You are on page 1of 9

Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.

1093/icesjms/fsab074/6231554 by Universidade Federal de Santa Catarina user on 17 April 2021


ICES Journal of Marine Science (2021), doi:10.1093/icesjms/fsab074

Alternative data sources can fill the gaps in data-poor fisheries


Alexandre M. S. Machado 1,2,3, Eduardo L. Hettwer Giehl 4
, Luiza Pacheco Fernandes1,
Simon N. Ingram 5, and Fábio G. Daura-Jorge 1,*
1
Laboratório de Mamı́feros Aquáticos, Departamento de Ecologia e Zoologia, Universidade Federal de Santa Catarina, CCB/ECZ, Florianópolis, SC
88010-970, Brazil
2
Programa de Pós-Graduaç~ao em Ecologia, Departamento de Ecologia e Zoologia, Universidade Federal de Santa Catarina, Florianópolis, SC
88010-970, Brazil
3
Department of Collective Behaviour, Max Planck Institute of Animal Behaviour, Konstanz, Germany
4
Laboratório de Diversidade e Conservaç~ao, Departamento de Ecologia e Zoologia, Universidade Federal de Santa Catarina, Florianópolis, SC
88010-970, Brazil
5
School of Biological and Marine Sciences, University of Plymouth, Plymouth PL4 8AA, UK
*Corresponding author: e-mail: daura.jorge@ufsc.br.
Machado, A. M. S., Giehl, E. L. H., Fernandes, L. P., Ingram, S. N., and Daura-Jorge, F. G. Alternative data sources can fill the gaps in data-
poor fisheries. – ICES Journal of Marine Science, doi:10.1093/icesjms/fsab074.
Received 25 August 2020; revised 7 March 2021; accepted 8 March 2021.

Assessing fish stocks harvested by small-scale fisheries is challenging. The lack of official fisheries data constrains the proper management of
such fisheries. Thus, alternative sources of information are crucial to enrich data-poor fisheries. Here, we evaluated different sources of data
for the mullet (Mugil liza) fishery, one of the most important but overexploited fisheries in Brazil. We gathered three alternative sources of
catch data by artisanal fisheries: 14 years of self-reported catches by artisanal fishers across 24 municipalities; 16 years of catches by traditional
beach seines mined from news outlets; and 13 years from a single community monitoring their beach seine catches. We tested whether alter-
native data sources follow the same trends of landing reports from systematic, official monitoring of the industrial fleet. We fitted Bayesian
time-series models to test if environmental changes and stock abundance can predict these data. We found that only self-reported catches
matched the official reporting trends, thereby improving our understanding of changes in the mullet stock. These findings reveal that self-
reported catches by fishers provide reliable additional data useful for management. Self-reporting data are cost-effective, deals with the com-
plexity of small-scale fisheries, and welcomes fishers as key stakeholders in management practices.
Keywords: Artisanal fisheries, Data-limited, Data mining, Mugil liza, Newspaper, Self-report, Self-sampling

Introduction 2020), making these stocks vulnerable to overexploitation


Effective fish-stock management depends on the challenging task (Costello et al., 2012).
of gathering reliable data. The assessments of commercially valu- Small-scale fisheries in developing countries are difficult to
able fish stocks in developed countries are often based on de- track and manage, thus invariably data-deficient. Artisanal fishers
tailed longitudinal datasets (e.g. de Lestang et al., 2012). typically use small vessels varying in technological capability, fish-
Fisheries scientists can therefore fit population-dynamics models ing gear, and target species. Landings can also occur anywhere
to these data to estimate stock trends, to determine stock status along the coast rather than being concentrated at large, regional
and to recommend sustainable harvesting quotas. However, fishing ports. These factors often result in the misconception that
gathering reliable data to evaluate population trends and status small-scale fisheries account for small and widespread catches
can be both expensive and time-consuming (Bland et al., 2015). with negligible consequences to the fish stocks (Pauly, 2006).
Hence, the assessment of many harvested fish stocks remains Partially as a consequence, small-scale fisheries are often over-
data-limited, thus potentially imprecise and biased (Free et al., looked by governments, while catches continued to increase and
C International Council for the Exploration of the Sea 2021. All rights reserved.
V
For permissions, please email: journals.permissions@oup.com
2 A. M. S. Machado et al.

have been underreported worldwide (Zeller et al., 2015; Pauly and (iii) catches from detailed logs of a beach seine fishing com-
and Zeller, 2016; Canty et al., 2019). Small-scale fisheries can col- munity were correlated with the official landing reports of the in-

Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsab074/6231554 by Universidade Federal de Santa Catarina user on 17 April 2021
lectively reach operational scales similar to industrial fisheries dustrial fleet, which was considered to be the most systematic in
(Alfaro-Shigueto et al., 2010), but because they tend to be con- the historical dataset. All these data sources covered fisheries har-
centrated in nearshore habitats they present potential threats to vesting the same stock, but each fishery has different dynamics in
sensitive coastal ecosystems and vulnerable species (Shester and terms of spatial scale, catchability and number of fishers involved.
Micheli, 2011). Neglecting underreported small-scale fisheries We then modelled how sensitive these data sources are to changes
can, therefore, misinform fish-stock management (Rudd and in environmental conditions and stock abundance, using sea-sur-
Branch, 2017), and mask the risks for biodiversity and successful face temperature (SST) and the catches per unit effort (CPUE) of
conservation (Alfaro-Shigueto et al., 2010). the industrial fleet as explanatory variables, respectively. Finally,
Recent effort lies in finding and validating alternative sources we used our fitted model—accounting for the influence of SST
to compensate for the absence of official data on small-scale fish- and CPUE of the industrial fleet—to forecast catches after 2013—
eries. To cite a few, historical ecology, ethnoecology, and citizen when official systematic monitoring of the industrial fleet became
science emerge as potential alternatives. Historical abundances only partial, but still provided CPUE for part of the total catches,
and distribution of marine species can be reconstructed from log-
as a sub-sample. A high correlation between official and non-offi-
book data and catch records (e.g. Lotze and Worm, 2009).
cial data would suggest that alternative data could fill the gaps
Historical baselines in fisheries can be identified from media
necessary to model population dynamics and generate predictions
reports (e.g. Francis et al., 2019). Catch reports can also be esti-
that can inform decision-makers.
mated by accessing traditional knowledge (e.g. Castello, 2004;
Damasio et al., 2015), or from self-management practices such as
Material and methods
local community records of their own catches (Mion et al., 2015).
Recently, the widespread adoption of technologies such as GPS,
Mullet
mobile phones, and social media has facilitated the collection, The mullet (Mugil liza) is a pelagic species that migrate to the sea
processing and sharing of information on fisheries (e.g. Venturelli to reproduce, but depend on estuaries and lagoons to feed and
et al., 2017; Sbragaglia et al., 2020). grow (Menezes et al., 2003). The southern stock shoals migrate
These alternative sources of information can be integrated from Uruguay and Argentina towards the waters of southern and
with quantitative techniques to enrich data on small-scale fisher- south-eastern Brazil (Sadowski and Dias, 1986; Mai et al., 2014).
ies (Kindsvater et al., 2018). In Brazil, small-scale fisheries are This reproductive migration occurs from May to July. The peak
overlooked, even though they provide livelihoods for many peo- of spawning occurs along with the southern Brazilian states (Rio
ple. However, since 2011, due to political and economic changes, Grande do Sul, Santa Catarina, and Paraná) and is associated
there has been no official fishery statistics for large or small-scale with the decrease of sea-water temperature and frequent southern
fisheries in Brazil (MPA, 2011). This lack of official monitoring winds (Herbst and Hanazaki, 2014; Lemos et al., 2014). Thus,
poses a significant challenge for fisheries management and hin- fishers are fishing the same cohort within a given year. Across
ders the reversal of declining fish stocks. A remarkable example is years, fishers are fishing different cohorts but at the same stage—
the mullet (Mugil liza) fishery. i.e. sexually mature mullets migrating to spawn (Garbin et al.,
The mullet fishery is one of the most important target species 2014). Most of the mullet catches (ca. 45%) occur along the coast
of the Brazilian fisheries. Every year during the mullet season of Santa Catarina (Lemos et al., 2016), where the mullet fishery is
(from May to July), both large and small-scale fisheries target the a recognized feature of cultural heritage (state law number
southern mullet stock along the south and southeast of Brazil. 15.922, 2012). Rules defining fishing zones can vary with vessel
Large-scale fisheries are dominated by the purse seine fleet, while size and location, but most catches occur in the sea water after
small-scale fisheries are more diverse, comprising gillnets, drift mullet shoals leave the estuaries and lagoons. In most estuarine
nets, beach seines, and even artisanal cast net fishers. Almost systems, only traditional artisanal cast net fishers are allowed (e.g.
10000 fishers rely on this fishery, with an estimated annual receipt Peterson et al., 2008; Santos et al., 2018). Such cultural aspects are
of over US$9 million (MPA/MMA, 2015). However, the mullet strengthened by media coverage in newspapers and on TV,
stock in southern Brazil has been considered threatened by over- reporting the start of the mullet season every year, and following
exploitation since 2004. A management plan for sustainable fish- catches in each season.
ing of mullet stocks was only established in 2015, proposing catch
quotas, specific fishing periods for the industrial fleet and arti-
sanal fisheries, spatial restrictions, and satellite vessel-tracking sys-
Datasets
tems for the industrial fleet (MPA/MMA, 2015). A recent study We assessed four datasets of mullet fisheries in the state of Santa
combined data from multiple fishing grounds and of different Catarina. First, we assessed the industrial and artisanal fisheries
fishing gear in a stock assessment (Sant’Ana et al., 2017). More datasets that comprised part of the data used in a recent mullet
than just raising a flag about the sustainability of the mullet stock, stock assessment (see Sant’Ana et al., 2017). Data from the indus-
this assessment raised concerns about a data crisis, as some data trial fleet comprise official landing reports from systematic moni-
from mullet fisheries are intermittent and unreliable (Sant’Ana toring conducted by the Fishery Studies Group from the
et al., 2017). Therefore, the reliability of alternative datasets must University of Vale do Itajaı́ (UNIVALI/GEP; available at http://
be assessed. pmap-sc.acad.univali.br/). Landings from purse seine, bottom
Here, we evaluated the validity of three alternative sources of gillnets and trawlers were fully assessed from 2000 to 2012, but
data on artisanal mullet fisheries in southern Brazil. We tested only partially reported after 2013. From this source of data, we
whether data from (i) self-reported catches by artisanal fishers, only assessed data from purse seines, which contribute to the ma-
(ii) catches by traditional beach seines mined from news outlets, jority of industrial fleet catches.
Filling data gaps with self-reported catches 3

Artisanal fisheries data came from three sources: self-reporting the mean SST per day, from May to July (92 days). Then, we cal-
catches, media reports, and logs of a beach seine fishing commu- culated the spatial and temporal suitability for each year, consid-
ering the number of days at the optimal condition (19–21 C; see

Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsab074/6231554 by Universidade Federal de Santa Catarina user on 17 April 2021
nity. The first source was an official effort maintained since 2003
by the Federation of Fishermen of Santa Catarina (FEPESC) that Lemos et al., 2016). Mullet migration is regulated by SST; sea-wa-
integrates catches self-reported by fishers. From 2003 to 2016, ter temperature triggers the ovarian maturation by affecting vitel-
fishers from 24 municipalities reported catches from beach seines logenesis, which stimulates the spawning migration of the species
and small vessels—lengths varying from 5 to 10 m, powered by (Whitfield et al., 2012; Lemos et al., 2014). Hence, the amount of
10 to 50 HP engines—along the coast of the state of Santa mullet available for fishing may depend both on the time window
Catarina, southern Brazil. In each municipality, individual fishers of optimal SST conditions and the extent of the coast under opti-
reported daily catches directly to the FEPESC during the mullet mal conditions. To account for this, we summarized the influence
season, from May to July. These raw data were provided by of SST in both space and time by summing the daily sea-surface
FEPESC upon request. area within the optimal conditions for each year (hereafter opti-
The second alternative source of data came from data mining mal SST). Next, we used CPUE from the industrial purse seine
media reports of beach seines found in local newspapers at the fleet as our proxy for trends in stock abundance. CPUE of the in-
Florianópolis Public Library, Santa Catarina. We focused on the dustrial fleet was calculated as catches over the number of fishing
newspaper with the highest state-wide circulation—Diário trips and converted to tonnesyear1. Due to our short time se-
Catarinense—searching for articles on beach seining catch ries, and the absence of significant technological innovations or
reports. We manually searched for articles in hard copies of news- changes in how the industrial fleet behaves, we assumed a con-
papers published from 2000 to 2016, in every issue from the first stant catchability throughout the study period.
day of April to the last day of July, encompassing the mullet sea- Trends in the artisanal catch data were related to optimal SST
son in southern Brazil (Vieira and Scalabrin, 1991; Lemos et al., and CPUE of the industrial fleet using Bayesian structural time-
2016). When available, we retrieved information on catch sizes series models, a useful class of forecasting time-series models that
(number of fish or biomass in tons), date, and location. generalizes many standard time-series processes in a Bayesian
The last alternative source of data were daily logs maintained framework to account for uncertainty (Scott and Varian, 2014).
and provided by fishers from a traditional beach seine fishing We fitted models using the annual total catch of each alternative
community at the Estaleirinho beach, in the state of Santa dataset as the response variable, and optimal SST and CPUE from
Catarina (27.048 S, 48.587 W). This community self-orga- the industrial purse seine fleet as predictors. Since all datasets had
nized in maintaining daily records of beach seine catches from different start dates, we adjusted predictors to match the length
2004 to 2016. of each response until 2012. In addition, we included terms for
autoregressive and for a local linear trend (moving average) in
Data analysis the models. Autoregressive terms aim to account for the depen-
As the industrial fleet and self-reporting data were reported in dence of stock abundance on the abundance of previous years.
terms of weight, to correlate all datasets, we converted the num- Here, we used only a 3-year-lag because of our short time series.
ber of fish captured in each of the alternative datasets to catch in However, as mullet reach sexual maturity at around 6 years
metric tonnes, considering each mullet to weigh 1.6 kg on aver- (Garbin et al., 2014), other values should be tested when longer
age. This figure was based on 11 newspaper articles that reported time series become available in the future. The local linear trend
both the number of fish and tonnes captured. Although this sim- was included as a proxy for stochastic or unknown processes that
plified conversion presumes a time-invariant individual average could cause further fluctuations in stock abundance.
weight, the mean size of mullets caught by the industrial fleet in Combining either predictors, autoregressive or local linear
the last decade was consistent across years (SEAP/PR and MMA, trends resulted in seven models for each of the three artisanal
2018). Then, we summarized catch data to tonnes per year, using fishing data. Model terms were combined as follows (i) predictors
only catches that occurred from May to July, when mullet migrate þ autoregressive þ local linear trend; (ii) predictors þ autore-
northwards to spawn—the peak of the mullet season (Herbst and gressive; (iii) predictors þ local linear trend; (iv) autoregressive
Hanazaki, 2014; Lemos et al., 2016). To check for correlation be- þ local linear trend; (v) predictors, (vi) autoregressive; and (vii)
tween the three alternative catch datasets, we used Pearson corre- local linear trend. This set of models allow us to test the effect of
lation tests, truncating the time series from 2004 to 2012, the SST and our proxy of stock size (CPUE of the industrial fleet) on
period when all datasets overlapped. catch data of the three alternative small-scale datasets, accounting
Because all datasets reported catches from the same stock, for time-dependence and/or stochastic processes in stock abun-
trends in all sources should depend on the same changes in envi- dance. In all cases, we ran a Markov Chain Monte Carlo
ronmental conditions and stock abundance. Thus, we also fitted (MCMC) algorithm for 10000 iterations and discarded the first
models to account for artisanal fisheries trends related to changes 1000 as burn-in. We then compared models by their ability to fit
in SST and stock abundance. We obtained SST data from the response variables using cumulative absolute prediction errors
National Oceanic and Atmospheric Administration (NOAA) calculated step-by-step for the observation in the step ahead. The
Environmental Research Division’s Data Access Program server smallest cumulative errors indicate good candidate models. In ad-
(ERDDAP). We chose SST data to match time (mullet season, dition, we used the best model to forecast catches from 2013 to
from May to July) and space (the coastline of the state of Santa 2016, when only a subset of the industrial fleet landings was
Catarina) of all the four sources of mullet catch data. We calcu- reported. If observed empirical data fell outside the forecasted
lated the mean daily SST in grid cells of 0.25 latitude by 0.25 95% credible interval, then we considered the model to be inade-
longitude (25 km2), ranging from latitude 25.875 to 28.875 quate to predict trends in catches. For chosen models including
and longitude 47.875 to 49.875. For each cell, we calculated optimal SST and CPUE of the industrial fleet, the relative
4 A. M. S. Machado et al.

