Within the ICES baseline landings data, several years included catch statistics with very coarse taxonomic resolution, i.e., “Finfishes nei,” “Flatfishes nei,” and “Anarhichas” (wolffishes). These broad “nei” (or “not elsewhere included”) categories are often reported to a finer taxonomic resolution in national statistics, suggesting that “nei” categories may be an artifact of the statistical reporting or harmonization process at ICES (Pauly and Zeller, 2015). These uninformative taxonomic groupings were taxonomically disaggregated based on best-available information and conservative assumptions about which species should be included in these categories (Table 1). “Finfishes nei” was disaggregated into the top 10 species caught proportionally by weight, excluding major commercially targetted species (i.e., not cod, capelin, or haddock). Non-major species were chosen under the assumption that they are less likely to be identified to the species or genus level in records, while valuable or commercially important species with dedicated ICES stock assessment groups likely are. “Flatfishes nei” and “Anarhichas” were both disaggregated proportionally by weight of landed flatfish and wolffish species, respectively.

Russian statistics for the Barents Sea have only been reported as ICES statistical area I, without subarea reference (Figure 1). In order to assign them to more spatially explicit locations, catch was split into each subarea: Ia (High Seas), Ib (Russian EEZ), or Ib (Norwegian EEZ, including Svalbard; Figure 1). Catch was split proportionally by surface area (Table 2). In doing so, most catch was allocated to Russia's EEZ, as that is the largest area, followed by Norwegian waters, followed by High Seas (Ia). As Norway and Russia have numerous bilateral fishing agreements, Russian fishing is indeed occurring in Norwegian waters (Nakken, 1998; FAO, 2007; ICES, 2015a). Very little information could be found on more spatially explicit fishing locations that could be applied to all data, as catch statistics are generally reported in broad geographical regions by national authorities and by ICES.

Generally, Russian fishing activities in the Barents Sea can be divided into three main groups by fishing gear: trawl, purse seine, and longline. According to an examination of the cod fishing fleet in the Barents Sea in 2004 (WWF, 2005), ~150 trawlers under 15 m and 200 trawlers over 15 m were active. While nearly half of the fleet is small and could be considered artisanal in nature, we considered all fishing gears that are actively moved through the water or across the seafloor while using engine power as industrial gear (or “large-scale”) irrespective of vessel size, as defined in Martín (2012). Furthermore, given the heavy focus on offshore fishing by relatively large vessels throughout the Barents Sea, we considered the purse seiners and longliners as industrial as well. We therefore considered all landings reported to ICES as part of the industrial sector. Catches by artisanal (i.e., small-scale commercial) fleets were estimated as unreported catches as described below.

Official ICES catch statistics are not corrected for unreported catches that may be included in ICES stock assessment working group reports (ICES, 2017a). We considered discrepancies between ICES publicly reported statistics and ICES Working Group reports used for stock assessment as “unreported stock assessment landings.” Unreported landings were added for nine species using data from several ICES Working Group reports (ICES, 2001; 2015a; 2015b; 2015b; 2016). Unreported stock assessment landings added an average of 10% to the total reported landings.

Unreported illegal landings, such as obtained through poaching, reflect estimates of entirely unreported landings across the fishery and are criminal in nature. Unreported illegal landings, oftentimes to avoid state control (during the Soviet era) or as a result of organized criminal activity (poaching), occurred throughout nearly the entirety of the study period (O'Hearn, 1980; WWF, 2005; FAO, 2007; Burnett et al., 2008). Historical estimates of tons of underreported catch per ton of reported catch acted as “anchor points” for years where such estimates existed in the literature. In between anchor point years, these estimates of underreporting were linearly interpolated unless otherwise stated. A more in-depth historical context behind these anchor points is presented in the discussion.

