Representativeness and Efficiency of Bird and Insect Conservation in Norwegian Boreal Forest Reserves JOGEIR N. STOKLAND University of Oslo, Department of Biology, Division of Zoology, P.O. Box 1050–Blindern, N–0316 Oslo, Norway, email
[email protected]
Abstract: I quantified local species richness of birds in different forest types and of beetles in spruce forests at different altitudes. In both cases I quantified timber production as a measure of land acquisition cost and used the ratio between the species richness and timber production as a measure of conservation cost-efficiency. I found a positive correlation between timber production and local species richness of birds as well as beetles, indicating that the forests most valuable for forestry are also the ones most valuable for biodiversity conservation. I used different selection procedures for combining sites in a reserve network to find the minimum set of sites that included all vulnerable species. The minimum set of sites for birds was 30% spruce forest, 30% pine forest, and 40% broad-leaved forest (the three main forest types). The minimum set of sites for the beetles was uniformly distributed along the altitudinal gradient. Both minimum sets were most cost-efficient for species conservation. I suggest that equal coverage of different productivity classes is more efficient for optimizing biodiversity conservation than over-representing low productivity sites. Less than 1% of Norwegian boreal forests have been protected as nature reserves. The reserve network is fairly representative with respect to altitude, but it is seriously skewed toward low productivity sites. The current network is suboptimal with respect to forest type representativeness, species protection, and cost-efficiency. This is a result of an inefficient strategy of selecting reserve sites and an unfortunate combination of selection criteria. Representatividad y Eficiencia en la Conservación de Aves e Insectos en las Reservas de Bosque Boreal de Noruega Resumen: Se cuantificó la riqueza local de especies de aves en diferentes tipos de bosques y la riqueza de coleópteros en bosques de coníferas a diferentes altitudes. En ambos casos, se cuantificó la producción maderera como una medidad del costo de adquisición de la tierra, así mismo se utilizó la relación entre la riqueza de las especies y y la producción maderera como una estimación de conservación costo-eficiencia. Se encontró una correlación positiva entre la producción maderera y la riqueza local de especies de aves y coleópteros, indicando que los bosques más valiosos para actividades forestales, lo son también para conservación de la biodiversidad. Se utilizaron diferentes procedimientos de selección para combinar sitios en una red de reservas para encontrar el mínimo juego de sitios que incluyeraon especies vulnerables. El mínimo juego de sitios para aves fue 30% de bosque de Picea, 30% de Pinar y 40% de bosque caducifolio (los tres tipos de bosques mas importantes). El mínimo juego de sitios para coleópteros se encontró uniformemente distribuído a lo largo de un gradiente altitudinal. Se sugiere que una cobertura igual de las diferentes clases productivas es mas eficiente en la optimización de la conservación de la biodiversidad que cuando se sobre-representan sitios de baja productividad. Menos del 1% de los bosques boreales de Noruega han sido protegidos como reservas de la naturaleza. La red de reservas es razonablemente representativa con respecto a la altitud, pero se encuentra seriamente sesgada hacia sitios de baja productividad. La actual red no es óptima con respecto a tipos de bosque representados, protección de especies y costo-eficiencia. Esto es resultado de una ineficiente estrategia en la selección de sitios de reserva y a una desafortunada combinación de criterios de selección.
Paper submitted March 8, 1995; revised manuscript accepted December 19, 1995.
101 Conservation Biology, Pages 101–111 Volume 11, No. 1, February 1997
102
Efficiency of Bird and Insect Conservation
Introduction Creation of nature reserves is an important measure for stemming the accelerating extinction of species and degradation of natural habitats. Assessing effectiveness of alternative reserve networks has recently become an area of active research, and key topics include selecting complementary sites, identifying minimum sets, and designing spatial configurations for efficient species conservation (Bedward et al. 1992; Pressey et al. 1993). The cost of reserves for alternative land uses, which I consider here, has been addressed several times but has been only superficially investigated beyond considering cost as proportional to reserve area. The challenge of identifying a minimum set of sites representing maximum biological diversity was overshadowed during the SLOSS (single large or several small reserves) debate, which focused on extreme combinations of refuge numbers and areas in nature reserve design (e.g., Diamond 1975; Simberloff & Abele 1976; Whitcomb et al. 1976; Margules et al. 1982). Simberloff and Abele (1982) concluded that the theory of island biogeography gives no clear recommendation with respect to few large or many small refuges. Recently, Lomolino (1994) summarized that most empirical studies tend to show that higher species richness occurs in accumulations of small reserves. Lomolino further showed that accumulation from small to large refuges, as well as from large to small refuges, was inferior in maximizing species richness compared to the best combination of reserves found by an exhaustive sampling of several sets of candidate reserves. An alternative approach to reserve network design emphasizes the need to cover the maximum variation of natural features prior to considering reserve size and numbers (Kirkpatrick 1983; Margules et al. 1988; Pressey & Nicholls 1989a). This approach has developed new selection methods that put together highly complementary sites that represent the feature variation in a minimum or near minimum set of sites (Pressey et al. 1993). In empirical studies natural features have been land type attributes and plant communities (Purdie & Blick 1986; Pressey & Nicholls 1989b), species lists (Kirkpatrick 1983; Margules & Stein 1989; Rebelo & Siegfried 1992; Sætersdal et al. 1993), or a combination of the two (Margules et al. 1988). This approach has shown to be far more efficient in selecting representative reserve networks than traditional ranking procedures, which tend to duplicate similar sites (Pressey & Nicholls 1989a; Sætersdal et al. 1993). The motivation for identifying minimum networks that maximize biological diversity is the political or economic (often ill-defined) limits in setting aside areas for conservation purposes. Bedward et al. (1992) argued that the cost of networks should preferably reflect opportunity costs for alternative land use, but they were
Conservation Biology Volume 11, No. 1, February 1997
Stokland
unable to explore this beyond network area due to inaccessible data. The opportunity cost is particularly relevant in forest ecosystems because timber production varies greatly from less productive to highly productive sites. In Scandinavia timber production varies by a factor of 10, and productivity also imposes marked patterns in diversity of most species groups. This has important implications for strategies aimed at providing a balance between logging and conservation. A Norwegian forest reserve plan has currently been launched in which sites have been selected on the basis of a ranking procedure. I assess alternatives using updated selection procedures, and I quantify the efficiency of alternative networks by using the ratio between cumulated species richness and total potential timber production in the networks.
