SEARCH

SEARCH BY CITATION

Keywords:

  • camera-trap;
  • cost-effective;
  • density;
  • detectability;
  • habitat suitability model;
  • invasive species;
  • migratory;
  • occupancy model;
  • tiger

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Challenges in monitoring species abundance and distribution at the landscape scale
  5. Using landscape-scale species data to inform policy
  6. Scaling up from the landscape scale?
  7. Acknowledgements
  8. References

1. The abundance and distribution of a species are affected by processes which operate at multiple scales. Large-scale dynamics are increasingly recognized in conservation responses such as metapopulation management, transfrontier protected areas and softening the agricultural matrix. Landscape-scale monitoring is needed both to inform and judge their efficacy. In this Special Profile we address some of the challenges presented by monitoring at the landscape scale, how models of species distribution can be used to inform policy, and we discuss how monitoring at the global-scale could be approached.

2. Collecting data over a large area is inherently costly, so methods which can provide robust information at low-cost are particularly valuable. We present two papers which test low-cost approaches against more data-hungry methods (indices of abundance vs. direct density estimates, and species distribution models built from presence-only vs. presence/absence data).

3. Occupancy modelling is a useful approach for landscape-scale monitoring due to the relatively low-cost of collecting detection/non-detection data. We discuss challenges, such as non-random sampling locations and periodical unavailability for detection, in using detection/non-detection data for monitoring species distribution. Such data can also provide estimates of abundance and we show how existing models have been modified to allow the abundance of multiple species to be estimated simultaneously.

4. Models of species distribution can be used to project likely future scenarios and thus inform conservation planning where distributions are likely to change because of climate change or changing disturbance patterns. We also discuss how an optimization framework can be used to make efficient management decisions for invasive species management in the light of imperfect information.

5.Synthesis and applications. Monitoring is needed for many purposes including auditing past management decisions and informing future choices. Much monitoring data are collected at the site scale, although management authorities increasingly recognize landscape-scale dynamics. Recent global targets for conservation require monitoring which can report trends at the global-scale. Integrating data collected at a variety of scales to draw robust inference at the scale required is a challenge which deserves more attention from applied ecologists.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Challenges in monitoring species abundance and distribution at the landscape scale
  5. Using landscape-scale species data to inform policy
  6. Scaling up from the landscape scale?
  7. Acknowledgements
  8. References

Over the past two decades there has been a shift in emphasis among conservation biologists from managing populations of threatened species at a single site, to considering larger scale dynamics (Baillie et al. 2000; Brown, Spector & Wu 2008; Lindenmayer et al. 2008; Levi et al. 2009; Bailey et al. 2010). Such landscape (or seascape) scale approaches to conservation make sense because the drivers of biodiversity loss (such as habitat conversion and fragmentation, overexploitation and climate change) tend to operate at large scales, and a substantial body of evidence has demonstrated the importance of dispersal between subpopulations and source-sink dynamics in the persistence of a species in a landscape. Consideration of these large-scale dynamics is behind such conservation responses as metapopulation management (Esler 2000; Rouquette & Thompson 2007), transfrontier protected areas (Smith et al. 2008) and drives to soften the agricultural matrix around conservation areas (Donald & Evans 2006; Perfecto & Vandermeer 2010; Koh et al. 2010). Monitoring at the appropriate scale is essential both to judge, and to improve, the efficacy of such approaches (Radford & Bennett 2007).

The term ‘monitoring’ has many different definitions and usages. Here we follow Yoccoz, Nichols & Boulinier (2001) in defining monitoring as the process of gathering information about a state variable (such as the abundance or distribution of a species) to assess the state of the system and draw inferences about changes over time. Monitoring species abundance or distribution at the landscape scale presents particular challenges. (i) Data collection over a large area is inherently costly so methods for minimizing costs, always important in any monitoring study, will be particularly significant. (ii) Species detectability tends to vary over space or time. Accounting for variable detectability will be particularly important in landscape-scale studies as surveys will inevitably cover a variety of habitats, and may have to be carried out in different seasons; both of which will influence detectability. In this Special Profile we bring together five papers which address these challenges of monitoring species abundance and distribution at the landscape scale, and three which demonstrate novel ways in which models based on species distribution data can be used to inform policy.

