Stock dynamics
In haddock, the abundance of all age groups fell steadily in the population between 1928 and 1939. The gradients of this decline, however, get flatter as age increases (Fig. 2). Between 1939 and 1945, all age groups, except the age 0s (recruits), experienced a step-wise increase in abundance. In the younger fish (ages 2 and 3), this increase continued until the end of the series in 1958. Abundances of older fish (e.g. ages 4 to 8) dipped between 1947 and 1950, after they again began to increase. It is interesting to note the similarities in time trend between the trawl survey numbers per unit effort (Fig. 2) and the commercial landings per unit effort (Fig. 3) data where the older fish from the trawl survey mirror the U-shaped pattern seen in the commercial data. In cod (see S5), there was a peak in abundance in 1935 followed by a dip into the war period and a strong recovery between 1939 and 1945. After the war when fishing recommenced, abundances fell just like haddock with the trough occurring in the early 1950s. Unlike haddock, recruitment (age 0s) in cod also increased during the war, although this appears to have been part of an overall long-term decline (1928–1958). The commercial data for cod show rising landings per unit effort between 1928 and 1938, an abrupt increase in 1947, followed by a fall and then recovery (see Fig. 3). The trawl survey data for whiting (see S6) have parallel time trends for ages 3–9: there is a peak in the early 1930s followed by abrupt increases between 1939 and 1945. After the war, increased fishing effort again caused a fall, followed by rising abundance into the late 1950s. Recruitment in whiting (age 0s) also appeared to ‘dip’ into the early part of the war period and increased thereafter. The commercial data (see Fig. 3) suggest overall rising abundance between 1928 and 1958.
Consider the cohort trajectories (1928 to 1958) from the modelled indices from Buchan for the three gadoids plotted in Fig. 4a–c. The change in gradients between the pre, post and WWII periods has been extracted from linear models fitted to each cohort and plotted in Fig. 5. The gradients of these lines represent a measure of total mortality (Z). A component of Z is due to mortality by fishing (F), and another unknown component is due to mortality by predation and disease (natural mortality, M). Estimates of Z (see Fig. 5) peaked for all species during the period immediately before WWII, mirroring the change in fishing effort (see Fig. 3). The effect of the cessation of fishing during WWII was dramatic, with sharp increases in the rate of decline in total mortality (Z) being observed in all three of the gadoids examined. Lowest mortalities were observed in 1942 (haddock), 1944 (cod) and 1945 (whiting), after Z began to rise again as fishing recommenced in the post-war period. Mortality rates never again reached the high levels observed in the late 1920s and 1930s, presumably since similar levels of effort (300,000 h) were not attained in the aftermath of WWII because much of the fleet capacity and its manpower had been negatively affected by the general war effort. In Fig. 6, the indices for haddock have been re-scaled for each age category so that all previous and subsequent observations in each time series reflect change relative to 1939, i.e. to the start of hostilities and to the cessation of commercial fishing.
The Mexican wave in the haddock population suggested by first principles is clear. The youngest fish (age 0s) responded negatively, levels falling by 83% between 1939 and 1945, whereas the older age categories all responded positively. The mean level of age 10 haddock in the population rocketed, increasing nearly 12 times between 1939 and 1945. Increases by the younger age groups were found to be arranged in an ordered sequence between these two extremes; for example, age 8s increased by 747%, age 6s by 579%, age 4s by 427% and age 2s by 170%. It should be noted that this wave is seen twice but travelled in the reverse direction after 1945 when levels of older fish decreased further and fastest when fishing recommenced in the post-war period (Fig. 6). A similar phenomenon was observed in both cod, G. morhua, and whiting, M. merlangus, although the numbers of younger fish in the population also rose in those species.
Both mean length and mean biomass have been suggested as useful indicators for assessing conservation and fisheries-related effects of MPAs (Pelletier et al. 2008), and this study supports that view. The mean length of haddock in the trawl survey sampled Buchan population was 24.8 cm in 1939 and 28.7 cm in 1945, while the catch per unit effort (in weight) of haddock rose from 11 to 59 kg h−1 of fishing. Similarly, cod average length in the trawl catches rose from 45.8 cm just before the war to 51.2 cm just after, while average whiting lengths increased from 26.5 to 27.8 cm. The catch per unit effort of cod increased nearly ten times from 4 kg h−1 in 1939 to 36 kg h−1 in 1945. Changes in whiting catch per unit effort were more modest, trebling from 5 kg h−1 in 1939 to 16 kg h−1 in 1945. Substantial increases in the catch per unit effort were also seen in the commercial data.
Environmental effects on stock dynamics
We hypothesised that the changes observed in gadoid abundance in Buchan, between 1939 and 1945, might plausibly have been caused by environmental factors such as changes in temperature or the volume of Atlantic inflow rather than by fishing. Above, we examined how the entire age structure of each of the three most important gadoids changed, between 1928 and 1958, using population abundance indices from trawl surveys. To examine the possible impact of changes due to the physical environment as opposed to those due to commercial fishing, we opted to simplify the problem by examining only ‘recruits’ (age 0s) and spawing stock biomass (SSB) (all the fish thought to be old enough to reproduce estimated using maturity-at-length data), comparing them to a range of environmental covariate data we were able to assemble.
