Management by proxy? The use of indices in applied ecology

Journal of Applied Ecology encourages contributions that can influence environmental management, policy or both, with evidence based on the most robust science possible. Natural resource management is often contentious, and any perceived weaknesses in the underpinning science are easily exploited by interest groups to undermine the wider endeavour (see, e.g. the experiences of the Intergovernmental Panel on Climate Change, Ravindranath 2010). Thus, the robustness of science designed to underpin management and policy is particularly important. Unfortunately, robust and unambiguous results are difficult to obtain in ecology. In particular, causal pathways in ecology are seldom linear, but are part of a ‘vast web of cause and effect’ of which, typically, we can study only a small part (Peters 1991; p. 134). Meaningful spatial and temporal scales for ecological processes often defy experiments, controlled manipulations and adequate replication; most modern ecological science is reliant on observational data and correlation is far easier to demonstrate than causation (cf. Sugihara et al. 2012). Formal experimentation is clearly impossible in the context of many of the most pressing questions of societal relevance, such as those regarding the impacts of climate change, large-scale agricultural intensification and habitat loss. The conceptual frameworks that inform our understanding of these processes often rely on entities that can be hard to measure directly and, thus, ecologists often depend on proxies or surrogates for those entities. In natural resource management, for example, common proxies include measures assumed to capture the conservation status of species, as well as measures assumed to provide information on ecosystems’ distribution, structure, functioning and service delivery (Mace et al. 2008; Ayanu et al. 2012; Pettorelli et al. 2014). Several studies have already pointed out issues and pitfalls related to the use of proxies in ecology and evolution (see, e.g. Nieberding & Olivieri 2007; Sekar 2012; Best & Stachowicz 2013) but less consideration has been given to the impacts of proxies on the robustness of management recommendations at a range of scales (but see Eigenbrod et al. 2010). This is despite the fact that proxies are a vital part of an applied ecologist's toolkit and are frequently used to link science with management and policy (Turnhout, Hisschemöller & Eijsackers 2007). Here, we discuss the role of proxies in applied ecology and management by examining three recent examples of problematic issues: proxies for the abundance of animal populations; proxies for habitat quality; and proxies for ecosystem functions and ecosystem services delivery. While this list is far from exhaustive, these are issues we frequently see as journal editors. They provide a useful starting point to discuss possible directions to improve our use of proxies in natural resource management. Population size is a key requirement for supporting local decision-making in species-based management and is also used to inform several national and international policies and regulations. For example, population size and the rate at which it changes are fundamental to species listing and delisting decisions under the Endangered Species Act (U.S. Senate Committee on Environment & Public Works 1973) and for the IUCN Red List (Mace et al. 2008); these decisions have knock-on implications for the international trade of organisms (CITES 1973). The use of indices as proxies for population abundance is perhaps one of the most widely recognized areas of contention in applied ecology. Across vast geographic areas, the relative abundances of many wildlife populations of cultural or economic importance are assessed using surveys that yield indices, rather than estimates, of abundance (Schwarz & Seber 1999). These include, for example, track counts used to monitor the abundance of many mammal species, camera trapping rates used to index the abundance of elusive or low-density species, pitfall traps to assess invertebrate populations, and hunting offtake used as an index of the abundance of many game birds and mammals. These proxy methods have all been the focus of considerable concern (McDonald & Harris 1999; Jennelle, Runge & MacKenzie 2002; Pollock et al. 2002; Webbon, Baker & Harris 2004; Keane, Jones & Milner-Gulland 2011; Saska et al. 2013). The central criticism relates to the generally unverified assumption that the relationship between the index and the true abundance is direct and constant (Pollock et al. 2002). One basic issue underpinning this criticism is that all of these methods depend on encounters (between observers, cameras, traps or hunters and tracks, animals or quarry) which, in turn, depend on the levels of activity of individuals of the focal populations, the detectability of their signs and the extent of efforts to locate them. These factors are not expected to be static in either space or time (Pollock et al. 2002) and could be affected by behavioural changes along ecological gradients of management interest (such as agricultural intensification). In spite of these criticisms, indices of abundance remain in widespread use for a combination of reasons. First, collecting indices tends to be a lower-cost approach than