Challenges in monitoring species abundance and distribution at the landscape scale

  1. Top of page
  2. Summary
  3. Introduction
  4. Challenges in monitoring species abundance and distribution at the landscape scale
  5. Using landscape-scale species data to inform policy
  6. Scaling up from the landscape scale?
  7. Acknowledgements
  8. References

Minimizing the costs, while ensuring data quality

Monitoring is costly and in a resource-limited world conservationists will seek to maximize the cost-effectiveness of monitoring, freeing up resources to spend on other activities (Murray et al. 2009). There will often be a trade-off between the cost of a monitoring method and the quality of the information it provides and if a low-cost method provides noisy data that does not allow trends to be detected robustly, it will be of little use for decision making and the resources invested will be wasted (Legg & Nagy 2005). Validation of low-cost methods against more data-intensive methods is therefore an essential step but is seldom done (Jones et al. 2008; De Barba et al. 2010). In this issue, two papers explicitly validate the use of lower-cost methods against more data-hungry and therefore costly approaches. The first (Jhala, Qureshi & Gopal 2011), looks at the cost-effectiveness of indices of abundance relative to obtaining direct estimates of density for a threatened species across a landscape. The second (Gormley et al. 2011), compares the value of presence-only to presence/absence data for modelling the distribution of an invasive species.

Counts of some easily measurable sign that is assumed to correlate with population density are often used in place of density estimates because the data collection is less costly. However, the use of such indices of abundance has been widely criticized as they are seldom calibrated with direct estimates of density, or tested for precision in detecting change in population size (MacKenzie & Kendall 2002). The tiger Panthera tigris is a highly charismatic and threatened species with populations scattered across 13 countries. The importance of robust monitoring was highlighted when it emerged that tigers had gone extinct from the Sariska Tiger Reserve, India, during a period when the official census continued to report a healthy tiger population in the area (Narain et al. 2005). Appropriate methods for monitoring the tiger have been vigorously debated. Work by Karanth et al. (2003) showed the weakness of the widely used pugmark census technique. However, although camera-trap based mark–recapture can provide precise estimates of tiger densities (Karanth et al. 2006), it is expensive and more suited to use at relatively small scales (Linkie et al. 2006). There is therefore a need for a quantitative assessment of cheaper and easier methods that can be deployed at the landscape scale. In this Special Profile we present the work by scientists from the Wildlife Institute of India and the National Tiger Conservation authority calibrating indices of abundance (pugmarks and scats encountered per kilometre searched) against estimates of tiger density made at the same time using state-of-the-art camera-trap mark–recapture methods. Their results (Jhala, Qureshi & Gopal 2011) are encouraging and suggest that indices of tiger abundance can reliably indicate tiger density, across a range of habitats and densities, and at much lower cost than camera-trap mark–recapture. Of course these results will not negate the need for detailed and precise studies in targeted areas. However, the work will be valuable for developing a cost-effective landscape-scale tiger monitoring programme.

Cost-effective monitoring is equally important in the study of invasive species. Knowing the current distribution of an invasive species, and being able to predict its potential distribution, is important information for those seeking to eradicate or control it (Hulme 2006). Opportunistic sightings (presence-only data) are often available at relatively low-cost but due to inherent biases may be expected to give less valuable information than data from carefully designed field surveys. In this Special Profile Gormley et al. (2011) compare estimates of the current and potential distribution of the introduced sambar deer Cervus unicolor in Australia, from models built using presence-only data, and presence/absence data. They were able to robustly estimate the current and potential distribution of sambar deer from either data set and could identify priority areas for surveillance monitoring to detect an expansion of the species range even from relatively low-cost and easily obtainable presence-only data.

Accounting for variable detectability (in space and time)

In 2004 the Journal of Applied Ecology published a Special Profile on new paradigms in species distribution modelling. Since then approaches such as information-theoretic-based model selection and incorporating remotely sensed data into habitat suitability models have, as predicted in the accompanying editorial (Rushton, Ormerod & Kerby 2004), become mainstream. The methods used to model species distributions continue to advance, providing ever more powerful approaches for targeting monitoring and management of threatened or invasive species (see also Gormley et al. 2011; Singh & Milner-Gulland 2011).

Habitat suitability models are often constructed using data on occupancy (the fraction of sampling units in a landscape where a species is present). However a species that is present in an area may go undetected, whatever method is used for surveying. Unless such false absences can be accounted for, apparent changes in occupancy and estimates of habitat preference will only be valid if detectability of the species remains stable over time and across different habitats (Mackenzie & Royle 2005); an unrealistic assumption. Occupancy modelling makes use of repeat surveys of sampling units to explicitly model and estimate detection probability. The approach was developed more than 20 years ago but was not well used until advances in computing made it possible to fully exploit its potential. Occupancy models, and extensions such as dynamic occupancy models (MacKenzie et al. 2003) and multi-scale occupancy models (Mordecai 2007; Nichols et al. 2008), are now used with different types of data including that provided by camera-traps (Rowcliffe et al. 2008) and from multiple detection methods in one study (Nichols et al. 2008), to allow for temporary emigration or immigration (Rota et al. 2009) and to identify predictors of detectability as a guide for focusing future monitoring effort (Guillera-Arroita et al. 2010). Occupancy modelling is a particularly useful approach for landscape-scale monitoring due to the relatively lower cost of collecting detection/non-detection data relative to surveys estimating abundance (Joseph et al. 2006) and because larger-scale studies may cover a greater range of habitats resulting in more variation in detectability than in smaller-scale studies. Two papers included in this Special Profile present further advances in the use of this flexible and useful modelling approach.