importance of each variable was calculated from posterior proba- 2000. From 2001 to 2016, newspaper articles reported a total of
bility distributions for their coefficients. 885.4 t of mullet landed (mean ¼ 55.3 6 46.4 SD t per year),

Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsab074/6231554 by Universidade Federal de Santa Catarina user on 17 April 2021
We carried out all the analyses in the R environment (R ranging from 4.9 in 2014 to 154 t in 2003. Throughout the years,
Development Core Team, 2019). We used the corrplot R package most of the newspaper articles were published in May (36%;
(Wei and Simko, 2017) to test the correlation between datasets n ¼ 25) and June (60%; n ¼ 42). Only one catch was reported in
and the bsts R package (Scott, 2019) to fit Bayesian structural April (25 April 2003) and two in July (05 July 2001 and 17 July
time-series models. 2002). Catches from 11 municipalities were reported in newspa-
per articles. Most of the catches reported occurred in Santa
Results Catarina Island (64%; n ¼ 45), with other municipalities ranging
The industrial fleet landed a total of 39555 t (metric tonnes) of from 1 to 6 reports. Fishers from the beach seine fishing commu-
mullet in the state of Santa Catarina, from 2000 to 2016 (mean ¼ nity registered 273 beach seines from 2004 to 2016. During this
2326 6 1409 SD t per year). Fishing operations had a median fleet period, fishers landed 109.5 t (mean ¼ 5.1 6 8.9 SD t per year),
size of 82 (6 31.14 SD) and a median of 139 (6 56.56 SD) fishing ranging from 0.6 in 2011 to 34.66 t in 2016.
trips per year. From 2000 to 2012, when the industrial fleet was Catches reported by the self-reporting approach correlated
systematically assessed, the average catch per unit of effort with catches from the industrial fisheries fleet (r ¼ 0.77, p < 0.05).
(CPUE) was 16.27 t (6 5.87 SD) per fishing trip, ranging from All other correlation coefficients were not statistically clear
6.92 in 2002 to 27.71 t per fishing trip in 2007. (Figure 2). The validity of this self-reporting approach was rein-
Self-reported data to FEPESC amounted 1316 monthly reports forced when accounting for spatial and temporal suitability of en-
from 2003 to 2016, comprising catches landed in 24 municipali- vironmental conditions and stock abundance (Table 1). All
ties in the state of Santa Catarina, including 24 fishing communi- artisanal fisheries datasets were predicted by optimal SST and
ties in the Santa Catarina Island (Figure 1). Fishers self-reported a stock abundance, but we found industrial fleet CPUE had the
total of 16224.9 t of mullet landed (mean ¼ 1158.9 6 811.3 SD t highest posterior inclusion probability (0.67) in the model for the
per year), ranging from 425.7 in 2013 to 3543.2 t in 2016. Media self-reporting data. We found higher coefficients and posterior
reports comprised 288 newspaper articles published from 2001 to inclusion probabilities for the proxy of stock abundance than for
2016, of which 70 reported mullet beach seines (median ¼ SST, but this relationship reverses in the beach seine fishing com-
3.5 6 2.6 SD, ranging from 1 to 11 newspaper articles per year). munity dataset (Table 1). These models with autoregressive terms
We found no media reports of mullet beach seining published in and predictors accumulated errors at a lower rate than other

Brazil

Self−reporting Media reports Beach seine community


25.875°S

N
0 50 100km
Latitude

ean
tic Oc
Atlan
28.875°S

100 12 40
50 6 20
0 0 0
2004 2010 2016 2004 2010 2016 2004 2010 2016

49.875°W 47.875°W 49.875°W 47.875°W 49.875°W 47.875°W


Longitude
Figure 1. Spatial and temporal scales of the alternative catch data sources. Each map indicates fishing spots in the state of Santa Catarina.
The self-reporting dataset comprises mullet catches from 24 municipalities, from 2003 to 2016 (each orange dot is a report in a given month
in each year); the media reports dataset was mined from 70 media reports from 2001 to 2016 (each dark blue dot is a media report); the
community self-monitoring dataset is a daily log of a traditional, artisanal beach seine fishing community (light blue dot) from 2004 to 2016.
Density plots on the left side of the panels show the distribution of self-reports and media reports along the coast of the state of Santa
Catarina. Bar plots show the number of self-reports, newspaper articles and successful beach seines reported by year.
Filling data gaps with self-reported catches 5

2007
2000

Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsab074/6231554 by Universidade Federal de Santa Catarina user on 17 April 2021
Self−reporting
2006
2009
1000
2004
2010
2008 2005
2011
2012 r = 0.77
0 p = 0.0159

2011 2009 2009


2011
100 2010
2010
Media reports

2007 2007
2004
2004
2008
2008 2006
50 2006

2012 2012
r = 0.54 r = 0.45
2005 2005
0 p = 0.1309 p = 0.2194
Beach seine community

2004
2009
2004 r = 0.41 2004
2009
2009
p = 0.2706
2007
10 2007
2007
2008 2008 2008
2006 2006
2006 2010
5 2010 2010
2012
2012 r = 0.30
2005 2012
r = 0.47
2005

0 2011 p = 0.4388 2011 p = 0.1964 2005


2011

0 2000 4000 6000 0 1000 2000 0 50 100


Industrial fleet Self−reporting Media reports

Figure 2. Correlation between official and alternative mullet catch data (tonnesyear1) in southern Brazil. The official catches come from a
systematic monitoring of the industrial fleet, while the alternative sources include self-reporting by artisanal fishers, data mined from media
reports, and community-based self-monitoring from a traditional beach seine fishing community. For comparison, the datasets were
truncated to the period during which data were available from all sources (from 2004 to 2012). Orange line and black dots indicate the only
significant correlation, between the official data (industrial fleet) and self-reported catches.

