Study area

The Galápagos Islands lie in the Eastern Tropical Pacific Ocean, approximately 970 km away from the continental coast of Ecuador. The archipelago comprises 19 major islands and more than 200 islets and rocks [35]. The oceanographic conditions around the Galápagos can be divided into two seasons: a cold season with temperatures between 18–20°C (from June to November), and a warm season with temperatures > 25°C (from December to May) [36, 37]. To evaluate the spatial distribution of fishing effort by zone, we used the marine ecoregions described by Spalding, Fox [38]: The Northern Galápagos Islands Ecoregion (northern ecoregion), the Eastern Galápagos Islands Ecoregion (eastern ecoregion) and the Western Galápagos Islands Ecoregion (western ecoregion) (Fig 1 and Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Study area location.

Ecoregions on the map are based on Spalding, Fox [38]. The GMR includes the entire marine area within a 40 nautical mile (76 km) strip, measured from the archipelago baseline (dashed line) [39]. The MPA shown in the map are those created until 2021.

https://doi.org/10.1371/journal.pone.0282374.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of Galápagos Islands Ecoregions based on Spalding, Fox [38].

Oceanographic characteristics taken from: Dirección del Parque Nacional Galápagos [40], Liu, Xie [37] and Sullivan and Bustamante [41]. Area estimated in km² using R Core Team [42].

https://doi.org/10.1371/journal.pone.0282374.t001

Data extraction

We downloaded daily fishing effort information from the Global Fishing Watch database version v20201001 [43]. Global Fishing Watch analyzes more than 50 million AIS messages daily and identifies fishing vessels and their fishing activities using two machine learning algorithms for image classification called convolutional neural network (for more information see supplemental materials of Kroodsma, Mayorga [44]). The first model analyzes all vessels AIS positions and predicts the vessel type (fishing or non-fishing vessels) incorporating a set of official registries of vessels with information on vessel identity and characteristics (length, engine power, tonnage). The current model is relatively accurate to identify fishing vessels from non-fishing vessels (99%) [44]. For fishing vessels, gear type is assigned when the information in the official registry matches the classification of the model [12]. Among the fishing gears identified by the model are drifting longlines (LONG), pole and line (POLE), set gillnets (GILL), set longlines (SETL), squid jiggers (SQUID), other purse-seines (PURS), tuna purse-seines (TUNA) and unknown (UNKN). Set gillnets and set longlines corresponds to gear fixed to the seabed [45]. “Unknown” gear is a generic category assigned to those fishing vessels for which the type of fishing gear used cannot be clearly determined.

The second model identifies the apparent activity of a vessel in a specific location based on the speed of the vessel, the change of direction within a defined area, spatio-temporal patterns of movement, and duration of the fishing event. The accuracy scores vary depending on the fishing gear, but in general range above 90% [44]. To convert the apparent fishing activities into fishing hours, an amount of time is assigned to each AIS detection (which is assigned half the time to the previous and next AIS position) and all positions in each grid are summed [12]. The final product of this model identifies the location and duration of fishing activity, which can be aggregated per grid cell (hereafter “hours”), flag state and gear type and summarized by day. The data used in this study was aggregated on a grid with 0.01 decimal degrees (approx. 1.23 km²) resolution [43].

For this investigation, we filtered the data downloaded from Global Fishing Watch to obtain only information of fishing vessels within the IEEZ and GMR areas from January 2012 to December 2021. To avoid confusion with the legal artisanal fisheries that can operate in Galápagos, we filtered the Ecuadorian-flagged vessels and used information from Ecuador’s National Public Registry of Vessels to identify which vessels were industrial and artisanal [46]. For those vessels that were not in the public registry, we reviewed their range of movement within the Global Fishing Watch iterative map [47]. Those vessels that limited their movements within the GMR were categorized as artisanal vessels. Finally, we categorized data with missing flag information as ‘Unknown flag’ regardless most of its fishing effort was allocated within the GMR (S1 Fig). Since the lack of a Maritime Mobile Service Identity (MMSI) number prevented assigning it to any of the local, national or foreign fleets, we only used this data for comparative purposes within the descriptive statistical section and eliminated it from further analysis.

