Correspondence Timothy C. Bonebrake, Department of Biology, 371 Serra Mall, Stanford University, Stanford, CA 94305-5020, USA. Tel: (818) 642-4093; Fax: (650) 723-5920. E-mail: firstname.lastname@example.org
Scientific and historical knowledge of worldwide animal-population decline is fragmentary at best. However, understanding historical population trends is essential for informing best efforts to preserve species. We reviewed the literature of long-term studies of population declines across a set of animal taxa and found that only 15% of the studies used data older than 100 years, and 58% of the studies lacked continuous data. Based on our review, we describe five general approaches to studying population declines: counting, correlative, evolutionary, geochemical, and historical. The most common method of population assessment was a census/counting approach (75% of studies) followed by a range mapping/correlative approach (17% of studies). Evolutionary, geochemical, and historical approaches are used less often but, in combination with traditional counting and correlative methods, they hold great potential as tools for conservation. The multidisciplinary approaches we identify and advocate here will be useful for understanding and potentially reversing population declines and ultimately stemming the tide of extinctions currently underway.
Many species have suffered large population declines as a result of anthropogenic influence on ecosystems worldwide (Balmford et al. 2003; Ehrlich & Pringle 2008). Some species have been driven to extinction, while many others teeter on the brink. While these population declines are so clear that conservation scientists have concluded that Earth faces a sixth mass-extinction crisis (Wake & Vredenburg 2008), quantifying population declines is a difficult, time-consuming, and research-intensive task. As a result, studies of population declines employ widely varying methods and estimates of biodiversity loss and species imperilment differ wildly (Ehrlich & Wilson 1991; Stork 1997).
Historical population records for most species are fragmentary and of questionable quality, if they exist at all (Balmford & Bond 2005). Observers, including most conservation biologists, typically, have not begun monitoring populations unless declines are already evident, so data for population sizes before declines began are rare. Long-term monitoring of populations over multiple decades is also expensive, so such efforts are difficult to sustain. More mundane administrative problems and technological changes can also cause loss of data. And, ultimately, there are simply not enough researchers or trained citizen scientists even to begin to attempt to document the changing status of billions of populations of plants and animals around the globe (Hughes et al. 1997).
The resulting dearth of historical population-trend data presents a number of conceptual problems for urgent efforts to understand the state of biodiversity and save species threatened with extinction. First, without reliable data on decline or growth in many species, estimates of the extent and urgency of the current extinction crisis will remain highly debatable, impeding decision making about the most effective conservation efforts.
Second, without knowing the time period during which a population declined, it is difficult, if not impossible, to elucidate why it declined. And without understanding the causes of decline, it is difficult to stem the decline by curtailing threats and creating conditions to promote population recovery (Caughley 1994). Full knowledge of a population's history cannot guarantee a population will not go extinct (e.g., the small-population paradigm; Caughley 1994) or even guarantee that a firm determination can be made regarding the cause of a decline. However, it is difficult to uncover causal mechanisms behind a population's trajectory without an accurate depiction of that population's history.
Finally, the fragmentary and sometimes contradictory nature of historical population data leads to questions about “shifting baselines” (Pauly 1995; Baum & Myers 2004; Saenz-Arroyo et al. 2005; Papworth et al. 2009) or “natural” population sizes (Willis & Birks 2006). Ultimately, the time designated as when a population was at a “natural” size is arbitrarily defined and the target population size of a restoration effort is always constrained by feasibility as well as past population sizes. However, the more historical data that are acquired and the more managers understand the past, the more informed the decisions about recovery and restoration targets can be (Swetnam et al. 1999).
In this review, we examine studies of population decline to answer the following questions: (1) How far back in time do conservation studies typically go to establish a baseline from which to measure decline? (2) What methods are used to obtain historical population data? (3) What are some of the advantages or disadvantages of each of the methods? We end with recommendations for improving investigation of population declines by combining available methods to describe and diagnose the causes of the declines and inform conservation efforts.
