Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
1. Biodiversity assessments usually rely on indicators as surrogates for direct measures. Although the ecological validity of indicators has been extensively studied, their economic feasibility and cost-effectiveness have seldom been assessed.
2. Here we present a novel generic framework for analysing the cost-effectiveness of biodiversity indicators and the effect of budget allocations on the quality of biodiversity surveys. We sampled a suite of environmental and biological indicators in a Mediterranean ecosystem and calculated their cost-effectiveness using measures of species richness, rarity and composition.
3. Environmental indicators were the cheapest indicator for richness and rarity but not for composition patterns, and they conveyed low accuracy (<70% of the variation in diversity patterns). For higher accuracy, plants and a combination of plants and insects provided the most cost-effective indication of species richness, rarity and composition. Representation of composition patterns conveyed higher representation accuracy per given budget than richness patterns.
4. Marginal costs of improving the survey’s ecological performance were high, making a taxonomically extensive sampling strategy non-cost-effective. Taxonomic identification of species-rich invertebrate taxa is the major cost component in surveying these groups, and the availability of taxonomic expertise is a critical factor in determining their cost-effectiveness.
5. We further illustrated the effects of socio-economic context on the cost-effectiveness of indicators by comparing the expected costs of conducting this survey in California and Morocco, two Mediterranean-type regions at opposite socio-economic extremes. Labour costs and the need for taxonomic out-sourcing were the main sources of differences between regions, showing that cost-effectiveness of indicators is, to a great extent, context-dependent, and that the availability of in-house taxonomic expertise is a major determinant.
6.Synthesis and applications. The acquisition of reliable data on biodiversity distribution is often a major limiting factor in effective conservation planning and management. We show that biodiversity representation and site prioritization can be conducted efficiently with limited funds by explicitly incorporating costs into the selection of indicators. The generic framework developed here for cost-efficiency analysis of indicators can improve the quality and scope of biodiversity surveys and subsequently improve conservation decision-making.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Acquiring reliable data on biodiversity distribution is often a prerequisite for effective priority setting and management of conservation areas (Margules & Pressey 2000). However, due to limited time, money and taxonomic expertise, surrogates commonly replace direct biodiversity assessments. These surrogates are broadly categorized as environmental indicators, i.e. physical characteristics of the environment, and biological indicators – subsets of taxa expected to reflect wider patterns of diversity (Oliver et al. 2004). Two fundamental requirements of indicators are that they be ecologically reliable, i.e. adequately reflect biodiversity patterns, and economically favourable, i.e. less costly than a full biodiversity survey (Lawton et al. 1998; Wilson 2000). To date, most of the research on biodiversity indicators has focussed on evaluating their ecological performance but has largely neglected their cost-effectiveness, i.e. the ratio between their ecological performance and the cost of surveying them (Gardner et al. 2008; Grantham et al. 2008).
Biodiversity surveys are an important tool in conservation decision-making (Margules & Sarkar 2007). Generally, higher levels of protection will be given to regions that are physically and biologically more diverse and/or unique. Such decisions require detailed knowledge of the different taxa inhabiting the area, their relative abundance, and the temporal and spatial variation in their distribution. However, a severe shortage of funding and limited time lead to partial often biased surveys of a narrow set of well-known and easily surveyed indicators that do not necessarily reflect wider diversity patterns, e.g. various vertebrate groups (Balmford & Whitten 2003). Choice of the most appropriate biodiversity indicator is a hotly debated topic, as using unreliable indicators can lead to erroneous decisions (Grand et al. 2007). Many selection criteria have been proposed and a large body of literature exists on the topic (e.g. Pearson 1994; McGeoch 1998; Noss 1999; Hilty & Merenlender 2000).
The cost structure of a survey – the relative cost of its major components, labour, field and laboratory equipment and supplies, and travel and lodging – also plays an important role in the choice of indicator(s), and thus on diversity mapping. Cost structure may differ considerably between eco-regions and between countries within eco-regions, depending on their: (i) level of economic development, which affects labour and lodging expenses, and can be expressed by per capita GDP (Gross Domestic Product) (Diener & Suh 1997); (ii) availability of local taxonomic expertise, i.e. whether taxonomic identification can be done by local experts or needs to be out-sourced by either sending specimens abroad or hosting foreign experts; and (iii) availability of research infrastructure, e.g. laboratories equipped for taxonomic work, reference collections, field gear, etc. These socio-economic factors may affect the cost-effectiveness of indicators, and lead to the application of different indicators in different countries. Therefore, conservation decisions ultimately depend on survey budgets and cost structure. The potential effects of socio-economic factors on survey cost structure and consequently, on conservation decision-making have received only limited attention to date (Lawton et al. 1998; Balmford & Gaston 1999).