Table 1. Coefficients and inclusion probabilities (in brackets) from Discussion


Bayesian structural time-series models with the lowest accumulated We found that self-reported catches of mullet by small-scale fish-
error for each dataset. eries can enrich formal fisheries data. The self-reported catches
Dataset Intercept SST CPUE are mostly explained by our proxy of stock abundance—the
Self-reporting 819.945 (0.376) 10.038 (0.124) 52.061 (0.67) CPUE from the industrial data—but also by greater extents in
Media reports 16.665 (0.182) 1.041 (0.19) 3.52 (0.529) space and time of optimal SST—an important trigger for the
Beach seine community 7.603 (0.305) 0.155 (0.189) 0.275 (0.139) mullet migration. Modelling self-reported catches led to accurate
predictions, following the same trends of the industrial fleet offi-
cial data. Combined, these findings demonstrate that self-
reported catches can provide robust information about mullet
stocks and compensate the lack of institutional data collection at
candidate models (Figure 3). When forecasting with models in- the regional scale.
cluding both autoregressive terms, and optimal SST and CPUE, Mullet fisheries have noticeable variations in space and time.
catches from 2013 to 2016 were mostly well-predicted (Figure 4). The industrial fleet can follow the migrating shoals northwards
This was found for self-reporting (Supplementary Figures S1–S3 while optimal climatic conditions push the migration (Lemos
and Supplementary Table S1) and media reports (Supplementary et al., 2016), and still concentrate landings at large, regional fish-
Figures S4–S6; Supplementary Table S2) datasets, with values ing ports. Conversely, small-scale fisheries must wait for the mi-
falling within the credible interval, with only one marginal grating shoals to approach the coast (Herbst and Hanazaki,
value. The beach seine fishing community dataset was, however, 2014). When assessing coastal fishing communities, conventional
poorly predicted (Supplementary Figures S7–S9; Supplementary monitoring systems in developing countries fail to address such
Table S3). spatiotemporal variability of small-scale fisheries (Gill et al.,
6 A. M. S. Machado et al.

Self−reporting
6000

Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsab074/6231554 by Universidade Federal de Santa Catarina user on 17 April 2021
5000
4000
3000
2000
1000
0 Model
Cumulative absolute error

Media reports Linear Trend + AR + PRED


500 Linear Trend + PRED
400
Linear Trend + AR
300
200 AR
100 PRED
0
Linear Trend
Beach seine community
AR + PRED
50
40
30
20
10
0
2002 2004 2006 2008 2010 2012

Figure 3. Cumulative absolute one-step-ahead errors for each set of candidate Bayesian structural time-series models. Lines show the
cumulative absolute errors for each model, colour coded by the combination of terms—linear trend (moving average), autoregressive (AR),
and predictors (PRED)—included in the model. Small cumulative errors indicate good candidate models. In models with predictors, annual
total catch of each alternative data from artisanal fisheries were fitted as the response variable (see original series in Figure 4) to the number
of days in optimal SST (19–21 C) and CPUE of the purse seine industrial fleet as predictors.

2019). Alternative catch data from small-scale fisheries can cap- Although a local community can provide detailed, fine-scale tem-
ture such variation at the fine scale and provide valid information poral data, it reports only the catches of a single fishing spot,
on mullet stocks, as well as details on fishers’ behaviour and making this data source unable to account for spatial variation.
the fisheries dynamics. Among the three alternative catch data That is, catch data from a fine spatial scale (here a single commu-
sources we explored here, self-reporting showed the most similar nity) might be disconnected from environmental processes at a
pattern to the official data. Self-reported catch data better large scale. However, this bias could be reduced by incorporating
accounted for the variations in space and time, as it relies on networks of multiple, local communities along the coast into a
reports from fishers using beach seines and small vessels that to- single, integrated, citizen-science project. A citizen-science ap-
gether account for a significant fishery harvest along the 500 km proach could broaden the spatiotemporal scale of this alternative
coast of the state of Santa Catarina. In contrast, our findings sug- data source, especially if coupled with new technologies and in-
gest neither newsworthy catches nor fine-scale data from a single formation systems, such as webcams in fishing spots, and online
beach seine fishing community can capture the variation in space social networks and reporting apps whereby fishers could share
and time of mullet fisheries. information on fishing activities along the coast (Devos et al.,
Media reports provide limited information to be used as a 2019).
proxy of stock abundance. Despite covering catches during a sim- Self-reporting by small-scale coastal fisheries provides the most
ilar period and along with a similar extent of the coast than the reliable information, partially because self-reporting includes doz-
self-reporting dataset, media reports are only a small sample of all ens of fishing communities across a large geographical range and
beach seines. Catches around more urbanized and accessible areas can effectively account for environmental conditions that influ-
may be disproportionally more reported in the media. Also, not ence stock abundance. Therefore, self-reporting data coped with
every catch is newsworthy, with unusually large catches more the challenges of sampling small-scale fisheries (e.g. variation in
readily making headline news. Media reports may also be biased landing spots and fishing gear), often overlooked by fisheries
by logistical limitations (e.g. reporters are close to specific fishing management and stock-assessment models. This is notably im-
spots) or editorial decision (e.g. publish only large or early portant for the maintenance of the mullet fishery in southern
catches). Hence, media reports are unlikely to represent a reliable Brazil, which since 2004 has been under threat due to poor man-
sample of mullet catches. But media reports might still play a role agement, lack of law enforcement and intermittent data collection
in reinforcing the regional cultural relevance of mullet fisheries (Abreu-Mota et al., 2018). Despite such difficulties, fishers main-
and shaping peoples’ perceptions of wildlife (Silk et al., 2018; tained their continuous self-report monitoring since 2003 provid-
Francis et al., 2019). ing long-term data. Fine spatial and temporal-scale data provide
The detailed catch logs from a single community also present a sufficient resolution for improved detection of changes in mullet
key limitation: catch data are highly spatially constrained. abundance and distribution, and fisheries effort at local and
Filling data gaps with self-reported catches 7

