Nature Conservation: Cost Effective Biological Surveys and Data Analysis

Costs and Compromises

CHAPTER 1

Cost Constraints on Surveys for Nature Conservation

Andrew A. Burbidge

A cost breakdown of three contemporary surveys that address study areas of very different sizes and complexities is presented. Their overall costs are related to costs of single species studies and the size of the budget of the Western Australian Department of Conservation and Land Management. There is an increasing demand for surveys but little chance of significant increases in staff and financial resources to pay for them. Thus, cost-effectiveness is now an important consideration during the planning and execution of surveys. Those planning surveys need to ensure that only those data required are collected and that costs are kept to a minimum.

Biological surveys have developed rapidly in Australia in recent years. They have become an essential prerequisite for many land use decisions and for planning the management of nature conservation lands. They have also become increasingly expensive, partly because more and more detailed data are collected. In order to discuss cost constraints on surveys it is first necessary to define ‘survey. Many of the activities of both professional biologists and other conservation Department staff working on aspects of nature conservation in the field can be described as survey; so can many of the observations made by naturalists. Usually, definitions of survey equate it with biogeography (including descriptive ecology) and exclude experimental ecology.

Biogeographic studies describe the present and/or past patterns in the geographic distribution of organisms over periods ranging from decades to millions of years, usually over areas at least as large as a natural district. Descriptive ecology, in contrast, is used to distinguish non-experimental studies confined to relatively small study areas such as a single reserve, or along a corridor between reserves (McKenzie 1988). Both can be termed survey and are discussed here. Most studies of field ecology in Australia include elements belonging to the ‘descriptive ecology-biogeography’ continuum (McKenzie 1988), because they contribute to our knowledge of:

1.  Patterns in the distributions of species or in the composition and diversity (e.g. species richness) of communities.

2.  The status (e.g. commonness, rarity, endemism, relictualism) of species and communities.

3.  The geographic and temporal scales at which various species operate.

4.  The influence that physical and biological factors have on ecological boundaries, the relative abundance of species, and the complexity and dynamic nature of communities at various localities at various times. Examples include: disturbers (fire, drought, exotic species), environmental gradients (scalars such as substrates, altitude, climate), barriers to dispersal, and connections with other organisms (food chains, competition for resources).

It is easy to demonstrate that nature conservation would be better served by a more comprehensive database on the continuing changes in abundance and distribution of plants and animals. For instance, the Western Australian Government recently announced a new policy on mining in national parks and nature reserves. One of the key points in this policy is that mining will be excluded from areas ‘of the highest biological or landscape value’. In deciding what is of the highest value the State’s Environmental Protection Authority is required to take into account, inter alia, the presence of endangered species, communities or habitats, the presence of significant wilderness or important wetlands and the importance of the area in terms of its role in protecting representative ecosystems. Clearly, unless there is a reasonably comprehensive survey database it will be difficult to provide definitive statements on what is endangered and what areas protect representative ecosystems.

The major limitations on the number and thoroughness of biogeographic and descriptive ecological surveys are staff and money. Surveys by professional biologists and naturalists are expensive. In nature conservation and other scientific research agencies they compete for scarce financial and human resources with many other activities. How much should be spent on surveys is probably not a realistic question to ask. What proportion of available nature conservation resources should be allocated to surveys is, perhaps, a better question.

Resources for surveys are unlikely to expand significantly by general budgetary growth in the near future because Governments are currently preoccupied with cutting public expenditure and reducing taxes. Thus, growth in resources for surveys is likely to come only from a larger share of existing budgets or from external grants. Even with growth, resource allocation for surveys will rarely reach a level that a scientist would consider ideal. In this context we should examine carefully costs and benefits and what trade-offs should be made between the amount of information that is desirable and the amount of money an agency can afford to, or is prepared to, spend.

Costs and Benefits

The cost of surveys varies with:

1.  The size of the area being examined.

2.  The number and types of groups of organisms being surveyed.

3.  The taxonomic status and ease of identification of the various groups.

4.  The type of survey design employed, especially the number of visits to an area, the amount of time required to collect sufficient data and the amount of data analysis required.