Data analysis

Due to the annual increase of the number of fishing vessels transmitting AIS and the number of satellites receiving the signals [12] temporal coverage of the Global Fishing Watch data varies over the years, we aggregated date over the years to reduce the bias. The dist2Line function of the ‘geosphere’ package [48], available in the R computer software version 2022.02.0 [42], was used to determine the boats’ entry distance into the study area. This function calculates the shortest perpendicular distance between points and polygons and generates a matrix with the distance and longitude/latitude of the point closest to the polygon (S2 Fig) [48]. We calculated the entry distance to the IEEZ from each AIS detections to the edge of the IEEZ polygon, while we calculated the entry distance to the GMR from AIS detections to the edge of the GMR polygon. Boxplots were used to compare the foreign and Ecuadorian vessels’ entry distance per flag state, gear type, seasons and ecoregions.

A utilization distribution model using Kernel Density Estimators (Kernel UD) was used to assess the core effort areas and spatial extent of fishing activities by vessel type (foreign and national), fishing gear and seasons. Kernel UD analysis defines the minimum area in which a point has the specific probability of being located according to its coordinates [49]. We used the kernelUD function available in the ‘adehabitatHR’ package of the R software version 2022.02.0 [42, 50], using the reference bandwidth, the extent of the IEEZ (longitude: - 85.90 to—95.34 and latitude: - 4.73 to 4.99) and a grid size constructed with a resolution of 0.01 degrees. We used 50% Kernel UD to estimate core-use areas and 95% Kernel UD to estimate the spatial extent of fishing. Two graphs were made for each variable analyzed: one integrating fishing data from all the known foreign fishing fleets, and another just for the Ecuadorian fleet.

Fishing effort by ecoregion

Our results highlighted that the eastern ecoregion exhibited the highest amount of fishing hours, with vessels operating more evenly distributed within the IEEZ yet clustered within the GMR. Differently, the western and northern ecoregions exhibited constant clustering in two areas: purse-seines preferring the western ecoregion near the border next to Isabela and Fernandina Islands, and drifting longlines the northern ecoregion in the area north to Darwin Island. Hu, Harrison [52] reported these two fishing fleets target the same tuna species (T. albacares, T. obesus and K. pelamis), yet there are differences in the habitat preferences of younger tunas (e.g. limited O₂ concentration, shallow mixed density layer, and high productivity conditions) targeted by purse-seiners compared to those of older tunas (e.g. wide ranging O₂ concentration, deeper mixed density layer, and low productivity conditions) generally targeted by drifting longlines. The GMR lies in the confluence of several surface and underwater marine currents whose interactions boost the overall biological productivity around the islands, particularly in the west of the reserve where the upwelling of nutrient-rich cold waters occurs [53]. These conditions create an area of relatively stable species-habitat relationship that favors the overall diversity of the area [54], which is taken advantage by foreign and Ecuadorean purse-seine fleets targeting tuna, billfish and swordfish [21, 22]. Notwithstanding, the Galápagos upwelling is known to expand towards the west during cold conditions but to concentrate near the west coast of Isabela and Fernandina Islands during the warm season or El Niño events [53, 55]. Since juvenile and subadult bigeye (T. obesus) and skipjack tunas (K. pelamis) targeted by purse-seiners are reported to prefer waters of high primary productivity [52], the strength and extent of the Galápagos upwelling is likely the underlining reason on the seasonal shift in fishing effort allocation by fleets targeting those species.

The northern ecoregion represents a unique "hotspot" for sharks [56], particularly over the seamounts found along the Cocos Ridge [57]. While previous analysis of AIS data did not depict this ecoregion as of high fishing interest [34], Martínez-Ortiz, Aires-da-Silva [30] denoted the area is an important fishing ground for the pelagic thresher shark (A. pelagicus) and blue marlin (M. nigricans). This is also supported by Cambra, Lara-Lizardi [57], who reported the presence of the scalloped hammerhead shark (Sphyrna lewini) and pelagic thresher shark aggregating over seamounts found along the Cocos Ridge. The high abundance of sharks over this region could be drawing the interest of the Ecuadorian and foreign drifting longline fleets during the cold season to fish on these species, as reported by Martínez-Ortiz, Aires-da-Silva [30]. Based on these findings, it is suggested the observed fishing patterns around the GMR from AIS data are tied to the seasonality of the oceanographic setting linked to the resource been targeted. This research did not use environmental data (e.g., biological productivity, sea surface temperature, currents), therefore we recommend additional studies to deepen the understanding of the fishing effort allocation in relation to these variables around the GMR by pairing AIS data with other sources of fishing effort and catch composition information, particularly to assess threatened and protected species. This type of information can also be used to make predictions of fishing activities based on environmental data to facilitate the monitoring and surveillance by enforcement authorities.

