This study was conducted in the Spermonde Archipelago (Figure 1), located in SW Sulawesi, off the coast of its capital city Makassar. It comprises up to 150 coral cays and reef islands in 400,000 ha of coastal waters with a maximum depth of 60 m (Pet-Soede et al., 1999; Jompa and Yusuf, 2012; Schwerdtner Mánez et al., 2012). These islands are on top of a submerged platform extending north-south and parallel to the Makassar Strait, where depth reaches 2,500 m (Imran et al., 2013). There are more than 45,000 people living on 54 of these islands, which are often extremely densely populated. For example, in 2010, 216 and 325 people per ha inhabited the islands of Badi and Barrang Caddi, respectively (Schwerdtner Mánez et al., 2012). Most of the populated islands are situated close to the mainland of SW Sulawesi (Pet-Soede et al., 2001a). Land run-off has been linked to poor water quality and localized eutrophication (Edinger et al., 1998; Teichberg et al., accepted). Households depend largely on fisheries or fishing-related activities for subsistence and income generation. Alternative jobs, such as petty trading or construction play a minor role in the local economies (Ferse et al., 2014; Miñarro et al., 2016). The anthropogenic pressure in Spermonde related to the heavy development of the coastal region has caused a widespread degradation of the coral reefs and heavy exploitation of its natural resources (Pet-Soede et al., 2001b; Ferse et al., 2014; Glaser et al., 2015).

The Spermonde Archipelago supports one of the largest reef fisheries of Indonesia (Ferse et al., 2014). The artisanal fishery is conducted by functional units consisting of groups or individual fishermen, plus the different boats and gears they use (Pet-Soede et al., 2001a; Deswandi, 2012). Large boats are used to install nets, often combined with light fisheries (pers. obs.). Launches are used to transport traps to be placed on the sea bottom. These boats can also be used with illegal fishing methods, such as blast and cyanide fishing (Deswandi, 2012). Small boats are mainly used with hook-and-line gears (Pet-Soede et al., 2001a). Overall, more than 20 types of fishing methods using different gears are used in Spermonde, which are often changed according to the season or the weather conditions (Breitkopf, 2014). In contrast, boat types are a rather stable feature of the fishing units (Deswandi, 2012).

The intra-annual dynamics in Spemonde's fisheries are related to the occurrence of two weather seasons: the SE Monsoon or dry season from May to September, and the NW Monsoon or wet season from December to March, with transitional weather in April, October, and November. Deswandi (2012) relates the two seasons to different spatial patterns of fishing effort distribution. During the dry season, there are good conditions for sailing, and fishing activities occur mainly toward the western sector. The wet season is characterized by heavy rainfall when fishing activities are limited to the inner areas of the Archipelago, and many fishermen emigrate toward regions with better weather conditions in S and SE Sulawesi (Ferse et al., 2014).

Data collection for this study took place from 25 November 2014 to 25 March 2015 on Badi Island (4°58′03″ S, 119°17′16″ E, Figure 1). Badi is located 24 km NW of Makassar and has an extension of 10 ha. According to the demographic statistics published in the communal meeting room, on December 2014 there were 1944 inhabitants and 518 households. Only fishermen using small boats and staying for the entire wet season were chosen as the study group (132 male adults). The boats were in average 6.3 (1.7) m long, made of wood or fiber glass, with a stabilizing side bar and an improvised engine consisting of a 5–10 hp water pump adapted to an axis with a helix. These boats are called kapal katinting, or lepa-lepa and are mainly used by hook and line fishermen (Pet-Soede et al., 2001a; Miñarro et al., 2016).

Participants of the study were grouped according to the type of information they were asked to provide, following the selection procedures listed in Table 1. Key informants included the gate keeper (a former bomb- and now trap-fisherman who cooperates with researchers from University of Hasanuddin and patrols the fishing grounds surrounding Badi), the Community Leader, one experienced fisherman who used to manufacture and sell octopus bait as alternative source of income, and two fish collectors (i.e., patrons). This research was conducted in accordance with all ethical standards outlined in the Amended and Updated White Paper on Safeguarding Good Scientific Practice by the German Research Foundation [Deutsche Forschungsgemeinschaft (DFG), 2013]. At the time of the fieldwork, no ethics approval was needed as per German or Indonesian law or regulations besides those approved in the research permit and by Hasanuddin University. In all cases, the request to provide information included an oral Prior Informed Consent procedure (Wiles et al., 2005) with no economic incentives. In case of questions and concerns that may arise after the study, contact details of the researcher were provided to the Community Leader.

An initial interview for participants of group 1 (Table 1) included general questions regarding fishing practices and personal background. Participants were asked to take GPS trackers on their fishing trips. Ten trackers Conrad GT-750 and five trackers i-gotU GT-120 were set to record coordinates and speed in a 5 s interval. Trackers were not given to the same person on consecutive days to minimize data correlation as described by Pet-Soede et al. (2001a). Trackers were collected the evening after the fishing trip at the participants' residence. A brief interview was conducted to collect data about the number of fishermen onboard during the trip, names of the fishing grounds visited along the route, employed gears in order of use, and their associated catch data (species, numbers of catches per species and their body weight).