5.  The distance and ease of access to the survey area from the research base.

To work out the real cost of surveys, ‘hidden’ or indirect costs have to be considered. These include salaries, depreciation, specimen identifications by other agencies, museum and herbarium curation of specimens, publication and administrative overheads. Table 1 compares the costs of three recent surveys carried out in Western Australia.

The surveys were selected to provide examples of:

1.  a broad biogeographic regional survey (the Nullarbor Region, McKenzie & Robinson 1987),

2.  a survey confined to a particular community type throughout a region (Kimberley rainforests, McKenzie et al. 1988), and

3.  a survey confined to a single reserve (descriptive ecology of Yanchep National Park, Allan Burbidge & Jim Rolfe, unpublished).

It is easy to see that surveys are expensive. Table 1 also shows the extra costs of working in remote, inaccessible places: the figure for vehicles, fares and charter is very high for the rainforest work because of the high cost of air fares between Perth and the Kimberley and the need to use helicopters for access to study sites.

The benefits of surveys are more difficult to quantify, particularly because some of them may not yet be realized. Many benefits are long term; the values of base-line data frequently increase with time. Benefits and limitations of well-designed surveys were discussed by McKenzie (1988). Benefits may include:

1.  Identification of endangered and/or endemic species or ecosystems.

2.  Resolution of land-use conflicts, including the explicit design of representative systems of reserves for nature conservation.

3.  Clarification of status of species or ecosystems (rarity, specialized habitat requirements, etc.), thus enabling research and management resources to be directed to areas of real need.

4.  Better management of nature conservation reserves, including setting priorities for resource allocation between reserves, resolution of some fire management questions including the positioning of firebreaks and the relationship between time since fire and presence/absence of some species, proper placement of recreational and management facilities, identification of problems such as the presence or absence of feral animals and invasive weed species, fungal disease or insect damage, and identification of trends in distribution and abundance of species and ecosystems.

5.  The setting of priorities for more detailed research studies.

Surveys often are not capable of answering all questions that arise from the data generated. The most obvious unanswered questions relate to the identification of causes for demonstrated changes in distribution and abundance. More specific and detailed ecological studies are needed to resolve such questions.

That surveys can answer important management questions is undoubted. In a recent workshop held by the West Australian Department of Conservation and Land Management and the CSIRO Division of Wildlife and Ecology, managers and planners were asked to consider practical examples of decisions they regularly had to make, and to identify the sorts of information they thought would be most useful for setting priorities for management (McKenzie 1988). A request to relinquish a reserve was used as an example. Four discrete sets of data were identified during the discussion that followed. All were of the ‘what and where’ sort.

1.  Are any rare or endangered species present?

2.  What is the species diversity (especially in richness and composition) of the communities found in the reserve and in the region?

3.  What are the major geographic patterns in diversity (especially richness and composition)?

4.  Is each ecosystem/community/species protected in another reserve?

These categories of information are all available or potentially available through well-designed biogeographic surveys. For these and other reasons surveys have become an important prerequisite for land use decisions, for land management planning and for environmental impact assessment.

Paying for Surveys

‘He who pays the piper calls the tune’. Who pays for surveys and what sorts of data do they want? Most surveys are paid for and carried out by Governments. Some surveys required for environmental impact assessment or for monitoring the effects of developments on the environment are paid for by private companies. A few surveys are carried out and paid for by private clubs or individuals.

TABLE 1. Cost (in 1988 dollars) of three recent surveys in Western Australia

Since most surveys are paid for by Government it might be instructive to see what proportion of funds are allocated to surveys by Government Departments. Unfortunately, such data are not readily available. In the case of my own Department the 1987/88 figures are roughly as follows:

Thus, while 7% of the total Departmental budget is allocated to the Research Division, only 0.5% is spent on biogeographical and descriptive ecological surveys. Additional funds spent on special purpose surveys, e.g. searches for rare species, bring the total proportion of the budget spent on ‘surveys’ to 0.6%. Only the ability to attract external grants has allowed expensive broad-scale surveys, such as those of the Nullarbor Region and Kimberley rainforests, to be carried out in recent years. In contrast, the cost of detailed ecological research on an endangered species, the Numbat (Myrmecobius fasciatus), has cost about $769800, including external grants of $216350, since 1981 (past costs adjusted for inflation at 10% per annum to approximate 1988 dollars).

