raster maps required to calculate an integrated map of the studied water area exposure to AFP for the entire monitoring period, namely the number of AFP cases in a distribution map (see Figure 4a) and the number of surveys at acceptable wind speeds (see Figure 4b), as well as an integrated map of the studied water area exposure to AFP (see Figure 4c).

An analysis of Figure 4 allowed us to find out the following.

The map of the number of AFP cases distribution for the entire monitoring period in Avacha Gulf (including Avacha Bay), shown in Figure 4a, demonstrated the almost ever-present AFP in the waters under study. The largest number of AFPs was registered near the eastern and northern coasts of Avacha Bay, as well as in the center of the bay, where the number of registered AFPs in some areas reached 7.

The map of a number of surveys under acceptable wind conditions (2 < V < 9 m/s) in Figure 4 demonstrates significant spatial variability of survey coverage for the studied water areas over the entire monitoring period. Four main regions can be distinguished in this map (see Figure 4b), namely the following:

-

A red area in the northwest of the studied water area of Avacha Gulf and in Avacha Bay, corresponding to areas with the largest number of surveys (varied from 476 to 540);

-

Two yellow–orange areas in the northeast and southwest of Avacha Gulf, corresponding to areas with an average number of surveys (varied from 274 to 368);

-

A green–blue area in the east of the studied water area, within which relatively few surveys were carried out (162 or less, down to 10).

Four such zones with quite different numbers of surveys under acceptable wind conditions are due to the peculiarities of the Sentinel-1A/B acquisition program, focused mainly on coastal waters. The variability in the number of surveys within each zone (visible in Figure 4b as large ~10 km cells, corresponding to large pixels of NCEP information products) is due to the variability of wind conditions. It can be noted that within the vast “yellow-orange zones” the number of surveys under acceptable wind conditions gradually decreased, moving away from the coast, which is explained by stronger winds in the open ocean.

An integrated map of the studied water areas’ AFP exposure for the entire monitoring period, shown in Figure 4c, obtained with a 3 km cell size using the proposed approach, demonstrates the following.

A significant share of all the studied water areas were subject to intense AFPs; the e values reached level of 4000 ppm and more in many cells both in the Avacha Bay area and in the western, southern, and central parts of Avacha Gulf. The highest values of exposure to AFP were recorded in Avacha Bay, where they could reach a level of more than 16,000 ppm.

It should be noted that, unlike Figure 4a, which represents a typical information product of AFP monitoring derived from satellite SAR images, Figure 4c allows for a formal classification of sea surface areas by the level of AFP exposure. In addition, as already mentioned above, such an information product minimized distortions in understanding the true spatiotemporal AFP distribution, inherent in typical AFP maps (for example, as in Figure 4a or Figure 3c), associated with the variability in the number of observations.

3.3. Interannual Dynamics of the AFPs

Figure 5 shows graphs of the interannual dynamics of area-averaged values of AFP exposure (e) for the period from 2014 to 2023 and annual maps of AFP exposure for the entire water area (left) and the bay (right).

As follows from the analysis of Figure 3a, the number of radar observations and detections of AFPs in 2014–2016 was too small to obtain representative annual estimates. Therefore, to analyze the interannual AFP dynamics, the data for these three years were combined and processed together. The data obtained in the period from 2017 to 2023 were processed separately for each year.

The interannual dynamics of the area-averaged AFP exposure was analyzed for different regions (see the graph inserts in Figure 5). The graph for Avacha Bay is shown in red, for Avacha Gulf it is in blue, and for all the studied water areas (Avacha Gulf and Avacha Bay) it is in green.

The analysis of Figure 5 allows us to conclude that the obtained spatial distributions of the AFP exposure index for the studied water areas can be characterized as complex, heterogeneous, and undergoing significant temporal changes from year to year (see inserts in Figure 5).

Analysis of the red graph in Figure 5 shows that the highest average values of the e index were recorded in the waters of Avacha Bay, which changed significantly throughout the monitoring period, reaching its maximum (up to 2032 ppm) in 2021. At the same time, in the rest of the waters of Avacha Gulf, the average values of the e index for the entire monitoring period did not exceed 124 ppm (see the numerical values of the graph nodes in Figure 5).

The analysis of Figure 5 allows us to reveal the following.

(1) In Avacha Bay from 2014 to 2023, the interannual dynamics of e were characterized by minima that appeared in 2014-2016, 2019, and 2022 (67, 455, and 444 ppm, respectively), as well as maxima in 2018, 2020, and 2021 (920, 1907, and 2031 ppm, respectively).

