Reply to: Shark mortality cannot be assessed by fishery overlap alone

Many shark species worldwide are vulnerable to overexploitation due to fishing. Using only the horizontal spatial overlap between the space use of 23 satellite-tracked shark species and the fishing distribution of pelagic longline fisheries tracked using an automatic identification system, Queiroz et al. concluded that sharks are at high risk when substantial horizontal overlap occurs. This approach to estimate fishing susceptibility, coupled with limited tag-based shark distributions to estimate fishing exposure index (FEI) hotspots, severely limits their findings and, therefore, conclusions. We challenge several assumptions made by the authors and argue that horizontal overlap alone is an unreliable indicator of susceptibility because other factors contribute considerably to catch risk, as shown https://doi.org/10.1038/s41586-021-03396-4

Landings from FAO total capture production (http://www.fao.org/fishery/statistics/global-capture-production/query/es) and shark FEI were calculated as described in Queiroz et al. 1  to account for zero catches; (ii) using data from 2012-2016 for which fishing effort data from the automatic identification system were available for estimating the average annual landings; and (iii) including Sphyrna spp. and/or hammerhead NEI in the hammerhead group because they may comprise S. zygaena, S. mokarran or S. lewini. NEI, not otherwise identified. a P value was statistically significant at the 5% level of significance.
by well-established ecological risk assessment methods 2,3 . Moreover, we consider that the locations of FEI hotspots are biased towards areas for which tagging data are available. These limitations may misdirect urgent management actions that are needed to mitigate-globally and holistically-true fishing risks for sharks in all ocean regions. Our criticisms comprise the four main points below. First, horizontal overlap does not provide a robust risk estimate. The two-dimensional horizontal overlap between the distribution of a species and fishing effort (that is, 'availability') is only one component of susceptibility explaining risk 4 , which also includes encounterability, selectivity and post-capture mortality. Encounterability is the potential for a species to interact with fishing gear within its depth range. Selectivity is the propensity for an organism to be caught once it encounters the fishing gear 4 . For example, even with 100% horizontal overlap between the distribution of a mesopelagic shark and a surface pole-and-line fishery, encounterability and selectivity are negligible, and thus the species would have low catchability and risk. The species asessed by Queiroz et al. 1 occupy different depth ranges and undergo diel vertical migrations 5,6 that result in different encounterability 7 ; however, this was not considered. Furthermore, shark species have different life-history traits, behaviours and mouth morphologies that differentially affect their selectivity to baited longline hooks 8,9 , which should also be included in risk assessments.
Current handling and release practices can reduce at-vessel and post-release mortality 10 . For example, at-vessel and post-release survival of pelagic sharks-including the great hammerhead and tiger shark species analysed by Queiroz et al. 1 -ranges from 33% to 100% 6,11 ; information that was also omitted from the risk estimation of Queiroz et al. 1 .
Widely used ecological risk assessments 2,3 include all susceptibility (and productivity) components 3 to estimate risk. Therefore, risk may not necessarily be high with high horizontal overlap if encounterability, selectivity and/or post-capture mortality is low (for example, for tiger sharks 7,9 ).
Second, the fishing exposure index (FEI) developed by Queiroz et al. 1 -relative shark density multiplied by fishing effort-is not a robust proxy for fishing-induced shark mortality as it is fundamentally another measure of geospatial overlap. The authors claim that FEI "reflects fishing-induced shark mortality" based on a linear relationship between FAO landings data for the North Atlantic and FEI values for eight shark species (extended data figure 5 of Queiroz et al. 1 ). We accessed the FAO statistics, following the authors' description of the data used, to calculate the relationship between shark landings and FEI. We tested different options of (1) landing periods; (2) all versus positive years; and (3) including non-identified hammerhead landings as hammerhead landings, and found no significant relationships (P > 0.1) in all combinations, except the single case cited by Queiroz et al. 1 (Table 1). Moreover, the relationship between FEI and catch including all species does not reflect fishing mortality, unless the abundance of each species is the same (catch = fishing mortality × abundance), which is not the case. For example, if abundance is low-as for white sharks-even low catches could reflect high fishing mortality and, vice versa, high catches could indicate higher shark abundance but not necessarily higher fishing mortality. Thus, FEI is not a reliable proxy for fishing-induced shark mortality. Because the conclusions of Queiroz et al. 1 hinge on FEI representing risk and fishing mortality, their conclusions lack support. Third, the use of 'exposure risk plots' between spatial overlap and FEI by species using the mean overlap and mean FEI across all species in a region as a reference point to delineate risk is misleading. High risk (red quadrants in figure 3 of Queiroz et al. 1 ) means that the risk of a species is above average, which may occur when exposure risk is low (for example, blue shark (Prionace glauca), eastern Pacific; great white shark (Carcharodon carcharias), Oceania). Worryingly, high-risk species can be considered low risk in a region if most sharks show high overlap and FEI.
Moreover, based on FEI hotspots, Queiroz et al. 1 concluded that "high fishing effort is focused on extensive shark hotspots globally". We disagree as there is a significant mismatch between FEI hotspots and shark density hotspots and fishing effort hotspots ( Table 2), revealing that shark hotspots are not related to main fishing effort areas (Fig. 1b). For example, the true-positive rate when FEI hotspots correctly identifies a fishing effort hotspot is 9% (Table 2). Furthermore, using their methodology, FEI hotspots cannot be identified in regions for which no fishing data from the automatic identification system are available (for example, neritic regions within Exclusive Economic Zones (EEZs)) or in areas with no shark tagging information (Fig. 1).
Finally, although the size of grid cells did not affect species risk exposure plots and species occurrence within the high-or low-risk zones, the absolute values of overlap and FEI are greatly affected by grid cell size. Supplementary table 9 of Queiroz et al. 1 shows that the mean overlap decreased from 21.6% at a resolution of 1° × 1° to 5.03% overlap using a resolution of 0.10° × 0.10°, whereas FEI decreased from 3.0 × 10 −5 to 3.9 × 10 −8 , respectively. The concomitantly large decrease in overlap and FEI may therefore affect FEI hotspots and, thus, compromise the results of Queiroz et al. 1 .
Queiroz et al. 1 concluded that limited spatial refuges for sharks exist in Areas Beyond National Jurisdictions (ABNJs). Of the total FEIs in their data (https://github.com/GlobalSharkMovement/GlobalSpatialRisk), 36% and 64% lie in ABNJs and EEZs, respectively. Furthermore, 56% of ABNJs (7,856 km 2 ) and 67% of EEZs (8,325 km 2 ) have FEI values of zero, thus clearly identifying possible refuge areas (Fig. 1). Although Queiroz et al. 1 underestimate refugia due to limited tagging, their results do not support the conclusion of "limited spatial refuge" in ABNJs.
To conclude, we agree with Queiroz et al. 1 about the need for improved conservation and management measures for sharks as mounting evidence suggests that their populations are being subjected to increasing pressure globally by fishing 12 . We also agree that 'industrial' pelagic fisheries have an important role in these impacts, but note that regional fishery management organizations for tuna have made some progress by adopting several shark non-retention and mitigation management measures 13 . There is also growing evidence 14 that the fleet size and impact of the often less regulated and monitored artisanal coastal fisheries-which primarily use longlines and gillnets-can be as large as those of industrial fleets that fish the ABNJs 15 . The magnitude of total shark catches by these fisheries must be better understood to determine the true global risk for sharks.
The analysis by Queiroz et al. 1 defines risk based only on horizontal overlap, equates FEI to fishing mortality and estimates FEI only on the basis of areas for which shark tagging data are available. It therefore identifies FEI hotspots that are not necessarily the areas in which sharks are at greatest risk from fishing. Therefore, using the hotspots identified by Queiroz et al. 1 to define spatial management measures may not only focus protection in sub-optimal areas, but could also detract from management efforts across 100% of shark distributions to mitigate mortality by reducing fishery encounterability, selectivity and post-capture mortality. Such management approaches, in collaboration with regional fishery management organizations for tuna and small-scale fleets, are essential to achieving meaningful reductions in risks from fishing for sharks.