The Mediterranean biome, a global biodiversity hot spot of highest conservation priority (Myers et al. 2000; Brooks et al. 2006), is under intense development pressure (Hoekstra et al. 2005). Conservation practitioners and planners are often faced with local-scale decision-making, especially in densely populated areas where only limited land is still available, and habitat loss and fragmentation are intense. Indicators are needed mostly for local-scale biodiversity assessments, as many land-use conflicts are confined to an area of a few square kilometres (Vogiatzakis, Mannion & Griffiths 2006). Tools developed for large spatial scales (hundreds to thousands of square kilometres) may not be efficient for finer-scale diversity assessments because the changes in patterns are more subtle and spatial autocorrelation may be high. For example, the region of the present study was given high conservation value in a country-level evaluation (TAHAL 2004). However, detailed surveys were subsequently conducted when various local-scale development projects were proposed and could not be adequately scrutinized using the coarser evaluation method (Kaplan, Kimhi & Choshen 2000). The Mediterranean biome is an interesting case study for exploring the effects of the biodiversity survey’s cost structure on cost-efficiency of indicators, as countries within this climatic region differ significantly in per capita GDP, availability of taxonomic expertise, and institutional infrastructure for biodiversity research. At one end there are countries/regions with developed economies (relatively high per capita GDP) such as California, some Southern European countries and Australia, where standards of living are high and well-established research institutions provide taxonomic support and research infrastructure. At the other extreme are the countries/regions with a developing economy (relatively low per capita GDP), such as some Middle Eastern and North African countries, where standards of living are lower and taxonomic expertise and research infrastructure are largely lacking.
In this study, we investigate the cost-effectiveness of biodiversity indicators for local-scale diversity assessments in a Mediterranean ecosystem and explore how our results are affected by socio-economic factors, reflecting other Mediterranean countries/regions. Specifically, we ask the following questions: (i) What is the shape of the cost-efficiency curve for different indicators and sets of indicators in the studied Mediterranean ecosystem? (ii) What is the optimal choice of indicator(s) under different budget constraints? (iii) How does budget allocation for biodiversity surveys affect site prioritization? (iv) How may socio-economic context generally affect the survey cost structure and ultimately the cost-effectiveness, and optimal choice of indicators?
Materials and methods
In 2003–2004, a comprehensive biodiversity survey was conducted in the Jerusalem Mountains and Judean foothills, a Mediterranean ecosystem in central Israel. The survey encompassed forty 1000-m2 plots representing the typical vegetation formations of the region (see Mandelik 2005; Mandelik et al. 2007). Five taxa were selected, based on the availability of standardized survey methodologies, local taxonomic expertise, sensitivity to local-scale habitat changes, and proven indicative abilities in other ecosystems on local and wider spatial scales (see review by Hilty & Merenlender 2000; Rodrigues & Brooks 2007): annual and perennial vascular plants, ground-dwelling beetles, moths, spiders, and small mammals. Moths were sampled in 25 plots that were not adjacent to light-contamination sources to avoid biased sampling. Additional species-poor taxa were sampled (scorpions, five species; diplopods, six species; reptiles, six species) but they were excluded from the analysis due to their low species numbers and abundance which limited the power of the analysis. The sampling effort needed to achieve a representative sample in this ecosystem was previously investigated (Mandelik et al. 2002) and set accordingly. In addition, a set of coarse- and fine-scale environmental variables (altitude, slope, aspect, foliage cover and heterogeneity, ground cover and heterogeneity; see Mandelik 2005) were recorded. All taxa were identified, mostly to the species level, by local expert taxonomists. The survey accounted for the major seasonal and spatial variation components in the studied ecosystem. Detailed species lists and diversity analyses are reported in Mandelik (2005) and Mandelik et al. (2007).
