2.1 Study area
The Reloncaví Fjord (41° 44' S and 72°32' W), has mussel beds in practically all its extension, and it is estimated that around 60% of the seed of M. chilensis demanded by Chilean mussel farming are obtained here (Segura et al. 2022). The fjord space is highly fragmented for the following uses: i) aquaculture concessions for mussel and salmon farming, ii) management and exploitation areas of benthic resources (MEABR), iii) areas with special permissions for mussel seed collection (Fig. 1), as well as iv) fishing in open access areas.
This fjord is a highly stratified system given its morphology (width ~3-5 km, length 60 km, maximum depth ~450 m) and high freshwater inflows (Fig. 1) (Valle-Levinson et al. 2007).
The main freshwater tributaries in this system are the Petrohué (280 m3/s), Cochamó (100 m3/s) and Puelo (650 m3/s) rivers, in addition of numerous other basins of smaller size and flow (Fig. 1) (León-Muñoz et al. 2021; https://chonos.ifop.cl/mosa/?domain=chiloe ). Depending on these conditions, in the upper fraction of its water column (0 - 20 m depth) there are strong spatial-temporal variations in temperature (9-18ºC) and salinity (2-33) (Castillo et al. 2015; Daneri et al. 2009; Molinet et al. 2015; Pickard, 1971). Another relevant characteristic of this fjord is the tidal amplitude, which varies between 6 to 7 m during spring tides and 5 to 6 m during neap tides (Valle-Levinson et al. 2007). Primary productivity of the Reloncaví Fjord is highly seasonal, with higher biomass values in spring (September) and autumn (March-April) (Iriarte et al. 2007; Lara et al. 2016).
The sampling sites were distributed considering the spatial division of the fjord made by Daneri, et al. (2009), who identified 4 sections according to their hydrodynamic and morphological characteristics: “Boca” at the mouth of Reloncavi fjord, two areas in the central gorge of the fjord; “Marimelli” and “Puelo”, and Cochamó, at the head of the fjord (Fig. 1). In each section, 2 to 4 benthic sites were sampled (Fig. 1, Appendix, Table A).
2.2 Collection of information
2.2.1 Estimation of seed harvesting pressure
Taking advantages of mussels' ability to settle on filamentous substrates (Fuentes and Molares, 1994; Tamarin et al. 1976), mussel seed collection takes place using a double or single long line system which sustain lines of ropes with a weight on the rope deeper end (Molinet et al. 2021; Segura et al. 2022). Ropes, approximately 4 m length and separated by 0.2 m, are installed on the long lines to collect seeds between 1 and 8 m depth along the fjord.
To have an indirect evaluation of the fishery intensity on M. chilensis seed, we used a long line pressure index (LPI) for the previously defined zones (Boca, Marimelli, Puelo and Cochamo). This was obtained from the ratio between the total length of long lines that collect seeds per zone and their corresponding shoreline length. To do this, the long lines installed in the Reloncavi Fjord were georeferenced and measured, and also the coastline for each zone was estimated, using images available from Google Earth between 2005 and 2020. These images have a resolution that allows identifying the length of each long line. The images were exported to ArcGIS 9.3 in kml format, where the information was displayed, the number of long lines was counted, and the length of each cultivation line was calculated. The type of long lines (double or single) installed was confirmed during the sampling activities. Finally, the long lines and shorelines were quantified by zone along the fjord (Boca, Marimelli, Puelo, Cochamó).
2.2.2 Variation in vertical distribution of M. chilensis and A. atra
The vertical distribution (~20 to 0 m depth) of M. chilensis and A. atra populations was studied at 12 to 14 sampling sites distributed along the Reloncaví Fjord (from the head to the mouth of the fjord), in three periods (2015, 2020 and 2022).
At each sampling site, 3-6 video transects (25 - 35 m long) were recorded along cross-shore gradients, in vertical profiles from 12-15 m depth to the upper limit of the intertidal (Table 1), which were separated by distances of 20 to 50 m. In each of these transects, videos were taken using a Seaviewer underwater camera (900 and 6000) and a Gopro camera. This method allows us to confidently identify individuals >20 mm in length.
Both cameras were mounted on a stainless-steel sled (0.35 m high; 0.3-0.5 m wide), at the base of which a 0.1 m accurate depth gauge (in the camera's field of view) was installed. The sled was kept perpendicular to the bottom with the support of a diver, who verified the boundaries of M. chilensis beds. During sampling at low tide, the sled was towed through the intertidal to the identified barnacle’s boundary, estimating the height between sea level and the video-transect boundary.
Complementarily, in each transect a sample was collected in the intertidal and another in the subtidal (using a 0.25x0.25 m quadrat) to count and measure the shell length (SL) of M. chilensis specimens. We identified individuals recruited during the previous year, as those whose SL was less than 26 mm (Asencio, 2015, and no published data from C. Molinet). Also, we recorded individuals larger than 39.7 mm SL, size at which 50% of female are sexually mature (SL50), following Molinet et al. (2016).
