Estimating the sustainability of towed fishing‐gear impacts on seabed habitats: a simple quantitative risk assessment method applicable to data‐limited fisheries

Impacts of bottom fishing, particularly trawling and dredging, on seabed (benthic) habitats are commonly perceived to pose serious environmental risks. Quantitative ecological risk assessment can be used to evaluate actual risks and to help guide the choice of management measures needed to meet sustainability objectives. We develop and apply a quantitative method for assessing the risks to benthic habitats by towed bottom‐fishing gears. The method is based on a simple equation for relative benthic status (RBS), derived by solving the logistic population growth equation for the equilibrium state. Estimating RBS requires only maps of fishing intensity and habitat type – and parameters for impact and recovery rates, which may be taken from meta‐analyses of multiple experimental studies of towed‐gear impacts. The aggregate status of habitats in an assessed region is indicated by the distribution of RBS values for the region. The application of RBS is illustrated for a tropical shrimp‐trawl fishery. The status of trawled habitats and their RBS value depend on impact rate (depletion per trawl), recovery rate and exposure to trawling. In the shrimp‐trawl fishery region, gravel habitat was most sensitive, and though less exposed than sand or muddy‐sand, was most affected overall (regional RBS = 91% relative to un‐trawled RBS = 100%). Muddy‐sand was less sensitive, and though relatively most exposed, was less affected overall (RBS = 95%). Sand was most heavily trawled but least sensitive and least affected overall (RBS = 98%). Region‐wide, >94% of habitat area had >80% RBS because most trawling and impacts were confined to small areas. RBS was also applied to the region's benthic invertebrate communities with similar results. Conclusions. Unlike qualitative or categorical trait‐based risk assessments, the RBS method provides a quantitative estimate of status relative to an unimpacted baseline, with minimal requirements for input data. It could be applied to bottom‐contact fisheries world‐wide, including situations where detailed data on characteristics of seabed habitats, or the abundance of seabed fauna are not available. The approach supports assessment against sustainability criteria and evaluation of alternative management strategies (e.g. closed areas, effort management, gear modifications).

Estimating the sustainability of towed fishing-gear impacts on seabed habitats: a simple quantitative risk assessment method applicable to data-limited fisheries.
1. Impacts of bottom fishing, particularly trawling and dredging, on seabed (benthic) habitats are commonly perceived to pose serious environmental risks. Quantitative ecological risk assessment can be used to evaluate actual risks and to help guide the choice of management measures needed to meet sustainability objectives. 2. We develop and apply a quantitative method for assessing the risks to benthic habitats by towed bottom-fishing gears. The method is based on a simple equation for relative benthic status (RBS), derived by solving the logistic population growth equation for the equilibrium state. Estimating RBS requires only maps of fishing intensity and habitat type -and parameters for impact and recovery rates, which may be taken from metaanalyses of multiple experimental studies of towed-gear impacts. The aggregate status of habitats in an assessed region is indicated by the distribution of RBS values for the region. The application of RBS is illustrated for a tropical shrimp-trawl fishery. 3. The status of trawled habitats and their RBS value depend on impact rate (depletion per trawl), recovery rate and exposure to trawling. In the shrimp-trawl fishery region, gravel habitat was most sensitive, and though less exposed than sand or muddy-sand, was most affected overall (regional RBS=91% relative to un-trawled RBS=100%). Muddy-sand was less sensitive, and though relatively most exposed, was less affected overall (RBS=95%). Sand was most heavily trawled but least sensitive and least affected overall (RBS=98%). Region-wide, >94% of habitat area had >80% RBS because most trawling and impacts were confined to small areas. RBS was also applied to the region's benthic invertebrate communities with similar results. 4. Conclusions. Unlike qualitative or categorical trait-based risk assessments, the RBS method provides a quantitative estimate of status relative to an unimpacted baseline, with minimal requirements for input data. It could be applied to bottom-contact fisheries worldwide, including situations where detailed data on characteristics of seabed habitats, or the abundance of seabed fauna are not available. The approach supports assessment against sustainability criteria and evaluation of alternative management strategies (e.g. closed areas, effort management, gear modifications).  3. The status of trawled habitats and their RBS value depend on impact rate (depletion per trawl), recovery 36 rate and exposure to trawling. In the shrimp-trawl fishery region, gravel habitat was most sensitive, and 37 though less exposed than sand or muddy-sand, was most affected overall (regional RBS=91% relative to un-38 trawled RBS=100%). Muddy-sand was less sensitive, and though relatively most exposed, was less affected 39 overall (RBS=95%). Sand was most heavily trawled but least sensitive and least affected overall (RBS=98%). 40 Region-wide, >94% of habitat area had >80% RBS because most trawling and impacts were confined to small 41 areas. RBS was also applied to the region's benthic invertebrate communities with similar results.  Here, we develop a simple, widely applicable quantitative level-3 ERAEF method for assessing relative benthic 92 status (RBS) in areas fished with towed bottom-contact gears. As an example application, we assess RBS for 93 seabed habitats and benthic invertebrate taxa in a tropical trawl fishery. 94