We classified three main cost categories: labour, equipment/supplies, travel and lodging, and four main stages in biodiversity surveys: field collection, laboratory processing (including taxonomic identification), museum curation (including labelling and data entry, and preservation), and data analysis. We calculated the monetary cost of surveying each of the indicators by summing the different cost components at each stage (see Table S1, Supporting information for detailed cost calculations).
Labour costs in Israel were calculated by multiplying the number of hours spent working on each indicator, during the four described stages, by the hourly cost to the employer. We classified three professional levels of personnel involved in the survey: unskilled field and laboratory assistants (6 USD per hour), trained technicians (including graduate students; 10 USD per hour), and expert biologists (32 USD per hour). Hourly payment rates were obtained from the standard salary tables of academic institutions in Israel. Perishable materials and supplies included field and laboratory items used for field sampling and processing of specimens, such as traps, chemicals, insect pins, etc. Non-perishable equipment included items used for specimen identification and long-term storage such as stereoscopes and storage cabinets. Travel costs were calculated based on the distance of the field site from the academic institution (Tel-Aviv University), lodging costs were added when an overnight stay was required, and food costs were calculated on a per diem basis. Labour costs for each taxon were estimated for the survey as a whole, while perishable materials and supplies and travel and lodging were calculated for a single sampling and multiplied by the number of samplings conducted.
When conducting field surveys, some of the costs, such as travel and lodging, are shared among taxa. Similarly, when using non-selective traps, non-target taxa will be sampled. Thus the combined costs of sampling sets of indicators are usually lower than implied by a simple additive calculation of single-taxon survey costs. To test the cost-effectiveness of using single vs. suites of indicators, we calculated the cost of sampling sets (combinations) of indicator taxa, taking into account these ‘shared costs’. In total, we had 26 different indicators and sets of indicators – five single taxon, environmental variables (regarded as one type of indicator), 10 possible pairs and 10 possible triplets of indicators.
Ecological performance of indicators
Each indicator was tested for its ability to reflect species richness, rarity (number of rare species) and composition (β diversity using Sorensen’s qualitative similarity index) of all taxa combined, i.e. the indicator taxon was a subset of whole biodiversity measured (Rodrigues & Brooks 2007). Naturally, the ecological performance of each taxon will be affected by the number of species it adds to the overall species pool. However, representation of total diversity patterns is generally the ultimate goal of biodiversity surveys and we wanted our analysis to reflect this. In addition, there are no external sources for diversity-pattern comparisons in this region. The indicative ability in our analysis is comprised of the number of species the indicator taxon contributes to the total species pool and its correlative relation with the other taxa. Rare species were classified as those in the first quartile of the abundance distribution of each taxon (Gaston 1994) corresponding to 50–76% of the species of each taxon. For all analyses, we included only the 25 plots in which moths were sampled. We use the term ‘diversity’ to refer collectively to the three components analysed (species richness, rarity and composition).
Cross-taxon congruence in species richness and rarity was analysed by regressing the richness and rarity of each indicator (using linear regression) and set of indicators (using stepwise multiple regression) against the richness and rarity of all taxa combined, respectively. Cross-taxon congruence in species composition was analysed using the average Sorensen’s similarity index of each plot with all other plots, calculated for each indicator and set of indicators. These values were regressed against the average Sorensen’s index for all taxa combined. We further examined correlations between the regression coefficients for the analyses of species richness, rarity and composition in order to test for congruence between these parameters.
We used a principal component analysis (PCA) on the environmental variables to extract main axes of environmental variation while accounting for correlative variables (see Mandelik 2005). Ground and vertical cover variables were arcsine transformed; altitude, aspect, and slope were square-root transformed. We conducted a forward stepwise multiple regression to test the relationships between the environmental variables (four main PCA axes, accounting for over 79% of the variation in the environmental variables) and species richness, rarity, and composition. We obtained consistent results when applying forward vs. backward selection procedures (Mandelik 2005) and further reduced potential inconsistencies by using non-correlated PCA axes (Whittingham et al. 2006).
