ArticlePublisher preview available
To read the full-text of this research, you can request a copy directly from the authors.

Abstract and Figures

106,107 ✉ replying to A. V. Harry & J. M. Braccini Nature https://doi.org/10.1038/s41586-021-03463-w (2021) Our global analysis 1 estimated the overlap and fishing exposure risk (FEI) using the space use of satellite-tracked sharks and longline fishing effort monitored by the automatic identification system (AIS). In the accompanying Comment, Harry and Braccini 2 draw attention to two localized shark-longline vessel overlap hotspots in Australian waters, stating that 47 fishing vessels were misclassified as longline and purse seine vessels in the Global Fishing Watch (GFW) 3 2012-2016 AIS fishing effort data product that we used. This, they propose 2 , results in misi-dentifications that highlight fishing exposure hotspots that are subject to an unexpected level of sensitivity in the analysis and they suggest that misidentifications could broadly affect the calculations of fishing exposure and the central conclusions of our study 1. We acknowledged in our previously published paper 1 that gear reclassifications were likely to occur for a small percentage of the more than 70,000 vessels studied, however, here we demonstrate that even using much larger numbers of vessel reclassifications than those proposed by Harry and Braccini 2 , the central results and conclusions of our paper 1 do not change. In our use of a third-party dataset such as GFW 3 , we stated clearly 1 that the dataset is undergoing continuous refinement to correct for acknowledged contamination of some gear types with others in some regions (for example, drifting longlines with bottom-set longlines off New Zealand 1). The characterization of GFW vessels (gear) is under-taken using two convolutional neural networks that were trained 3 on 45,441 marine vessels (fishing and non-fishing) that identified six Published online: xx xx xxxx Check for updates Q1 Q2 Q3 Q4 Q5 A list of affiliations appears at the end of the paper.
This content is subject to copyright. Terms and conditions apply.
Nature | Vol 595 | 8 July 2021 | E17
Matters arising
Caution over the use of ecological big data
for conservation
Alastair V. Harry1,2 ✉ & J. Matias Braccini1
arising from N. Queiroz et al. Nature https://doi.org/10.1038/s41586-019-1444-4 (2019)
Highly collaborative and data-intensive ecology studies are at the fore-
front of innovative solutions to global issues in conservation and natural
resource management
1,2
. In their spatial risk assessment of industrial-
ized fishing, Queiroz etal.
3
use big data and collaborative science to
outline a global conservation blueprint for pelagic sharks. In Austral-
ian waters, their analysis incorrectly identified global risk ‘hotspots’
in areas that are not subject to fishing and where spatial closures and
other management measures are already in place to protect sharks.
We highlight the potential for large-scale global analyses to misdirect
conservation efforts if not aligned with regional needs and priorities.
Although ecologists have enthusiastically adopted collaborative,
data-driven approaches in recent years, limited attention has been
given to the challenges in this emergent field, including the potential
for these often highly impactful studies to confound management and
conservation actions4. We applaud the collaborative effort by Quei-
roz etal.3 in assimilating satellite tagging data on 1,800 large pelagic
and neritic sharks generated by 153 authors. However, we also caution
against the use of data-intensive methods for guiding policy at the
global scale without proper acknowledgement of their risks, complexi-
ties and limitations.
In their paper, Queiroz etal.
3
identify Australia’s North West Shelf
(NWS) as a global fishing exposure hotspot for sharks on the basis
of spatial overlap with purported drifting longline and purse seine
fishing vessel movements, despite no such fishing having occurred
during the past two decades in this area. When we downscaled the
approach of Queiroz etal.3, we found errors in the data used to evalu-
ate fishing exposure in these waters that were derived using a machine
learning approach applied to vessel automatic identification system
(AIS) location data5.