We assumed unreported illegal landings were zero from 1950 to 1959. This reflects the conservative assumption that all landed catch was reported during the years of Stalin's rule and immediately following his death. From 1960 to 1975, unreported landings (as a percentage of reported landings) rose steadily from 0 to 33%, to reflect an estimate reported in O'Hearn (1980). This 33% rate was kept steady from 1976 until the last year of the Soviet Union (1990) and was increased thereafter to 40%, to reflect an estimate by the Norwegian Directorate of Fisheries that underreporting had reached a rate of “almost 50 per cent” (Burnett et al., 2008). As the estimate of almost 50% comes from a 2008 report, we assumed 2008 to be the last year of such high underreporting. An underreporting rate of 5% for the year 2014 was chosen given the 2015 Arctic Fisheries Working Group report estimating little to no underreporting (ICES, 2015a); for years between 2008 and 2014, the rate was linearly decreased from 40 to 5%.

The above rates of underreporting were applied to the reported baseline landings of all fish except Atlantic salmon (Salmo salar). According to the Working Group on North Atlantic salmon (WGNAS), illegal poaching of salmon is a “considerable” problem in the Barents and White Seas, particularly after the 1990s (ICES, 2015b). The report goes on to say that this high level of underreporting continued into the 2000s. Independent estimates of salmon poaching in the region indicate that poaching may reach underreporting levels as high as 50% (Spiridonov and Nikolaeva, 2005). As such, unreported catches for salmon followed the above unreported rates until 1991, at which point underreporting increased to 50% of reported landings and remained at that level until 2014.

While all ICES reported landings for Russia were categorized as industrial, Russian national data from the Russian Federation Federal State Statistics Service (фeдepaпьнaя cлyжбaгocyдapcтBeннoй cтaтиcтики) included catch statistics by species for the White Sea separately from the Barents Sea. These data do not seem to be included in the ICES data. Russian national data for only the Barents Sea generally matched ICES statistics very well, and were thus considered to be comparable to the ICES dataset; there was no comparable match between national data and ICES data for the White Sea. Therefore, we assumed ICES baseline statistics did not include catches for the White Sea, and added the national White Sea data as unreported landings. Because the White Sea is a relatively small, sheltered, shallow body of water that is likely being fished by a smaller coastal fleet, we assumed that all landings from within the White Sea were artisanal in nature. Federal statistics were only available for the years 2010-2013, which were averaged and converted into a percentage of reported Barents Sea landings per year (i.e., 0.2%). For all years from 1950 to 2014, we therefore assumed that White Sea artisanal landings were equivalent to 0.2% of reported Russian Barents Sea landings, broken down by taxa as reported in the national data for the White Sea. These landings were designated as unreported artisanal. While artisanal fishing activity has existed in the Barents Sea region since 1950 (Shevelev et al., 2011), our estimate for the whole area is likely not a very comprehensive representation of artisanal fishing in the entire Barents Sea. We consider our approach to provide a very conservative minimal estimate of artisanal activities in these waters, and we would like to encourage further research on non-industrial fishing in the wider Barents Sea area.

Discards are unwanted fish (bycatch) that are caught in the process of actively targeting a more desirable species, and are especially common in non-selective fishing gears such as bottom trawls. Bycatch from industrial gears are often discarded overboard and generally experience high mortality rates. While discards happen in nearly all industrial fisheries, there is as of yet no official reporting of Russian discards by fishery within the Barents Sea (ICES, 2015a).

Discards were calculated by associating a fishing gear with a primary commercially targeted species, then using independently published estimates of bycatch rates for that gear to calculate the tonnage of discards per tonnage of landed catch. Discards were therefore calculated as a percentage of total landings (reported plus unreported) by major target taxa and gear associated with that target fishery. Only the largest commercial fisheries with the best available information on gear types were chosen to calculate discards; thus, our discard estimates may be underestimating other discards, as discards likely exist for all other fish caught and reported in the Barents Sea. As gear types and discard rates change over the years with improvements in technology and with changes in regulation, the gear type and discard rates associated with each fishery varied by decade. For example, the installation of sorting grids throughout various fisheries meant that discards decreased as time went on. Not only did discards decrease overall as the decades passed, but improvements in trawling technology and better targeting also meant the overall species composition of discards changed.

The exception to calculating discards by fishing gear was the crab fishery, where a flat discard rate was applied to all crab and miscellaneous marine invertebrate catches, regardless of how they were caught. While there was insufficient information available to associate the crab fisheries with specific gears, there were multiple independent estimates of rates of bycatch within the crab fishery in general.