The proportion of budgets spent on surveys by other similar Departments in Australia will vary with responsibilities and direct comparisons are not possible even if the data were available. However, the general picture is unlikely to be very different. It could be argued, therefore, that nature conservation Departments spend too small a proportion of their resources on surveys. If this is so, they must not realize the benefits of such work. How, then, do you justify getting a larger slice of the cake? The following strategy is one that might bear fruit.

1.  Work out priorities for data acquisition.

2.  Define the target for the data.

3.  Educate the public and those people making decisions as to the value and benefits of survey data.

4.  Demonstrate that surveys are cost-effective. Obtaining more resources is only one way of increasing the amount of survey data collected. Another method, that should run in parallel with efforts to get a bigger slice of the cake, is to develop cheaper and more efficient ways of collecting and analyzing survey data. These include:

1.  Collecting only those data that are relevant and useful in answering the questions being asked. The ‘shotgun’ approach to surveys rarely is cost-effective.

2.  Using automated data capture techniques, e.g. bat and bird call recorders, automatic weather stations.

3.  Using computers to record and store data in the field (i.e. reducing the number of steps between an observation being made and report presentation), to analyze data and produce reports.

4.  Ensuring that publication costs are kept down by including analysed and tabulated data rather than raw data such a annotated lists of species.

5.  Using desktop publishing.

6.  Using volunteer labour.

7.  Tapping into non-research staff in the agency.

Surveys have rapidly become more popular and more sophisticated in recent years. In most areas of endeavour, particularly in relatively new areas with developing technologies, the trend is to demand more and more detail leading to more and more expense. That biological or ecological surveys have followed such a trend can easily be demonstrated by a quick look at the recent literature (see Burbidge 1984). Obviously this trend cannot continue if surveys are to continue to provide information across the broad spectrum of problems that face us. Thus, good survey design is becoming increasingly important. Careful planning of surveys is needed to ensure that the resources available are used efficiently. Important questions that should be asked during planning are:

1.  Why are you planning this survey? Is it more important than other studies?

2.  What exactly is the question being asked?

3.  What is the minimum data set required to answer the question? How many organisms or groups of organisms should be surveyed?

4.  What is the most efficient and cost-effective method of collecting the necessary data? Does the sampling strategy provide data that are suitable for formal quantitative analysis and statistical appraisal?

5.  Who else may be able to use the data? Will they contribute to the costs of collecting the data?

6.  What is the most cost-effective method of communicating the results?

Conclusions

Biological surveys are becoming an increasingly important part of the process of nature conservation. Data produced by surveys are needed for making land use decisions, assessing the environmental impact of development proposals, developing land and species management plans and indicating the direction of detailed research. In recent years costs of surveys have increased sharply, partly because of the greater quantity and detail of data collected and partly because of inflation. There is a need, therefore, to consider the cost of surveys in the planning and design phase more than has been the case sometimes in the past.

Australia is a large country with enormous nature conservation problems. We still lack basic information on the distribution and abundance of a wide variety of organisms and on the changes taking place in their assemblages, so well-designed surveys have the potential to provide considerable benefits. In order to realise this potential, surveyors must plan scientifically sound surveys with two main thoughts in mind; overall nature conservation requirements and cost-effectiveness.

Acknowledgements

I thank Norm McKenzie, Allan Burbidge and Jim Rolfe for providing the cost data in Table 1, Tony Friend for providing the cost of Numbat research and Norm McKenzie and Roger Underwood for constructive criticism of a draft of this paper.

References

Burbidge, A.A. (1984). Selecting and managing parks and reserves : interpretation and communication of survey data. In: Survey methods for nature conservation. (Eds.) K. Myers, C.R. Margules and I. Musto. pp. 337–402. CSIRO : Melbourne.

McKenzie, N.L. (1988). Biogeography and its use for setting priorities for management. In: Ecological theory and biological management of ecosystems. (Eds.) D.A. Saunders and A.A. Burbidge. pp 39–44. Occasional Paper No. 1/88 Dept. Cons. Land Manage.: Perth.