We searched the literature for papers describing long-term population dynamics by querying “conservation,”“population,” and “decline” (i.e., papers that contained each of these terms) in the BIOSIS and ISI Scisearch databases between the years 1987 and 2007. The motivation behind the use of each of the terms was as follows:
1The term “conservation” constrains the results largely to studies with a conservation goal. Consequently, this analysis does not necessarily include research from the “natural resource fields” (fisheries, wildlife management, etc.) that are related to but distinct from conservation; the former primarily emphasizing economic concerns (Soulé 1985; Meine et al. 2006).
2The term “population” reflects our interest in populations as a critical unit for study in conservation (Hughes et al. 1997).
3We chose “decline” as the population parameter of interest given the term's historical and paradigmatic importance in conservation (Caughley 1994). While terms such as “trend” and “change” would have been more inclusive, the use of “decline” is sufficient for our purposes.
This search yielded 8,885 papers. We scanned titles and abstracts from each of these papers to select studies that described a population trend over a 10-year period at minimum. For purposes of feasibility, we refined our search by selecting studies published in nontaxon-specific ecology and conservation journals. Ultimately, we used papers from 31 journals but over half of the studies were published in either Conservation Biology or Biological Conservation (see Appendix S1 and Table S1 for details). When two papers described the same population or system, only the more recent paper was used.
With respect to the conservation literature as a whole, the papers we selected are likely to be biased in a couple of ways. First, our definition of “long-term” (10 years) is arbitrary and there is no reason why a study taking place over 9 years could not also be called long term. Therefore, the results are likely conservative with respect to the average temporal scale of conservation studies. For example, had we selected all studies of 5 years or longer instead of 10, then of course the percentage of long-term studies spanning over a century would be much lower than reported here. Second, conservation studies that did not have the word “conservation” in the abstract (e.g., fisheries, wildlife management studies) were not used, and so there is a bias toward populations that is not of direct economic importance. Finally, the explicit use of “decline” in our initial query logically biases these studies toward populations that are declining or have declined. Our literature search virtually ignores the many populations in nature that are not declining but are still of interest to conservation biologists.
For each paper, we collected data about the geographic region or country, organism type, and the time-period covered. Taxonomic coverage was biased toward birds (37%), mammals (27%), amphibians (11%), insects (8%) and reptiles (6%) while other animal groups each made up less than 5% of the papers. Geographic coverage was also biased: 65% of the studies were done in either North America or Europe. We then recorded whether or not the population data were continuous (recorded annually over the period of study) or snapshots (recorded in distinct, discontinuous periods). Finally, we examined the method used to determine the long-term population trend. We defined 13 historical approaches used in the papers (Table 1) and, for each paper, we noted which approach or approaches were used.
Table 1. Methods of population decline inference categorized by general approach and briefly defined
Systematic counting of an organism or species either directly or indirectly (such as the counting of nests or feces)
Mark Release Recapture
Capture and marking of an organism and subsequent recapture of that organism allows for estimates in population size (via Jolly-Seber, Lincoln-Peterson, etc.) and survivorship
Use of range maps to determine the dynamics in spatial extent of a population or species through time
Inference of population dynamics through change in the extent of the habitat used by a population; dependent upon strong correlation between population size and habitat
As in the case of habitat relationships, inference of population dynamics through changes in weather; dependent upon strong correlation between population size and weather
Use of the notes of explorers or naturalists to provide insight into the abundance or presence of an organism at a particular location
Use of hunting/fishing records to extrapolate population size
Consulting knowledgeable agents (residents, game wardens, hunters, etc.) for population size or extent information
Use of photographs, be they aerial or otherwise, to provide information on habitat extent and/or numbers of organisms or spatial extent
Use of locality or relative abundance data collected from museums or herbariums
Using human associated historical evidence (such as the bones of eaten organisms) of presence or abundance
Using the fossil record to determine presence or abundance through time
Inference of past population sizes from the genetic record
Use of isotopic signatures to provide information on population trends if a sufficiently powerful trend is evident contemporarily and historical items (museum specimens, sediment, etc.) available for analysis
Approaches to understanding population declines
Our systematic review of the scientific literature on population declines identified 265 papers that described population trends extending over at least 10 years, which we defined as long-term trends (Appendix S1 and Table S1). These 265 papers serve as a characteristic example of the state of peer-reviewed conservation literature on long-term population declines. Few of these studies, however, are truly long term, and most represent snapshots rather than continuous population studies. Only eight studies (3%) examined population sizes beyond 5,000 years; only 41 studies (15%) spanned more than a century (this value includes the studies looking beyond 5,000 years); and 154 (58%) lacked continuous population data, with some relying on as few as two measures to infer population declines.