determining absolute estimates of abundance (Jones 2011). Second, as Caughley (1977) observed, ‘absolute estimates of density [are often] unnecessary luxuries’, and management practices such as threshold harvesting or studies of relative habitat utilization are not reliant on estimates of absolute abundance. Third, in some situations, evidence suggests that indices can provide a reasonable proxy for abundance. A recent applied example relates to a critical issue in Australian wildlife management. Specifically, the extent to which track counts reflect true abundance is a key concern about studies of the ecological impacts of the dingo Canis lupus dingo (Hayward & Marlow 2014). In defence of those studies, Nimmo et al. (2015) cite work showing that track counts were linearly related to the true abundances of various carnivores in a number of specific studies. Moreover, they point out that, for elusive species, indices based on abundant field signs might well be associated with lower uncertainty than population estimates based on sporadic but infrequent direct sightings (Nimmo et al. 2015). In summary, proxies for population abundance are widely used and often criticized. They might be adequate in some situations but that will usually depend on the specific management question being addressed. More often, indices of abundance are vulnerable to criticism and will require careful calibration. Identifying potential sites suitable for species translocations, or sites at which we should prioritize restoration efforts, are important management activities world-wide. Habitat monitoring is also strongly encouraged by international conventions: the Convention on the conservation of migratory species (CMS), for example, explicitly states that ‘parties that are range states of a migratory species shall endeavour to conserve and, where feasible and appropriate, restore those habitats of the species which are of importance in removing the species from danger of extinction’ (www.cms.int). Habitat is usually defined as the resources and conditions present in an area that produce occupancy – including survival and reproduction – by a given organism (Morrison, Marcot & Mannan 1992). The spatial extent and resolution at which habitat is defined tends to be a function of several variables, such as the number of individuals considered, the body size of the species considered, and the spatial resolution at which habitat selection patterns are explored (itself shaped by the availability of environmental and species occurrence data; Pettorelli 2013a,b). Within this context, habitat quality is sometimes perceived as some function of fitness or per-capita population growth (Van Horne 1983; Johnson 2007), where these are expected to be higher in high-quality habitats. The problem with this definition of habitat quality is that measuring per-capita contributions to population growth in a given habitat requires intensive monitoring, usually over a long period. As a result, attempts to capture variation in per-capita population growth often use information that is less costly and time consuming to obtain, such as abundance, survival, body condition or corticosterone levels of the focal species (see, e.g. Van Horne 1983; Marra & Holberton 1998; Johnson 2007). It has also been assumed that habitat suitability, usually determined from occurrence or presence/absence data using species distribution models (SDMs), will provide a useful indication of habitat quality (see, e.g. Larson et al. 2004; Martin et al. 2012). Working with the giant kangaroo rat Dipodomys ingens, Bean et al. (2014) compared these different proxies for habitat quality, using SDMs generated at a range of scales. They found that habitat suitability from SDMs correlated with abundance (especially when SDMs were parameterized at a fine scale) but not with survival or body condition. Life-history theory is reassuring in this regard, providing good reasons why neither survival (Pilastro, Tavecchia & Marin 2003) nor body condition (Houston & McNamara 1993) need be strongly related to habitat quality. More troubling, however, is that ecology also provides reasons to question the relationship between population abundance and habitat quality, especially where despotic behavioural processes can result in large numbers of individuals being excluded from high-quality (and into marginal) habitats (Van Horne 1983). These considerations call into question a number of widely used proxies for habitat quality. Habitat quality is unquestionably related to the availability of resources, but also to levels of predation. Which of these predominates in determining the quality of habitat for a given population is often described using the terminology of ‘bottom-up’ vs. ‘top-down’ control (e.g. Hunter & Price 1992). The proportion of mortality attributable to predation is sometimes used as a proxy for the extent of top-down control (Sinclair, Mduma & Brashares 2003; Hopcraft, Olff & Sinclair 2010). Paradoxically, however, behavioural modelling of risk-sensitive foraging shows that an increase in predation may well result from a decline in food availability (McNamara & Houston 1987). These results appear to be borne out in empirical systems (Sinclair & Arcese 1995), undermining the use of predation rate as an index of top-down control and emphasizing the difficulty