McKenzie, N.L. and Robinson, A.C. (Eds) (1987). A biological survey of the Nullarbor region, South and Western Australia, in 1984. Dept. Environment & Planning: Adelaide.

McKenzie, N., Kenneally, K. and Winfield, C. (1988). W.A.’s rainforests. Landscape, 3, 16–22.

Department of Conservation and Land Management, W.A. Wildlife Research Centre, P.O. Box 51, Wanneroo, W.A. 6065

CHAPTER 2

Mapping the Environment at Different Scales:

Benefits and Costs for Nature Conservation

R.L. Pressey and M. Bedward

The benefits of mapping are quantified and related to map scale. The primary benefit of mapping is the delineation of units which are more homogeneous in terms of physical attributes or species than the landscape at large. A secondary benefit is the representation of fine scale unmapped diversity, in this case species, by reserving examples of map units. Analyses of costs of soil and land use surveys in relation to map scale are reviewed and assumed to apply to vegetation mapping. Costs of reserving units at different scales are simulated. Cost-effectiveness curves for homogeneity/survey cost, species reserved/survey cost and species reserved/reservation cost all peak at coarse map scales and decline with finer scales. If the benefits of mapping are to be maximised then mapping and reserve selection are necessary at a relatively fine scale, despite low cost-effectiveness.

Maps of natural resources are an important basis for nature conservation. Maps allow natural features to be identified, located and described and their relative extent and conservation status to be assessed. Production of a resource map requires a choice of appropriate scale. Mapping exercises vary from low intensity, extensive surveys termed exploratory or reconnaissance (small scale), to highly intensive surveys of restricted areas (large scale) (e.g. Kuchler 1967, Dent and Young 1981). Key factors in the choice of map scale are the size of the area to be covered, the information already available and the detail of new information required. Such factors can be considered by balancing the benefits and costs of the mapping exercise. Attempts to quantify the benefits, costs and cost-effectiveness of mapping at different scales have mainly been focused on soil surveys. In this paper, we review this literature on soils and extend the analyses to vegetation mapping.

The benefits and costs of mapping and reservation are not related to scale, per se, but to the intricacy of mapping (length of boundaries, extent of units) which different scales allow to be represented. There is a close relationship between map scale and the intricacy of the patterns produced (Bie and Beckett 1971) with maps at larger scale allowing the landscape to be subdivided more finely. It is intricacy which confers benefits and involves costs. Accordingly, we refer to scale in this paper as coarse (e.g. 1:1000000, 1:250000) and fine (e.g. 1:25000, 1:10000).

Benefits of Mapping

There are two major benefits for nature conservation which come from mapping natural resources: the subdivision of the landscape into units which are relatively homogeneous and, related to this, the role of these units as a basis for planning a representative reserve system.

Homogeneity in Soil Mapping

To be of value, a map of soils or any other natural resource should allow more precise statements to be made about natural features than could be made about the landscape at large (Beckett and Burrough 1971a). Mapping should describe the variation in the landscape and produce units which are more homogeneous, in terms of at least some attributes, than the area as a whole.

Measures of homogeneity of soil units as indices of the utility of soil maps have been proposed by Beckett (1967), Webster and Beckett (1968) and Beckett and Burrough (1971a). The latter authors derived an index called Relative Variance (RV) which measures the fraction of the total variation of any single soil attribute described by a mapping exercise: RV = (pooled within-unit variances) / (variance over whole survey area). Their measure of utility or homogeneity was (1-RV). A value of 1 indicates a perfect map in which all spatial variation of an attribute has been accounted for by the map units.

Beckett and Burrough (1971a) calculated (1-RV) for several attributes at several map scales in three trial areas. Their data provide some substantiation for Beckett’s (1971) idealised curves of homogeneity against map scale (Fig. 1). Beckett (1971) proposed that, for any soil mapping procedure except one involving intensive grid sampling of a single attribute, (1-RV) would reach a ceiling which could not be raised by mapping at finer scales. Two related factors have been proposed to account for this ceiling of homogeneity. First, mapping which relies on air-photo interpretation and general field observations reaches a scale beyond which some or most soil attributes have no external expression (Beckett 1968). Second, general purpose soil surveys account for most attributes indirectly, for example by mapping soil profile classes rather than isolines of each attribute. The variability of map units produced by these surveys cannot be reduced below that inherent in the profile classes (Beckett and Burrough 1971a).