Reporting summary
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.

Data availability
To prepare Table 1 and linear regressions between North Atlantic annual shark landings (FAO total capture production) and shark FEI as calculated by Queiroz et al. 1 , FAO statistics available from http:// www.fao.org/fishery/statistics/global-capture-production/query/es were used following the description of the data by Queiroz et al. 1 . To produce Table 2 and Fig. 1, data from Queiroz et al. 1 were used from https://github.com/GlobalSharkMovement/GlobalSpatialRisk. Our previously published paper 1 provided global fine-scale spatiotemporal estimates (1° × 1°; monthly) of overlap and fishing exposure risk (FEI) between satellite-tracked shark space use and automatic identification system (AIS) longline fishing effort. We did not assess shark mortality directly, but in addition to replying to the Comment by Murua et al. 2 , we confirm-using regression analysis of spatially matched data-that fishing-induced pelagic shark mortality (catch per unit effort (CPUE)) is greater where FEI is higher. We focused on assessing shark horizontal spatiotemporal overlap and exposure risk with fisheries because spatial overlap is a major driver of fishing capture susceptibility and previous shark ecological risk assessments (ERAs) assumed a homogenous shark density within species-range distributions [3][4][5] or used coarse-scale modelled occurrence data, rather than more ecologically realistic risk estimates in heterogeneous habitats that were selected by sharks over time. Furthermore, our shark spatial exposure risk implicitly accounts for other susceptibility factors with equal or similar probabilities to those commonly used in shark ERAs 3,5 .
First, actual depth distributions are seldom incorporated in shark ERAs and full vertical overlap with an encounterability probability of one is often applied 3,5 . This is an implicit assumption in our FEI as the pelagic species that we tracked exhibit vertical movements that overlap with depths of pelagic longlines (for example, 18-267 m) 6 during both the day and night 7 . Second, we account for selectivity by focusing our fisheries-independent spatial estimates directly on individuals that were actually caught by the focal fisheries. The majority of the 1,804 sharks tagged were caught on commercial-type longline hooks before release. This is equivalent to a selectivity probability of around one as used in shark ERAs 5 . Third, the commercially valuable sharks that we tracked are seldom discarded by major high-seas longlining fleets 8 , indicating that an implicit assumption of a fishing mortality probability of one does not substantially overestimate the mortality that occurs. Murua et al. 2 overlook that fact that although some species with fishing prohibitions (such as silky and great hammerhead sharks) may be released alive, reported hooking mortalities are high (for example, 56% for silky sharks and 96% for great hammerhead sharks) 9,10 in addition to at least around 50% post-release mortality 11,12 . Collectively, this indicates 78-98% total mortality even of prohibited species. The similar assumptions between our analyses and previous assessments result in comparable susceptibility estimates that will not alter our FEI. For example, we estimated that shortfin mako, blue and porbeagle sharks as the highest exposure risk species in the North Atlantic, which were also the shark species with the highest estimated susceptibilities to longline fishing in a recent Atlantic shark ERA 4 .
Regarding FEI being related to fishing-induced shark mortality, we stated 1 that the significant positive relationship between Food and Agriculture Organization (FAO) fishery landings data and individual-species mean FEI "implies that the index reflects fishing-induced shark mortality". Our conclusion was appropriately cautious because we recognized that FAO landings data were limited in quality, aggregated at regional scales and subject to high levels of unreported or underreported data 13 , and are potentially unrelated to shark relative abundances. Murua et al. 2 confirm the result presented in our paper and also show nine further data combinations that we did not test resulting in eight non-significant positive relationships. However, having few data points (n = 8 species per test) when comparing the spatial complexity of FEI (1° × 1° grid) to non-spatially explicit FAO datasets-given the high variability in the quality of landings data-biases results towards non-significance. To address this, we tested linear-regression models for spatially matched data, including longline CPUE (a relative measure of abundance) of pelagic sharks as the response variable and FEI, fishing effort and number of longline sets as explanatory variables, including interactions with year or month (Supplementary Information). The best model when testing interactions with month was for fishing effort (Akaike information criterion weights (wAIC) = 1), but the deviance explained was similar between this model (46%) and those models that included FEI (42%) or the number of sets (43%). When testing interactions with year, the best model was FEI (wAIC = 0.89), showing a significant and positive relationship with CPUE (n = 523, r 2 = 0.11, F 9,513 = 7.17, P < 0.0001). Bootstrapping tests randomly by removing 1-25% of data confirmed that the best model alternates between fishing effort and FEI as an explanatory variable of shark CPUE. For spatially matched data, therefore, pelagic shark CPUE is significantly greater in areas in which FEI is higher and is as good an explanatory variable of CPUE as fishing effort itself, corroborating our previously published result 1 that FEI reflects fishing-induced shark mortality.