Recreational fishing has historically been popular in Russia (FAO, 2007), particularly for salmon on the Kola Peninsula in the Barents Sea region (Figure 1; ICES, 2012, 2015b). The ICES planning group on recreational fishing indicates that on average, recreational fishing accounts for 2–8% of a country's total reported landings (ICES, 2010). However, as there is little to no data on recreational fishing in the Barents Sea region prior to 1990, a likely conservative recreational fishing rate of 0.5% of reported landings was applied for the period 1950-1990. 1990 was likely the year when recreational fishing was first “officially” opened to foreign tourists, and it is assumed that recreational fishing increased in popularity with the increase in tourism to the region after the fall of the Soviet Union. Thus, it was assumed that from 1991 onwards, the lower end estimate of 2% from ICES (2010) was chosen for calculating recreational catch.

The exception to the data-poor recreational sector in Russia's Barents Sea waters is recreationally caught Atlantic salmon (S. salar). Recreational catches of salmon after 1991 are exceptionally well documented by the Working Group on North Atlantic salmon (WGNAS; ICES, 2015b). Data on the number of salmon that were caught and then retained each year in the recreational fishery were obtained from the 2015 Working Group report (ICES, 2015b). The average annual mean fork lengths and whole weights of Atlantic salmon, for all sea ages, for each year from 1991 to 2012, were published by the National Oceanic and Atmospheric Administration (NOAA) and used to convert the number of salmon caught to the weight of salmon caught (Sheehan et al., 2013). For the years 2013 and 2014, the same weight as in 2012 was used.

After calculating the total recreational landings per year, the estimated catch of salmon for that year was subtracted from the total. The recreational catch without salmon was then split evenly between seven commonly targeted recreational species described on numerous Russian fishing websites: cod (G. morhua), navaga (Eleginus nawaga), polar cod (B. saida), wolffish (Anarhicas lupus), haddock (M. aeglefinus), saithe (P. virens), and pollack (Pollachius pollachius).

While the Soviet Union had public cafeterias to ensure nobody went hungry, food shortages were common and diets were supplemented with home cooking. However, obtaining groceries often involved waiting in long lines and paying exorbitant prices. It is likely that non-commercial catching of fish for family consumption (subsistence) complemented the rural diet, particularly in coastal communities.

Catch from subsistence fishing was calculated in two steps: first, the Russian rural population of the Barents Sea was estimated; second, this population estimate was multiplied by per-capita estimates of fish consumption in the USSR and Russia. Per-capita consumption rates were adjusted through time to reflect changes that are noted in the literature. Population data for the period from 1950 to 2001 were obtained from Populstat³, while from 2002-onwards Russian census data for the years 2002 and 2010 were used. For the years between 2002 and 2010, population data were interpolated; for the years 2010 to 2014, estimates of total population size provided by the Federal State Statistics Service⁴ were used. For both per-capita consumption rates and population, estimates in years between data points were linearly interpolated. See Supplementary Materials for more details.

In order to estimate the amount of fish consumed that were actually caught via subsistence fishing (as opposed to being purchased at market), we relied on the conservative estimate derived in the Russian Black Sea fisheries reconstruction (Divovich et al., 2015). Thus, it was assumed that 5% of all fish consumed was caught via subsistence fishing until just after the dissolution of the USSR (1992), thereafter increasing to 20% by 1995 and 26% by 2002 to reflect a decreased reliance on government food services and the increased food costs associated with the collapse of state subsidies (Divovich et al., 2015). The derived per-capita subsistence catch rate was then applied to the estimated rural population around the Barents Sea to estimate a likely total tonnage of subsistence fishing. The species disaggregation for subsistence fish was kept the same as in recreational fishing, as both are small-scale fisheries that employ similar methods of fishing. We recognize that there may be overlaps between recreational and subsistence fishing.