We found that these studies used 13 different methods to document or infer historical population sizes (Table 1). From these 13 methods, we defined five general approaches to documenting and inferring historical population sizes: counting, correlative, evolutionary, geochemical, and historical. The methods were by no means exclusive. Some studies used more than one method and often derived appreciable insights from combining methods.
The counting approach consists of estimating population sizes through direct censuses (for our purposes here we do not necessarily mean total count) or mark-recapture. A direct census can take many forms, including organism counts, transects, or other indirect methods such as counting nest sites (e.g., Homewood et al. 2001; Estes et al. 2006). The counting approach is the most direct and likely the most accurate method. It was the most common method and used by 195 (74%) studies to document or infer population trends (Fig. 1).
Censuses and mark-recapture studies are expensive and time-intensive. Furthermore, without foresight or luck, population census data are not usually obtained before declines begin, particularly in species of little economic or cultural importance. Also, only 82 (42%) studies included continuous population censuses. Some discontinuous studies of population histories consisted only of two snapshots, for example, one transect done in the 1960s or 1970s repeated once again in the 1990s (Cortes et al. 1998; Strayer & Fetterman 1999; Sullivan 2000; Fischer & Linsenmair 2001). Particularly when using relative abundance data, detection probability must also be incorporated into any trend analysis (MacKenzie & Kendall 2002).
Some studies used mark-recapture methods, in which individual organisms are captured and counted, marked, released, and then counted again if they are recaptured in order to estimate a population's size (e.g., Francis et al. 1992; Thomson et al. 1997). One advantage of mark recapture is that it can also track individuals across time and space allowing for estimation of demographic and dispersal parameters, enhancing the ability to tease out potential causes of population declines (White & Burnham 1999).
In the absence of population data or estimates to infer trends, many studies use available data believed to be correlated with population size. Correlated data are also sometimes used to fill in temporal gaps between periods when population sizes were estimated by counting and to estimate population sizes before counting began. Correlative models used to reconstruct population histories can take a variety of forms including the use of contemporary data and knowledge about dynamics to “hindcast” historical abundances or estimate historical abundances through the use of historical data and a known relationship between population size and those historical data.
We found that range size was used as a correlate for population size in 44 (17%) studies to infer population declines (Fig. 1). This inference assumes that a range contraction is coincident with declines in population size. For example, if an amphibian species is not present at a pond or other site at the end of its range when a study was done but was known to exist there in a previous period, then a reasonable inference can be made that some range contraction has occurred (e.g., Fisher & Shaffer 1996; Lips et al. 2004). While some studies have found an adequate correlation between population size and range size (Channell & Lomolino 2000), there is little consensus with regard to the accuracy of such inference (Skelly et al. 2003; Goehring et al. 2007). However, recent advances in occupancy modeling (e.g., Bayesian state-space models) provide improved inference and lowered uncertainty to range map data (Royle & Kéry 2007; Tingley & Beissinger 2009).
Other correlates with population size can also offer insight into population trends in the absence of data derived from counting organisms. Habitat data are often more readily available. If population size and some habitat parameters are tightly correlated for a particular organism, then reasonable inferences about population size can be made based on appropriate habitat data (McCulloch & Norris 2001; Nystroem et al. 2007). Similarly, Roy et al. (2001) used a well-studied correlation between weather and population sizes of 23 species of British butterflies to estimate historical population sizes. Establishing reliable correlations among populations and habitat conditions provides one of the most promising, and currently underutilized, approaches to estimating historical populations.