Tables 1, 2 and 3 show pair-wise correlations (1928–1958) between two indices (i.e. R_age0 = numbers of age 0 fish which is also known as ‘recruitment’ and SSB = numbers of sexually mature fish) representing changes in the haddock, cod and whiting populations and some potentially explanatory oceanographic variables: temperature, salinity and the North Atlantic Oscillation Index [Note: some sea surface temperature data are available for the war time period]. The haddock recruits (Table 1) were positively correlated with SSB (r = 0.35) and negatively correlated with temperature in quarter 3 (r = −0.43). There was also a negative relationship between haddock SSB and salinity in quarter 3 (r = −0.36). Cod recruitment (Table 2) was negatively correlated (−0.41) with salinity in quarter 3, whereas SSB was not related to any of the variables we examined. Recruitment in whiting, on the other hand, was negatively related to salinity in quarter 1 (−0.46), while SSB was positively related to sea surface temperature in quarter 1 (Table 3). It should be noted here that there are also close connections between the oceanographic variables themselves: salinity and temperature being positively related.
Table 1 Haddock: correlation coefficients between recruits, spawning stock biomass (SSB) the year before, sea surface temperature (SST), salinity (SAL) and the North Atlantic Oscillation index (NAO)
Table 2 Cod: correlation coefficients between recruits, spawning stock biomass (SSB) the year before, sea surface temperature (SST), salinity (SAL) and the North Atlantic Oscillation index (NAO)
Table 3 Whiting: correlation coefficients between recruits, spawing stock biomass (SSB) the year before, sea surface temperature (SST), salinity (SAL) and the North Atlantic Oscillation index (NAO)
These data have been plotted as time series in Fig. 7. Recruitment fell steadily between 1928 and 1939 in haddock, cod and whiting. In the case of haddock, recruitment fell further during the war and then rose again between 1945 and 1958. In both cod and whiting, recruitment rose during the war and in cod continued to rise thereafter (Fig. 7). Levels of recruitment in whiting, however, remained fairly stable after the war.
SSB in all three gadoids fell between 1928 and 1939. During the war, SSB levels of all three species rose. Immediately after WWII, spawner abundance dipped again, presumably when fishing recommenced, after spawner abundance rose again gradually.
In order to investigate these relationships in more detail, we extended the simple pair-wise correlation analysis (e.g. Tables 1, 2 and 3) by using a combination of multiple linear regression models on log-transformed data, together with traditional nonlinear stock–recruitment models (Ricker 1954). In both cod and whiting, spawning stock biomass was not clearly related to any of the environmental data: the best predictor being the SSB the year before (see Tables 2 and 3). In haddock, there is an indication (r = −0.36, Table 1) that SSB might be negatively related to salinity during quarter 3 (SAL Q3). When this was tested with an analysis of variance (Table 4), however, it was not found to be statistically significant (p = 0.1).
Table 4 Haddock: analysis of variance summarising the relationship between spawners and salinity in Buchan during quarter 3
An examination of environmental factors (e.g. temperature, North Atlantic Oscillation Index) thus failed to convincingly explain the abrupt changes in either the abundance of sexually mature fish or their changing age structure, and we are confident that they can be attributed to the virtual absence of fishing mortality.
Stock–recruitment relationships
It is usually assumed that recruitment (R_age0) is directly related either to SSB in year t or to SSB the year before (SSBt − 1). To explore these relationships, we fitted R_age0 to the abundance of spawners, together with a range of environmental variables, using linear models and compared the output using nested analysis of variance tests. In haddock, spawner abundance the year before (SSBt − 1) was quickly eliminated, recruitment (R_age0) being more strongly related to SSB in the same year. S7 shows that both SSB and temperature during quarter 3 (given that SSB is in the model) are related significantly to recruitment. The coefficients from model 3 are the following: intercept = 4.2, SSB = 0.5 and SST Q3 = −0.64. Hence, rising late summer temperatures will result in falling recruitment. Between 1939 and 1945, average quarter 3 temperatures rose (see also Fig. 7), indicating that temperature could have caused the fall in recruitment that we see during the war. We also fitted a Ricker stock–recruitment curve to these data, adding quarter 3 temperature data as a covariate, according to the methods outlined in the FLR package (Kell et al. 2007). In this formulation, temperature in quarter 3 was not significant.
In the case of cod, neither SSB nor SSBt − 1 were useful for predicting cod recruitment in Buchan between 1928 and 1958 (see S8). Salinity in quarter 3 had a negative effect (p = 0.06) on cod recruitment, when SSB was also included in the model. Quarter 3 salinity fell between 1939 and 1946 (Fig. 7) and could, therefore, have been a factor in the increasing recruitment we observed during the war.
The patterns observed in haddock and cod were similar for whiting. Again, neither SSB nor SSB the year before could explain the pattern in whiting recruitment in Buchan between 1928 and 1958. Salinity in quarter 1 had a positive effect when SSB was also in the model (see S9), and so salinity in early spring might, therefore, have contributed to the increasing recruitment we observed during the war, since it fell quite sharply between 1938 and 1948 (Fig. 7), although the exact mechanism cannot be deduced from this study.