(a) Industrial fleet CPUE


50 90

Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsab074/6231554 by Universidade Federal de Santa Catarina user on 17 April 2021
CPUE (ton/fishing trips)
Fleet size

Days in optimal SST


40 25
75
60
30 125

20
30

10
Partial
monitoring
0 0

(b) Self-reporting
4000 90
Predicted catches

Days in optimal SST


Credible interval
3000
Yield (ton)

60

2000

30
1000

0 0

(c) Media reports


300 90
Predicted catches

Days in optimal SST


Credible interval
Yield (ton)

200 60

100 30

0 0

(d) Beach seine community


50 90
Predicted catches
Days in optimal SST

Credible interval
40
Yield (ton)

60
30

20
30

10

0 0
2000 2003 2006 2009 2012 2015

Figure 4. Catch time series of mullet fisheries in southern Brazil. Background histograms show the number of days (maximum ¼ 92) in
optimal conditions of mean daily SST from 19 C to 21 C. Temperatures from 19 C to 21 C are considered the best conditions for the mullet
catches. (a) Data generated by an official landing monitoring program of the industrial fleet. Solid line and points show catch per unit of
effort (CPUE), calculated as tonnes by fishing trips. Landings were only partially assessed after 2013, represented by the grey dashed line in
panels (a)–(d). In panels (b)–(d), lines and squares show annual catches from 2000 to 2016, colour coded by each alternative data source.
Lines and diamonds in black show the predicted catches and shaded areas the credible interval estimated from each Bayesian model (see the
full posterior distribution in Supplementary Figures S3, S6, and S9). (b) Catch data from an integrated self-reported approach conducted by
fishers from 24 municipalities in the state of Santa Catarina, southern Brazil; (c) Catch data mined from media reports in the state of Santa
Catarina; (d) Catch data from self-organized and maintained daily records by fishers of only one traditional beach seine fishing community.
8 A. M. S. Machado et al.

regional spatial scales. This is of particular relevance given climate References


change, since increasing SST might disrupt the timing and routes Abreu-Mota, M. A., Medeiros, R. P., and Noernberg, M. A. 2018.

Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsab074/6231554 by Universidade Federal de Santa Catarina user on 17 April 2021
of the mullet migration, forcing changes in fishers’ behaviour and Resilience thinking applied to fisheries management: perspectives
in their ability to follow high biomass concentrations. These fine- for the mullet fishery in Southern-Southeastern Brazil. Regional
scale data will also help to identify local changes in mullet abun- Environmental Change, 18: 2047–2058.
dance and the consequences for their trophic interactions. As Alfaro-Shigueto, J., Mangel, J. C., Pajuelo, M., Dutton, P. H.,
Seminoff, J. A., and Godley, B. J. 2010. Where small can have a
mullet is a key prey species for top predators, such as the bottle- large impact: structure and characterization of small-scale fisher-
nose dolphins Tursiops truncatus gephyreus (Teixeira et al., 2021), ies in Peru. Fisheries Research, 106: 8–17.
this needs to be accounted for in fisheries management and in Bland, L. M., Orme, C. D. L., Bielby, J., Collen, B., Nicholson, E., and
conservation plans. McCarthy, M. A. 2015. Cost-effective assessment of extinction
We recommend implementing policies that encourage the self- risk with limited information. Journal of Applied Ecology, 52:
reporting by artisanal fishing communities and include self- 861–870.
reporting data into a more comprehensive, dynamic, and inte- Canty, S., Funes, M., Box, S., Zylich, K., Derrick, B., Divovich, E.,
Lindop, A., et al. 2019. The hidden value of artisanal fisheries in
grated management approach that could lead to a very-welcomed Honduras. Fisheries Management and Ecology, 26: 249–259.
paradigm shift in fisheries management. For instance, the cultural Castello, L. 2004. A method to count pirarucu Arapaima gigas: fish-
aspects of the mullet fishery are the key to this paradigm shift and ers, assessment, and management. North American Journal of
to the success of self-reporting, or even other alternative Fisheries Management, 24: 379–389.
approaches relying on data from artisanal communities. Costello, C., Ovando, D., Hilborn, R., Gaines, S. D., Deschenes, O.,
Reinforcing such cultural aspects will likely motivate fishers to and Lester, S. E. 2012. Status and solutions for the world’s unas-
sessed fisheries. Science, 338: 517–520.
engage in data collection (Newman et al., 2017), thereby facilitat-
Damasio, L. d. M. A., Lopes, P. F. M., Guariento, R. D., and
ing their inclusion in more sustainable and participative manage- Carvalho, A. R. 2015. Matching fishers’ knowledge and landing
ment plans. Therefore, given that the mullet is a culturally data to overcome data missing in small-scale fisheries. PLoS One,
important species in southern Brazil, by integrating alternative 10: e0133122.
data sources, managers not only benefit from a more detailed de- de Lestang, S., Caputi, N., How, J., Melville-Smith, R., Thomson, A.,
scription of the fishing system, but this also welcomes fishers to and Stephenson, P. 2012. Stock assessment for the west coast rock
lobster fishery. Fisheries research Report, 217. Department of
become key stakeholders in co-management schemes (Freitas
Fisheries, Western Australia. 200 pp.
et al., 2020), boosting their sense of stewardship and their role in
Devos, R. V., Barbosa, G. C., and Vedana, V. 2019. Fish news: percep-
the species conservation for future generations. tual skills, technique and distributed cognition in mullet fishing.
Vibrant: Virtual Brazilian Anthropology, 16: 1–23.
Supplementary data Francis, F. T., Howard, B. R., Berchtold, A. E., Branch, T. A., Chaves,
Supplementary material is available at the ICESJMS online ver- L. C. T., Dunic, J. C., Favaro, B., et al. 2019. Shifting headlines?
sion of the manuscript. Size trends of newsworthy fishes. PeerJ, 7: e6395.
Free, C. M., Jensen, O. P., Anderson, S. C., Gutierrez, N. L., Kleisner,
Acknowledgements K. M., Longo, C., Minto, C., et al. 2020. Blood from a stone: per-
formance of catch-only methods in estimating stock biomass sta-
We thank Ivo da Silva and the collaborators of the Federation of
tus. Fisheries Research, 223: 105452.
Fishermen of Santa Catarina (FEPESC) for sharing the data; the Freitas, C. T., Lopes, P. F. M., Campos-Silva, J. V., Noble, M. M.,
staff from the Public Library of the State of Santa Catarina for the Dyball, R., and Peres, C. A. 2020. Co-management of culturally
help with newspapers archives; Leonardo Viera and Jo~ao Campos important species: a tool to promote biodiversity conservation
for sharing the detailed logs from Estaleirinho beach seines. We and human well-being. People and Nature, 2: 61–81.
thank Mauricio Cantor for thoroughly reviewing the manuscript Garbin, T., Castello, J. P., and Kinas, P. G. 2014. Age, growth, and
text. This study was financed in part by the Coordenaç~ao de mortality of the mullet Mugil liza in Brazil’s southern and south-
eastern coastal regions. Fisheries Research, 149: 61–68.
Aperfeiçoamento de Pessoal de Nı́vel Superior—Brasil (CAPES) -
Gill, D. A., Oxenford, H. A., Turner, R. A., and Schuhmann, P. W.
Finance Code 001 (A.M.S.M.; E.L.H.G). Figures were designed 2019. Making the most of data-poor fisheries: low cost mapping
with icons from flaticon.com. of small island fisheries to inform policy. Marine Policy, 101:
198–207.
Author contributions Herbst, D. F., and Hanazaki, N. 2014. Local ecological knowledge of
F.G.D.J. and A.M.S.M. conceived the study; L.P.F. performed fishers about the life cycle and temporal patterns in the migration
data mining on newspapers; A.M.S.M. and E.L.H.G. explored and of mullet (Mugil liza) in Southern Brazil. Neotropical
Ichthyology, 12: 879–890.
analysed the data; A.M.S.M., E.L.H.G. and F.G.D.J. interpreted
Kindsvater, H. K., Dulvy, N. K., Horswill, C., Juan-Jordá, M.-J.,
the data; A.M.S.M., F.G.D.J. and S.N.I. wrote the original draft of Mangel, M., and Matthiopoulos, J. 2018. Overcoming the Data
the manuscript. All authors reviewed the manuscript and gave fi- Crisis in Biodiversity Conservation. Trends in Ecology and
nal approval for publication Evolution, 33: 676–688.
Lemos, V. M., Varela, A. S., Schwingel, P. R., Muelbert, J. H., and
Data and code availability Vieira, J. P. 2014. Migration and reproductive biology of Mugil
The raw data underlying this article were provided upon request. liza (Teleostei: mugilidae) in south Brazil. Journal of Fish Biology,
85: 671–687.
Data will be shared on request to the corresponding author upon
Lemos, V. M., Avila Troca, D. F., Castello, J. P., and Paes Vieira, J.
permission of the third parties. The formatted dataset and R code 2016. Tracking the southern Brazilian schools of Mugil liza during
used here are available at https://github.com/machadoams/mul reproductive migration using VMS of purse seiners. Latin
let-alternative-data. American Journal of Aquatic Research, 44: 238–246.
Filling data gaps with self-reported catches 9

Lotze, H. K., and Worm, B. 2009. Historical baselines for large ma- state-space models with multiple CPUE data: the case of a mullet
rine animals. Trends in Ecology and Evolution, 24: 254–262. fishery. Scientia Marina, 81: 361.
Mai, A. C. G., Mi~ no, C. I., Marins, L. F. F., Monteiro-Neto, C., Santos, M. L., Lemos, V. M., and Vieira, J. P. 2018. No mullet, no