In Western Australian state waters—an area larger than the Bering
Sea—99.8% of longline and 100% of purse seine AIS data were incorrectly
classified by the machine learning algorithm (Table1 and Fig.1). Incor-
rect classifications included movement data from other types of com-
mercial fishing vessels as well as non-fishing vessels. For example, 95%
of the data for purse seines in Western Australia waters were attributed
to the movements of the research vessel of our agency (which, inci-
dentally, does not undertake purse seine or drifting longline surveys).
The area of the NWS identified as highest risk falls within a spatial
closure of 0.8millionkm
2
in which directed shark fishing has been
prohibited since 2005
6
. Although an area to the northeast remains
open to shark fishing, none has occurred since 20096 and a network of
State and Commonwealth marine reserves has since been implemented
over much of that area. Fishery-independent surveys carried out over
a 17-year period confirm stable or increasing relative abundance and
size of large sharks in the region
6
. Historically, the waters adjacent
to the NWS shelf were indeed important fishing grounds for foreign
drifting longline vessels before their exclusion from Australian waters
in 19977, and for Australian vessels in the subsequent years8. Contem-
porary longlining by a domestic tuna and billfish fishery still occurs,
although these vessels were absent from the AIS data used by Queiroz
etal.3. Since 2005, the intensity of this fishery has decreased and its
footprint shifted to the southwest9.
The approach of Queiroz etal.3 fared better at the scale of the entire
Australian Exclusive Economic Zone and offshore territories (10.2mil-
lion km2), where the tuna and billfish longline fleet operating off east-
ern Australia was correctly classified (Fig.1). However, 51% of drifting
longline data were still incorrect (Table1) and, notably, several demersal
trawlers were also misclassified as being part of the longline fleet. Data
from these vessels led to the incorrect identification of another pelagic
longline risk hotspot within the Great Barrier Reef Marine Park (Fig.1),
where this fishing method is not permitted. In the case of both the
NWS and Great Barrier Reef, the fishing exposure hotspots identified
were due to fewer than five vessels being misclassified, highlighting a
presumably unexpected level of sensitivity in the analysis.
As illustrated here, although patterns identified in global analy-
ses may be broadly informative, they can also be incorrect or misin
-
formative at regional levels where there is the scope for misallocating
resources for conservation and management. Framed alternatively,
https://doi.org/10.1038/s41586-021-03463-w
Received: 6 November 2019
Accepted: 16 March 2021
Published online: 7 July 2021
Check for updates
1Fisheries & Agriculture Resource Management, Department of Primary Industries and Regional Development Western Australia, Hillarys, Western Australia, Australia. 2Centre for Sustainable
Aquatic Ecosystems, Harry Butler Institute, Murdoch University, Murdoch, Western Australia, Australia. e-mail: alastair.harry@gmail.com
Table 1 | Summary of machine-learning classiied ishing
effort data
Western Australia Australia and offshore
territories
Total area (millionkm2)2.27 10.2
Gear type Longline Purse seine Longline Purse seine
Total classified vessels 11 376 15
Incorrectly classified
vessels 9 3 24 11
Fishing hours 41,074 2,650 190,355 7,511
Incorrect fishing hours (%) 99.82% 100% 51% 82%
The machine-learning classiied ishing effort data used by Queiroz etal.3 to evaluate the
risks to sharks from ishing in Western Australian and Australian maritime jurisdictions. The
table shows the total number of vessels classiied as using longlines or purse seine, and their
respective ishing hours, along with the number of vessels and percentage of ishing hours
found to be incorrect. Australia and offshore territories includes all offshore and sub-Antarctic
territories and the Australian Antarctic Territory.
Content courtesy of Springer Nature, terms of use apply. Rights reserved
... Big data has 5 V characteristics: volume, velocity, variety, value, and veracity. Big data itself is an abstract concept [6][7][8]. Relying on the development of the Internet and cloud computing, big data plays an increasingly important role in all walks of life [9]. The Internet of Things is the Internet of everything, namely, the Mobile Internet, which is no longer the link between people but machine to machine, device to device, and system to system [10]. ...