Beckett et al. (1967) recognised that an overall assessment of the value of a soil map would be a function of the homogeneities of many attributes, together with weightings for their relative importance. Such a function was not, however, applied.

Homogeneity in Vegetation Mapping

The major purpose of vegetation mapping is to describe spatial variation in vegetation and, indirectly, that of the underlying physical and chemical factors. Homogeneity of map units is also an important consideration for vegetation surveys. Of particular interest is the homogeneity of units in terms of plant species which are entities of major concern in the conservation of vegetation. The floristic homogeneity of vegetation types can be assessed in two ways: by collecting and analysing floristic data from previously mapped vegetation units of different scales or by simulating varying map scales through classification of existing site data into varying numbers of groups. The latter approach is taken here.

Two data sets were used for the numerical classification analysis: O’Hares Creek (56 sites, 253 species) and Yengo National Park/Parr State Recreation Area (143 sites, 490 species). Alternative hierarchical site classifications were derived from both raw and standardised species data using two association measures (dissimilarities) — Kulczynski (1928) and Bray-Curtis (Bray and Curtis 1957) — and the Unweighted pair-Group Arithmetic Average (UPGMA) clustering strategy with flexible sorting (Belbin 1987). The trend in number of classification groups from one (all fused) to maximum (all sites separate) is generally analogous to a gradient from coarse to fine scales in vegetation mapping (Fig. 2). However, map units may have a component of spatial heterogeneity which is not present within site groups derived from a numerical classification. In that case the analogies drawn here between site groups and map units will tend to overstate the homogeneity of map units.

The relative homogeneity of vegetation types was measured at each level of the classification (i.e. for each number of site groups in the hierarchy). First, the average within-group association between sites was calculated using the dissimilarity measure. The ratio of this figure to the average pairwise association for the data set as a whole is a measure of heterogeneity analogous to the RV of Beckett and Burrough (1971a). Homogeneity is defined as (1-heterogeneity). The use of average pairwise comparison within a group as a measure of group homogeneity was discussed by Lance and Williams (1967) and attributed by them to Sokal and Michener(1958).

The curve for homogeneity using O’Hares Creek raw data and the Kulczynsi association measure with UPGMA clustering (Fig. 3) has three sections which reflect different trends in homogeneity as groups are progressively split. At relatively coarse scales (section 1), homogeneity rises rapidly as the main discontinuities in the data are delineated (or major vegetation types identified on a map, perhaps in relation to obvious geological, soil and topographic differences). In section 2, the gain in homogeneity is slower as further groups are formed (or the broad mapping units subdivided). Section 3 is an artifact of the analysis — homogeneity rises steeply when nearly all groups are single sites and so completely homogeneous — and is not analogous to vegetation mapping. Similar curves were obtained for the Yengo data set and for standardised data and classification with the Bray-Curtis association measure.

Figure 1. Idealised relationship between map scale and homogeneity of soil map units (modified from Beckett 1971). (1-RV) is an index of homogeneity derived from relative variance which measures the fraction of the total variation of any single soil attribute described by a map.

Figure 2. Dendrogram from classification of O’Hares Creek raw data using the Kulczynski measure of association and UPGMA clustering. The thick black lines indicate how varying numbers of groups can be defined to simulate varying map scales. (Data by courtesy of David Keith, NPWS).

Figure 3. Homogeneity analysis from O’Hares Creek raw data classified with the Kulczynski measure of association and UPGMA clustering, showing an initial rapid rise in homogeneity (section 1), a subsequent gradual rise (section 2) and a final rapid rise (section 3) which is an artefact of the analysis. (Data by courtesy of David Keith. NPWS).

Figure 4. Relationship between relative reservation status of map units and their scale (degree of subdivision) (Pearson’s correlation coefficient, r = –0.73, p < 0.001). Relative reservation status= % fine types reserved / % coarse types reserved. Degree of subdivision = 1- (no. coarse types / no. fine types). See Table 1 for raw data.