The maximum values of AFP exposure in Avacha Bay, identified in 2020 and 2021, could have been caused by emergency discharges of industrial and household wastewater into the sea, as well as ship spills. In 2020, the largest AFP in terms of area was registered in the bay (the GIS-based calculated area was ~11 km²), and a large accumulation of AFP was recorded (32 pcs., see Figure 3a). In 2021, the largest number of AFPs was recorded in Avacha Bay—39 cases (see Figure 3a)—for the entire monitoring period. The inserts shown in the upper right of Figure 5 testify to such a feature. In these inserts, clusters of dark red cells indicating the maximum levels of AFP exposure can be seen. The increase in e values in 2021 could have probably been influenced, among other things, by the recovery of shipping activity, which was reduced during the COVID-19 pandemic.

The interannual positive trend was revealed, indicating that over 10 years the averaged exposure of Avacha Bay to AFP increased by ~3.5 times (approximately from 400 to 1400 ppm; see the dashed red line in Figure 5).

(2) The interannual dynamics of e in Avacha Gulf (excluding Avacha Bay) from 2014 to 2023 was characterized by minima in 2017, 2019, and 2022 (58, 87, and 37 ppm, respectively), as well as maxima in 2018 and 2020 (135 and 188 ppm, respectively).

The cause of the identified maxima, as in the case of Avacha Gulf, could be ship spills (they probably prevail here) and discharges of wastewater, which, due to the vast area of the gulf, do not have such a strong effect on the level of e values. No significant interannual trend was identified (see the blue dashed line in Figure 5).

(3) The interannual dynamics of e for the entire studied water area are largely determined by the AFPs of the relatively small Avacha Bay. The average value of the AFP exposure for the entire water area under study was 93 ppm, which indicates a high level of pollution by anthropogenic films (comparable to areas of the Black Sea with intensive shipping, for which e from ~90 to ~130 ppm was previously determined [22]).

It should be noted that in the previous study, which was carried out in the Black Sea using three independent cases in order to calculate e, the absolute error in estimating e using SAR data was on the order of ~10 ppm [22]. In this study, a similar technology for processing SAR data was used over the water area with nearly the same dimensions, so we believe that the accuracy of determining e for Avacha Gulf was approximately at this order.

Avacha Bay is the most navigable and isolated from the ocean part of the studied water area. There are many sources of anthropogenic impacts on its shore. It is subject to AFP approximately ~14 times more than the open part of Avacha Gulf (the average value of the e for the open part of the bay was 63 ppm, and for Avacha Bay it was 894 ppm).

3.4. Preliminary Physical Interpretation of the Quantitative Results Obtained Using the Proposed Approach

Let us consider the physical interpretation of the quantitative results obtained using the proposed approach.

For example, the value e ~93 ppm obtained for the entire studied water area can be interpreted as follows: at an arbitrary point in the studied water area at an arbitrary time, the probability of the AFP presence is 93 ppm = 93 × 10⁻⁶~10⁻⁴. Taking into account the area of the Avacha Gulf (10⁴ km² including Avacha Bay), such a probability is equivalent to the constant (at each time moment) presence of ~1 km² AFP in Avacha Bay. Using such a result, it is possible to estimate the volume of incoming pollutants, taking into account the approximate average thickness and lifetime of films. Thickness and lifetime depend on the nature of the pollutants, needing additional investigation such as water sampling [1,2]. As a result of further development of such studies, the obtained data can be converted into quantitative estimations of pollutant concentrations and used for environmental impact assessments and policymaking.

In the above example, an integral assessment was obtained that characterizes the general condition of the water area in terms of AFP exposure. Depending on the specific region (cell) of the water area, the level of AFP exposure varies significantly (see Figure 4 and Figure 5).

Let us consider another example concerning the situation in local regions of the water area. Let us assume that in a separate cell (with an area of 3 × 3 = 9 km²) the value of e was 10,000 ppm. In reality, in Avacha Bay the values of e reached even higher values, up to ~16,000 ppm, see Figure 4c and Figure 5, top right. The exposure to AFP e ~10,000 ppm = 0.01 in a cell means that at an arbitrary point in the cell of the studied water area at an arbitrary moment in time the probability of AFP is 0.01, which can be interpreted as the presence of permanent pollution constituting 1% of the cell area (i.e., permanent pollution with an area of 0.09 km²).

Let us conduct a thought experiment to assess the danger of such a situation from an ecological point of view. Let us assume that this cell covers a section of water adjacent to the shore with an area of 3 × 3 km², which is a recreational area near the beach, within which a section with an area of 0.09 km² (i.e., more than ~10 football fields) is constantly polluted. Obviously, such a level of pollution is a serious environmental threat.