Fishing by foreign flags

The observed fishing patterns by flag and gear type reported by our study align with previous work using AIS data (2012–2016) that examined fishing activity in Ecuadorean waters, yet found Colombia as the leading foreign flag operating within the IEEZ [34]. These differences may respond to the use of a bigger dataset processed with refined algorithms into our analysis, an increased number of fishing permits allocated to other countries (such as Panama), and, in the case of China, the steady increase in operations around Ecuador’s IEEZ since 2016 [58]. Our study also found the IEEZ and the GMR were fished by 13 foreign flags, of which seven do not belong to the coastal states of the Eastern Tropical Pacific Ocean. During 2012–2018, AIS data revealed that Eastern Tropical Pacific fishing effort was dominated mostly by Chinese and Taiwanese flag squid jiggers (64%), followed by tuna purse-seiners (17%) and drifting longlines (15%), both flagged to a wider range of countries [59]. Similar results have been reported in other EEZs, where distant-water fleets dominate fishing in these areas and often approach protected areas [60, 61]. The presence of distant-water fleets in the IEEZ may be due to Ecuador’s fishing agreements and permits with these countries. The United Nations Convention on the Law of the Sea (UNCLOS) gives nations exclusive rights to their open waters EEZ, meaning foreign vessel can operate within these under specific agreements through their State Departments [62]. During this research, a list of foreign vessels approved to fish in the IEEZ under this framework, and for what fisheries and time periods, could not be accessed. Care is thus advised while assessing this data to infer into potential illegal practices, since it is impossible to filter out the names and registry of the foreign vessels legally operating within Ecuador. This caution is also relevant to the observed fishing pressure from Chinese and Portuguese vessels in the limits of the IEEZ. While studies have reported Chinese vessels could be involved in illegal entries within the IEEZ [58, 63, 64], our method of analysis gridded the spatial information and used the centroid of the calculated grids to estimate fishing activity per cells. This could produce some records to be depicted as within the IEEZ, particularly for Portugal since it only recorded one fishing record with 2.4 fishing hours. A further in-depth analysis is highly recommended to (1) differentiate the foreign vessels with and without fishing permits, and (2) to assess the detailed track of each vessel without permits for understanding their patterns and level of incursion within the IEEZ. This complementary analysis and the subsequent adoption of cost-effective methods to tackle illegal fishing (e.g., increase monitoring patrols and radars in areas where fishing effort is higher) could help Ecuador regain confidence from the European Commission, who categorized Ecuador as a fully non-cooperative country in the fight against IUU fishing with a yellow card in 2019 [65].

Fishing effort within GMR

A different scenario is observed for foreign vessels detected by Global Fishing Watch algorithms fishing within the GMR. For this MPA, only local (to Galápagos) artisanal fishers are allowed to fish and under specific gear type and fishing power characteristics [66]. Yet, since the creation of the GMR in 1998, there have been multiple of seizures of boats from Ecuador (mainland), Costa Rica, Colombia and Japan illegally operating within the marine reserve [20, 67–70]. Testimonies from Galápagos fishermen corroborate that unauthorized fishing by industrial vessels is recurrent within 40 nautical miles of this MPA [71]. Our study flag suspicious fishing activities from Colombia, Japan, Nicaragua, Panama, USA, Venezuela, Vanuatu, and ‘Unknown flags’ within the GMR. It is noteworthy that most of these records are in the southcentral region at an entry distance of nearly 100 km in the reserve. Out of these, Vanuatu and Japan are the only ones with a few fishing events registered close to port areas of the GMR (1 and 2 events, respectively), and not for the IEEZ. Since it is unlikely these vessels were fishing right in front of the main navy and park ranger offices, a potential misidentification by the algorithm could be the reason why those vessels were flagged as fishing.