During those interviews the fishing grounds were identified only by their names instead of showing a map to the fishermen because the time for collecting the trackers was short (20 min) and not many fishermen were familiar with those types of maps (Deswandi, 2012). To compensate, a bathymetric map was used to ask key informants about the location of the most frequently mentioned fishing grounds. These maps already included local names of individual reefs and islands. Anecdotal information relevant for interpreting the digital spatial data was added to the forms after a pilot study and related to whether the fishermen spent time resting, selling their catch or looking for landmarks instead of fishing (if so, when/where).

The histogram of boat speed values showed a bimodal distribution with one peak at 0 km/h and another at 7 km/h (Figure 2). This confirmed that some gears were used with static boat and others with the boat in movement, as described by the fishermen. Three classes of boat speed (slow, moderate, and fast) are defined and characterized in Table 2.

The total fishing area was divided in seven fishing grounds (Figure 3). Three grounds, called Badi, Tintingan, and Malla'bang (No 1, 4, and 6 in Figure 3), concentrated 84% of the fishing actions. They form a route starting in the shallow reefs (5–10 m) of Badi and Tintingan, until reaching the deeper (50 m) Malla'bang. Those routes were further used for the spatial analyses. The fishing grounds closer to the western border of the Archipelago (Figure 3) were visited before the beginning of the wet season, but were avoided by fishermen later due to safety reasons as the weather conditions worsened. The furthest fishing action was located 36.3 km NW of Badi Island.

The main fishing gears represented in the spatial data were “trolling line” and “octopus bait.” In the contingency table results, these gears were significantly associated to specific fishing grounds (Chi²: 58.328, d.f.: 6, p < 0.01). Trolling line was more frequent in Badi, Tintingan, and Malla‘bang, while octopus bait only occurred in Badi and Tintingan. The average fishing time per action varied between 3 and 4 h for both gears in most fishing grounds (Figure 4). However, the fishing time per action of trolling line in Tintingan was with 1.2 h significantly lower than Octopus fishing in the same area (U = 1081, Z = 4.3, p < 0.01, r = 0.63) and trolling line in Malla'bang (U = 1540, Z = 4.6, p < 0.01, r = 0.62), while the differences to Badi were not significant.

The taxonomic composition was similar between the catch data subsets obtained from interviews and catch landings surveys (Table 3). The highest percentage of biomass was made up by the Octopoda order. Species diversity was dominated by the Serranidae family, with 20 out of the 83 fish species identified. Within Serranidae, the most abundant species were Plectropomus leopardus, Plectropomus areolatus, and Pseudanthias sp. with 52.5, 23.5, and 9.1% of the taxon's catch landings, respectively. The Plectropomus species are sold in the LRFFT while Pseudanthias is used for local consumption.

CPUE estimates per fishing gear are presented in Table 4. Octopus bait had the highest CPUE, and exceeded almost 10 times the CPUE of the gear with the lowest CPUE (i.e., Hook and line squid). Trolling line for live grouper had a relatively low CPUE compared to other gears to target fishes.

The method used to identify fishing grounds in this multi-gear fishery consisted in dividing the trip in shorter periods, and the seascape in fishing zones based on the fishermen's own classification, attempting to capture the decisions they made throughout the fishing day (i.e., what, where, and for how long to fish), and quantify the effort invested. A key element in this approach is boat speed due to its ability to differentiate fishing from non-fishing periods. The potential use of speed for this purpose was already pointed out by Daw et al. (2011) in tracking artisanal boats from East Africa. More recently, Alvard et al. (2015), presented a Partial Sum Method to automatically identify fishing activity based on speed, complementing their boat tracking data with onboard observations in a single-gear artisanal fishery from the Caribbean. In our case, working with speed-based criteria alone was not enough to quantify, because there were multiple fishing gears involved per trip, including cases where a single speed class could involve more than one gear. Therefore, we consider that the inclusion of anecdotal information remains essential in the data analysis for multi-gear artisanal fisheries, and this can be easily included in the sampling protocols by doing brief interviews after the fishermen have rested from work and are willing to talk.

Another reason to include traditional interview methods in boat tracking studies is that artisanal fisheries are very diverse, and GPS setting up and data processing may require different configurations. For instance, Alvard et al. (2015) worked with hand-line fishermen in Dominica, and identified three speed ranges far above the ones described in this study (Table 5), probably because in that case the boats had more powerful engines, and perhaps also for differences in the behavior of target species because they target tuna (Scombridae), marlin (Istiophoridae), and dolphinfishes (Coryphaenidae) that swim faster than the groupers. Therefore, research plans should still include procedures to characterize the fishing gears and fishermen's knowledge on how to use them.