The planned nature of the former Soviet economy was designed to allocate goods and services as effectively as possible across all sectors of Russian society. During the Stalin years in the first half of the 1950s, harsh authoritarian rule likely prevented the underground economy from thriving. After Stalin's death, however, widespread corruption and a weak economy led to the steady rise of the Soviet “second” or “shadow” economy—i.e., the black market. The black market was so important to maintaining Russians' access to goods and services that unregulated and illegal economic activities were pervasive in all sectors of the economy. Nearly a third of all food purchased for the home was done so via this black market (Sampson, 1987), and nearly a quarter of the fish produced entered this black market (O'Hearn, 1980).

Soviet fisheries were not immune to this pervasive corruption. Illegal underreporting of catch began to steadily rise following Stalin's death in 1953 and the subsequent loosening of authoritarian control. The government, however, was not ignorant of underreporting; in fact, as one Soviet official noted, “the government knows exactly who is dealing in what—arrests are only made when there is some larger political reason,” and data on fish and game in particular was “very good” (O'Hearn, 1980). In reports O'Hearn found in the Soviet press from the 1980s, Soviet observers lamenting the lack of environmental oversight commented on the “painfully large number” of poachers using the black market for personal gain. Small fines for poaching, along with the opportunity to fetch up to 4 to 10 times the “official” Soviet price, meant that poaching was commonplace throughout Soviet fisheries (O'Hearn, 1980). It was standard practice for fishers to first offer catch on the black market and then officially hand in the rest, and it was estimated that by the 1980s, roughly 25% of total commercial catch was meant for the black market. Official Soviet reports note that unreported landings may have been as high as 267% of reported catch (O'Hearn, 1980). To remain conservative, the lower estimate of 25% of total catch, or one third of reported catch, was chosen in the present study for calculating unreported commercial catch during this time period. It is possible that our study therefore underestimated actual catches during the Soviet period.

Following the dissolution of the Soviet Union, underreporting of catch only increased further in all of the former Soviet republics. Rates of illegal fishing increased as a result of the sudden collapse of Soviet regulations and controls and the opening of the market to the outside world as the iron curtain lifted (FAO, 2007). It is during this period from the late 1990s to the early 2000s that the Norwegian Directorate of Fisheries began apprehending Russian fishing vessels in the Barents Sea in order to enforce bilateral quotas and reported their conservative estimate of overfishing “of almost 50 per cent” by Russian ships in the Barents Sea (Burnett et al., 2008).

More recently, reports of illegal catch and underreporting have substantially decreased. This is mostly a result of greater cooperation and enforcement on the part of the Joint Norwegian-Russian Fisheries Commission. In fact, one of the most recent Arctic Fisheries Working Group reports estimates that there has been little to no illegal or unreported commercial catch in recent years (ICES, 2015a). In the case of cod in particular, reports from Norwegian-Russian analysis groups indicate that actual catches of cod have roughly matched officially reported landings of cod since 2009 (ICES, 2015a).

Our conservative estimates of unreported black-market landings during the Soviet era and immediately post-Soviet collapse has interesting implications for discards. Because discards are calculated as a proportion of both reported and unreported landed catch, calculated discards would have more than doubled for the time period if a higher rate of landings underreporting would have been applied, without any change in fishing effort. The subsequent collapse of certain fisheries (such as cod and capelin) could then potentially be attributed not only to the high rate of discarding in Soviet fisheries, but to a high rate of unreported catch as well. Future adjustments and improvements upon this catch reconstruction should aim to refine this estimate of underreporting in Soviet fisheries.

The Barents Sea has experienced considerable fluctuations in both stock abundances and catch that has been documented by the Norwegians as early as 1860, if not earlier (Alexseev et al., 2011). In particular, the capelin stock has been known to be highly variable (Gjøsæter, 2009). Indeed, these large natural variations in capelin are one of the primary reasons fisheries scientists first came to research the Barents Sea (Alexseev et al., 2011). Similar fluctuations in cod, haddock, and saithe catches in the White Sea have been documented as well (Alexseev et al., 2011). Typically, these fluctuations are the result of abiotic factors unique to the biogeography of the region, such as the nutrient load of the system after winter (Matishov et al., 2004). These historical trends appear to have continued into the present day given the fluctuations in capelin and cod catch (Figure 3A).