Methods Study Area and Forest Conservation The study was conducted in southeastern Norway, a region of 40,500 km2 and 60% of the productive forests in Norway (Tomter 1993). Most of the area is boreal forest, but hemiboreal forests comprise a transition zone to a small amount of nemoral forest in the south. The coniferous tree line in the region varies between 900 and 950 m, above which there is a 50 to 100–m belt of sub-alpine birch forest. The region is dominated by low and medium productive forests (Fig. 1b). About 50% of the productive forest area is dominated by spruce, 40% by pine, and slightly more than 10% by deciduous (hardwood) forest. The proportion of deciduous forest decreases to 5.4% in mature forests (logging class V; Tomter 1993). Boreal forest productivity is primarily determined by the length of the growing season (closely correlating with altitude and latitude) and the soil supply of water and nutrition (both reflected by the vegetation). Because the latitudinal scope of this study was quite narrow, forest produc-
Figure 1. Altitudinal distribution of the total forest area and the forest reserve network in southeastern Norway (a), and the proportion of total forest area and the forest reserve network in southeastern Norway in three forest productivity classes (b).
Stokland
Figure 2. Timber production in relation to altitude and forest understory vegetation in southeastern Norway. The dark-shaded vegetation type was used in the elevation study of beetle species. Blank cells indicate missing data (Source: Anonymous 1981).
tivity is primarily a result of altitude and vegetation type (Fig. 2). Clear-felling has commonly been the adopted logging method in Norway for the past 50 years. The logging history is, however, several hundred years old, and today only 0.6% of the productive forest area is older than 160 years (Tomter 1993). During 1990–1994 the Norwegian reserve network was extended to cover 0.9% of the productive forest area, and it comprises 152 km2 boreal forest and 3–4 km2 nemoral forest in southeastern Norway (Haugen 1991; I. Haugen personal communication). The network is fairly representative with respect to altitude, although there is a slight under-representation of areas below 200 m and a slight over-representation above 500 m (Fig. 1a). There is, however, a serious skew toward reserves of low productivity and a corresponding underrepresentation of highly productive areas (Fig. 1b). Data on tree composition are currently available for about half of the reserve area. This information indicates that spruce forests are somewhat over-represented and pine forests somewhat under-represented, whereas the proportion of deciduous forest is about 5% in the reserves.
Species and Forest Data Data on birds and beetles were collected from sites in the boreal and hemiboreal part of the region. Birds were censused at 40 sites stratified in different types of mature forest stands using 36–ha plots. The sites, purposely selected to cover all tree species compositions, were distributed between 40 to 600 m above sea level. Because the tree composition varied internally in some census
Efficiency of Bird and Insect Conservation
103
sites, I divided each census plot into subplots of 4 ha to achieve more homogenous units when I related local species richness to tree composition and altitude. In each subplot I recorded the vegetation type and the soil depth at eight different points. Altogether, I encountered 56 species or 84% of all boreal forest birds in southeastern Norway (omitting nocturnal species). I defined a subset of 32 species, which were more or less confined to mature forest, as potentially vulnerable to forestry (hereafter referred to as vulnerable species). This set should not be confused with red-listed species (although a few of the species are red-listed) because it includes all woodpeckers and tits, of which several are quite common. These species have experienced extensive habitat loss due to forestry, and population decline is documented for several of these species in Finland (Järvinen & Väisänen 1977; Väisänen et al. 1986). I collected beetles using 10 window traps in 0.5–ha plots at 17 different sites in mature or old-growth myrtillus type spruce forest at elevations from 150 to 870 m (i.e., all forested altitudes). A total of 309 species was collected, which is 10%–20% of the forest dwelling beetle species in Norway. I defined a subset of 91 species as potentially vulnerable to forestry (referred to as vulnerable species). These species need medium to very decayed wood for fulfilling their life-cycle. Like the vulnerable bird species, several beetle species are quite common. The forestry is nevertheless removing a substantial part of their habitat when trees are logged and taken away instead of being allowed to go through the decay process in the forest. I calculated forest productivity in the bird census plots and at the beetle trapping sites from altitude, vegetation type, and soil depth using regression equations developed by Nilsen and Larsson (1992). Annual timber production was further calculated from Tveite and Braastad (1981) who showed the empirical relationship between productivity and timber production when optimal logging is adopted. For birds I calculated the timber production for each census plot, and for beetles I calculated the timber production per hectare at the sample site.
