The commercial fisheries of the Canary Islands are composed of small-scale boats which fish within the EEZ of the Canaries, as well as industrial vessels operating mostly in the fishing grounds of NW Africa, hence making it difficult to identify what part of the whole fleet [i.e. number of boats, GRT and horsepower (hp)] was dedicated to the artisanal fishery in each period.

Nonetheless, grey literature has provided some data on the artisanal fleet, and existing data gaps were resolved by interpolating catch for the missing years. This resulted in 1390 artisanal boats in 1968, increasing to 1709 boats in 1983 (Fig. 2). From 1980 to 2010, data on the number of artisanal tuna boats were available (Delgado de Molina et al. 2012), and for 1950 to 1979 we assumed that the number of bait boats followed the same trend as that of total artisanal boats (Fig. 2). We believe that this estimate is appropriate because of the existence of tuna canneries since the 1920s and 1930s, indicating that a substantial and consistent tuna catch must have been available since then to operate. Large boats (GRT>50) normally fish further from the islands and hence focus primarily on tuna all year round. Small and medium-sized boats with GRT of less than 50 catch tuna near the shore or close to the islands in the summer season, and in the remaining seasons they focus on bentho-demersal species.

Data on the number of artisanal fishers in the Canary Islands were available by island for 1969, 1987, 1995, 1997-2002, 2005, 2008 and 2012. In 1950, the crew required to operate the same number of boats was 60% higher than in 1970, as motors had not yet been introduced and most of the boats at this time were row-boats. We interpolated data between years with missing data. The total amount of fishers has declined over time (Fig. 3), as has their relative representation in the resident population of the Canary Islands, from 1.74% of the resident population in 1950 to 0.10% by 2010.

Artisanal catch data were available for several key landing sites, starting with 1968, when data were available by island with varying levels of detail on the catch taxonomic composition (García-Cabrera 1970). Cumulatively, there were 3098 t of bentho-demersal species, 19900 t of pelagic species and 6850 t of species for which the taxonomic identification was unclear. After 1968, bentho-demersal catches were available for 1980, 1981, and 1982 from a compilation of various sources (Gafo-Fernández et al. 1984a, b, Barrera-Luján et al. 1983a). Given the rate of increase in catch from 1968 to the early 1980s, we estimated total catch of these species for all islands. According to a partial survey during eight months of 1982 (Barrera-Luján et al. 1983b, Delgado de Molina et al. 1983, La-Roche Brier et al. 1983), the catch of bentho-demersal species (including cephalopods, sharks and rays, and crustaceans) was only 893 t compared with the minimum estimate of 12995 t of catch. Even if we scale this to include the entire year, resulting in 1340 t of catch, this amount is still over seven times lower than the catch for only three of the seven islands (Fuerteventura, Gomera and Tenerife). Thus, we excluded this landing site, but we did use this information for other components of the present catch reconstruction.

Data for recent years were available from 1999 to 2004 (Canarian Government 2006). We depicted these data next to the reported data (Popescu and Ortega Gras 2013) from 2006-2010 (Table 1).

The fishing data available before 2006 were obtained from data recorded by two fishers of Gran Canaria (from 1971 to 2009) and La Palma (from 1975 to 2012), and commercial fish transactions between all the small-scale fleet of Mogán and a wholesale fishmonger (from 1989 onward). We also reviewed documents and grey literature, from which it was possible to obtain survey-based information regarding the fleet and gears used (González et al. 1988, 1991, García-Santamaría et al. 2001, among others), as well as data on the trends in the abundance of fishing resources from changes in the CPUE.

Since about 30 different gears exist, to create a comparable CPUE time series that would be ideal for comparison, we chose to model the development of the trap gear, as traps are used extensively all year, around all islands (except in El Hierro after 1994, Fuerteventura between 2000 and 2010, and La Gomera after 2015). The use of other gears, i.e. longline and gillnets, is only permitted during certain periods and in specific areas of the islands. Therefore, trap CPUE is a more homogenous index of changes in fish abundance.