DEVELOPMENT OF THE RBS METHOD 96
The dynamics of the abundance of seabed communities are assumed to be described by a Schaefer where δB/δt is the rate of change in abundance B in time t, R is recovery rate, K is carrying capacity, D is trawl 100 depletion rate (specific to different gear-types) and F is trawling effort as swept-area ratio (the total area 101 swept by trawl gear within a given area of seabed, divided by that seabed area). This model has been used for The usual implementation of the logistic equation is dynamic, with trawling-induced mortality input as a time-111 series and abundance output as a time-series. However, for data-limited situations, an approach that does not 112 rely on a time series of inputs is desirable. If the question about risk is framed as "will the current level of 113 fishing lead (or has it led) to habitat status that compromises a defined management objective?", then a 114 simpler approach can be used to assess status. This involves solving the logistic equation for the equilibrium 115 state (i.e. δB/δt=0), in which case eqn 1 has the solution: 116 where B/K represents relative benthic status (RBS). Thus the equation can be used when K is unknown, or 117 cannot be clearly defined. The method assumes that the current (or future) level of trawl effort F has been (or 118 will be) applied indefinitely. An analogous approach, based on this assumption, was used to project long-term 119 biomass of benthic species under constant F (Appendix C in Ellis, Pantus & Pitcher 2014). 120 Estimation of RBS (eqn 2) requires relatively few parameters: habitat type, trawl effort, depletion rates and 121 recovery rates. Regional application of RBS requires maps of habitats and trawl effort; both should be 122 determined for grid cells at a scale that adequately captures within-region heterogeneity of habitats and trawl 123 effort. Grid cells of areas ~1-5 km 2 typically are small enough that the distribution of fishing effort within those intensity may be derived from fishing vessel logbooks and/or vessel monitoring systems (VMS); typically as 126 hours of effort. These data need to be gridded at a suitable cell resolution, and converted to trawl swept-area 127 ratio (using information on gear swept-width, tow speeds, and grid-cell area). 128 Trawl impacts differ among gear types and habitats, and recovery rates differ among habitats. Typically, 129 habitats in stable environments are dominated by longer-lived and more sensitive biota that recover slowly, 130 while habitats exposed to high levels of natural disturbance (e.g. mobile sediments) tend to be dominated by  type, and benthos taxa. Estimates of i for gear-by-habitat and for taxa-by-habitat (for otter trawl) were 150 inferred assuming additivity on the log scale and ignoring the possibility of interactions (Table 1). Impact values 151 were assumed, conservatively, to represent the effect of a single trawl pass, although this may not have been 152 the case in all studies included in the meta-analysis. The impact values (Table 1) where B 0 is the abundance immediately after experimental impact. B 0 is a function of depletion rate d per 167 trawl and the number of experimental trawls T; thus, B 0 =K(1−d) T and the complete model is: 168 presented their figure 5 on a log(response ratio) scale (i.e. relative to 1). T was assumed to be unity because, in 170 this instance, d was separately estimated by eqn 4 and to estimate r it was only necessary for the model to fit 171 abundance immediately after impact. If, in future, eqn 6 was used to simultaneously estimate both r and d, the 172 actual value of T would be important. 173 The recovery information in Collie et al. (2000) was for habitat and taxa main effects only. Habitat-by-taxa 174 recovery rates for 3 taxa in 4 habitats were inferred in the same manner as those for impact effects. The 175 experimental scale r estimates were adjusted, using eqn 3, to grid-scale R.  The Gulf-wide status of habitats, accounting for their different sensitivity and exposure to trawling, was 206 quantified by plotting the distribution of RBS values against proportion of habitat area, by mapping their 207 spatial distribution and by the region-wide average RBS value. 208 RBS was also assessed for three benthos taxa. In addition, their absolute status was estimated using 209 information on their distributions (see Appendix S1). 210

DEPLETION AND RECOVERY RATES 212
The status of trawled habitats, and hence their RBS score, depends on their depletion rate, recovery rate and 213 exposure to trawling. Gravel and Malacostraca have the highest depletion rates in response to otter trawling, 214 whereas Mud and Bivalvia have the lowest (Table 2). Sand and Polychaeta have the highest grid recovery rates 215 (R), whereas Gravel and Bivalvia have the lowest (Table 3). The sensitivity of habitats or taxa to trawling is 216 given by the ratio D/R and the critical level of F that would drive their equilibrium status to 0 is R/D. Hence, 217 Gravel is the most sensitive habitat and has critical F=4.6, whereas Sand is least sensitive. Malacostraca are the 218 most sensitive taxa and have critical F=5.7 (pooled across habitats), whereas Bivalvia are least sensitive. Most trawling footprint, by area, occurred on Sand, followed by muddy-Sand, Gravel and Mud (Table 4). 233 However, relatively, muddy-Sand was proportionally more exposed to trawling followed by Sand and Gravel 234 ( Figure 4); there are few areas of Mud and these were least exposed. A similar proportion (~10%) of each 235 habitat, except Mud, was exposed to high effort (swept-ratio >~2). 236