Population declines are often studied on a short-term ecological time scale, but it can also be informative to examine population trends on long-term evolutionary time scales through fossils and genetics (Willis & Birks 2006). The fossil record is notoriously incomplete and is rarely used to infer historical population sizes (Fig. 1). If the record is available, however, powerful inferences can be made (Edgar & Samson 2004).
Genetic studies, on the other hand, are rapidly becoming more widely used to infer historical population sizes (Roman & Palumbi 2003). Laboratory techniques and analytical frameworks for genetic analysis have advanced at a startling pace (Briggs et al. 2009). In particular, recent application of Bayesian inference has allowed for thorough analysis of genetic data and detailed reconstruction of demographic histories (Pearse & Krandall 2004). Although there are a variety of assumptions in such genetic models (e.g., mutation and dispersal rates) and error and confidence intervals are often large, careful application of genetic analysis can provide the basis for reliable inferences about population trends (Goossens et al. 2006; Yeung et al. 2006).
Although rare, geologic and chemical traces may also provide evidence of population sizes of organisms. In one instance, concentrations of phosphorus and copper in the sediment of a penguin colony in the Antarctic were used as a “penguin proxy,” as concentrations of the metals were hypothesized to change as a function of guano quantity (Zale 1994). Stable isotopic signatures might also be useful for inferring population sizes given confidence in relationships between chemical signatures and population dynamics. Norris et al. (2007) correlated carbon and nitrogen isotopes in marbled murrelets with population sizes in existing colonies around the Pacific over 40 years and then used that correlation to estimate population trends in the 19th Century using isotope analysis from museum specimens of the species. Geochemical approaches are not commonly used to infer historical population sizes, but could hold promise for future efforts to infer long-term population trends in special cases such as these.
Human beings are prodigious, often enthusiastic, but highly selective observers of nature. Their historical observations are passed down through a variety of mechanisms including letters, diaries, notes, books, business records, oral communication, maps, paintings, photography, films, and increasingly a wide array of electronic forms (e.g., Milnergulland & Beddington 1993; Lloyd & Powlesland 1994; Lotze & Milewski 2004; Beissinger & Peery 2007; Jameson & Ramsay 2007). We have categorized some of the sources we found used in the conservation literature in Table 1; but the number of potential historical sources certainly exceeds this list.
Information about populations is increasingly being inferred from all of these kinds of sources. When a 16th Century explorer noted that the Manx shearwater (Puffinus puffinus) was “extremely abundant” throughout the Azores archipelago and modern surveys show that the bird is now extremely rare (see Monteiro et al. 1996), the utility of the historical observation for inferring a population decline is obvious. Useful population data have often also been derived from hunting and trapping records to infer, for example, populations of Eastern spotted skunks over the past 60 years in the Midwest (Gompper & Hackett 2005). These kinds of harvest or catch data have been most widely and effectively used to infer historical population sizes for large marine animals (see Lotze & Worm 2009 for a review).
The disadvantages of such data are two-fold. First, they are almost always gathered for other purposes and rarely collected for the purpose of establishing a population baseline for future studies. Records, therefore, tend to be spotty and often lack important information such as sampling effort. Second, they are frequently gathered in different historical, social, political, economic, and cultural contexts, which make interpretation of the data difficult and potentially misleading without taking these differences into account.
Counter-balancing these disadvantages, historical data can often be far reaching in temporal and spatial extent. People have been interacting with nature and recording their observations for many centuries across much of the Earth. We must use caution in the interpretation of such data and be cognizant of the uncertainties and difficulties inherent in making inferences from data derived from historical sources. However, because these sources are so widely available, if often fragmentary, they also provide powerful possibilities for using historical sources to corroborate and to raise questions about inferences derived from other sources of data.
Combining forces to understand population trends
The results of the literature review emphasize and quantify two major problems within conservation biology. (1) Few studies take a long-term perspective on population declines, even of the conservative subset we examined (datasets of 10 years or longer), only 15% looked beyond 100 years. (2) More than half of studies that do take a long-term perspective do not have continuous datasets. Both of these issues hamper our ability to accurately judge the extent of deterioration of the natural world. A long-term perspective based on deficient data can complicate conservation efforts by muddying our understanding of the ultimate causes of population declines and disguising natural baselines (for whenever that baseline might be defined).