of teasing apart these processes. Overall, theoretical considerations suggest that proxies for habitat quality and the processes that determine it will often be flawed. As with proxies for abundance, however, proxies for habitat quality might be suitable for addressing management questions at some spatial scales. The challenge remains to identify those situations under which proxies are useful and to determine alternative approaches where they cannot be used. With a policy agenda increasingly focused on ecosystem service provision (Perrings et al. 2010), understanding the ecology of ecosystem functioning and its implications for the delivery of ecosystem services has never been more important. It can be extremely difficult to measure ecosystem function or service directly, and spatial proxies, such as small-scale above-ground biomass in plants, are often employed (e.g. Tilman et al. 2001). Perhaps less widely appreciated, however, is the fact that certain easily measured ecosystem functions are sometimes used as proxies for more difficult to measure ecosystem functions and services. For example, most biodiversity–ecosystem function experiments measure productivity and assume that this parameter correlates with a broader suite of ecosystem functions and services. Yet recent work has shown that ecosystem productivity can be quite divergent from functions and services such as methane consumption, pest control and pollination (Werling et al. 2014). A vast array of approaches to measure ecosystem functions and/or ecosystem services is currently available to ecologists and managers. Large-scale attempts to assess ecosystem productivity might rely on proxies such as climatic and structural variables that correlate with productivity (Ruiz-Benito et al. 2014), or vegetation indices derived from remote sensing information (Pettorelli 2013b). Identifying which methodological approaches and proxies are likely to support robust decision-making in a given context is a particular challenge. The European Union's Water Framework Directive (WFD), for example, uses an ecological status score for aquatic habitats to represent the amount of ecosystem service provisioned. One recent assessment of the WFD's use of ecological status found that status correlates with some ecosystem services and taxonomic diversity (Tolonen et al. 2014), suggesting that this proxy is useful for quickly assessing overall service delivery (though caution is required when considering single or few services). By contrast, recent analyses have shown that land cover type (e.g. Ayanu et al. 2012) is a poor proxy for ecosystem service, with potential to mislead management strategies (Eigenbrod et al. 2010). Species-based proxies for assessing ecosystem function and service delivery have also been heavily discussed. Two decades of research have now been devoted to testing the hypotheses that higher species diversity leads to greater ecosystem function and service delivery, with the view to establishing species richness as a proxy for these ecosystem attributes (Balvanera et al. 2006; Cardinale et al. 2006). Results so far, however, seem to indicate that functional diversity (Petchey & Gaston 2006; Mouchet et al. 2010; Cadotte, Carscadden & Mirotchnick 2011), rather than species richness itself, directly affects ecosystem function and the delivery of ecosystem services (Tilman et al. 2001; Cadotte, Carscadden & Mirotchnick 2011; Woodcock et al. 2014). Species loss may be functionally random, with the result that certain extinction scenarios may have minimal effects on functional diversity (Sasaki et al. 2014), rendering species loss an unreliable indicator of ecosystem function loss. Interestingly, both functional and phylogenetic diversity have actually been shown to explain variation in ecosystem function (Flynn et al. 2011; Cadotte 2013). However, there are cases where phylogenetic and functional diversity may not be strongly correlated and actually show different responses to disturbance or management (Cadotte, Albert & Walker 2013; Bässler et al. 2014). Clearly, more work is required to understand how well these facets of diversity correspond to the important ecological differences that influence the ecosystem functions that may attract management interventions. Overall, the validity of different proxies for ecosystem function and service provision has been the focus of much research but few proxies have been shown to work well across a wide range of scenarios. Proxies should be used cautiously in situations in which they have been shown to work, or subjected to further validation beyond those circumstances. The examples above highlight the challenges of drawing robust management-relevant inferences from ecological science, providing insights into the appropriate use of proxies and ways to move forward. Here, we suggest three essential considerations when using and/or developing proxies. As with so much in ecology, the relevance of proxies often depends on the scale of enquiry. As illustrated by Bean et al. (2014), spatial resolution and spatial extent are correlated entities in most habitat quality assessments. Unsurprisingly, assessments focused on smaller areas or utilizing environmental data collected at greater spatial resolution are capable of detecting greater variability in apparent quality than coarser, large-scale