The curve in Figure 3 does not reach a ceiling as in Figure 1. Its tendency to rise continually indicates no limit to the benefit of mapping vegetation at progressively finer scales, even though the greatest rate of increase in homogeneity is at relatively coarse scales. For these data sets, the maximum benefit, in terms of homogeneity, is to be gained from a map drawn at the finest possible scale.

Mapping as a Basis for Representative Reserves Primary conservation goals are to establish reserves to protect the full range of natural environments and all native species. Map units are often used as a basis for both goals because of the difficulty of obtaining comprehensive data on species relative to that of producing resource maps. Land classification is therefore a central problem in assessing and achieving the representativeness of reserves (Austin and Margules 1986).

TABLE 1. Relationship between scale and reservation of classification or map units (the more numerous units in each group are finer geographic subdivisions)

The extent to which map units are represented in reserves is inversely related to their fineness of definition (Table 1, Figure 4). Representation is more comprehensive for coarse, heterogeneous units, at which scale much internal variation may be unrepresented, than for finer, more homogeneous units. Should the goal be to represent units at the finest end of the scale or is some intermediate scale adequate? One way of answering this question is to determine what percentages of species are reserved when reserves are located to contain parts of map units drawn at different scales.

At each level of classification of the two data sets used for the homogeneity analysis, one and five sites were selected randomly from each group, assumed reserved, and the total number of species in these nominal reserves recorded. Results for the Yengo data set, based on medians of ten random selections of the required sites from each group, are shown in Figure 5. The vertical distance between the curves indicates, predictably, that more species are reserved at any particular scale when more sites per group (greater areas of vegetation types) are reserved. Less obviously, reservation of very large numbers of groups is required to reserve all species: 90 groups in the case of five sites/group and 142 groups for one site/group. These group numbers correspond to a level of detail beyond that which would feasibly be produced in a vegetation map. The assumption that all species are adequately protected when examples of all map units are reserved is therefore unwarranted.

Figure 5. Percentage of species represented by reserving equal numbers of sites per group at different scales and different intensities. (From classification of Yengo raw datausing the Kulczynski measure of association and UPGMA clustering).

Figure 6. Percentages of species reserved in one site per group at different scales using classification to structure reservation (gross) and reserving sites arbitrarily (random). (From classification of Yengo raw data using the Kulczynski measure of association and UPGMA clustering).

The curves for species reserved against scale each include two components: the effect of increased reserve area, regardless of the classification (or map), and the effect of using the classification (or map) to locate the reserves. The two can be separated by plotting the ‘gross’ curve — species reserved due to both factors — with a ‘random’ curve — species reserved solely because of increasing reserve area (Fig. 6). The random curve was produced by randomly selecting sites from the whole data set to give a number of sites equivalent to the total reserved from site groups at each level of the classification. The curve shows median values of ten random selections of each number of sites. Random selection of sites from the whole data set (or locating reserves randomly, regardless of map units) is clearly the most important component of the gross curve, although selection from groups (or locating reserves to sample each map unit) generally provides a net benefit.

Costs

An assessment of cost-effectiveness requires that the benefits to be gained from mapping or classification of data into environmental types must be weighed against costs. In this section, two types of costs are considered: costs associated with the mapping or survey exercise and costs associated with the area required to reserve map units.

Costs of Survey and Mapping

The relationships between the costs of mapping exercises and their scales have been demonstrated in several analyses of soil and land use surveys. The results of these studies are assumed here to be indicative of those associated with vegetation mapping. Bie and Beckett (1971) produced a comprehensive comparison of survey effort (person days) per unit area in relation to map scale for 66 Australian soil and land use surveys. They found an exponential increase in effort (closely related to total costs) as map scales became finer. Analyses of subsequent Australian surveys show the same trend (Hallsworth 1978) while other studies of overseas surveys also support these findings (e.g. Veenenbos 1957, Bie and Beckett 1970, Burrough and Beckett 1971, Western 1978).

