Is Overfishing the Main or Only Factor in Fishery Resource Decline? The Case of The Magdalena River Fishery and Its Correlation with Anthropic Pressures

Overfishing has been historically considered as the main cause of fish stock depletion worldwide. This paradigm has oriented fishery management towards a classical approach, under which externalities to fisheries were not considered as they were difficult to assess and measure. The aim of this study is to describe the dynamics of different environmental, economic, and demographic variables (water flow, forest cover, gold production, population growth, stored water volume, and sediments) in relation to the behavior of the fishery production in the Magdalena-Cauca river basin from 1980 to 2015. Generalized Additive Models were used to determine the variables that best explain fishery production. The findings confirmed that environmental deterioration of the Magdalena River basin explained at least 60% of the reduction in fishery production. Thus, we concluded that the traditional approach of making fishers responsible for the decline of fish production was a misguided argument, and before implementing restrictions on fishing activity, a better understanding of the overall system is crucial. Hence, fishery management should involve the economic and social sectors that affect the offer of ecosystem services within the basin, including fishing.

. These values were estimated for the area above and below 1,200 m 171 a.s.l., as this altitude is considered as the limit for migratory fish distribution. To 172 analyze mercury concentration associated with gold mining, we considered the total 173 annual gold production in the department of Antioquia in kg·year -1 since more than 60% With the collected and estimated information, we built a 36 x 9 matrix (years x 208 variables) (S1 Text). We first scanned the data to establish the distribution pattern of the 209 variables (with or without normal distribution). Datamining established non-parametric 210 models as a path, so we decided to use the Generalized Additive Models (GAM) of 223 Thus, with the response variable (fish production) and the different predictor variables, 224 we created 203 combinations allowing us to run 35 models (S1 File). We selected the 225 model that best explained fish production behavior in the Magdalena River basin, based 226 on the deviance and the AIC values. For precision purposes and considering the variable 227 stored water volume, the five models that simultaneously crossed "water volume stored 228 below 1200 m a.s.l., and total stored water volume for the entire basin" were discarded, 229 as the second includes the first. The correlation between fisheries production and the differences in the 251 maximum amplitude of water flows showed a coupling between both variables (Fig 3).
252 In general terms, the fishery production curve presented the same shape and pattern as 253 the curve generated by the moving averages; that result suggests a strong influence of 254 water flow regime on fish abundance in the system (Fig 3). In 1992, after significant 281 to the present day (Fig 4b and c). These increases are related to the progressive 282 implementation and operation of the main dams in the basin. In turn, the first main 283 increase in the total water volume stored and the one below 1,200 m a.s.l. occurred in 284 1986 and concurs with the reduction in fish production, suggesting a change in the 285 environmental dynamics of the river that prevented fish populations from recovering 286 their previous abundance (Fig 2 and 4c).

288
Gold production showed a rapid increase between 1980 and 1986, followed by 289 a steady decrease until 1994, before showing a continuous but fluctuating upward trend 290 (Fig 4d). It should be noted that fish production decreased right after gold production 291 reached its highest value (Fig 2), suggesting a slightly delayed effect of mercury 292 contamination due to gold mining on fish stocks.

294
The reduction of the forest area was continuous and progressive during the 295 entire study period (Fig 4e). The same can be observed for demographic growth (Fig   296 4f).

298
Concerning the sediment fluctuation over time, the change in the magnitude of 299 the maximum values after 1986 is noteworthy (Fig 4g). From that year onwards, the 300 maximum values are approximately 30% higher than in previous years, corresponding 301 also with the moment when the first dam started operations and with the decrease in 302 fishing production (Fig 2). 303 304 Figure 5 presents the functional correlations between fish production and the 305 predictor variables. Of the 30 models that were selected, 27 reported synergy between 306 the different variables and presented an explained deviance of more than 60% on fish 307 production behavior (Table 1)  The environmental variable that best forecasted the fishery production was 323 forest cover, meanwhile at a productive level, it was the total volume of stored water, 324 and, at a demographic level, it was population growth (p < 0.05, higher explained 325 deviance and lower AIC) ( Table 2). Fishing production was high since there was a large 326 forest cover, low total water stored volume levels, and a low population in the basin.   In terms of demographic growth, we confirmed the hypothesis that the higher 420 the population growth, the lower the fish production; therefore, it can be used as a proxy 421 for the alteration of water quality in the basin. This is because urbanization and human 422 activities lead to large urban wastewater discharges into the rivers, which affect the 423 water quality and life of aquatic organisms. Fishing production in the basin had its 424 largest records in the 80s when less than 26 million people were living in the basin; 425 however, currently, the basin has about 6 million more inhabitants.  . We agree that a basin-wide approach 464 is necessary, including cumulative effects and also climate variability, which merits 465 immediate and coordinated intervention within the framework of strengthening inter-466 sectoral governance of fisheries. It is clear that the decline in ecosystem services and the 467 associated severe socio-economic and environmental impacts will be increasingly 468 challenging to reverse or mitigate these, affecting thousands of coastal inhabitants 469 whose livelihoods depend or not on fisheries.

471
To summarize, the results obtained allowed us to conclude that the decrease in 472 fish abundance was in large proportion due to environmental causes. We consider that 473 fishing activity and landings responded more to the environmental state of the 474 ecosystems than to any sort of approach in fisheries management. We consider that 475 fishers in recent years have self-regulated towards new levels of abundance and that 476 fishery authorities should be more supportive towards good fishing practices that the 477 fishers have adopted for their survival, as a result of the reality they perceive every day.
478 Surely, making fishers responsible for the decrease in fish production is a misguided 479 argument, and, before trying to implement restrictions on fishing activity, a better 480 understanding of the entire system and its dynamics is necessary. Recently, different 481 approaches have been discussed, like the concept of balanced exploitation (46), which 482 considers that fishing pressure should be distributed in proportion to the natural 483 productivity of ecosystems, forcing, in our case, a response to environmental dilemmas. 484 We consider that the system has already been adjusted to a lower level and found a new 485 balance. Therefore, the implementation of classical fisheries management based on the 486 overfishing paradigm is no longer sustainable, and managers of artisanal fisheries can 487 no longer avoid external factors. Moreover, fisheries management must involve the