Article
Full-text available
In view of the uncertainty and diversity of the metabolic grey prediction model in the prediction process, resulting in poor prediction effect, a metabolic grey prediction model based on big data and Internet of things technology is constructed. Establish the aerobic metabolic process of human telecontrol and put forward the detection index of aerobic metabolic cycle function in human telecontrol; on this basis, use the metabolic grey prediction model analysis algorithm to determine the active intrusion intention of complex network, establish the intrusion intention attack behavior set function, establish the internal operation architecture under the technology of big data and Internet of things, and realize the construction of metabolic grey prediction model. The experimental results show that the constructed model can realize data prediction, with high confidence level and good effect.
Article
Full-text available
The COVID-19 pandemic has led to reduced anthropogenic pressure on ecosystems in several world areas, but resulting ecosystem responses in these areas have not been investigated. This paper presents an approach to make quick assessments of potential habitat changes in 2020 of eight marine species of commercial importance in the Adriatic Sea. Measurements from floating probes are interpolated through an advection-equation based model. The resulting distributions are then combined with species observations through an ecological niche model to estimate habitat distributions in the past years (2015–2018) at 0.1° spatial resolution. Habitat patterns over 2019 and 2020 are then extracted and explained in terms of specific environmental parameter changes. These changes are finally assessed for their potential dependency on climate change patterns and anthropogenic pressure change due to the pandemic. Our results demonstrate that the combined effect of climate change and the pandemic could have heterogeneous effects on habitat distributions: three species (Squilla mantis, Engraulis encrasicolus, and Solea solea) did not show significant niche distribution change; habitat suitability positively changed for Sepia officinalis, but negatively for Parapenaeus longirostris, due to increased temperature and decreasing dissolved oxygen (in the Adriatic) generally correlated with climate change; the combination of these trends with an average decrease in chlorophyll, probably due to the pandemic, extended the habitat distributions of Merluccius merluccius and Mullus barbatus but reduced Sardina pilchardus distribution. Although our results are based on approximated data and reliable at a macroscopic level, we present a very early insight of modifications that will possibly be observed years after the end of the pandemic when complete data will be available. Our approach is entirely based on Findable, Accessible, Interoperable, and Reusable (FAIR) data and is general enough to be used for other species and areas.
Article
Full-text available
Effective ocean management and conservation of highly migratory species depends on resolving overlap between animal movements and distributions and fishing effort. Yet, this information is lacking at a global scale. Here we show, using a big-data approach combining satellite-tracked movements of pelagic sharks and global fishing fleets, that 24% of the mean monthly space used by sharks falls under the footprint of pelagic longline fisheries. Space use hotspots of commercially valuable sharks and of internationally protected species had the highest overlap with longlines (up to 76% and 64%, respectively) and were also associated with significant increases in fishing effort. We conclude that pelagic sharks have limited spatial refuge from current levels of high-seas fishing effort. Results demonstrate an urgent need for conservation and management measures at high-seas shark hotspots and highlight the potential of simultaneous satellite surveillance of megafauna and fishers as a tool for near-real time, dynamic management.
