In total, 2575 tows carried out over six decades across the Adriatic Sea detected 33 small, demersal, meso-predatory elasmobranch species (average trophic level: 3.9, SD: 0.1231
), including 12 sharks, 20 rays and one chimaera (included for their evolutionary and ecological similarity, ). Of these, 11 species ceased to be detected during the period of observation (no more occurrences after the year 2000, ), while 6, mostly deep-water species and small skates, were only recently detected by the MEDITS surveys which expanded to greater sampling depths ().
Species caught in the trawl surveys
Across all five surveys, species-specific frequency distributions were very skewed (), with few dominant species and many occurring only sporadically. The Hvar survey detected the highest diversity (23 species, Shannon Index [SI
]: 3.39, ), dominated by small-spotted cat sharks and thornback skates with unstandardized densities of 426.8 and 76.8 individuals/km2
, respectively, and frequencies of occurrence (FO) of 0.76 and 0.71. The other 21 species were caught in < 21% of the tows with densities below 11 ind./km2
. Over time, moving to the most recent survey, richness and abundance decreased toward more flattened and truncated distributions. In MEDITS (SI: 1.96), the small-spotted catshark was still the most abundant species, but with a density of 62.1 ind./km2
and FO of 0.20 (), a 6.8- and 3.8-fold decline, respectively, compared to Hvar. Overall, 21 out of 27 species had FO < 0.021 and densities < 4 ind./km2
. The high elasmobranch abundance and diversity characterizing the central Adriatic during the Hvar survey in 1948–49 disappeared (). Yet, species richness and abundance were higher in the eastern coastal areas than elsewhere (). Elasmobranch abundance in Croatia was almost one order of magnitude higher than in Italy (Supplementary Fig. S1
), where sharks and rays were largely absent except for a relatively high-density zone in the upper Adriatic (above the 50 m isobath, MEDITS) mainly composed of spurdogs (Supplementary Fig. S2
), smooth-hounds, and eagle rays.
Frequency of occurrence (FOi) and mean density of elasmobranchs caught in the analyzed surveys, and Adriatic fishing effort.
Spatial comparisons of catches between Hvar and MEDITS surveys.
Standardized catches generally confirmed the above patterns. Although the fitted models had considerable selection uncertainty - in each survey, more than half of the species had a selected best model with <10% chance (Akaike weight <0.1) of being the most plausible among a 95% confidence set of models (see methods, Supplementary Tables S2–S6
) – spatial covariates (depth, latitude and longitude) resulted as the best predictors of species abundance ( and S2–S6). In the largest surveys available (Hvar and MEDITS), these predictors pointed to an increase in catch rates from the Italian to the Croatian coasts, and for most species from deep to shallow waters (i.e. from offshore to inshore, Supplementary Fig. S3
). Temporal covariates (year, time of the year) were less important than spatial variables, but many species had significant short-term temporal trends within surveys ().
Analyses of catches in the Jukic survey (1963–1971), located in the central eastern Adriatic, identified nine species with reliable short-term trends in abundance, and all except the thornback skate showed an increase in standardized catches (three species were significant, ). In comparison, between 1994 and 2005 across the Adriatic (MEDITS surveys), 16 species had reliable estimates of population change. Of these, nine species, mainly sharks, showed declines (3 statistically significant, ), while increases were mostly shown by meso-pelagic rays and small skates; yet these trends were not significant except that for the eagle ray, which increased by 3.74 times (Confidence Interval, CI: 1.06, 13.45) in 11 years.
Over longer periods of time, changes in abundance were larger and more significant. Comparing Hvar and MEDITS surveys showed that elasmobranchs declined by 94.5% over 57 years (). Sharks declined more than rays (−95.6% vs. −87.7%) with small-spotted catsharks (−96.2%, CI: −97.8%, −93.5%) driving most of the patterns. Rays shifted in species composition. In particular, the thornback skate, the most abundant ray in the 1940s, recorded the steepest decline (−97.2%, CI: −98.4%, −95%), whereas brown skates increased by 2.36 times (CI: 1.05, 5.3) becoming the most abundant skates (). Significant long-term increases were also detected for eagle rays (111 times, CI: 17.05, 735), marbled torpedoes (75.9 times, CI: 5.07, 1135.64) and spurdogs (3.1 times, CI: 1.05, 9.27) ().