STATUS ASSESSMENT 237
The RBS (B/K) of each habitat type as a function of trawling effort shows that Gravel would be most affected 238 by trawling at all levels of effort (Figure 4), reflecting the higher depletion rates and slower recovery rates 239 (Table 2, Table 3). At swept-area ratios >4.6, the fauna of Gravel were estimated to be fully depleted, with 240 RBS=0 in 18 cells (~2.1%). Most Gravel was not exposed to trawling and ~93.4% of Gravel had RBS >50%. The 241 distribution of RBS values by habitat area (Figure 5) can be used to define other status thresholds; e.g. ~86% of 242 Gravel had RBS >80%. The Gulf-wide average RBS over all Gravel was 91%. Muddy-Sand was relatively more 243 exposed to effort but was less sensitive; the minimum RBS of muddy-Sand was 57% and ~93% had status >80%  Figure 5). The Gulf-wide RBS of muddy-Sand was 95%. Sand had most exposure to high effort but was the 245 least sensitive habitat (Table 2, Table 3); its Gulf-wide RBS was >98% and >99% of Sand had status >80%. Mud 246 had limited exposure to effort and no exposure to high effort (Table 4); its Gulf-wide RBS was >99% and all 247 Mud cells had status >80%. The spatial distribution of habitat RBS (Figure 6) effectively matches that of trawl 248 effort but with differences in trawled areas due to differences in sensitivity among sediment types. For 249 example, the lowest RBS values were for Gravel in moderate-high effort areas, while neighbouring Sand 250 habitat exposed to similar or greater effort levels had higher RBS values. 251 The regional average RBS values of the three benthos taxa were similar to those for habitats, in the range ~91-252 96%. Malacostraca were most affected and Bivalvia least. The absolute status results for taxa differed from 253 their RBS, because they accounted for their distributions. Nevertheless, the Gulf-wide absolute status 254 estimates were similar to average RBS because the abundance of each taxon was about average in trawled 255 areas (Appendix S1). 256

257
The development of the RBS method is timely because it addresses needs arising from national legislation that 258 incorporates the ecosystem approach to fisheries (FAO 2003) driven by international policy commitments (Rice 259 2014) and requirements from certification organisations (e.g. MSC 2014) to take account of the impacts of 260 towed bottom-fishing gears on seabed habitats in management plans and fishery assessments. RBS provides a 261 simple quantitative tool for assessing benthic impacts of bottom trawls and other towed fishing gears. The 262 method is widely applicable, including to fisheries where trawl impacts have not yet been assessed, because it 263 requires relatively few data inputs: 1) effort maps that can be derived from commonly collected VMS or tow 264 data; 2) habitat maps that may be available from local regional surveys, or alternatively national or global 265 geoscience databases of sediments provide first-order mapping of habitats (e.g. dbSeabed); 3) impact and 266 recovery parameters, ideally from local experiments linked to habitat classifications used for the seabed where 267 available, but with meta-analyses (as used herein) providing a more widely applicable alternative.  status with high-resolution at large spatial scales. Geographically, RBS can be applied most broadly for habitats 272 classified by sediment type, because sediment maps are more widely available than maps of other habitat 273 characteristics. RBS can enable assessments of risk framed as: will (or has) the current level of fishing lead to 274 habitat status that compromises a defined sustainability criteria (such as our example: proportion of habitat 275 with RBS>50%) or management objective (if set, such as our example: regional RBS>80%)? This flexibility of 276 application cannot be achieved with qualitative or categorical trait-based scoring type assessments and/or 277 non-spatial approaches, which only provide ranking of sensitivity or potential risk (e.g. low, medium, high). other than sediment type, such as temperature and/or primary production -which may enable recovery 297 parameters to account for regional variations in environment. One potential bias when applying RBS to mobile 298 fauna is the possibility that experimentally measured recovery rates reflect movement of individuals into the 299 impacted area, as well as population growth. This bias was accounted for, to an extent, by the adjustment of 300 experimental r to grid-scale R. In future, meta-analysis of faunal abundance across quantified gradients in 301 trawling intensity may be used to estimate grid-R directly. 302 In our assessment of Exmouth Gulf, habitat RBS and faunal absolute status were affected little at the regional 303 scale, with status ≥90% for all habitats and faunal taxa assessed. This was because <2-7% of the region was 304 trawled sufficiently intensely to yield RBS values <50% and most of the area was either not trawled or trawled 305 lightly. Further, most high-intensity trawling occurred on Sand, which was relatively resilient. Nevertheless, in 306 regions where trawl effort is more intensive and more widely distributed, larger impacts may be expected.