Characterizing fishing effort and spatial extent of coastal fisheries is important for effective management of marine resources because fishing areas are often overestimated (Stewart et al., 2010), which hinders the design of spatial use regulations. Like in other cases (Daw et al., 2011; Alvard et al., 2015; Metcalfe et al., 2016), the map obtained in this study (Figure 3) clearly shows that fishing effort is not evenly distributed in the area around Badi but focused on specific fishing grounds. This type of information can be useful for spatial zonation for management planning, but should be complemented with GPS tracking data from other islands and seasons, and combined with knowledge of the informal institutions ruling the access to these fishing grounds (Deswandi, 2012; Gorris, 2016). Successful management strategies found in Sulawesi include No-Take Areas in Wakatobi (Unsworth et al., 2007) and area-specific stewardship for a marine territory in Spermonde's Langkai Island (Gorris, 2016), among others, which are routinely monitored by local enforcement mechanisms. However, implementing spatial management strategies in the two main areas identified in this study (Tintingan and Malla'bang) will be extremely challenging because there are no inhabited islands nearby where communities could engage in the enforcement of informal rules to manage these resources and protect them from destructive fishing practices (Glaser et al., 2010).

The distance traveled by participants (up to 35 km) was relatively high compared to the estimates in places like Seychelles, Indian Ocean (< 5 km; Daw et al., 2011) and the Congo (2–20 km; Metcalfe et al., 2016). Pet-Soede et al. (2001a) reported a travel distance of 6.2 km for the gear trolling line for live grouper, but their average represented data collected from fishermen from all over the Archipelago and their sampling methods were different (i.e., based on a research vessel), therefore it is uncertain whether there was an actual increase in travel distance.

Our results on fishing effort allocation among fishing grounds suggest that fishermen associate each area to a group of target species and their corresponding gears. When using the appropriate gear for a certain area, they spent a “standard” amount of time fishing there (in our case, 3–4 h). When they chose a less fit gear in a fishing ground, they spend less time doing so (1 h). The last was the case of fishermen trolling line for live grouper in a shallow reef area like Tintingan (5–10 m), where groupers abundance should not be expected to be optimal since the habitat of this group has been found to be located below 10 m with an abundance peak at 15–20 m depth (Unsworth et al., 2007). Correspondingly, there were no attempts to catch octopi in the deep area of Malla'bang because, according to fishermen, Spermonde's species inhabit shallow waters (Pers. obs.). Considering that Tintingan area is on the way to the hotspot for trolling line (Malla'bang), it might make sense for fishermen to try trolling line in Tintingan despite having low success chances. Perhaps this reflects that other drivers, such as market demand, also affect fishermen's effort allocation decisions. Market information is transferred to fishermen via a chain of exporters, transport middlemen and patrons. Based on this information, fishermen exploit particular fishing grounds in order to obtain the resources, which are then passed through the same chain of traders in reverse order until reaching consumers (Ferse et al., 2014).

In our study case, catch composition was dominated by species sold in international markets. On one hand, groupers have been target species in Spermonde for decades (Radjawali, 2012; Ferse et al., 2014). We found the same CPUE for trolling line for live groupers (1.7 kg boatday⁻¹) as reported for Spermonde Archipelago by Pet-Soede et al. (2001a). However, the cited study also reported spatial and temporal differences in CPUE throughout the year, and the area where our participants actually fished was reported to have 2.1–3.3 kg boatday⁻¹ CPUE during the rainy season (Pet-Soede et al., 2001a). A decrease in CPUE may be due to several factors (e.g., increase in the number of fishermen and changes in gear efficiency, among others), but it could also reflect a local decrease in groupers' abundance, which would not be surprising given the extended use of cyanide as fishing method to target this group (Radjawali, 2011). To continue supplying the international demand for LRFF, good fishing practices such as the use of trolling line instead of cyanide diving should be encouraged among fishermen.

On the other hand, octopus has only recently appeared in the scientific literature of this region (Ferse et al., 2014; Miñarro et al., 2016) and this study provides the first quantitative report of CPUE for this target group in the area. This reflects the dynamic nature of artisanal fisheries' social-ecological systems in terms of adaptation to new market demands and/or resource availability (Schwerdtner Mánez and Ferse, 2010). Octopus demand from Asian fisheries has increased in recent years, potentially associated with the decline of the Mediterranean octopus production (Pauly and Zeller, 2015). In parallel, increase in octopus abundance worldwide (Doubleday et al., 2016) might be also propelling this fishery. The identification and conservation of the fishing grounds where these species are fished is a key issue for the sustainability of this emerging fishery and can benefit from monitoring programs using GPS tracking technology and the speed ranges reported in this study.

The present study represents a step forward in applying tracking technology to the study of artisanal fisheries. This approach can be used in impact assessment studies. Potentially, it could also be used to assess social and economic influences on fishing behavior that are matter of debate in Spermonde, such as the influence of fish traders on the fishing behavior of the fishermen with whom they establish a “patron-client relationship” (Foster, 1961; Chozin, 2008; Deswandi, 2012; Radjawali, 2012; Miñarro et al., 2016). Our findings also point at the importance of conducting stock assessments of commercial species such as groupers and octopi to estimate the fishing effort that could be exerted on these resources without compromising their sustainability. The collection of length-frequency data is particularly suitable in the context of tropical artisanal fisheries and would allow applying length-based stock assessments, as well as investigating the length preferences in the length-selective fisheries for live fish trade (Mildenberger et al., 2017).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.