Our results also point to opportunities, however. Counting approaches are relied upon heavily but all other methods are underutilized (Fig. 1). Of course limitations exist, but we argue that, whenever possible, alternatives to counting approaches should be sought out and used if appropriate. Using a wide variety of approaches for a more complete picture of population trends and acknowledging that “data can come from anywhere” (Sagarin & Pauchard 2009) will ultimately enhance our ability to diagnose and reverse declines.
Specifically, we suggest expanded use of nontraditional approaches for inferring population trends including the evolutionary, geochemical, and historical approaches. Increased use of sequencing and genetic markers in addition to advances in the population genetic analyses of these data in recent years provide an excellent conservation opportunity. It is becoming possible to derive long-term and accurate pictures of population dynamics with genetic data collected from living animals. One geochemical method increasing in popularity in ecology and potentially conservation is the application of isotopic data to arrive indirectly at historic population estimates. Finally, the historical approach and use of data gathered by people of the past may still be an under-appreciated source for historical population numbers despite its growing use and sometimes temporally far-reaching potential.
Some models for success exist for such multifaceted and interdisciplinary approaches, particularly in the context of marine conservation and fisheries management. The History of Marine Animal Populations as part of the Census of Marine Life initiated a collaboration between scientists and historians to synthesize the multiple sources of historical data on marine populations to produce innovative management recommendations (Starkey et al. 2008). Other recent syntheses in marine conservation have used multiple historical records to document extensive declines in marine fisheries (Jackson et al. 2001; Pinnegar & Engelhard 2007; Lotze & Worm 2009). Additionally, global efforts working to assess species statuses (e.g., IUCN Red Lists, Encyclopedia of Life, US Fish and Wildlife Endangered Species Program) often incorporate many types of historical data and deftly manage the balance between data reliability, accessibility, and temporal extent. These interdisciplinary efforts are a useful guide when trying to determine population trends or measuring changes in natural diversity.
Also, the importance of long-term population data is increasingly being recognized in ecology and infrastructures are being developed to address some of the difficulties in maintaining research projects over time. For one example, the U.S. National Science Foundation's Long-Term Ecological Research network is explicitly designed for study of long-term ecological phenomena. Such efforts and networks serve to extend the temporal perspective of counting and census approaches that have been limited traditionally.
Multiple methods should be used to infer population data, but assumptions made through the use of a particular method should also be made explicit. When possible, analysis of population data should accommodate the limitations of those methods as well. For example, once data have been compiled, trend descriptions can be improved by the use of analytic methods that incorporate the uncertainty in the population-inference approach. Bayesian inference can accommodate some data limitations such as imperfect detection and census gaps (e.g., Royle & Kéry 2007; Brooks et al. 2008; Kéry et al. 2009).
Rarely in conservation, do we have all the data required to accurately quantify past population dynamics or trends. In this review, we have highlighted the scope of this problem. Conservation studies tend to be short term in perspective, and the majority of studies do not monitor populations continuously. The result is a fragmentary and potentially biased view of population trends and vulnerability.
We also provide recommendations for addressing this situation. To bridge this data gap, we recommend using and integrating multiple data sources through counting, correlative, evolutionary, geochemical, and historical approaches. As evident in this review (Fig. 1), evolutionary, geochemical, and historical approaches in particular are underrepresented in the literature despite their utility, especially on longer time scales. Better understanding of the past is prerequisite to attempting restoration of populations to secure levels, and at best to something close to their natural baselines. We believe our understanding of the past can be improved through use of these multiple approaches and that thanks to recent methodological advances in all of these fields; such multidisciplinary efforts will become easier, less costly, and more accurate in the future.
We are grateful for the generous funding provided by the Woods Institute for the Environment at Stanford University for making this work possible through an Environmental Ventures Project. We thank Corey Bradshaw, Tim Caro, Tim Coulson, and two anonymous reviewers for insightful comments.