Thus, it was shown that the generated quantitative information products using the introduced AFP exposure concept can be interpreted and used, for example, for making management decisions.

3.5. Relationship of AFP with Possible Sources of Anthropogenic Impacts

Figure 6 shows a comparison of the spatial distribution of the AFP exposure of the studied water area over the years from 2014 to 2023, with data on marine traffic, which is considered to be one of the main potential AFP sources.

The map shown in Figure 6a shows a typical annual distribution of shipping density in Avacha Gulf (the case of 2022) with conditional color gradation indicating zones with relatively low, medium, and high density. The maximum level of shipping density in this water area corresponds to a value of ~20 thousand routes per km² per year [37].

Figure 6b–d show subsets of the spatial grid cells (with color-coded e values), coinciding with low-, medium-, and high-density shipping zones, respectively, as well as these zones themselves, shown in different colors.

For each zone with varying level of shipping density, the percentage of area covered by cells with a non-zero e values was computed (see

Table 1. The percentage of the area covered by cells with values e > 0 in zones with varying levels of shipping density.

	Low Shipping Density	Medium Hipping Density	High Shipping Density
The percentage of the zone area covered by cells with non-zero AFP exposure (%)	25	28	47

).

An analysis of Figure 6 and Table 1 confirms the assumption that the observed AFPs are largely due to intensive shipping. This is demonstrated quantitatively by the fact that 47% of the zone with high levels of shipping density is polluted, whereas this indicator varies between 25% and 28% in the zones with low and medium levels of shipping density, respectively. In addition, an analysis of Figure 6b shows that in the zone with high levels of shipping density, the e values reach higher values more often than in other zones, which is especially clearly visible within Avacha Bay. This indicates that intensive shipping is a significant source of AFP in the region under study.

The demonstrated experimental joint analysis of spatial distributions of AFP exposure and shipping density shows that the proposed approach made it possible to obtain a quantitative information product that soundly confirmed the assumed significant contribution of ship traffic in the generation of AFPs within the waters of Avacha Bay near the Kamchatka Peninsula.

It should be noted that in addition to shipping, numerous coastal sources, such as ports of various purposes and oil storage facilities, as well as ship repair, fish processing, and other plants, also exert a significant anthropogenic load on the studied water area. The evidence of activities of these sources in the form of AFPs is clearly visible in some SAR images, including those shown in Figure 3c (see examples of SAR images for 20 October 2021 and 4 April 2020, where AFPs are directly adjacent to the coast).

4. Conclusions

One of the indicators of negative impacts on coastal water areas is anthropogenic film pollution (AFP). The appearance of AFPs on the sea surface is caused, first of all, by ship operational or accidental spills (including oil spills). A large number of works are dedicated to the problems of detection of AFPs and determination of their types and sources by using satellite data. However, at present, there is another problem consisting of the insufficient development of methods for obtaining quantitative assessments characterizing the level (intensity) of impact of sea waters by AFPs based on the results of long-term satellite monitoring.

This study offered an approach to the registration and quantitative analysis of spatiotemporal AFP distributions, based on the processing of long time series of SAR images. This approach takes into account inhomogeneous observation coverage, as well as the spatiotemporal variability of hydrometeorological conditions. Taking these factors into account allows us to come closer to obtaining representative estimations of the probability of AFP occurrence in marine waters based on the processing and analysis of satellite SAR monitoring data. The approach to analyzing the AFP spatiotemporal dynamics is based on the calculation and further analysis of annual spatial distributions of a certain value, e_AFP (exposure to AFP). It is determined for each regular spatial grid cell as the ratio of the total area of registered AFP to the total area of performed SAR observations in this cell under acceptable hydrometeorological conditions.

The proposed approach allowed us to perform a quantitative assessment of spatiotemporal dynamics of marine anthropogenic film pollution in Avacha Gulf (Kamchatka), with an area of about ~10 thousand km², exposed to quite intense pollution. There were 318 cases of AFP detected in 2014–2023 in Avacha Gulf, covering 332 km² in the total area (~3% of the water area) based on the 1134 processed radar Sentinel-1A/B scenes. The average value of AFP exposure, e, is about 93 ppm, which evidences the high level of AFP in the studied water area. An interannual positive trend was revealed, indicating that over the 10-year period under study, exposure of the waters of Avacha Bay (the most polluted part of Avacha Gulf) to AFP increased ~3-fold. An analysis of AFP spatial distributions and marine traffic maps indicates that this type of activity is a significant source of anthropogenic film pollution in Avacha Gulf (including Avacha Bay).

It was shown that the generated quantitative information products using the introduced AFP exposure concept can be interpreted and used, for example, for making management decisions.