Reserve Selection Methods Research on reserve selection has developed several methods for maximizing the natural variation in reserve networks. The ultimate purpose of such methods is to assist applied conservation work, and they should be tested on several data sets to evaluate their efficiency. I have tested the performance of three methods: • Iterative selection is a stepwise procedure in which all unselected sites are examined and the one which adds most new features is chosen each time a new site is included in the network (Kirkpatrick 1983; Margules et al. 1988). Note that many site combinations remain
Conservation Biology Volume 11, No. 1, February 1997
104
Efficiency of Bird and Insect Conservation
unexamined when using this method because selected sites are not later dropped out in favor of others. There are several variations of the iterative selection method (see Pressey et al. 1993 for a review), and the features used as criteria may be land type attributes (including plant communities) as well as species lists. Thus, the approach strongly emphasizes complimentarity among the selected sites. I used the number of (1) all new species and (2) new vulnerable species as reserve selection criteria. In cases of tied alternatives at a step (i.e., several sites with the same number of new species), the site with the highest species richness was selected. • Constrained selection is an approach that maximizes feature variation among the selected sites under the constraint that total area (or acquisition cost) should not exceed an upper limit. In addition, all site combinations are examined, ensuring that the true optimal is found. This approach was first introduced in reserve selection by Cocks and Baird (1989) who adopted the technique of integer programming. I used an analogous algorithm that runs through all possible combinations of sites and picks out the most species rich combination of sites in sets from one to n sites. Each run was subjected to the constraint that total timber production (reflecting acquisition cost) should not exceed a user-specified upper level. Thus, a series of several runs produced an array of maximum cumulated species richness attainable at each constraint level. • Ecological coverage selection is a novel approach that facilitates site selection based on site features in order to cover or represent an ecological gradient (e.g., altitude) in a pre-determined manner. This enables comparisons of one gradient coverage with alternative, presumably better, coverages. A target attribute determines the position of candidate sites along the gradient, and the sites are split in two groups with values of the attribute above and below a chosen target value. A set of n sites is established by random selection of sites from the two groups such that the resulting average of the target attribute ends up near the chosen target value. The cumulated species richness is calculated for 100 replicated sets and averaged over these replicates. This process is carried out for different set sizes. The procedure also allows selection from one of the two groups. As the target attributes in this study, I chose the percentage of deciduous trees for birds and altitude for the beetles. I assessed the performance of the three selection methods by using a simple efficiency index. The implementation cost of a reserve network, in terms of acquisition cost or lost logging opportunity (income), is proportional to the accumulated timber production for the selected set of sites:
Conservation Biology Volume 11, No. 1, February 1997
Stokland
P 5 ∑ area (i) 3 production (i), where i denotes the sites. Referring to R as the cumulated species richness in a selected set, I used R/P or “richness-to-production ratio” as an index to measure the conservation cost-efficiency of single sites as well as for sets of several sites. Throughout, I have used the term cumulated species richness for overall species richness in a reserve network comprising several sites and local species richness for species richness of the individual sites.
Results Forest Productivity and Local Species Richness The local bird species richness increased significantly with higher proportion of deciduous trees (Fig. 3a, pdecid. , 0.001, two-way ANOVA). In deciduous forests the species richness decreased with increasing elevation, whereas coniferous forests (less than 40% deciduous trees) had similar species richnesses at different altitudes (Fig. 3a, p , 0.01, interaction term in two-way ANOVA). There is a systematic relationship between ground vegetation and tree composition, and high proportions of deciduous trees also indicate more productive sites. The local species richness to timber production ratio was fairly similar for most forest types and altitudes, but it increased at higher elevations in coniferous forests (Fig. 3c, pdecid. , 0.001, pelev. , 0.05, two-way ANOVA). The ratio increased be-
Figure 3. Species richness of all bird species (a) and vulnerable bird species (b) in subplots of 4 ha in forests with different proportions of deciduous trees (0–100%) in three altitude intervals (0–200 m, 201–400 m, 401–600 m) (error bars indicate SD), and the corresponding ratios between species richness and timber production (c) and (d) (numbers above columns indicate number of subplots).