Cost-efficiency analysis of indicators
We used the adjusted R2 of the regression models of the richness, rarity and composition analyses as measures of ecological efficiency. We plotted these coefficients against the cost of sampling each indicator and set of indicators to obtain a cost-efficiency curve. The indicators/sets of indicators that have the highest ecological performance under different budgets collectively form what we define as a ‘cost-efficiency frontier’. All indicators not on the frontier have equal or lower ecological performance relative to those on it, but cost more. We applied the Bonferroni method to account for multiple testing.
Travel and lodging costs are highly case-specific, e.g. dependent on the distance between study sites and the research institution and logistics. However, since this component may comprise a substantial portion of the survey costs, we tested the correlation between the costs of conducting the survey with and without travel and lodging for each indicator/set of indicators.
We further explored the effect of budget allocation on the probability of erroneous richness and composition mapping and consequent erroneous site prioritization. Plots were ranked according to their total species richness and composition similarity. The latter was based on average values of the similarity index for the 24 possible pairs of plots for each plot. These were referred to as the ‘correct rankings’. For each indicator and set of indicators, we performed a 25-step process starting with the plot with the highest number of species, or lowest similarity values, i.e. the most unique species composition. This would be the first plot to be set aside for conservation according to this indicator/set of indicators. Next, we added the remaining 24 plots, one at a time, according to their number of species, or similarity index values. We then compared the ‘correct ranking’ with the ranking obtained for each indicator and set of indicators. The number of plots from the ‘correct ranking’ that were not included in each step of the indicator ranking were referred to as mistakes in site prioritization. We compared these figures to: (i) average mistakes – average number of erroneous plots chosen compared to the ‘correct ranking’, for each indicator and set of indicators, at each step separately, and for all 25 steps together, and (ii) null (baseline) probability of making such mistakes – the number of erroneous plots from sets of randomly chosen plots, regardless of their rank.
Finally, we explored the effect of socio-economic factors on the survey cost structure of indicators by analysing two opposing case studies – California, among the highest per capita GDP in this biome (41 663 USD; US BEA 2008), with ample well-established research institutions harbouring broad taxonomic expertise, and Morocco, among the lowest per capita GDP in this biome (2145 USD; World Bank 2008), where taxonomic expertise, as well as infrastructure for field surveys are largely lacking.
Since the main purpose of this analysis is to provide a general illustration of the impact of socio-economic context on cost-effectiveness of indicators and since no data are readily available on the exact cost structure in Morocco or California, we applied the aforementioned gauges to evaluate the different costs in these countries. We evaluated the costs of conducting our survey in California and Morocco based on the per capita GDP ratios in comparison to Israel [used for calculating expected labour and lodging costs (see Diener & Suh 1997); per capita GDP of Israel 19 927 USD (World Bank 2008)], travel expenses compared to Israel (based on cheapest car rental and petrol rates), and the need for taxonomic out-sourcing and non-perishable equipment in Morocco but not California (see Table S2, Supporting information for detailed cost estimations). We evaluated the accuracy of using per capita GDP ratio as an indicator for the difference in labour and lodging costs by comparing the ratio between wages of biological scientists in California (Bureau of Labor Statistics 2008) and Israel (no exact wage values were available for Morocco). The ratio between wages and per capita GDP in California and Israel is 2·02 and 2·09 respectively, corresponding to the ratio between wages of experts in the two regions ($48·57 in California, $24 in Israel). Similar results were obtained for technicians. We assumed that taxonomic out-sourcing in Morocco would be needed for the species-rich taxa for which no species-level field guides are available (i.e. beetles, moths and spiders) and would include labour costs (calculated at a rate of 3 USD per specimen, based on previous studies conducted in our region; Y. Mandelik, unpublished data) and costs of shipping specimens to European countries where many of the relevant reference collections and much of the expertise for the fauna of this region are found.
A total of 420 plant species (2800 specimens), 424 beetle species (12 656 individuals), 111 moth species (10 397 individuals), 102 spider species (8119 individuals), and 8 mammalian species (544 individuals) were sampled (see Mandelik 2005 for detailed species lists). Surveying these cost 117 806 USD, of which labour cost 83 884 USD, field and laboratory equipment/supplies cost 23 262 USD, and travel and lodging cost 10 660 USD.
