The Long-Term Annual Datasets for Azov Sea Basin Ecosystems for 1925–2024 and Russian Sturgeon Occurrences in 2000–2024

Abstract

The abundance of the Russian sturgeon population in the Sea of Azov declined many times in the XX–XXI centuries. This paper presents long-term annual and spatial occurrence datasets to create statistical and machine learning models to better understand the distribution patterns as well as biological and ecological features. The annual dataset provides annually averaged results of environmental and biotic population estimates obtained by in situ observations for 1925–2024. The spatial occurrence dataset contains raw survey observations with a bottom trawl over the period of 2000–2024. Preliminary diagnostics of the annual dataset reveal no evidence of non-stationarity or significant outliers that cannot be explained by biological parameters. The published dataset allows any researcher to perform statistical and machine learning-based analysis in order to compare and describe the population abundance or spatial occurrence of Russian sturgeon in the Sea of Azov.

Dataset: https://zenodo.org/records/15099528.

Dataset License: license under which the dataset is made available (CC-BY-SA 4.0).

1. Summary

Sturgeon fish species (Acipenseridae) are among the most ancient aquatic life forms that are currently living. However, as a result of human activity (mainly due to fisheries and dam construction), the abundance of these fish species has significantly decreased, but they are rarely almost extinct [1,2]. In many river and sea basins, the abundance of sturgeons is maintained by artificial reproduction in hatcheries.

In the Sea of Azov, rapid population collapse has happened many times. Using archeological and historical materials as a reference, the impact of anthropogenic factors on sturgeon populations may be chronologically divided into four stages. The first longest stage was associated with the development of agriculture (cutting down of forests along watersheds and floodplains, plowing of large land areas, shallowing rivers, and gradual reduction in spawning areas) [3]. The second stage began at the end of the XIX century, and it was associated with the development of industry and the intensification of fishing [4,5]. During this period, the reproduction of sturgeon populations still occurred only due to natural spawning. The third stage began after the regulation of the river’s flow by damming of the main spawning rivers (starting when the Tsimlyansk Hydroelectric Complex was built on the Don River in 1952), which led to a sharp reduction in the natural reproduction of sturgeons and other species [6,7]. The fourth stage in the late 1980s–early 1990s was associated with the high pressure of overfishing and IUU (Illegal, Unregulated, and Unreported) fishery in a difficult socio-economic situation, shortly before and after the collapse of the Soviet Union [8]. Up until now, no sturgeon populations in the Sea of Azov have been able to recover their abundance and commercial status.

For these purposes, it is important to study the sturgeon distribution, biological aspects, and the influence of the environment and human activity. To achieve this goal, long-term historical datasets for future statistical analysis or modeling (including a “big data” approach within machine learning techniques) should be published. This paper presents and describes long-term annual data collected by the Azov-Black Sea branch of VNIRO and detailed in situ survey data.

The dataset consists of two parts: The first table is a long-term annual (yearly averaged) post-processed water ecosystem dataset for 1925–2024. The second one is a detailed in situ spatial occurrence of sturgeon species and environmental parameter dataset in the Sea of Azov for 2000–2024.

Some parts of this dataset have been previously published as graphs or tables in other studies, mostly in Russian, namely, stock assessments of sturgeons [8], stock assessments of other demersal species like sea roaches and gobies [8,9], a summary of hydrological data [10,11], and zooplankton and zoobenthos estimates [8,12].

2. Data Description

2.1. Descriptor Table

2.1.1. Long-Term Annual Dataset

The long-term annual dataset table (

Table 1. Column descriptions of the long-term Azov Sea ecosystem dataset.