In the coastal area off the eastern Adriatic (Zupanovic area, ), between 1957 and 2005 elasmobranchs increased significantly by 2.12 times (CI: 1.59, 2.83). Rays increased faster than sharks (3.79 vs. 1.88 times) with brown skate and common eagle rays recording fold increases of 17.7 (CI: 9.9, 31.7) and 19.73 (CI: 7.08, 55.03) respectively. As for sharks, the patterns were largely driven by the almost 2-fold increase of small-spotted catsharks (1.92, CI: 1.1, 3.4). The smooth-hound recorded the largest increase (21.13, CI: 8.23, 54.25), and the thornback skate was the only elasmobranch showing a significant decline (−39.8%, CI: −63.14%, −1.77%) ().
In the Jukic area, comparing catch rates in the period 1948–1971 (Hvar-Jukic comparison), elasmobranchs declined by 61% in 23 years. These were mainly sharks (−61%), while rays recorded a moderate and non-significant decline (−38%), although brown skates suggested some increase, and the common eagle ray increased by 128 times (CI: 46, 358). The spurdog was the only shark increasing significantly (4.43-fold, CI: 1.43–13.72), while 4 of the other 5 species we could model recorded declines between −63% (small spotted cat-shark, CI: −83.6%, −16.6%) and −96.8% (the smooth-hound, CI:−98.7, −92%). In contrast, in the period 1963–2005 (Jukic-MEDITS comparison), sharks and rays as a group declined comparably and significantly by more than 90%. Spurdogs reversed their earlier increase with a decline of −87.6% (CI: −95.2%, −68%) and the increase of common eagle rays became non significant. This time, smooth-hounds recorded the only significant increase (356 times, CI: 6.8, 18534; ).
All these results were robust to mis-specification and uncertainty of trawl performance (in terms of swept area), to sampling and estimation error of sediment composition, and to the surveys' sample sizes (see Supplementary Methods
Our projected distribution of current fishing intensity () reflected the patterns in abundance and distribution detected in MEDITS (except for the upper Adriatic Sea, ), and was consistent with the spatial parameter estimates of the standardized catches (Supplementary Fig. S3
). The eastern Adriatic (mainly fished by Croatia) has a much lower level of fishing pressure than the western side (exploited by Italy). Italy records about twice the amount of otter trawlers (1541) than Croatia (855) with an average trawler having about 2.25 times the Horse Power (199 HP, sd = 149) of an average Croatian one (88.5 HP)32
. We predicted a higher fishing intensity all along the northwestern Adriatic, especially between Fano and Pescara, and around Chioggia (). While in Croatia, heavily fished areas would be the southeastern part of Istria, and above the Dalmatian channels between Sibenik and Split. Note that the distribution of fishing effort was coarser in Croatia because our source data was aggregated by major fishing districts, and we assumed that the allocation of fishing boats per port was proportional to the port population density (Supplementary Materials
). Dividing the number of boats fishing in these different Croatian sectors by the sectors' trawable surface (i.e. excluding 3 nautical miles from the shores) showed that most of the Croatian channel region would be free from trawling, which instead would take place mostly in the coastal areas (between 3 and 6 nautical miles) of Istria and the outer channel areas (Supplementary Fig. S4
). If the number of otter trawlers longer than 18 meters is a good index of offshore fishing intensity (between 3 and 40 nautical miles), fishing exploitation is an order of magnitude higher in Italian than Croatian waters (0.005 vs. 0.0003 boats per square kilometer, Supplementary Fig. S4
The intrinsic vulnerability of species did not explain the variability of the observed rates of change. We found no significant relationships between r and βys estimated from MEDITS (slope 0.996, p-value: 0.504, R2: 0.035) Jukic (slope = 3.499, p-value 0.152, R2: 0.27), historical comparisons in the Hvar area (slope = 14.020, p-value: 0.51, R2: 0.039), Jukic area between Hvar and Jukic (slope = 5.91, p-value: 0.75, R2: 0.013), Jukic area between Jukic and MEDITS (slope = 14.53, p-value = 0.45, R2: 0.072) and in the Zupanovic area (slope = 23.61, p-value: 0.18, R2: 0.21).