Using spatial exposure risk plots between overlap and FEI to indicate higher or lower than average exposure risk (that is, potential capture susceptibility) is not misleading because the categorization relates specifically to areas in which shark species were tracked and overlap with fishing effort occurred. We previously showed 1 the FEI maps alongside the exposure risk plots to make this point clear. Higher exposure risk can be driven by high FEI when it occurs in specific space-use areas, even if spatial overlap appears relatively low in a region (for example, for white sharks in Oceania). Correct interpretation of our exposure risk estimates requires reference to the areas over which shark hotspots and fishing effort occurred.
FEI hotspots driven by shark hotspots in large-scale ocean ecosystems (for example, the Gulf Stream) led us to conclude that high levels of fishing effort are focused on extensive hotspots of shark space use 1  effort areas. However, we did not calculate fishing effort hotspots nor relate them to shark density hotspots or FEI hotspots because this approach ignores key drivers of FEI hotspots (see below) and is selective of available data. We did not equate high levels of fishing effort solely to fishing effort hotspots because sharks are often caught and retained by fishing vessels that did not specifically target sharks, so shark relative density or FEI hotspots should not be expected to correctly predict fishing effort hotspots in the majority of cases. Rather, we showed that FEI hotspots arise from shark relative density hotspots, high fishing effort levels (not only the highest fishing effort levels considered by Murua et al. 2 ), a combination of both, and some (<2%) are driven by lower shark densities or fishing intensities (Extended Data Table 1).
Consistent with our conclusion, vast areas with higher-than-average fishing effort extend across major shark density and FEI hotspots (Fig. 1). For example, FEI hotspots overlap with shark density hotspots in 56% of grid cells globally, and overlap with higher-than-average fishing effort in 81% of grid cells (Fig. 1). That shark density hotspots and higher-than-average fishing effort together drive 39% of FEI hotspots supports our original conclusion. This is even more clearly seen for individual species (Fig. 1b-e and Extended Data Table 2). For example, globally, blue shark hotspots and high fishing effort together drive 50% of blue shark FEI hotspots (Fig. 1b, d) and, regionally, white shark hotspots and high fishing effort in the northeast Pacific together drive 67% of FEI hotspots (Fig. 1c, e). The claim by Murua et al. 2 that shark hotspots are not related to main fishing effort areas is not supported when all drivers of FEI hotspots are considered.
Furthermore, large reductions in grid cell size do not affect FEI hotspots. We previously provided results showing, as expected, that reductions from 2 × 2° to 0.1 × 0.1° lowers absolute overlap and FEI values but relative exposure-risk plots remain unchanged (extended data figure  4 and supplementary figure 4 of ref. 1 ). It is possible that our results and conclusions could be affected if the spatial positions and extent of FEI hotspots-indicating potential changes in relative drivers that affect overlap and FEI estimates (see above)-were substantially altered as the size of the grid cells decreases. However, the position and extent of FEI hotspots remain largely unchanged as grid size decreases (Fig. 2a, b), indicating that the results and conclusions concerning FEI hotspots are highly unlikely to be affected.
Lastly, we disagree that our analyses do not support our conclusion of limited spatial refuge for pelagic sharks from current levels of fishing effort in Areas Beyond National Jurisdictions (ABNJs). Globally, only about one third of ABNJ shark hotspot grid cells were free from AIS-tracked longline fishing effort, indicating that fishing effort overlapped with the majority of shark hotspots (Fig. 2c, d and Extended Data Table 3). Some heavily fished regions showed even lower levels of spatial refuge, only 13% and 20% of Indian Ocean and North Atlantic shark hotspot grid cells, respectively, were free from fishing effort. Hotspots are areas of preferred habitat where sharks spent most time 1 , thus it was justified to conclude that for the results presented there was limited spatial refuge in ABNJs. The percentage of spatial refuge for sharks in ABNJs decreases to <25% of shark relative density hotspots when additional AIS data that were not previously available are included (Extended Data Table 4), indicating that our original spatial refuges were actually overestimated.
In summary, we think that the arguments presented neither call into question our results and conclusions nor misdirect management efforts as our exposure risk estimates are spatially and temporally explicit. We do not dispute that regional fishery management organizations for tuna have put management measures in place; these were described in our paper 1 . Nevertheless, pelagic sharks have declined globally over many decades [13][14][15] , strongly indicating that additional measures are still required to conserve populations effectively, including more complete data reporting, catch quotas and greater enforcement 13,15 . The data and analyses in our paper 1 contribute to this goal. Indeed, regional fishery management organizations for tuna state that data on biologically important areas, spatiotemporal distributions of shark stocks and interactions with fishing fleets 8 are needed to aid management. We have provided a first step by making available fishery-independent data 1 on shark spatial density and hotspot locations to complement current assessment approaches.