Stokland
Efficiency of Bird and Insect Conservation
105
Figure 5. The cumulated number of bird species in 14 out of 40 sites chosen with the iterative selection method (a), and the running average in tree composition of these sites (c). The cumulated number of vulnerable bird species in 11 out of 40 sites chosen with the iterative selection method (b), and the corresponding tree composition of these sites (d).
tio remained fairly stable with increasing altitude (Fig. 4b, NS, ANOVA). Maximizing Species Richness in Multiple Sites Figure 4. Local species richness of all beetle species and vulnerable beetle species in spruce forest in three altitude intervals (a) (error bars indicate SD), and the corresponding ratios between species richness and timber production (b) (sites are identical to those in a).
cause species richness did not change with altitude, whereas the timber production decreased. In deciduous forests the local species richness and timber production decreased more or less in parallel with increasing altitude and the ratio remained stable. This general pattern was also found among the vulnerable birds (Fig. 3d, pdecid. , 0.01, pelev. 5 0.07, two-way ANOVA). For beetles the local species richness was highest at low and intermediate altitudes (Fig. 4a, p , 0.05, ANOVA). This tendency was even more pronounced for the vulnerable species (Fig. 4a, p , 0.001, ANOVA). Although the species richness decreased with increasing altitude, the richness-to-production ratio seemed to increase (Fig. 4b, p 5 0.07, ANOVA) because timber production decreased more rapidly. The species richness of the vulnerable species decreased more or less in parallel with timber production, and their richness-to-production ra-
There are clear patterns in local species richness and species turnover along forest gradients. I used iterative selection to find the combination of forest habitats that maximized the cumulated species richness in a minimum set of sites. This was done for all species including vulnerable species. When I selected sites to cumulate the total bird species richness in a minimum set, all species were encountered in 14 out of 40 sites (Fig. 5a). The tree composition in the selected sites rapidly stabilized to around 35% spruce and deciduous trees and about 30% pine (Fig. 5c). There was no altitudinal skew among the selected sites. The average timber production in the selected sites was 141 m3/year compared to 110 m3/year in the remaining 26 sites, but the difference was not statistically significant (Mann-Whitney U-test). All vulnerable bird species were encountered in a minimum set of 11 sites (Fig. 5b). In these the tree composition was quite similar to that which maximized total species richness, but the proportion of deciduous trees was perhaps slightly higher (Fig. 5d). The average production in selected and nonselected sites was 4.1 and 3.1 m3/ha/year, respectively ( p , 0.1, Mann-Whitney U-test). The accumulation of total species richness was almost as good as when the sites were selected on the basis of vulnerable species. From the sixth site the selection based on vulnerable species
Conservation Biology Volume 11, No. 1, February 1997
106
Efficiency of Bird and Insect Conservation
Figure 6. The cumulated number of all beetle species at 17 sites chosen using the iterative selection method (a), and the corresponding altitudes of the selected sites and running average of the altitudes (c). The cumulated number of vulnerable beetle species in 11 out of 17 sites chosen with iterative selection (b), and the corresponding altitudes of the sites (d).
performed exactly as well for total species richness as when total species richness itself was maximized. The reverse was also true. All vulnerable species were encountered from the thirteenth site when the sites were selected on the basis of all new species. All beetle sample sites had unique species and were consequently necessary in order to include all species, but more than 90% of the species were encountered in the first 10 sites (Fig. 6a). When vulnerable species was used as the selection criterion, all these species were encountered in 11 sites (Fig. 6b). Sites at low altitudes were on average selected before those at high altitudes, which was reflected in the increasing running average ending at 561 m (Fig. 6d), and the elevation of all unselected sites was above the average of the selected sites. The average production of all sites was 2.5 m3/ha/year. The sites that included all vulnerable species had an average timber production of 2.7 m3/ha/year, whereas sites that included 90% of the vulnerable species had an average production of 3.1 m3/ha/year. Neither of these sets differed significantly from the remaining sites with respect to timber production. The selection sequence was characterized by an alternating between low and high altitudes, which adds most new species to those already selected. The accumulation of total species richness was nearly as good when sites were selected on the basis of vulnerable species, and vice versa. Efficiency of Different Strategies The efficiency of various selection strategies becomes comparable when the solutions are related to a common production scale. For the iterative selections above and
Conservation Biology Volume 11, No. 1, February 1997
Stokland
Figure 7. The cumulated richness of vulnerable bird species (a) and vulnerable beetle species (b) as functions of cumulated timber production of the selected sites. The upper (bold) curve is the result of constrained sampling, the next uppermost curve is the result of iterative selection. The three lower curves are results from different ecological coverages, which represent sites with averages of 10%, 15%, and 30% deciduous forest (a) and sites with average altitudes of 500, 700, and . 600 m (b).
the two other selection strategies, I related the cumulated species richness at each selection step to the cumulated timber production of the sites. Only the selection efficiency for vulnerable species was evaluated in detail. The optimal solution, in the sense of maximum efficiency, was achieved by constrained selection. This is shown Fig. 7, where the constrained selection exhibits the upper bound of species richness at each production level. Iterative selection was somewhat less efficient for birds. All vulnerable bird species were found by iterative selection and by constrained sampling in sets of 11 sites (7 sites were identical in the two solutions), but the cumulative timber production of the iterative solution was 13% higher (1630 m3 versus 1435 m3, Fig. 7a). The tree composition in the minimum set using constrained selection (proportions of pine, spruce, and deciduous trees were 27%, 32%, and 40%, respectively) was almost identical to that found with iterative selection (Fig. 5d). For the vulnerable beetle species the iterative selection gave exactly the same minimum set as the constrained selection (Fig. 7b), which, coincidentally, also comprised 11 sites. Iterative selection closely approached and matched the maximum efficiency twice along the selection sequence. In many real situations reserve selection is based upon the ecological attributes of sites because no information is available about species composition. I evaluated the performance of selecting sites based on tree composition when considering birds, and based on altitude for beetles. For the vulnerable birds the ecological coverage selection was substantially less efficient than either iterative or constrained selection, which were based on spe-
Stokland
cies composition (Fig. 7a). When I reduced the average deciduous trees percentage from 30% to 10%, the efficiency decreased markedly (Fig. 7a), despite the fact that the average production of the selected sites decreased as well. For the vulnerable beetles the ecological coverage selection method was also less efficient compared to the other selection strategies. Because there were few sample sites, it was impossible to explore significantly different alternatives. There was little effect produced by changing the average altitude from 500 to 700 m, but when only sites above 600 m were selected, the efficiency dropped markedly (Fig. 7b). I also calculated the efficiency of different ecological coverages for all beetle species. It then turned out that the efficiencies of the three alternatives were very similar, including the alternative of selecting sites above 600 m only.