From fishing research surveys (Barrera-Luján et al. 1982, 1983b, Pérez-Artiles et al. 1987, Caldentey-Morales et al. 1988, among others), the CPUE for fish traps targeting demersal species can be grouped into four well-defined periods: (i) a period of high abundance in the 1950s and early 1960s; (ii) a period of relatively lower CPUE in the late 1960s, according to García-Cabrera (1970), who indicated that fishing grounds shallower than 100 m depth were already overfished; (iii) a period of intermediate-low abundance during the 1980s; and (iv) a period of low abundance from the late 1990s to the present day. Such data enabled us to partially rebuild the temporal changes in CPUE over the time period from 1950 to 2010 (Fig. 4). Prior to 1969, we assumed that CPUE increased at half the rate after the late 1970s. A progressive 93.3% decrease in CPUE values has been observed over 60 years (1950-2010).

Data on tuna catch were collected by the Spanish Oceanography Institute (IEO) and reported to the ICCAT between 1970 and 2010 (Fig. 6). Prior to this, data on tuna catches were not reported, yet given the longstanding history of tuna fishing and indications from other sources (e.g. García-Cabrera 1970), it can be concluded that a substantial tuna fishery existed. Therefore, we reconstructed catch for the period 1950 to 1969 from various sources.

Catches reported by ICCAT for the bait boat tuna fishery from 1962 to 1969 were quite low, and since data by García-Cabrera (1970) indicate catches in 1968 at least three times as high (and at most nine times as high), we believe they are underestimated significantly. Additionally, while there was an upward trend in catches from 98 t in 1962 to 3298 t in 1969, we did not believe this trend was representative, as this simply represented an increase in reporting capacity rather than any significant changes in tuna fishery. Nonetheless, we utilized the ICCAT data to generally understand the relative change of catch from year to year while using the magnitude suggested by the 1968 data point as well as IEO data from the early 1970s.

For 1968, high tuna catches were reported. In La Gomera most of the 11000 t of catch was tuna, so we assumed 90%, resulting in 9900 t. In Tenerife, 10000 t of catch were reported to be large- and medium-sized pelagic fish. Without a clear indication we simplistically assumed 50% was large pelagic species (tunas) and the rest were other pelagic species. Catches from La Palma, Gran Canaria and Lanzarote totalled 6850 t, with no indication as to what species were caught. While most of them was assumed to be bentho-demersal species, the remaining 822 t of catch were pelagic species and we again assumed a 50% split between tuna-like and other pelagic species. This resulted in an estimate of 15311 t of tuna catch in 1968.

Prior to 1968 we assumed that catch was approximately the same as the average catch from 1968 and 1970-1972, as effort in terms of the number of boats was constant for this time period. Hence, from 1950-1961 we assumed catches averaged the catch of 1968, 1970, 1971 and 1972. Starting in 1961, there are data on catch at 98 t, increasing to 491 t in 1963 and declining to 144 t in 1964. Finally, we maintained catch at the average used for the 1950-1961 and the 1965-1967 periods. For 1969 an average was taken between the 1968 and 1970 data.

Pole and line fishing requires a substantial amount of bait to lure tuna and tuna-like species. The most common species used for bait are the Atlantic chub mackerel (Scomber colias), followed by the European pilchard (Sardina pilchardus). Other common species used as bait were bogue (Boops boops), longspine spinefish (Macroramphosus scolopax), European anchovy (Engraulis encrasicolus), sand smelt (Atherina presbyter), jack and horse mackerels (Trachurus spp.) and sardinellas (Sardinella spp.) (Canarian Government 2006, Herrera-Perdomo 2017).

Logbook data from the IEO suggest that a medium- to large-sized bait boat uses about 2300 to 2500 kg of live bait per month, or on average 28.8 t per year. Since bait boats range from 1 to 200 GRT, we assumed that medium to large bait boats were boats of more than 50 GRT and that small bait boats were those of less than 50 GRT. Herrera-Perdomo (2017) estimated that 626.4 t of live bait was used by the tuna fleet during the period 2014 to 2016. We made the conservative assumption that bait boats of less than 50 GRT used half the amount of bait used by medium to large boats, averaging 14.4 t of bait annually.

Landing sites of catch of pelagic species were available for 1968, 1982 and from 2000 to 2010. We excluded using the landing site for 1982, which reported an overall catch of 4644.6 t of pelagic species (less tuna) in 8 months, due to the severe underestimation of bentho-demersal species compared with other data. Likewise, we excluded the data from 1999-2004 (Canarian Government 2006), which also included captures in the African fishing grounds. Thus, we interpolated the catch between the 1968 landing site of 5411 t and the 2008 landing site of 1222 t, thereafter following the trend of official reported data (Fig. 7).