2.3. Image processing
To classify the type of substrate (as an indicator of sampled frame) and to quantify and identify the mussel species present, freeze-frames were taken in 30-50 cm wide quadrats. The recording time (hour: minute: second) and depth were recorded for each frame. The geographic coordinates of each sampled frame were obtained by aligning the video recording time and GPS recording time. From the processed images, a database was elaborated considering video number, sector, depth, species observed, their abundance, and coverage.
The processing of the video-transects was validated by an expert observer, who reviewed 10 seconds of each video. In cases where the review had objections, the video was reviewed again paying attention to the sources of error (e.g. sampler, video quality, sector, other). The application of this methodology was done assuming that the observed density of mussels could be underestimated because these organisms can be arranged in layers, where even the juveniles adhere to the byssus of the adults, so the density estimated is considered a proxy of the real density according to this image recording methodology (Molinet et al. 2015).
2.4 Puelo River flow and water column salinity records
To characterize the monitoring sites through the study periods, salinity data of the water column of the Reloncaví Fjord and the flow of the Puelo River, the only tributary course with continuous records of this system, were collected and systematized.
The flow data of the Puelo River were obtained from the repositories of the Chilean Water Directorate (DGA), specifically from the Carrera Basilio gauging site, which is located upstream of its outlet in the Reloncaví Fjord (https://dga.mop.gob.cl/Paginas/default.aspx).
Salinity was obtained from sampling sites of the Fisheries Development Institute (IFOP), who has been recording salinity profiles (CTD Seabird 19), with approximately monthly frequency, at 3 points in the Reloncaví Fjord (one site in the Puelo area and two sites in the Marimelli area (north and south), (Fig. 1C) since 2013 (Segura et al. 2022). In addition, the Caicura site was considered as a reference since it is located near the river discharge to the Reloncaví Fjord (Fig. 1 C).
Considering that M. chilensis and A. atra beds are distributed between 0 and 20 m depth, only the layer between 0 and 20 m depth of the CTD records was considered for this analysis.
2.5. Data analysis
Since M. chilensis is distributed in the intertidal and subtidal, the first step of the data analysis was to standardize the depths of the surveys carried out during 2015, 2020 and 2022. For this, a reference of the highest tide level (HTL) was defined, correcting the time and height for the Reloncavi Fjord relative to the Puerto Montt Pattern Port (SHOA Pub.9000, www.shoa.cl; https://www.ioc-sealevelmonitoring.org/). The depth of each frame of the analyzed video-transects was referenced to HTL using the following formula:
Frame depth= Depth0+(HTL-TLS)
where: Depth0 is the depth or height observed in the sampling, TLS is the tide level at the time of sampling (according to the tide table) and HTL, as mentioned, is the reference level of the highest tide.
Once the sampling depths were standardized, the depth variation of the mussel beds was analyzed, considering in each vertical profile and video-transect the presence (at least 1 mussel) /absence of the species A. atra and M. chilensis. For each video-transect, site and species, the average, minimum (upper limit referred to the HTL, with at least a mussel) and maximum (lower limit referred to the HTL, with at least a mussel) depths were estimated.
The variation of the three depths (minimum, average and maximum) was evaluated using linear mixed models and generalized linear mixed models, when the linear models did not meet the assumptions of normality and homoscedasticity of the residuals (Crawley, 2007; McCullagh and Nelder, 1989; Venables and Ripley, 2002). Year, area, species, and density (Individuals/m2) were used as fixed effect predictor variables. The sampling site and zone (Boca, Marimelli, Puelo, Cochamo) variables were evaluated as random effects because their hierarchical structure. To analyze the sources of variation, a type II ANOVA was applied, considering that the samplings were unbalanced (different number of transects per year, zone and site). In addition, the effect of the Long lines pressure index on the variation of the minimum, average and maximum depth of the M. chilensis and A. atra beds was analyzed, using linear and non-linear models (Crawley 2007). For this analysis we considered the LPI of 2014, 2018 and 2020, which were compared with the average, minimum and maximum depth of mussels in 2015, 2020 and 2022, respectively.
To identify salinity temporal trends, we used the data series available from IFOP (4 sites, between 1 and 20 m depth) between 2013 and 2022, the linear trend of mean, minimum and maximum salinity values were analized (Crawley 2007). The Puelo River flow data series for the same period was also explored.
The Akaike information criterion (AIC) (Akaike 1974; Burnham and Anderson 2004) was used to select the most informative models on the variation of the average, minimum and maximum depth of the A. atra and M. chilensis beds. All analyses were performed in R 4.3.0 (The R Development Core Team 2024), making use of the following libraries: car (Fox and Sanford 2010), nlme (Pinheiro et al. 2013), MASS (Venables and Ripley 2002), lme4 (Bates et al. 2015), ICcmodavg (Mazerolle 2023).