The correlation between species richness and rarity accounted for 85% of the variation in regression coefficients (r = 0·921, P ≪ 0·001). Correlations between species composition and species richness and rarity were lower, accounting for c. 46% and 54% of the variation in regression coefficients, respectively (composition-richness: r = 0·68, P ≪ 0·001; composition-rarity: r = 0·738, P ≪ 0·001). We therefore present the cost-efficiency analyses for richness and composition; cost-effectiveness for rarity was highly similar to that of the richness analysis.
Travel and lodging constituted on average 11·5 ± 5·5% of total survey costs in Israel for the different indicators and sets of indicators. Survey costs with and without travel and lodging were highly correlated (R2 = 0·996, P ≪ 0·001). We therefore present the analysis in which travel and lodging costs were excluded, to better focus on the other cost components analysed.
Labour was the major cost component when sampling fauna and flora (37·5–80% of total costs for the different indicators), but not environmental variables. The expenses for field vs. laboratory labour differed greatly among taxa and depended on the ease of their taxonomic identification (Fig. 1). When sampling beetles, moths and spiders – small-bodied species-rich arthropod taxa – laboratory work constituted c. 50–75% of total survey costs, mainly due to time-consuming specimen identification requiring high expertise. Due to the high cost of labour, total costs incurred in sampling these indicators were highest (Fig. 1). When sampling taxa which can be identified with relative ease by non-experts – small mammals and plants in this study – most of the costs were incurred in field collection, and total costs were much lower (Fig. 1). Sampling environmental variables was by far the cheapest option.
The richness and composition cost-effective frontiers were similar in terms of included indicators, but throughout much of the inspected range, the composition frontier was higher, i.e. it conveyed higher representation accuracy per given budget (Fig. 2; see Table S3, Supporting information for detailed regression results). Environmental variables represented 63% of the variation in richness patterns and constituted the lower end of the curve, but did not correlate significantly with composition patterns (Fig. 2). While an exhaustive representation of richness and composition patterns in this ecosystem cost more than 88 000 USD (using the combination of beetles, plants and moths), only 3% of this sum (3040 USD) was required for the minimal representation of richness patterns using environmental variables (Fig. 2). Only 11% of the total sum (c. 9600 USD) was required for the minimal representation of 77% of the variation in species composition by sampling plants. Plants were included in all sets of indicators along the richness and composition frontiers (Fig. 2).
Most indicators and sets of indicators decreased the number of erroneous plot rankings and improved site prioritization compared to null (random) site selection (Fig. 3). For the analyses of both species richness and species composition, plants had the lowest average mistakes per single indicator taxon, and beetles and plants had the lowest average mistakes per pair of indicator taxa (Fig. 3). The combination of beetles, plants and moths performed best for both richness and composition representation, having 0–2 erroneous plots per step (Fig. 3). However, the combination of beetles and plants, and even plants alone, had generally lower errors compared to the null and average number of mistakes using the 26 possible indicators and sets of indicators.
Labour contributed the most to total costs and to differences between countries/regions with different survey cost structures (Fig. 4). These differences were greatest for assistants and technicians. The estimated expenditure on expert taxonomists was similar in California and Morocco (106 195 USD and 101 916 USD, respectively), compared to 50 720 USD in Israel, but constituted 75% of the total estimated survey costs in Morocco, compared to 51% in California, and 43% in Israel. For all taxa except moths, sampling was predicted to be most costly in California, because although taxonomic expertise is available, labour costs are higher (Fig. 4). The estimated differences between countries/regions were greatest when sampling beetles, probably because of the expected high diversity and subsequent extensive taxonomic work. Sampling environmental variables exhibited similar low costs in all three countries (Fig. 4). Travel and lodging were estimated to constitute on average 9·5 ± 6·8% and 9·8 ± 5·8% of total survey costs in Morocco and California, respectively, for the different indicators and sets of indicators. Correlation between survey costs with and without travel and lodging were highly significant for both California and Morocco (R2 = 0·998, 0·999 respectively, P ≪ 0·001 for each).