Discussion The positive relationship between timber production and local species richness of birds and beetles exists because of their common link to forest productivity. In more detailed analyses I have shown that bird species richness is related to productivity through higher proportions of deciduous trees, but it is lower at high-productive sites when coniferous trees dominate (Stokland 1994a). The local species richness of vulnerable beetles seems more directly related to productivity through the abundance of decaying wood (Stokland 1994b). Thus, the forests most valuable for forestry are also the ones most valuable for biodiversity conservation. Research on systematic reserve selection has developed guidelines and well-founded principles for putting together reserve networks (Pressey & Nicholls 1989a; Pressey et al. 1993). This research has not had any influence upon the present extension of the Norwegian forest reserve network because the research field is unknown to decision makers and many biologists. Besides implications for the Norwegian reserve network, this study also has some points of more general interest. Reserve Selection Methods Pressey and Nicholls (1989a) showed that iterative selection is consistently more efficient than the ranking approach in maximizing representativeness in a minimum reserve network. The iterative approach, which is used in this study, seems to have obtained status as the best method for reserve selection. Recently, however, Underhill (1994) raised a valid point that this technique may fail to find the true optimal combination of sites because selected sites are not allowed to be dropped out later in the selection process. Instead, he advocated the use of linear programming, which does find the true optimum.
Efficiency of Bird and Insect Conservation
107
Sætersdal et al. (1993) tested linear programming as well as iterative selection for selecting nemoral forest sites based on species lists of vascular plants and birds, and they found identical solutions with both methods. In this study iterative selection and constrained sampling (analog to linear programming) produced identical minimum sets for the vulnerable beetles. The minimum bird set from iterative selection comprised the same number of sites as the true optimum (7 out of 11 sites were identical), but the timber production cost was 13% higher in the iterative solution. Although iterative selection and constrained sampling arrived at very similar minimum sets, they selected the sites in different sequences. Iterative selection tended to choose species rich sites first, whereas constrained selection favored low-productive sites first. Iterative selection and linear programming produce similar results because all sites with unique species must be in the minimum set; therefore, both methods seem to be good alternatives with particular strengths. In iterative selection one can easily add spatial design criteria (Bedward et al. 1992; Nicholls & Margules 1993), whereas linear programming guarantees that the optimal solution is found. In real conservation work it is often more important to explore near-minimum alternatives than it is to identify the numerical optimum. This is due to the need for compromise alternatives in conflict situations and the need for improvements in relation to spatial design criteria. Furthermore, the computing time for large data sets can be very long for optimizing algorithms, in the order of hours or days compared to seconds or minutes for iterative approaches. This is still an important argument for using iterative methods in realworld analyses. The conclusion is that both iterative selection methods and linear programming (or a similar optimizing method) have important roles to play in reserve network design. Spatial and Ecological Coverage Most studies have focused on the area needed in the minimum set. This spatial coverage varies from 6% to 75% when using iterative selection (Margules et al. 1988; Pressey & Nicholls 1989b; Rebelo & Siegfried 1992; Sætersdal et al. 1993). None of these results are comparable, however, because the spatial scales varied greatly and investigation efforts were different. Rebelo and Siegfried (1992) found that 6% of the total area was needed for representing all Proteaceae species at least once in the Cape floristic region, and 17%–24% was needed to protect all species at least 3–5 times. Pressey and Nicholls (1989b) found that 6% was needed in a minimum set to represent all land systems in a comparable large region. Margules et al. (1988) and Sætersdal et al. (1993) studied more restricted areas and concluded that 45%–75% was needed to represent all species at least once. Margules et
Conservation Biology Volume 11, No. 1, February 1997
108
Efficiency of Bird and Insect Conservation
al. (1988) argued that commonly cited minimums of 5%– 10% of the total land area would be inadequate in their case, but they did not discuss whether their reference area was perhaps too small. The sufficiency of 5%–10% is indeed a question of scale, and it is a challenge to suggest a biologically sound scale at which the minimum area for protection of all species should be calculated. Although it is important to determine the spatial coverage, it is equally important to determine the ecological coverage of the minimum set, i.e., how these sites are distributed relative to ecological gradient(s). I found that the minimum set of sites for the birds comprised 35%– 40% deciduous forest, 30%–35% spruce forest and 30% pine forest (Fig. 5). These figures, also reached by constrained sampling, are quite different from the 5% mature deciduous forest in the forest landscape. I found the same tendency for the vulnerable beetle species along the altitudinal gradient. The minimum set was evenly distributed along the gradient (Fig. 6d), but in this case the number of candidate sites was quite small (especially at low altitudes). I hypothesize that sites in minimum sets are very often uniformly distributed along ecological gradients, probably with a slightly higher frequency in the most species-rich segment(s), irrespective of the frequency of the segments in the landscape. This should, however, be tested carefully for various organisms and ecosystems in order to constitute a firm basis for recommendations. Uniform ecological coverage is not expected when species defining the minimum set exhibit a nested drop-out type of beta diversity along ecological gradients (sensu Harrison et al. 1992) instead of true species turnover.