A total of 16247 fishing licences, i.e. private individuals and charter boat captains, were issued in 2005, each one valid for three years (MAPyA 2006). This means that in 2005 the number of active anglers was approximately three time the number of professional fishing licences issued, ca. 49000 active anglers. In 2007, the number of valid fishing licences grew to approximately 60000, then to 100000 in 2009 (Pascual-Fernandez and de la Cruz Modino 2011), and peaked at 110000 in 2011 (MAGRAMA 2013). Recently, around 90000 licences have been active yearly (http://www.gobcan.es).

Additionally, Jiménez-Alvarado (2016) reported that approximately 10% of individual anglers fish without licences, which is equivalent to unreported fishing licences at 11.1% of reported ones. Ultimately, we created separate time series for recreational anglers and charter vessels, which would have different catch rates because charter vessels take many passengers at a time and fish all year round.

Of the recreational fishing boats registered in the Canary Islands in 2005, MAPyA (2006) indicated that 827 were under the “sixth list”, or recreational vessels which are for-profit, and 22619 vessels were for personal recreational fishing (“seventh list”). Hence, we assumed that of the 48741 active anglers in 2005, 827 were charter licences for taking tourists, and the rest were generally for residents.

We extended this division over time by creating a proxy variable: the number of charter licences divided by the tourist population, which in 2005 was 0.009%. We assumed that this ratio would be 0% from 1950 to 1959 when tourism had not yet expanded, interpolated to 0.009% in 2005 and then continued the linear trend to 2010. This time series of the proxy ratio was multiplied by the tourist population from 1950 to 2010 to obtain an estimated time series of the number of charter boats in operation.

Since we have data on the total number of active licences from 2005 to 2010, we subtracted the number of charter licences to obtain the number of recreational licences. For the years prior to 2005, we utilized a similar strategy as that for charter licences, obtaining a ratio of resident licences in 2005 to resident population in 2005 at 2.7%. Since 1950 to 1959 was a time of food shortage (Palmero 2001), indicating that subsistence fishing was more likely than recreational fishing among residents, we interpolated between 0% from 1950-1959 to the 2005 value. Also, the development of the touristic industry in the 1960s facilitated a better economic position for the local population and, thus, investment in recreational fishing equipment. This ratio from 1950 to 2004 was multiplied by the resident population to generate a complete time series of reported licences. Finally, we adjusted this time series to account for the 10% of recreational anglers fish without a fishing licences by multiplying the reported resident licences (not charter) by 11.1%.

From comparing boat capacity on recreational charters, it appears that on average, each charter takes between 3 and 4 people at a time. We assumed a very conservative number of trips at two per year per charter, resulting in each charter licence accommodating the equivalent of seven private licence holders. It is likely that fishing is far more common among charters, who fish the whole year, but we assume this conservative estimate.

From interviews with anglers, MAPyA (2006) reported a catch rate of 0.085 t fisher^–1 trip^–1 when fishing from a boat and 0.0085 t fisher^–1 trip^–1 when fishing from the shore. We weighted these rates by the number of fishers in each category to obtain one representative rate of 0.0099 t fisher^–1 trip^–1. Furthermore, the average number of trips taken was 43 trips annually (MAPyA 2006), so we adjusted this rate to obtain a total catch rate of 0.425 t fisher^–1 trip^–1 in 2005.

While this catch rate is appropriate for 2005, we varied catch rates over time using some simple assumptions about changes in the CPUE and technological improvements. These changes came first for artisanal fishers and then recreational fishers, and we assumed a five-year lag. Throughout the 1970s, most of the artisanal fleet became equipped with onboard engines and hydraulic fishing winches, and in the 1980s and 1990s other technological improvements such as radio, GPS, synthetic nets and echo sounders were also incorporated. We assumed a constant level of technology until 1975, thereafter increasing to 100% in 2005, and then remaining constant (Fig. 8A). CPUE was modelled in Figure 8B, where the 2005 value was also normalized to 100%. The merging of the two trends created a variable trend line of the catch rate before and after the 2005 catch rate (Fig. 8C).

The time series of catch rate was adjusted accordingly, and this time series was multiplied by the total amount of recreational anglers, counting seven anglers per charter licence to obtain the entire time series of recreational catches. For the species composition, we used the percentage of fishers who target certain species as a representative sample of catch (MAPyA 2006).