This study provides the first cost-efficiency analyses of environmental and biological biodiversity indicators for a Mediterranean ecosystem. By presenting the use of a cost-efficiency frontier and analysing how it is affected by different socio-economic factors, we provide a generic framework that can be instructive for other Mediterranean ecosystems, as well as other biomes. Though this study is prone to case-specific issues, such as the sampling effort and sampling techniques applied, our survey addressed main seasonal and spatial variation in the ecosystem, and we obtained comprehensive data sets (Mandelik et al. 2002; Mandelik 2005). An important finding was that a minimal representation of c. 70% of the variation in diversity patterns is feasible, even with limited funds (less than 10 000 USD), if a cost-efficient indicator is applied. Favourable conservation outcomes in other problems related to the economics of biodiversity conservation have been obtained upon incorporation of cost-benefit information (Naidoo & Adamowicz 2005; Naidoo & Ricketts 2006), especially when costs and benefits were strongly positively correlated (Ferraro 2003). In our study, plants were the cheapest cost-efficient indicator for richness and composition patterns. However, the marginal costs of representing the additional c. 30% of diversity variation are high, requiring c. 9 times the initial budget. Thus, the accuracy needed is a main factor in determining the budget requirements of biodiversity surveys. We further found that the cost-efficiency of biodiversity indicators is, to a great extent, context-dependent, and affected by socio-economic factors.
Environmental variables are the cheapest indicator for local-scale diversity assessments in the studied ecosystem (Grantham et al. 2008), but have two major drawbacks: they reflect richness (and rarity) patterns, but not composition, and they provide only coarse diversity assessments (<70% of the variation). Similar poor performance of environmental indicators has been found in other studies on larger spatial scales (regional and continental; see Hortal, Araújo & Lobo 2009 and references therein) and when selecting complementary conservation networks (Rodrigues & Brooks 2007), suggesting that they might be suitable for coarse-filter habitat classification but cannot replace the fine-filter biological indicators needed for local diversity mapping. When higher accuracy in representation is needed (>70% of the variation, as in most cases), as well as an indication of compositional patterns, plants are the cheapest indicator. Plants are a major component along the efficiency frontiers, and are also the most efficient single taxon for site prioritization. The cost-effectiveness of plants may stem from the lower costs of floral compared to faunal surveys (particularly of small-bodied, species-rich arthropod taxa), and the relatively high performance of plants as ecological indicators (Lawton 1983), including successful application of the higher taxa and similar approaches (Mandelik et al. 2007; Mazaris et al. 2010). Interestingly, ecological assessments of development projects (e.g. Environmental Impact Assessments, EIAs), are usually funded at or above 10 000 USD. However, an analysis of the ecological quality of EIAs in Israel has shown that most are of poor quality and fall short of this threshold (Mandelik, Dayan & Feitelson 2005).
The costs of representing richness, rarity, and composition are similar at the high end of the cost-efficiency curve (representing over c. 90% of the variation). At lower values, richness (and rarity) is more costly to represent than composition. Measures of species composition better reflect cross-taxon congruency in both diversity patterns and responses to disturbances than measures of species richness (Barlow et al. 2007) and are thus given preference in conservation applications (Margules & Pressey 2000; Margules & Sarkar 2007). Our analysis shows that the representation of composition patterns is also more cost-effective than richness.
As expected, the more taxa sampled, the higher the ecological performance achieved. However, the cost-efficiency of a taxonomically extensive sampling strategy is low, as evidenced by the relatively flat shape of the efficiency frontiers and the high marginal costs of improving the ecological performance: beyond the initial level of representation of c. 70–75% of the variation in diversity patterns, a 1% increase in the representation of richness and composition patterns costs an average 2475 USD and 3755 USD, respectively (Fig. 2). Highly similar diminishing returns on investment in surveys have been found in other studies (Grantham et al. 2008). Therefore, the choice of any of the indicator(s) appearing on the cost-efficiency frontier should ultimately be based on the accuracy needed in mapping biodiversity, and on the urgency of conservation action (Grantham et al. 2009).