Concepts of Representation Austin and Margules (1986) made a distinction between environmental and biological attributes when assessing representativeness. This distinction is also present in reserve selection studies because environmental attributes (e.g., climate, landform, soil and vegetation classes) are typically used in poorly known districts, whereas species lists are used in better known areas. Although there are obvious relationships between the two attribute classes, it is possible to select sites that represent the variation in environmental attributes without representing all species even approximately (Pressey & Bedward 1991), and also the reverse is possible (Margules et al. 1988). When species is the selection criterion, it is relevant to consider whether some species should have priority in the selection process. Margules et al. (1988) argued that “unique” species (occurring only once) should have first priority. If unique means that these species are biogeographically marginal to the region, Bedward et al. (1992) suggested that such species should not influence the location of reserves, but instead be disregarded in
Conservation Biology Volume 11, No. 1, February 1997
Stokland
the selection process. In ecosystems where natural habitats are modified to a large extent, it is important to check whether species that are most sensitive to modifications have typical habitat preferences or deviate from the total species pool. To account for this I used a vulnerability criterion to favor such species in an alternative network selection. I found that the vulnerable species were very well represented when maximizing total species richness and vice versa. Järvinen (1982) found a similar strong relationship between number of endangered and all vascular plants in a single site comparison, as did Rebelo and Siegfried (1992) for rare and all plant species in a network comparison. One reason for these relationships is the obvious lack of independence between nested data sets. More fundamentally, different subsets of one organism group probably respond quite similarly to environmental variation between sites. Different organism groups (i.e., vascular plants, birds, invertebrates), on the other hand, tend to show quite different patterns in local species richness and occurrence of rare species across the same sites (Emberson 1985; Pendergast et al. 1993; Sætersdal et al. 1993). Thus, it is more important to explore taxonomically different organism groups along several environmental gradients, rather than seeking confidence for a suggested reserve network by comparing various subsets of one organism group. Implementation Costs Several authors have stressed the importance of implementation costs in reserve selection strategies (Cocks & Baird 1989; Bedward et al. 1992), but no studies have included real costs beyond network area, which is a good first approximation. In forest ecosystems timber production increases substantially from low to high productive sites, and the richness-to-production ratio is a relevant cost-efficiency index for evaluating alternative selection strategies. The strong positive relation between vulnerable species richness (both birds and beetles) and timber production yielded fairly similar richness-production ratios in most forest types (Figs. 3 and 4). At the lowest productivity (low proportion of deciduous trees, high altitude), however, the richness-to-production ratio for vulnerable birds was about 1.5 times higher. For all species (birds and beetles) the ratio was about two times higher at low compared to high productive sites. The invariably high ratios when productivity was at the minimum primarily reflects that low-productive sites are of low value to forestry (because timber production approaches zero), whereas medium and highly productive sites seem relatively equally valuable to conservation and forestry. Nevertheless, this means that, in single site evaluation, the most efficient conservation (defined by the richness-toproduction ratio) seems to be at low productive sites. In multiple site selection it soon became a bad strat-
Stokland
egy to accumulate low-productive coniferous sites for the vulnerable bird species because species-poor sites then are duplicated while implementation costs increase (Fig. 7a). The efficiency of ecological coverage selection improved when the proportion of deciduous trees increased from 10% to 30%, but the complementary solutions from constrained and iterative selection were far more efficient. For vulnerable beetle species there was a very similar pattern. Complementary solutions were most efficient, characterized by a uniform ecological coverage of the ecological (i.e., altitude) gradient. Unfortunately, the beetle data was not sufficient to explore the full variation of various altitude coverages. However, when I sampled sites above 600 m only, the efficiency for the vulnerable species decreased (Fig. 7b). The efficiency of selecting sites above 600 m for all beetle species, on the other hand, was similar to those of sites with averages of 500–700 m. The general conclusion is that complementary reserve networks, which include several species-rich, high productive sites, seem to be the most cost-efficient strategy when timber costs are considered. This is analogous to the basic finding when ranking and iterative selection methods are compared. Evaluation based on static criteria (here, single site cost-efficiency) is useful in recommending single sites, but static criteria should not be applied in multiple site selection because the value of an additional site depends on the attributes of the sites already selected. It is premature to generalize that a substantial proportion of high-productive forest, together with medium and low-productive sites, is the most efficient conservation strategy for all organisms. The declining species richness of many organism groups with altitude (Hågvar 1976; Solhøy 1976) and latitude (Lahti et al. 1988; Väisänen & Heliövaara 1994; but see Kouki et al. 1994) indicate that this generalization may be valid, but a positive relationship between species richness and productivity is not sufficient to make this conclusion (cfr. the efficiency result of all beetle species). It seems more appropriate to conclude that a similar coverage of productivity classes is at least as efficient and probably more efficient than over-representing low-productive sites for maximizing representativeness of most organism groups. The efficiency results of this study also have some implications beyond reserve selection strategies. “New” or “multiple use” forestry has broadened the scope of forest management research to include biodiversity as new decision criterion (Kuusipalo & Kangas 1994). This development is welcomed because it indicates biodiversity effects of various forestry practices. This study shows that the efficiency of conservation measures (such as non-logging in small key biotopes) varies with site properties and forest productivity. Another point is that the biodiversity effect of a single management plan is scaledependent. This is analogous to the single site-multiple