Reported landings of cod in the earlier period of this study (1950-1980) before the stock collapse exhibit regular periodicity, which in turn drive the periodicity behind the majority of unreported landings and discards (Figures 3A,B). This periodicity has been well documented in the literature, where it has been found that the Barents Sea cod stock fluctuates in harmony with the Kola temperature cycle (Nakken, 1994; Yndestad, 2003; Ottersen et al., 2014). However, the regular rise and fall of reported catch appears to be unusual and does not closely match any cyclical trends of Norwegian cod catch in the region (Figure 6), and could potentially be driven by misreporting to ICES by the former Soviet Union. It remains to be determined if the substantial redirection of (unmonitored) catches to the black market may be a contributing factor to the unusually large fluctuations seen in USSR-reported data on cod catches in the first 3 decades of the present time series (Supplementary Figure S1).

Russian fisheries in the Barents are heavily dominated by bottom and pelagic trawling. Trawling had already been established by the 1920s but became especially well developed by the 1950s (Grekov and Pavlenko, 2011; Shevelev et al., 2011) and only intensified through the 1960s-1970s. Initially, trawls primarily targeted cod and other demersal fish stocks, while capelin and herring were only caught as baitfish. As trawling intensified, however, these conventional stocks declined and both Norway and the Soviet Union began to develop purse seine and pelagic trawl fisheries for the industrial targeting of capelin in the Barents Sea (Gjøsæter, 2009). Both pelagic and bottom trawling continued to dominate Russian fisheries, while Norwegians steadily developed longline fisheries at an industrial scale (Gjøsæter, 2009). The first Russian automated longliner, borrowed from well-established Norwegian technology, was only introduced into the Russian fishery in 1982 (Gjøsæter, 2009). By the 1990s, the collapse of the Soviet Union, with the associated collapse of the subsidies system and resulting economic downturn, meant that there was a reduction of trawling effort. Trawlers required more fuel and active fishing time, and were thus more expensive to operate. This led to an overall reduction in catch along with marginal increases in the number of longliners in the Russian Barents Sea fishery (Grekov and Pavlenko, 2011). However, even to the present day, trawlers still dominate the Russian fleet, accounting for around 90% of total Russian catch in the Barents Sea (Nakken, 1998; Grekov and Pavlenko, 2011; Wienerroither et al., 2011; ICES, 2015a).

Recent declines in catch beginning in the 1970s are strongly tied to overexploitation (Matishov et al., 2004; Pavlovich, 2016). Given that Norwegian and Russian researchers and government bodies alike acknowledge overexploitation and discarding of underage fish in the 1970s-1980s as a primary cause for the decline of both pelagic and demersal stocks, including cod (Nakken, 1994; Gullestad et al., 2014), haddock (Kiseleva and Nichols, 2016), redfish (McBride et al., 2014) herring (Gjøsæter, 1995), and capelin (Beverton, 1990; Hjermann et al., 2004), it is not unreasonable to assume other stocks were affected by the excessive fishing effort in the region as well (Figure 4). Modern trawling technology is notorious for scooping up entire schools of fish, with midwater and bottom trawl fleets equipped with immense nets that can reach over 100 m in width and several hundred meters in length (Morgan and Chuenpagdee, 2003) and electronics such as echosounders, gyro-compasses, and radio direction finders (introduced into the Barents Sea fishing fleets in the 1950s; Shevelev et al., 2011), all designed to catch as many tons of fish as possible in one trip. Unless considerably larger mesh sizes are introduced, nets capture all age-classes of a stock, preventing any stock recovery let alone growth in the coming years, as happened with the Barents Sea herring and capelin stocks (Beverton, 1990; Alexseev et al., 2011; Shevelev et al., 2011). Bottom trawling, in particular, is known to be highly destructive, decimating slow-growing deep-water stocks of fish and damaging the benthic habitat structures that they may depend on (Løkkeborg and Fosså, 2011; Norse et al., 2012; Puig et al., 2012). The high rates of both pelagic and bottom trawling in combination with high rates of discarding as compared to the Norwegian fleet (Nedreaas et al., 2015) may therefore have contributed to the rapid declines of stocks in the Barents Sea in the 1970s-1980s.