studies. Extent and resolution are, thus, important parameters that directly shape the usefulness of habitat quality assessments: broad-scale models are more likely to be adequate for identifying potential protected areas, whilst finer-scale models will be required to inform local management strategies such as habitat restoration. Spatial scale is also likely to be fundamental to debates about the usefulness of prey availability as an indicator of habitat quality for predators. Recent work on the sublethal effects of predation and the ‘landscape of fear’, shows that prey will often be most abundant in areas in which they are relatively inaccessible to predators and, thus, that the correlation between prey abundance and prey availability may be low (Laundré 2010). This suggests that, at a fine scale, prey availability might be a poor proxy for habitat quality – but, at a landscape scale, predators will still prefer areas with abundant prey. Thus, the spatial scale at which a question is posed has a strong bearing on the relevance of a given proxy. Temporal scale might also shape the relevance of a proxy. Owing to variation in activity among years, proxies for population density based on either sighting rates or spoor abundance might be poor indicators of interannual changes in population size. However, unless activity or detectability show long-term, deterministic trends, such indices will be useful to identify the existence of trends in abundance over longer time frames. Even over short time frames, indices might be useful, as long as the variation in detectability and activity is independent of abundance and of substantially lower magnitude than the variation in population size that researchers wish to or not this is the should be a focus of work at these proxies. studies and which an between spatial and the resolution of temporal While the recent use of occupancy and detectability models encourages temporal this can at a of understanding important spatial of species which, in many may be more important from a conservation et al. 2014). in a given proxy is directly related to the amount of evidence its Proxies to be to local so is often a requirement of various proxies for population abundance has been the focus of As Nimmo et al. (2015) a range of evidence to support the use of track abundance as a useful proxy for abundance of to large this not that track abundance will be a useful proxy in all work has focused on indices of abundance and for factors that determine how the relationship between the index and abundance across time and For example, et al. data on of in the to identify factors that In this they were to for time of when index to absolute Saska et al. assessed the of on the activity of to control for activity when pitfall trapping to relative abundance. for activity or can also of abundance indices across species, those indices are based on track counts or trapping rates et al. 2008; et al. are also available for for of when using indices of population density based on abundance. Specifically, & approaches for for from both observational and both of which can undermine the use of counts as proxies for abundance. assessing the validity of various proxies of habitat quality are also (see, e.g. Johnson 2007; Pettorelli 2013b). for proxies of habitat quality derived from habitat suitability modelling can be found in work that to the quality of This used by et al. for assessing the of that habitat suitability modelling can areas that these models produce useful proxies for habitat quality that is to the One to this is that habitat suitability modelling based on or presence/absence data cannot between and habitats. et al. measures of into habitat suitability however, those measures will often be proxies for As proxies for ecosystem function and service provision have also been subjected to but with several proxies have ecosystem service in particular, a wide array of processes Thus, further validation is likely to be required to show that proxies well across One of the fundamental for proxies from the spatial and temporal scale of most ecological However, and research can to and proxies and might also our on them. The for example, is a large-scale which will information on distribution with The is to over decades at an of sites and has the potential to provide direct of the link between and ecosystem to our understanding of this also ecological with a of sites across the are on scales. The for will include animal monitoring on a research of the This will remote monitoring of the and of large numbers of individuals of species far smaller than can be at many this will direct of habitat use and the of habitat on a scale beyond currently our understanding of the validity of proxies for habitat quality. Applied ecologists should to the that natural resource management and environmental policy the possible underpinning the lower of monitoring work, robust approaches to policy and management often require monitoring, to key ecological on a scale that the need for proxies. As Martin of the has we it for that will to provide vast to for on – but we still about the one where we already that not be about for monitoring that yield insights into the key ecological processes supporting on insights could our understanding of ecological systems and, the of and to with