Efficiency of Bird and Insect Conservation
109
site differences: a management plan yielding a reasonable local balance between timber harvest and species richness may produce undesired effects if adopted throughout an extended region. This regional biodiversity effect is far more important than the local effect. The Norwegian Forest Reserve Network The reserve network in Norwegian boreal forests comprises 0.9% of the productive forest area including the recent extension. Despite a scientific recommendation to protect at least 2% of the productive forest (Baadsvik 1988), the current spatial coverage is considered politically “reasonable.” In addition to the spatial deficiency, the reserve network also has some serious weaknesses in ecological coverage because low-productive sites have been over-represented. This is a result of an unfortunate combination of ranking criteria, an inefficient selection strategy, and a lack of relevant biological information about the candidate sites. The criteria by which all sites were ranked were biogeographical representativeness, area, naturalness, and species richness of vascular plants (Haugen 1991). Particularly large and minimally logged areas achieved a high rank as so-called “type areas.” Together, these independently good criteria constituted a poor combination, because low-productive sites that have been moderately utilized by forestry received the highest scores. Subsequently, the inefficient strategy of only selecting top-ranked sites (Pressey & Nicholls 1989a) was applied, and similar sites were repeatedly picked out in different biogeographical regions. This serious bias was somewhat counteracted by selecting assumed species-rich or vulnerable “special areas,” but so-called supplementary areas intended to balance the ecological coverage of different forest types were explicitly downweighted in the selection process (Haugen 1991). A further unfortunate factor was the misleading label “Coniferous forest plan” instead of the more appropriate “Boreal forest plan.” This moved the attention away from the important diversity element of boreal deciduous trees. Information about species composition and quantitative coverage of vegetation types as well as tree composition was critically missing for the candidate sites. Thus, it was impossible to perform any analysis of complementarity to suggest more representative networks. The insufficient survey work and the suboptimal selection strategy have probably caused a highly suboptimal use of money for biodiversity conservation. Sætersdal and Birks (1993) demonstrated a similar inappropriateness in the nemoral forest network in western Norway because of the failure to cover the ecological variation appropriately. They identified four forest communities based on numerical analyses of vascular plants from sixty sample sites. One community was not represented at all in the existing network of 12 reserves, whereas another was only represented in one reserve.
Conservation Biology Volume 11, No. 1, February 1997
110
Efficiency of Bird and Insect Conservation
The Value of Deciduous Forests In addition to the bird analyses above, I have found that near 50% deciduous forests was the optimum for beetles sampled in various mature forests including some nemoral stands (Stokland 1995). The importance of deciduous forests is further indicated by the host tree relations of insects. Among the minimum 900 Scandinavian beetle species living in dead wood, 46% live exclusively in deciduous trees (28% when nemoral specialists are omitted), 19% in coniferous trees, and 35% in both (calculated primarily from Saalas 1917; Palm 1951; Palm 1959). For several hundred phytophagous macrolepidoptera species deciduous trees host more than 10 times as many species as coniferous trees (Seppanen 1970). Thus, even if deciduous-dominated stands comprise a small proportion of the boreal forests, they are very important to the species richness. Recommendations regarding scales and spatial arrangements of reserves are difficult to make using this study. The conclusion about deciduous forests was based on birds censused in 36–ha plots. In another study (Stokland 1995) I sampled beetles using window traps in 0.5–ha plots and found (sub)populations of deciduousspecific beetles in isolated deciduous stands as small as 3 ha. The other extreme is represented by the endangered White-backed woodpecker (Dendrocopus leucotus) with a home-range of about 100 ha, mature, deciduousdominated forest for each breeding pair. These findings indicate that mature deciduous stands are valuable irrespective of their size. There are two types of boreal deciduous woodlands in Scandinavia: (1) small, permanent, highly productive patches dominated by aspen, birch, or alder on south-facing slopes or in riparian sites and (2) patches of all sizes and various productivity associated with successions following wildfires. In general, it is not feasible to match the criterion of deciduous dominance with large reserves (10–100 km2) because deciduous-dominated woodlands often are much smaller. Thus, in order to over-represent the proportion of deciduous-dominated stands in the reserve network, a substantial part of the network should be allocated to such forest types. Among these one should allow reserves smaller than the lower limit of about 1 km2. Furthermore, some large reserves should be managed as “forest fire reserves” in order to maintain succession stages of deciduous forest together with the other qualities of burned forests.