Column Header	Units (Metric)	Description
don_flow	km3	The total annual Don River flow volume
salinity_taganrogbay	‰	Average annual water salinity in the Taganrog Bay
salinity_azovsea	‰	Average annual water salinity in the Sea of Azov
sander_azovdon_totalbiomass_kt	kt	Sander (Sander lucioperca) total biomass estimation via the swept area method on the 1st of January of a given year
sander_catches_kt	kt	Sander (Sander lucioperca) annual fishery catch volume
gobidae_biomass_kt	kt	Goby (Gobiidae) total biomass estimation via the swept area method on the 1st of January of a given year
gobidae_catches_kt	kt	Goby (Gobiidae) annual fishery catch volume
rus_sturgeon_fisherybiomass_kt	kt	Russian sturgeon (Acipenser gueldenstaedtii) fishery biomass estimation on the 1st of January of a given year
zooplankton_mgm-3	mg/m3	Relative (to water volume) total zooplankton biomass during the summer period via the swept area method
zoobenthos_gm-2	g/m2	Relative (to bottom area) total zoobenthos biomass in autumn
zoobenthos_forage_gm-2	g/m2	Relative (to bottom area) forage fraction of zoobenthos biomass in autumn (suitable for sturgeon feeding)
rutilus_fishery_biomass_kt	kt	Sea roach (Rutilus rutilus Heckelii) annual fishery biomass estimation via the swept area method on the 1st of January of a given year
rutilus_catch_kt	kt	Sea roach (Rutilus rutilus Heckelii) annual fishery catch volume
temp_avg_summer	°C	Average sea surface temperature in summer from in situ observations
rus_sturgeon_numbers_ths	thousands ind.	Russian sturgeon population abundance (numbers)
rus_sturg_catch_legal_t	t	Russian sturgeon legal annual fishery catch volume
rus_sturg_catch_iuu_t	t	Russian sturgeon Illegal, Unreported, and Unregulated (IUU) annual catch volume estimation (if available)
sturgeon_release_juv	million juv.	Russian sturgeon hatchery annual release numbers for the Don and Kuban Rivers

) (“Annual Azov Sea benthic ecosystem dataset.csv”) contains averaged data which can be arranged in the following categories: (1) demersal fish biomass estimates and catch statistics, (2) hydro-biological specimen (zooplankton and zoobenthos) estimates; (3) hydrologic parameter evaluations; and (4) hatchery production.

According to the ecosystem approach, the presented time series of biotic and abiotic factors may be related, as previously shown [9]. An example of a theoretically possible relationship is between the volume of the Don River runoff and the reproduction potential of all represented fish species. The volume of Don River runoff directly affects the salinity of the water in the Taganrog Bay and the entire Sea of Azov. Air and water temperature affect the duration of the vegetation period and the timing of when fish begin to spawn and feed. Parameters such as salinity and temperature determine the structure of zooplankton communities and, to a lesser extent, those of zoobenthos.

In the classical ecosystem, the Russian sturgeon is considered a top consumer. However, due to its diverse diet, the traditional predator–prey relationship between Russian sturgeons and gobies is not always applicable. Sanders and roaches should also be considered food competitors for the sturgeon. The catch statistics of all fish species can be seen as an anthropogenic factor that influences their populations.

From a biological perspective, all time series can have an impact on the abundance (and biomass) of the Russian sturgeon population.

2.1.2. Spatial Occurrence Dataset

The spatial occurrence dataset contains two tables (

Table 2. Column description of the spatial in situ dataset.

Column Header	Units (Metric)	Description
Spatial data table description
weight	g	Total weight of Russian sturgeon fish in the catch by one trawl iteration
numbers	ind.	Total number of Russian sturgeon fish in the catch by one trawl iteration
is_fishery_length	boolean	Fish length (SL) is greater (1) or lower (0) than the commercial size (90 cm)
date	MM/DD/YYYY	Capture date
gear_square	km2	The size of the fishing gear area, calculated using the trawl mouth area
temp_air	°C	Air temperature
temp_water_surface	°C	Surface water temperature (at 2.5 m depth)
waving		Waving according to the Beaufort scale
latitude	° N	Latitude of trawling operation start (WGS-84, projection EPSG:4326)
longitude	° E	Longitude of trawling operation start (WGS-84, projection EPSG:4326)
survey_id		Relationship to survey description table (by ID)
Survey table description
id		Survey ID
date_start	MM/DD/YYYY	Date of survey start
date_end	MM/DD/YYYY	Date of survey ending

): the table with full spatial in situ data (“Azov Sea trawl survey sturgeon data.csv”) and the table with the survey and description (“Azov Sea trawl survey description.csv”).

2.2. Data Structure

All datasets are also formatted as comma-separated values (.csv). The column separator is a comma (“,”), and the decimal separator is a dot (“.”). No separation for thousands was used.

2.3. Spatial In Situ Survey Dataset Preview

The spatial dataset includes 973 rows.

Table 3. Spatial dataset filling summary.

Data (Column)	Non-Empty Cells
weight	651
numbers	968
temp_air	415
temp_water_surface	535
waving	904

describes summary diagnostics of non-empty cells. Figure 1 shows the spatial distribution of the available data.

2.4. Annual Dataset Diagnostics

In the previous paper [9], the tests of time series stationarity for this dataset (excluding the 2023–2024 years) were applied. No evidence of non-stationarity was found.