Personnel costs are generally the major component of a survey for a soil map: up to around 75% of the total survey and mapping budget (Bie and Beckett 1970, Dent and Young 1981). The density of ground observations tends to increase with the number and intricacy of boundaries to be surveyed (Bie and Beckett 1970). Together with the time required to analyse, compile and interpret the observations and samples, this increased field work is a major reason for the higher cost of fine-scale mapping.

Costs of Reserving Map Units

The inverse relationship between reservation status and fineness of map units has an important implication for the area, and hence monetary cost, of reserves. The comparisons between scales in Table 1 and Figure 4 are based on equal reserve areas. It follows that, for individual reserves of the same size, a much larger total reserve area is needed to represent all the fine units than all the coarse units in any region. This trend is demonstrated by the number (and area) of sites needed to represent increasingly larger numbers of site groups in nominal reserves (Fig. 7). Reserving one site per group (small parts of each map unit) is a less costly way of representing the groups (and the map units) than five sites per group (large parts of each map unit) but introduces problems of lower long-term viability of populations in small reserves. Figure 7 also represents the area costs of achieving the levels of species conservation shown in Figure 5.

Cost-effectiveness

Cost-effectiveness of surveys for mapping Beckett and Burrough (1971b) defined the cost-effectiveness of soil mapping in terms of the ratio of benefit/cost, i.e. (l-RV)/(cost per unit area at different map scales). They found that this ratio generally decreased at finer scales for most soil attributes involved in their analysis. Beckett (1971) proposed an idealised curve for the cost-effectiveness of a soil survey (other than that by grid sampling for single attributes) which rises to a peak and then declines (Fig. 8A). This curve is partly supported by their empirical results and rests on the assumption of an initial increase, at coarse scales, in benefit (1-RV) which is greater than the increase in costs. As the rise in (1-RV) slows and then levels out, with costs continuing to rise, cost-effectiveness peaks and declines. If this assumption applies, such cost-effectiveness relationships with scale might be derived from the homogeneity of vegetation types (Fig. 3) and the percentages of species reserved (Fig. 5). However, other trends in cost-effectiveness are possible. Benefits could rise at the same rate as costs at coarse scales before being overtaken by the sharply rising cost curve (Fig. 8B) or the rate of increase in benefits could be lower than that of costs at all scales (Fig. 8C).

Figure 7. Area costs of reserving equal numbers of sites per group at different scales and different intensities. (From classification of Yengo raw data using the Kulczynski measure of association and UPGMA clustering).

Whichever situation applies, but particularly for the situation in Figure 8C, cost-effectiveness is likely to decline at a scale beyond which there are still benefits to be gained for homogeneity and species conservation. The sacrifice of some additional homogeneity of map units in the interests of cost-effectiveness is an accepted procedure. A less obvious implication of mapping at a cost-effective scale is that it is probably too coarse to enable a reserve system to be structured to contain all the species in a region. This has important implications for the use of maps in conservation planning and for the types of complementary information necessary.

Cost-effectiveness of Reservation

The efficiency of reservation of map units at any given scale, in terms of the total area required to sample them, depends on the procedure for reserve selection (Pressey and Nicholls 1989). For any one procedure, all the broad units can be sampled in a smaller total area than all the fine units (Fig. 7). An important criterion in deciding on an appropriate map scale on which to base reserve selection is the percentage of species which could be reserved as a result.

Figure 8. Possible relationships between cost-effectiveness of mapping and map scale. A: idealised relationship modified from Beckett (1971); B,C: alternative relationships.

Figure 9. Cost-effectiveness of reservation (species reserved per unit area) at different scales and different intensities. (From classification of Yengo raw data using the Kulczynski measure of association and UPGMA clustering).

Using this criterion, the cost-effectiveness of reservation can be calculated by plotting species reserved per unit area against scale. The curves obtained for the simulated reservation of Yengo vegetation types (Fig. 9) follow the trend of Figure 8C, i.e. the benefit of number of species reserved always rises more slowly than the cost of reserving them. Cost-effectiveness is maximal at the coarsest scale. With the addition of more reserves, more species are reserved but at a decreasing rate. For most scales, the cost-effectiveness of reserving five sites per group (large areas