In all surveys, the location of sampling units would ideally be fully independent. However in many situations it is more practical to locate sampling units (point counts, transects, etc.) along an existing access route or transects than to site them randomly, and in studies of low-density and rare species non-random sampling may be done to increase detection. Both these reasons for non-random sampling may introduce biases. Van der Burg et al. (2010) use random effects in a Bayesian hierarchical model to account for spatial dependence of sampling units, allowing them to account for such biases. They use their resulting models of the distribution of a rare bird (the mountain plover Charadrius montanus) to investigate the efficacy of management of this species. Conservationists and environmental managers often have to deal with relatively sparse or poor-quality data (because a study has been poorly designed or because the species is rare). The work by Van der Burg et al. (2011) shows how careful modelling can maximize the value of such data to conservation and management.

Mordecai et al. (2011) focus on the challenges posed when studying mobile or episodic species (which are therefore only sometimes available for detection) in a landscape where survey locations need to be spatially clustered for logistical reasons. Although these challenges have been addressed in previous studies, they extend the occupancy modelling framework to account for both problems simultaneously. The approach they use builds on previous multi-scale occupancy models (Mordecai 2007; Nichols et al. 2008) and allows simultaneous estimation of occupancy (which they define as the probability that a site is occupied by a species at least once in the survey period), use (the probability that the species is available to be detected given that the site is occupied), and detection probability (the probability that the species is detected on a given visit given that the site is being used). In traditional occupancy models, use and detection are confounded into a single parameter. Mordecai et al. make a convincing case that there are many ecological problems where separating these parameters, as their model allows, is important. For example, where an animal may be absent from much of its home range at any given time and a researcher wants to investigate patterns of occupancy. Like Van der Burg et al. (2011), they also deal with issues of spatial dependency of sampling locations in their model using random effects. They argue that their approach has a wide application for studying clustered detection–non-detection data for elusive species across a range of spatial and temporal scales.

Ecologists often not only want to know a species’ distribution but how many individuals there are. Royle & Nichols (2003) showed how simple detection–non-detection data can provide information on abundance. In this issue, Yamaura et al. (2011) extend the Royle & Nichols model to estimate the abundance of a number of species simultaneously. They apply their model to multiple repeat bird surveys in a forest recovering after a fire. This model has enormous potential for application to landscape-scale monitoring problems as it can be used to extract information on community structure as well as the dynamics of individual species from relatively easily obtained data.

Using landscape-scale species data to inform policy

  1. Top of page
  2. Summary
  3. Introduction
  4. Challenges in monitoring species abundance and distribution at the landscape scale
  5. Using landscape-scale species data to inform policy
  6. Scaling up from the landscape scale?
  7. Acknowledgements
  8. References

A number of papers included in this Special Profile (Van der Burg et al. 2011; Gormley et al. 2011; Mordecai et al. 2011) suggest ways that models of species distributions can be improved or can make more efficient use of available data. Three further papers consider how such models can be used to make the best use of available information to inform policy. Singh & Milner-Gulland (2011) use a combination of long-term aerial survey data, remote-sensed habitat data, and projections of likely future scenarios of climate change and disturbance to inform landscape-scale conservation for a migratory ungulate. Regan, Chadés & Possingham (2011), and Baxter & Possingham (2011) show how optimization frameworks can be used to make efficient decisions for invasive species management in the light of imperfect information on species distributions.