Summary boxplot diagnostics [13,14] for every time series from the annual dataset are shown in Figure 2. The largest variations and significant outliers were found, in general, for the Don River flow volume (a), sander total biomass estimations (d), goby catch volume (g), sea roach catch volume (i), and zooplankton relative biomass (j), as well as for zoobenthos relative biomass (k). However, this is not surprising because the study period covers various conditions of the Azov Sea ecosystem after shifts [15,16]. The Don River flow was significantly reduced after 1952 when the Tsimlyansk Hydroelectric Complex was built [7]. This construction has caused, in general, changes in the salinity regime, and, afterwards, affected goby, sander, and sea roach populations. This leads to population collapses and, consequently, catch volumes [17]. Overfishing and IUU fisheries in the late 1990s also impacted the official catch statistic representativeness.

Large variations and “outliers” in zooplankton were due to the rapid ecosystem change in the early 1990s provoked by the invasion of jelly, Mnemiopsis leidyi, and, in general, by the degradation of trophic chains [18]. Against this background, some extremely high zooplankton biomass values in the previous period look like “outliers”; in fact, this is not quite accurate.

Some “extremely high” values of total zoobenthos biomass were caused by rapid growth in the population abundance of the invasive species. In recent years, the basis of the benthos biomass in the Sea of Azov has been formed by a bivalve mollusk from Southeast Asia—Anadara kagoshimensis (Tokunaga, 1906) [12]. High zoobenthos biomass was also caused by the Mediterranean mussel Mytilus galloprovincialis Lamarck, 1819; the population abundance growth was related to the increased sea salinity [19]. Against the background of the introduction of the alien species into the Sea of Azov, high zoobenthos biomass can be caused by a sharp decrease in benthivorous fishes and trophic chain degradation since the 1990s and due to some atmospheric macro processes [20].

A detailed explanation of the impact of invasive species can be seen in Figure 3. As can be seen, since 2014, the total zoobenthos biomass has started to grow, though the forage fraction has not. In general, this explains the detected outliers in the process of the diagnostics of zoobenthos time series.

Taking into account this discussion, it should be clear that “outliers” in statistical tests are not “outliers” in a biological way.

3. Methods

3.1. Annual Dataset

The annual dataset was built on top of summary estimations on in situ data by the Azov-Black Sea branch of VNIRO. The total or fishery biomass of every fish species was estimated using the swept area method [21], based on the bottom trawl survey in autumn. This method is quite simple but expensive: Every year, a survey on a standardized spatial grid is performed in the autumn period with a bottom trawl. Then, all catches in a trawl by species (weight of every species in one trawl iteration) are divided by the gear fishing area and categorized by the catchability coefficient calculated earlier. Finally, these averaged (or bootstrapped by median) density values are multiplied by the Azov Sea total area, allowing for obtaining total (or fishery) biomass estimation. Further, this estimation is recomputed from a survey moment (autumn) to the beginning of the next year (the first of January) by including natural and fishery mortality rates and expected recruit counts after spring or summer spawning.

Zooplankton relative biomass was calculated by averaging summer in situ observations relative to water volume. Samples were collected according to accepted methods [22]. A medium-sized Jeddy plankton net (inlet diameter, 24 cm; mill gas mesh size, 0.076 mm) was used.

Zoobenthos relative biomass was calculated in the same way as zooplankton, with the difference that the calculation was made relative to the area of the bottom water space. Samples were collected according to accepted methods [22]. The Petersen bottom grab with a sampling area of 0.1 m² was used as benthos sampling gear.

The air temperature was measured by an aspiration psychrometer MV-4M. Measurements were averaged for the summer period.

The water surface temperature measurements were carried out by a TM-10 mercury thermometer integrated into the GR-18 bathometer, as well as a platinum resistance thermometer, the sensor of which was integrated into the Vector-2 flow direction and velocity meter.

Sturgeon larvae numbers (juveniles) were counted at hatcheries before the release by partial sampling and recalculation to the basin areas.

3.2. Spatial In Situ Survey Dataset

The spatial dataset was collected during the regular bottom trawl surveys in the Azov Sea by the Azov-Black Sea branch of VNIRO. In surveys with an 18–27 m width (since 2017, only a 25 m width trawl has been used), a bottom trawl with a full mesh size of 13 mm was used. The default trawl iteration interval length is 1/2 h.

Russian sturgeon fish in catches were not dissected due to the fishing ban since 2000. After length measurement, weighing, and counting, all fish were released back into the seawater.

All Russian sturgeon fish were separated into two subsets: fishery and non-fishery. This splitting was carried out in accordance with a previously applied regulation measurement: all fish with a length of more than 90 cm (SL) were in the fishery group, and those with less than 90 cm were in the non-fishery group. This fishing measure is applicable only to the Russian sturgeon in the period before the 2000s, but it differs significantly for other sturgeon species.