Reporting summary
Further information on experimental design is available in the Nature Research Reporting Summary linked to this paper.

Corresponding author(s): David W. Sims
Last updated by author(s): 4th February 2021 Reporting Summary Nature Research wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency in reporting. For further information on Nature Research policies, see our Editorial Policies and the Editorial Policy Checklist.

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.

n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly The statistical test(s) used AND whether they are one-or two-sided Only common tests should be described solely by name; describe more complex techniques in the Methods section.
A description of all covariates tested A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient) AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals) For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted

Software and code
Policy information about availability of computer code Data collection No data collection software was used.

Data analysis
All analyses described were undertaken in R.
For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Research guidelines for submitting code & software for further information.

Data
Policy information about availability of data All manuscripts must include a data availability statement. This statement should provide the following information, where applicable: -Accession codes, unique identifiers, or web links for publicly available datasets -A list of figures that have associated raw data -A description of any restrictions on data availability Data used in linear regression modelling are available on GitHub (https://github.com/GlobalSharkMovement/GlobalSpatialRisk/derived_data/). Data and source code used for preparing figure maps (shark relative spatial density, longline-fishing effort and shark-longline-fishing overlap and FEI) are available on GitHub (https://github.com/GlobalSharkMovement/GlobalSpatialRisk).