The large-scale movements of migratory species, mean that monitoring and management at the landscape-scale is necessary (Sutherland 1998). Although many migrations follow relatively established routes, their precise location and timing may vary and may be influenced by climate change and human disturbance (Wilcove & Wikelski 2008). Such variability makes it difficult to monitor trends in abundance or distribution, or chose where to locate static conservation interventions. The saiga antelope Saiga tatarica is considered Critically Endangered following a 95% decline in population size over the last two decades (Milner-Gulland et al. 2003) but the understanding of trends is hampered by biases in the monitoring techniques which have been used (McConville et al. 2009). Saiga are currently the focus of considerable conservation attention in parts of its range and, given the level of threat facing the species, it is essential that this effort is effectively deployed. Information on current, and likely future, distributions is required to inform conservation planning, and monitoring that can robustly detect trends given the challenges posed by the species’ ecology is needed. A remarkable data set from 25 years of aerial surveys allowed Singh & Milner-Gulland (2011) to identify factors predicting the spring distributions of saiga in Kazakhstan. They used their model to predict the distribution under likely scenarios of climate change and human disturbance. They found that the distribution and density of saiga has changed over time and is likely to continue to change into the future. Their models of likely future distributions can be used to improve the placement of planned protected areas. The approach taken by Singh & Milner-Gulland (2011)– using species distribution models in conjunction with projections of future scenarios – is likely to be useful in targeting monitoring and conservation action at the landscape scale in a range of circumstances.

Another paper in this Special Profile (Regan, Chadés & Possingham 2011) considers how the detectability of a species (and therefore how costly it is to get reliable information on a species’ distribution) influences the optimal strategy for its management. Using the case study of broomrape Orobanche ramosa, an invasive parasitic plant in Australia, they show how the optimal strategy for eradicating the species depends, in part, on the species’ detectability. Using a partially observable Markov decision process (POMDP) they show that costly, effective actions for controlling the species (such as soil fumigation), should be used in preference to less effective but lower cost methods (such as reducing the availability of hosts) if detection rates are high. However when detection rates are low, it is optimal to continue managing the species using the lower cost methods even when the species is not detected, in order to buffer against this low detectability. Despite the great advances in using occupancy models to estimate detectability, this novel study may be the first to explicitly consider how variable detection might change the optimal management strategy for a species.

Staying with invasive species management in Australia, Baxter & Possingham (2011) consider the trade-off between investing in action to control the invasive ant Solenopsis invicta and further surveys to allow better predictive maps benefiting future searches. Their work shows the importance of investing in knowledge, as long as that knowledge acquisition has a clear purpose for informing management. Their work has relevance wherever ecologists seek to predict species distribution for the purpose of making management decisions; for example identifying priority areas for new protected areas. Given the difficulty and cost of gathering information over the vast range of the saiga antelope, this framework could perhaps be applied to the case study highlighted by Singh & Milner-Gulland to help identify the optimal investment in further monitoring to improve understanding of the species’ likely distribution vs. investment in conservation interventions on the ground.

Scaling up from the landscape scale?

  1. Top of page
  2. Summary
  3. Introduction
  4. Challenges in monitoring species abundance and distribution at the landscape scale
  5. Using landscape-scale species data to inform policy
  6. Scaling up from the landscape scale?
  7. Acknowledgements
  8. References

Natural ecosystems interact at all spatial scales. For example, the persistence and abundance of a species at a site may be influenced by the level of exploitation it is exposed to at that location, by the pattern of habitat fragmentation at the landscape scale, and by the spread of a damaging invasive species around the globe. Therefore, although monitoring and management at the landscape scale is important and justifiably attracting increased attention, it will not always be sufficient for understanding and combating biodiversity loss. Targets for conserving biodiversity are increasingly set globally (Perrings et al. 2011) and the failure to meet the first global biodiversity target was blamed partially on at the lack of measurable targets and appropriate monitoring (Butchart et al. 2010). Monitoring which can provide robust inference at the global scale is therefore needed to properly audit progress against these global targets (Jones et al. 2011). However this is not necessarily a call for a prescriptive, top-down global monitoring programme as that would be prohibitively expensive (Scholes et al. 2008). Most monitoring is carried out to meet local or regional management objectives and cost-effective global-scale biodiversity monitoring will have to make use of these data. As the papers in this Special Profile demonstrate, applied ecologists are grappling with the challenges of monitoring effectively and efficiently at the landscape scale. Interesting and important challenges remain in tackling how monitoring data, collected at a variety of scales, can be integrated to monitor biodiversity change globally.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Challenges in monitoring species abundance and distribution at the landscape scale
  5. Using landscape-scale species data to inform policy
  6. Scaling up from the landscape scale?
  7. Acknowledgements
  8. References

The author thanks E.J. Milner-Gulland, Aidan Keane, Neal Hockley, Brady Mattsson, Tracey Regan and Andy Royle for helpful discussion and comments.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Challenges in monitoring species abundance and distribution at the landscape scale
  5. Using landscape-scale species data to inform policy
  6. Scaling up from the landscape scale?
  7. Acknowledgements
  8. References