The air temperature was measured by an on-board mercury thermometer in the shade. The water surface temperature was measured at a 2.5 m depth by a digital water thermometer.

3.3. Data Homogeneity

Considering the period of the research data, it is important to check uniformity. The length of the time series is primarily due to the existence of the Azov-Black Sea Fisheries Institute, which was founded in 1937 as a research station. The first ship expeditions and surveys in the Sea of Azov began in the same year, but the most extensive research has been carried out since 1956, when trawl fishing gear was introduced. Subsequent expansions of expeditionary research occurred in 1965 (monitoring environmental parameters and bottom surveys in spring), 1973 (special surveys on sturgeon), and until the 1990s. In the 1990s, following the collapse of the Soviet Union, scientific research expansion was halted due to economic difficulties, but the monitoring and evaluation of fish populations continued. In 2016, due to reduced funding, some marine surveys had to be optimized, and the tasks of several surveys were combined into one. This led to some changes in the way research was conducted in the Sea of Azov during the period of 1991–2022.

During this period, Russia and Ukraine carried out research in the Sea of Azov after the collapse of the Soviet Union. Towards the middle of the 1990s, a Russian–Ukrainian agreement was reached to establish the Russian-Ukrainian Fisheries Commission for the Sea of Azov. Under this agreement, research in the Sea of Azov was conducted jointly, with data exchange between the two countries’ institutions.

Given these features, it should be noted that among the data presented, the most reliable are those for the period after 1956. The earliest available data include estimates of the annual runoff of the Don River (provided by the Agency for Water Resources), as well as estimates of the biomass of sanders and roaches (based on traditional fishing and research data from those years). In subsequent years, as expeditionary research expanded, additional estimates became available, including hydrological indicators such as water salinity and temperature (starting in 1960), zooplankton and zoobenthos biomass (since 1970), and data on the number of juvenile Russian sturgeons produced in hatcheries (since 1967), as well as parameters related to their legal and illegal fishing (since 1980).

To understand possible violations of measurement uniformity, it is important to consider approaches and methods of sample processing. As mentioned earlier, the temperature measurements were carried out using specialized equipment with high accuracy, which did not require calibration under the specific temperature conditions in degrees Celsius. Salinity measurements were also conducted using a similar approach. Additionally, the sampling methods remained consistent throughout the entire period of observations.

The approach to estimating relative biomass and the method of sampling zooplankton and zoobenthos have not changed since 1970. Only the systematic position of species has been updated in accordance with international studies. The methodology for collecting and processing samples and estimating relative biomass is described in detail in a methodological manual [22] and complies with the standards of hydrobiological research [23].

However, there may be some inaccuracies in the reliability of the estimates of stock biomass (commercial or total) or fish abundance. The swept area method, a traditional method for estimating fish stocks in the Sea of Azov, was used throughout the research period. Despite this, when using this method, one of the key parameters is the catchability coefficient and the representativeness of spatial data. When preparing these data, the same catch coefficient was applied to each data series for estimating stock biomass. Estimates of the fish stocks’ biomass were based on data from specialized bottom trawl surveys, which are conducted on the same spatial grid every year.

It should be mentioned that the dataset presented is sufficiently and satisfactorily homogeneous, to the best of our knowledge in the field of biological research. At the same time, there are no predefined methodological discrepancies in the presented dataset within each time series in the dataset.

4. User Notes

Shared datasets can be used to discover historical cycles of the Russian sturgeon population in the Azov Sea. One way of assessment is to train machine learning models for population abundance forecasting. The same approach was already applied on a small annual dataset to predict goby population biomass [9]. To enable predictions, the lags between predictor factors (environment, food chain, etc.) and the predictable variable were introduced, which allowed for finding a reliable model fitting up to the lag of +2 years.

Another way was to match the spatial occurrence dataset to environmental factors in order to describe relationships of sturgeon occurrence in relation to habitat conditions. Following this way requires an evaluation of a spatial hydrological dataset with temperature, salinity, and other factors, such as the “Copernicus Project” on the Black Sea [24] or other available research studies on the Azov Sea [25,26,27,28]. Such solutions can make it possible to determine the tolerance ranges of temperature, salinity, and other factors for sturgeon population distributions.

Ultimately, such solutions for forecasting in space and time can significantly increase the reliability of fishery management (including its regulation) and may help preserve this species in conditions of low abundance and insufficient data collection in recent periods.