Article
Full-text available
Population genetics has been increasingly applied to study large sharks over the last decade. Whilst large shark species are often difficult to study with direct methods, improved knowledge is needed for both population management and conservation, especially for species vulnerable to anthropogenic and climatic impacts. The tiger shark, Galeocerdo cuvier, is an apex predator known to play important direct and indirect roles in tropical and subtropical marine ecosystems. While the global and Indo-West Pacific population genetic structure of this species has recently been investigated, questions remain over population structure and demographic history within the western Indian (WIO) and within the western Pacific Oceans (WPO). To address the knowledge gap in tiger shark regional population structures, the genetic diversity of 286 individuals sampled in seven localities was investigated using 27 mi-crosatellite loci and three mitochondrial genes (CR, COI, and cytb). A weak genetic differentiation was observed between the WIO and the WPO, suggesting high genetic connectivity. This result agrees with previous studies and highlights the importance of the pelagic behavior of this species to ensure gene flow. Using approximate Bayesian computation to couple information from both nuclear and mitochondrial markers, evidence of a recent bottleneck in the Holocene (2,000-3,000 years ago) was found, which is the most probable cause for the low genetic diversity observed. A contemporary effective population size as low as 111 [43,369] was estimated during the bottleneck. Together, these results indicate low genetic diversity that may reflect a vulnerable population sensitive to regional pressures. Conservation measures are thus needed to protect a species that is classified as Near Threatened. K E Y W O R D S approximate Bayesian computation, bottleneck, effective population size, microsatellite DNA, mitochondrial DNA, tiger shark
Article
Full-text available
Overexploitation of large apex marine predators is widespread in the world’s oceans, yet the timing and extent of declines are poorly understood. Here we reconstruct a unique fisheries-independent dataset from a shark control programme spanning 1760 km of the Australian coastline over the past 55 years. We report substantial declines (74–92%) of catch per unit effort of hammerhead (Sphyrnidae), whaler (Carcharhinidae), tiger shark (Galeocerdo cuvier) and white sharks (Carcharodon carcharias). Following onset of the program in the 1960s, catch rates in new installations in subsequent decades occurred at a substantially lower rate, indicating regional depletion of shark populations over the past half a century. Concurrent declines in body size and the probability of encountering mature individuals suggests that apex shark populations are more vulnerable to exploitation than previously thought. Ongoing declines and lack of recovery of vulnerable and protected shark species are a cause for concern.
Article
Full-text available
Although fishing is one of the most widespread activities by which humans harvest natural resources, its global footprint is poorly understood and has never been directly quantified. We processed 22 billion automatic identification system messages and tracked >70,000 industrial fishing vessels from 2012 to 2016, creating a global dynamic footprint of fishing effort with spatial and temporal resolution two to three orders of magnitude higher than for previous data sets. Our data show that industrial fishing occurs in >55% of ocean area and has a spatial extent more than four times that of agriculture. We find that global patterns of fishing have surprisingly low sensitivity to short-term economic and environmental variation and a strong response to cultural and political events such as holidays and closures.
Book
Full-text available
This report is strictly confidential. It presents estimates of average daily Foreign Fishing Vessel (FFV) numbers for thirteen months of surveillance data. Data was provided by Border Protection Command (BPC) in the Australian Customs Service on all surveillance flights from June 2006 to June 2007. Apprehension data for the same period was provided by the Australian Fishing Management Authority (AFMA). The first three months of the project, after receiving surveillance data in January 2008, was spent cleaning and preparing the data for analysis using the methodology developed by Salini et al. (2007). This included the calculation of flight swath areas using a modified analysis method (utilising Oracle Spatial) described in Section 4, which dramatically reduced data processing time. Further methodological development was undertaken, namely developing a new estimator (nQ), which was hoped to decrease processing an analysis time. The data was processed in order to calculate: • the area covered for all surveillance flights overlapping Zones 1, 6 and 7 and most of Zone 2, and the areas immediately outside the Australian Exclusive Economic Zone (AEEZ) from those Zones, • the exact location of each legal and illegal foreign fishing vessel (FFV) sighted during those flights. Legal FFVs are those located in regions outside the AEEZ plus some regions inside the AEEZ where either fishing is allowed in general, or where particular types of FFVs are allowed. Illegal FFVs are those located in regions inside the AEEZ which do not have any such allowance. The processed data were organised into months according to the commencement time (in UTC time) of each flight, and the flight in which each FFV was sighted. This data was used as input to a process which provided estimates of the daily number of legal and illegal FFVs in each of these Zones and the individual regions within them. Six estimators are presented, including results for a new ‘quick’ estimator nQ . Although nQ performed reasonably well in comparison to the preferred estimator nC recommended by Salini et al. (2007), it tended to slightly overestimate when the density of FFVs was low and produce more variable results when FFV numbers were high. Key assumptions made in the analysis include: • the swath widths provided by BPC for Type3 vessels for each aircraft and each sea state are representative, • the detection rate within that swath width is 100%, • the reduced swath width for the smaller Type2 vessels (typically, a 15% narrower swath) does not have a significant impact on the estimates, since only 16% of all vessels sighted are not in the Type3 class, • there are few, if any, vessels at grid points which are not surveyed within a given month. Estimates of daily numbers of legal and illegal FFVs are provided for thirteen months from June 2006 to June 2007. Detailed estimates by month and region are provided in Appendix B but the key summary tables for illegal FFVs are shown in Table 6.4. We have also implemented the algorithm of calculating the new estimator, nD, which was not reported previously. This is reassuring that it produced nearly identical results to nC, indicating that our recommendation of using nC or nD is probably reasonable. We concentrated on reporting results using nC to estimate the daily number of FFVs present in a region for easier comparison with the results of Salini et al. (2007) and shows longer-term trends in FFV numbers and apprehensions. The analyses for the 13 months of surveillance data provide a useful temporal comparison between similar times of the year. The results may be summarised as follows: • the daily numbers of illegal FFVs in the study region was highest between August and October both in the present study and the preceding 18 month study period of Salini et al. (2007). This is likely to be due to more favourable weather conditions for smaller vessels travelling from Indonesia, particularly to Zones 6 and 7. • the daily numbers of illegal FFVs increased slightly overall during the first four months of the study to around 18 FFVs, after which time FFV numbers steadily declined to about 3 FFVs per day in June 2007. However, to place these results in context, FFV numbers had declined from about 60 FFVs per day in September 2005 to about 15 FFVs at the start of the present study in June 2006. Results for particular Zones can be summarised as follows (see Section 6.3.1): • The number of FFVs in Zone 1 was generally low, averaging around 2.5 FFVs per day but reaching 5-6 FFVs between January and March 2007. • For Zone 2, estimated daily FFVs were extremely low, with none seen in the last 8 months of the 13 month study period. • For Zone 6, the number of FFVs was highest in September 2006 (13-15), which was similar to FFV numbers in the previous 18 month study. However, after September 2006 numbers dramatically declined to less than one vessel in June 2007. It is important to note that the period of decline coincides with a seasonal decline in FFVs observed in previous years. However, the magnitude of decline clearly indicates a reduction in FFV numbers. • For Zone 7, the number of FFVs showed a significant and consistent decline from around 10 FFVs per day in June 2006 to less than 1 vessel per day in June 2007. To place these results in context with the results of Salini et al. (2007), FFV numbers declined from a peak of around 42 per day in January 2006. A number of assumptions are required to make any estimate of FFV fleet size. Some attempt is made at this in Section 7, but there is little information on which to base good estimates of either “residence time” per visit or numbers of visits per month by each vessel. A plausible range of values for these is used and leads to a coarse estimate of FFV fleet size in the range between 15 and 180 depending on the assumptions used. This demonstrates a significant reduction in fleet size estimated to be between 55 and 600 in 2005-2006. These assumptions can be refined as outlined under Recommendation 6 and would lead to more precise estimates. Based on the analyses of surveillance and apprehension data from June 2006 – June 2007 inclusive, we make the following recommendations for improving surveillance data gathering: Recommendation 1: We recommend that data for the months since June 2007 be analysed using the methods in this report, to provide estimates for at least one more year to confirm the decline in FFVs is real and not due to artefacts such as seasonal incursions. Recommendation 2: We recommend that BPC adopt the methods