Taxonomic identification of species-rich invertebrate taxa is expensive, and makes up a large part of the total cost of surveying these groups, as has been found in tropical ecosystems (Lawton et al. 1998; Gardner et al. 2008). Hence, the availability of taxonomic expertise is a critical factor in determining the cost-effectiveness of surveying most invertebrates, including those regarded as good biodiversity indicators such as some beetle and spider groups (McGeoch 1998; Hilty & Merenlender 2000; Pearce & Venier 2006). For the faunal taxa that require laborious identification, the higher the number of species, the higher the sampling costs, in contrast to prior assumptions (Rohr, Mahan & Kim 2007). Beetles, despite their good indicative ability in Mediterranean ecosystems (Mandelik et al. 2007; Zamora, Verdú & Galante 2007), appear only in the second half of the efficiency frontiers. Furthermore, reducing field expenses by using sets of indicators sampled by the same technique, such as beetles and spiders sampled with pitfall traps, does not improve cost-efficiency (but see Gardner et al. 2008).
In countries where labour is costly, species-rich taxa that require high expertise and time for their identification may not appear on the cost-efficiency frontier, despite having good ecological performance. In those countries, cost-effectiveness analysis might lead to the application of an indicator(s) with lower indicative abilities if its sampling is less labour-intensive. The ‘taxonomic impediment’– the severe shortage of taxonomic expertise in most parts of the world (Giangrande 2003), might further decrease the cost-efficiency of indicators needing expert identification, and may enhance the application of gauges that are ecologically less favourable. Furthermore, over-reliance on cost-efficiency analyses may limit the search for, and development of, new indicators that are currently less cost-efficient due to poor taxonomic knowledge and lack of sampling methodologies (Pawar 2003).
A need for taxonomic out-sourcing due to lack of in-house knowledge will affect mostly developing economies, as this may consume a large part of their total survey budget, as illustrated in our analysis for Morocco. Though travel and lodging costs might be higher than the cheapest rate we accounted for, they constituted c. 10% of all survey expenses, and thus have a limited impact on the survey’s cost structure. Naturally, some of our cost estimations might not always be fully realized; nonetheless, our analysis illustrates two contrasting extremes along a gradient of interconnected socio-economic factors characteristic of the Mediterranean biome. Our analysis further showed that the cost structure of biodiversity surveys greatly affects the total costs of surveying different indicators, and consequently the optimal selection of indicator(s). Hence, the accuracy of conservation decision-making is to a great extent context-dependent and will ultimately be dictated not only by overall funding allocation but also by socio-economic factors, mainly per capita GDP and availability of in-house taxonomic knowledge.
The development and application of DNA barcode technology may affect our results and conclusions. These technologies are likely to reduce the cost of identifying species-rich taxa (Hebert et al. 2003; Kress et al. 2005; Hajibabaei et al. 2007) and the need for taxonomic out-sourcing, placing additional taxa on the cost-efficiency frontier. The application of the higher taxa and similar approaches may similarly reduce cost of taxonomic identification and affect the cost-efficiency frontier (Mazaris et al. 2010).
Our generic framework may facilitate reallocation of survey funds to expand and/or better focus the spatial, temporal, and taxonomic scope of biodiversity surveys and to include gauges for functions and processes that are essential for long-term management of ecosystems (Kremen 2005). The data produced using cost-efficient indicators would ultimately improve the link between monitoring programmes and procedures of risk analysis, site prioritization and adaptive management (Cleary 2006). To achieve this goal, however, cost-efficiency analyses of indicators in other ecosystems, on different spatial scales and with different taxa are needed, so that general guidelines for the optimal choice of indicators can be formulated.
Data acquisition is only the first step in effective conservation. In light of ever-limited conservation budgets and intense development pressures, data acquisition is competing with subsequent conservation actions for time and money (Grantham et al. 2008). The trade-offs between the cost and time required to get more data vs. applying it in subsequent conservation actions and the urgency of doing so (rate of habitat conversion and fragmentation; Grantham et al. 2009) should ultimately dictate the allocation of time and money spent on the different stages of the conservation process. The strong diminishing-return pattern in acquiring additional data on biodiversity found here and in other studies (Bode et al. 2008b; Grantham et al. 2008) points to the need to move away from the traditional approach of trying to get as much data as possible to a more critical and holistic evaluation of the marginal value of additional data to the conservation process as a whole.
We thank J. Hortal, S. Meiri, M. Coll and three anonymous reviewers for their most thoughtful and valuable comments and E. Ungar for statistical advice. U.R. is supported by the Adams Fellowship Programme of the Israel Academy of Sciences and Humanities. This study was funded by the Hebrew University of Jerusalem Ring Center for Interdisciplinary Environmental Research.