Records from other vessels of known foreign flags represent a different scenario since most of them were registered operating near the western border or around islands but far from port areas. The GMR waters host important nursery and aggregation areas for commercial fish and protected shark species [72], making both the IEEZ and the GMR susceptible to infringements on fishery resources by national and foreign industrial fleets [69]. There is a possibility that vessels were found only in innocent passage, although the density of vessels transiting around the waters of the GMR is not as high as in other areas of the Eastern Tropical Pacific Ocean [73]. Either by fishing or transiting, the observed entries by our study represent suspicious activities within the reserve. A detailed analysis of the full AIS track data coupled with declared entry details to port authorities is highly recommended to detect common vessel behaviors that could lead to illegal fishing events within the reserve boundaries. This information could aid managers and authorities in developing tailored management and cost-effective surveillance plans for foreign and Ecuadorian industrial fishing boats entering the reserve under alleged health, mechanical or other emergencies [74].

Local artisanal fleet

The concentration of fishing activities within the GMR observed by ‘Unknown flags’, or by Ecuadorean vessels using gears including other purse-seines, set gillnets, set longlines and unknown fishing gears, are attributable to the local (to Galápagos) artisanal vessels that are legally operating within the reserve. The spatial distribution of fishing effort of Galápagos artisanal boats is influenced by several geographical (e.g., latitude, longitude, homeport), oceanographic (e.g., El Niño-Southern Oscillation), social and economic (e.g., target species, vessel type, revenues) variables [75]. Fishing grounds of these vessels are found all over the archipelago, but the most frequently fished sites are close to the fishing ports [24, 75]. The Galápagos fleet is required by law to use AIS, even in the case of small fiberglass boats, and are authorized to fish with trolling, deep sea and mid-water hook and lines, coastal gillnets, and Hawaiian slings [23]. Under these conditions, it is possible the assigned fishing gears and unknown flags recorded within the GMR do not represent potential illegal fishing events but errors by the algorithm when assigning a fishing gear [12], particularly since the used classification does not represent all the possible fishing gears used within the GMR [76]. To avoid any confusion, records from unknown flags were removed from data analysis and were not displayed in the results section. However, the presence of a high number of ‘Unknown flag’ vessels detections (> 2,400) shows the existence of large gaps in vessels monitoring, which may represent a risk to the GMR because these vessels have often associated with IUU fishing practices [77]. Therefore, it is necessary to assess the extent of ‘Unknown flag’ fishing activities by employing strategies such as detailed monitoring of their tracks and port arrivals.

Limitations of AIS data

Our study demonstrates the value of AIS data to analyze spatial and seasonal patterns of global fisheries in and around an MPA. Nevertheless, since this data does not provide information on fishing hours and the fishing gear used in itself [7], machine learning algorithms calculate a probability that a vessel is fishing based on the type of gear and its movements [10, 78, 79]. The identification of vessels and their fishing activities is limited by the fact that not all fishing vessels use this system [80], the ones that use it may present incorrect or incomplete identity data in the official vessel records [12], or because they choose to turn off the AIS to cover potentially illegal operations [81]. Another important limitation is that the algorithm cannot currently properly identify vessels that use multiple fishing gears or differentiate between their fishing activities [12].

The resulting gaps in AIS data are a substantial hurdle, and their use may be inappropriate for estimates of fishing intensity. To tackle this, Global Fishing Watch rates all designations of fishing effort or fishing activity as “apparent” and suggests considering the information as an estimate [82]. Catch reconstructions have shown that the Global Fishing Watch algorithm can provide reasonable estimates of fishing effort under optimal conditions (e.g., proper use and reception of AIS, the existence of vessel records with the type of fishing gear) [12], especially for areas where no or few prior data exist. Each year the coverage of global fishing activity is better due to the increase in the number of vessels that use AIS, the launch of more AIS-capable satellites [12], the latter approaching 100% global coverage, and the number of nations that publicly share their VMS data supported by Global Fishing Watch. Currently, ten countries including Ecuador, share their VMS data to this platform, but this information cannot be downloaded because it remains the property of the respective national authorities [46]. Previous research has used AIS data to map the fishing effort of fishing fleets [10, 44], assess fisheries management and marine planning [8, 9] and even analyze the overlap of fishing effort with threatened or endangered species [83, 84]. These studies have shown that AIS data, especially when combined with additional and complementary information such as VMS, can be a powerful tool for evaluating near real-time fishing activities as well as to support maritime spatial planning [11, 85].