Downloaded from https://academic.oup.com/icesjms/advance-article/doi/10.1093/icesjms/fsab074/6231554 by Universidade Federal de Santa Catarina user on 17 April 2021
Miranda, L., Schwingel, P. R., Lemos, V. M., et al. 2014. gain: cooperation between dolphins and cast net fishermen in
Microsatellite variation and genetic structuring in Mugil liza southern Brazil. Zoologia, 35: 1–13.
(Teleostei: mugilidae) populations from Argentina and Brazil. Sbragaglia, V., Correia, R. A., Coco, S., and Arlinghaus, R. 2020. Data
Estuarine, Coastal and Shelf Science, 149: 80–86. mining on YouTube reveals fisher group-specific harvesting pat-
Menezes, N. A., Buckup, P. A., Figueiredo, J. L., De, and Moura, R. L. terns and social engagement in recreational anglers and spear-
and De, 2003. Catálogo Das Espécies de Peixes Marinhos Do fishers. ICES Journal of Marine Science, 77: 2234–2244.
Brasil. Museu de Zoologia da Universidade de S~ao Paulo, S~ao SEAP/PR, MMA 2018. Relatório final do comitê de acompanha-
Paulo. 160 pp. mento das cotas de tainha - Safra 2018. Secretaria Especial da
Mion, M., Piras, C., Fortibuoni, T., Celic, I., Franceschini, G., Aquicultura e da Pesca da Presidência da República (SEAP/PR),
Giovanardi, O., Belardinelli, A., et al. 2015. Collection and valida- Ministério do Meio Ambiente (MMA): 91. Brası́lia, DF, Brasil.
tion of self-sampled e-logbook data in a Mediterranean demersal Scott, S. L. 2019. bsts: Bayesian structural time series. R package ver-
trawl fishery. Regional Studies in Marine Science, 2: 76–86.
sion 0.9.1. https://cran.r-project.org/package¼bsts.
MPA/MMA. 2015. Plano de gest~ao para o uso sustentável da tainha,
Scott, S. L., and Varian, H. R. 2014. Predicting the present with
Mugil liza Valenciennes, 1836, no Sudeste e Sul do Brasil.
Bayesian structural time series. International Journal of
Ministério da Pesca e Aquicultura (MPA), Ministério do Meio
Mathematical Modelling and Numerical Optimisation, 5: 4–23.
Ambiente (MMA): 238.
Shester, G. G., and Micheli, F. 2011. Conservation challenges for
MPA. 2011. Statistical Bulletin of Fisheries and Aquaculture from
Brazil (Boletim estatı́stico da Pesca e Aquicultura do Brasil). small-scale fisheries: bycatch and habitat impacts of traps and gill-
Ministério da Pesca e Aquicultura (MPA): 60. Brası́lia, DF, Brazil. nets. Biological Conservation, 144: 1673–1681.
Newman, G., Chandler, M., Clyde, M., McGreavy, B., Haklay, M., Silk, M. J., Crowley, S. L., Woodhead, A. J., and Nuno, A. 2018.
Ballard, H., Gray, S., et al. 2017. Leveraging the power of place in Considering connections between Hollywood and biodiversity
citizen science for effective conservation decision making. conservation. Conservation Biology, 32: 597–606..
Biological Conservation, 208: 55–64. Teixeira, C. R., Botta, S., Daura-Jorge, F. G., Pereira, L. B., Newsome,
Pauly, D. 2006. Major trends in small-scale marine fisheries, with em- S. D., and Sim~ oes-Lopes, P. C. 2021. Niche overlap and diet com-
phasis on developing countries, and some implications for social position of three sympatric coastal dolphin species in the south-
sciences. Maritime Studies, 4: 7–22. west Atlantic Ocean. Marine Mammal Science, 37: 111–126.
Pauly, D., and Zeller, D. 2016. Catch reconstructions reveal that Venturelli, P. A., Hyder, K., and Skov, C. 2017. Angler apps as a
global marine fisheries catches are higher than reported and de- source of recreational fisheries data: opportunities, challenges and
clining. Nature Communications, 7: 10244. proposed standards. Fish and Fisheries, 18: 578–595.
Peterson, D., Hanazaki, N., and Sim~ oes-Lopes, P. C. 2008. Natural re- Vieira, J. P., and Scalabrin, C. 1991. Migraç~ao reprodutiva da tainha
source appropriation in cooperative artisanal fishing between fish- (Mugil platanus Gunther, 1980) no sul do Brasil. Atlântica, 13:
ermen and dolphins (Tursiops truncatus) in Laguna. Brazil. Ocean 131–141.
Coast. Manag, 51: 469–475. Wei, T., and Simko, V. 2017. corrplot: visualization of a correlation
R Development Core Team. 2019. R: A Language and Environment matrix. R package version 0.84. https://cran.r-project.org/pack
for Statistical Computing. R Foundation for Statistical age¼corrplot.
Computing, Vienna, Austria. https://www.r-project.org/. Whitfield, A. K., Panfili, J., and Durand, J.-D. 2012. A global review
Rudd, M. B., and Branch, T. A. 2017. Does unreported catch lead to of the cosmopolitan flathead mullet Mugil cephalus Linnaeus 1758
overfishing? Fish and Fisheries, 18: 313–323. (Teleostei: mugilidae), with emphasis on the biology, genetics,
Sadowski, V., and Dias, E. R. A. 1986. Migraç~ao da tainha (Mugil ecology and fisheries aspects of this apparent species complex.
cephalus Linnaeus, 1758, sensu latu) na costa sul do Brasil. Reviews in Fish Biology and Fisheries, 22: 641–681.
Boletim Do Instituto de Pesca, 13: 31–50. Zeller, D., Harper, S., Zylich, K., and Pauly, D. 2015. Synthesis of
Sant’Ana, R., Gerhard Kinas, P., Villwock de Miranda, L., Schwingel, underreported small-scale fisheries catch in Pacific island waters.
P. R., Castello, J. P., and., and Paes Vieira, J. 2017. Bayesian Coral Reefs, 34: 25–39.

Handling editor: Sarah Kraak

You might also like