outlined in this report (using estimator nC or nD) into their normal practice. The new ‘quick’ estimator, nQ, is useful for obtaining ‘ball park’ estimates of FFV numbers, but should not be used for accurately estimating FFV numbers. Recommendation 3: We recommend that a further analysis of the current database be undertaken to provide empirical estimates of detection probabilities as a function of distance from the aircraft's track, in order to validate or refine the swath widths currently in use. Recommendation 4: We recommend that BPC routinely collect the time and position of the aircraft at the point where any FFV sighting takes place. Recommendation 5: The data field recording “Type of vessel” needs to be redefined to provide an unambiguous and clearer distinction between the different types of fishing vessels. Recommendation 6: More effort needs to be given to the estimation of “residence time” and “number of visits per month” in order to provide reliable estimates of the size of the fishing fleet involved. The extent to which apprehension effort may influence the residence time should also be explored. Recommendation 7: We recommend carrying out statistical evaluation of all the estimators and providing “hard” evidence on their performance, in particular which one is the most appropriate for the given dataset which has a biased nature of sampling because flight path is not random in the surveillance area. This will allow us to access to which extent the quasi-Poisson assumption is valid which is vital for our calculation of the standard errors. The study should include bias, variance, and mean square errors (MSE). This would require high statistical skills and extensive simulation studies. Recommendation 8: We recommend that a further study be undertaken to estimate the size of the catch of shark species from these illegal FFVs in order to improve the quality of stock assessment of sharks in northern Australia, and to improve current ecosystem models that can predict the ecological effects of IUU on the broader ecosystem.
Article
Reliable information for population assessments is rare for sharks. We quantified patterns in catch rates and mean size for numerous tropical and subtropical species from 15 years of fishery-independent surveys (2002–2017) in northwestern Australia. This study region represents an area of ~0.8 million km2 which was closed to commercial shark fishing from 1993 or 2005 onward due to the very high State-wide catches of sandbar (Carcharhinus plumbeus) and dusky (Carcharhinus obscurus) sharks. A total of 43 shark and ray species were sampled, with sandbar shark being the most commonly caught species, followed by milk (Rhizoprionodon acutus), spot-tail (Carcharhinus sorrah), tiger (Galeocerdo cuvier), blacktip (Carcharhinus limbatus and Carcharhinus tilstoni), dusky and sliteye (Loxodon macrorhinus) sharks, and scalloped hammerhead (Sphyrna lewini). For sandbar shark, catch rates increased between 2008 and 2017 whereas for other taxa catch rates were mostly stable (albeit fluctuating). Mean size at capture fluctuated across years with no particular trends. Unlike for other parts of the world, catch rates and mean size of northwestern Australian sharks have been stable or increased in recent years. Though most shark species have conservative life histories, when science, management and enforcement work synergistically, sustainable resource use, recovery and conservation outcomes can all be achieved.
Article
There have been efforts around the globe to track individuals of many marine species and assess their movements and distribution, with the putative goal of supporting their conservation and management. Determining whether, and how, tracking data have been successfully applied to address real-world conservation issues is, however, difficult. Here, we compile a broad range of case studies from diverse marine taxa to show how tracking data have helped inform conservation policy and management, including reductions in fisheries bycatch and vessel strikes, and the design and administration of marine protected areas and important habitats. Using these examples, we highlight pathways through which the past and future investment in collecting animal tracking data might be better used to achieve tangible conservation benefits.
Article
Illegal, unreported and unregulated (IUU) fishing can lead to irreversible ecosystem changes; however, it is also one of the most difficult forms of fishing to manage and deter. In northern Australia over the past decade there has been a large increase in small-scale IUU fishing. We suggest that this small-scale fishing has arisen because of long-term fisheries over-exploitation in South East Asia. This IUU fishing forms part of the expansion-and-displacement cycle that can rapidly reduce biomass and alter the trophic structure of local ecosystems. With increasing human populations in the region, the pressure to fish illegally is likely to increase. Regional responses are required to deter and monitor the illegal over-exploitation of fisheries resources, which is critical to secure ecosystem stability as climate change and other destructive human activities threaten food security.