Recommendations The skew in the Norwegian forest reserve network implies some obvious recommendations. First, more resources should be invested in survey work to establish a firm basis for guidelines before a further extension of
Conservation Biology Volume 11, No. 1, February 1997
Stokland
the reserve network is made. Such survey work should quantify tree composition and extent of different vegetation types and establish comparable species lists of relevant vulnerable organism groups. The surveys should be carried out in existing reserves as well as in promising supplementary sites. My analyses show that promising sites are various high-productive forest types and deciduous-dominated forests (riparian forests, forests on south-facing hillsides). Second, one should carry out a proper cost-efficiency analysis based on iterative and constrained selection to identify sites that best balance the existing reserve network. Third, future extensions should allow small forest refuges and actively include adjacent forestry-modified stands that will eventually regenerate into old-growth states.
Acknowledgments I thank S. Dale, R. A. Ims, N. C. Stenseth, and M. Sætersdal for valuable comments on the manuscript. I also wish to express my sincere thanks to E. Pierce who kindly improved the English. Financial support for the work was received from The Norwegian Research Council. Literature Cited Anonymous. 1981. The national forest survey 1964–76, Hedmark county. Norwegian Institute of Land Inventory, Ås. Austin, M. P., and C. R. Margules. 1986. Assessing representativeness. Pages 45–67 in M. B. Usher, editor. Wildlife conservation evaluation. London. Baadsvik, K. 1988. Forslag til retningslinjer for barskogsvern. DN-rapport 3 -1988. Directorate for Nature Management, Trondheim, Norway. Bedward, M., R. L. Pressey, and D. A. Keith. 1992. A new approach for selecting fully representative reserve networks: addressing efficiency, reserve design and land suitability with an iterative analysis. Biological Conservation 62:115–125. Cocks, K. D., and I. A. Baird. 1989. Using mathematical programming to address the multiple reserve selection problem: an example from the Eyre peninsula, South Australia. Biological Conservation 49:113–130. Diamond, J. M. 1975. The island dilemma: lessons of modern biogeographic studies for the design of natural reserves. Biological Conservation 7:129-146. Emberson, R. M. 1985. Comparisons of site conservation value using plant and soil arthropod species. Bulletin - British Ecological Society 16:16-17. Harrison, S., S. J. Ross, and J. H. Lawton. 1992. Beta diversity on geographic gradients in Britain. Journal of Animal Ecology 61:151–158. Haugen, I. 1991. Barskog i Øst-Norge. Utkast til verneplan. DN-rapport 1991-5. Directorate for Nature Management, Trondheim, Norway. Hågvar, S. 1976. Altitudinal zonation of the invertebrate fauna on branches of birch (Betula pubescens Ehrh.). Norwegian Journal of Entomology 23:61–74. Järvinen, O. 1982. Conservation of endangered plant populations: single large or several small reserves? Oikos 38:301–307. Järvinen, O., and R. A. Väisänen. 1977. Long-term changes of the North European land bird fauna. Oikos 29:225–228. Kirkpatrick, J. B. 1983. An iterative method for establishing priorities
Stokland
for the selection of nature reserves: an example from Tasmania. Biological Conservation 25:127–134. Kouki, J. P., P. Niemelä, and M. Viitasaari. 1994. Reversed latitudinal gradient in species richness of sawflies (Hymenoptera, Symphyta). Annales Zoologici Fennici 31:83–88. Kuusipalo, J., and J. Kangas. 1994. Managing biodiversity in a forestry environment. Conservation Biology 8:450–460. Lahti, T., A. Kurrto, and R. A. Väisänen. 1988. Floristic composition and regional species richness of vascular plants in Finland. Annales Botanici Fennici 25:281–291. Lomolino, M. V. 1994. An evaluation of alternative strategies for building networks of nature reserves. Biological Conservation 69:243–249. Margules, C., A. J. Higgs, and R. W. Rafe. 1982. Modern biogeographic theory: are there any lessons for nature reserve design? Biological Conservation 24:115–128. Margules, C. R., A. O. Nicholls, and R. L. Pressey. 1988. Selecting networks of reserves to maximise biological diversity. Biological Conservation 43:63–76. Margules, C. R., and J. L. Stein. 1989. Patterns in the distribution of species and the selection of nature reserves: an example from Eucalyptus forests in south-eastern New South Wales. Biological Conservation 50:219–238. Nicholls, A. O., and C. R. Margules. 1993. An upgraded reserve selection algorithm. Biological Conservation 64:165–169. Nilsen, P., and J. Larsson. 1992. Bonitering av skog ved hjelp av vegetasjonstype og egenskaper ved voksestedet (Site index estimation from vegetation type and site properties). Rapport Skogforsk 22:1– 43. Palm, T. 1951. Die Holz- und Rinden-käfer der nordscwedischen Laubbäume. Meddelanden 40 (2), Statens Skogsforskningsinstitut, Stockholm. Palm, T. 1959. Die Holz- und Rinden-käfer der süd- und mittelscwedischen Laubbäume. Pages 1–374 in Opuscula Entomologica Suppl. XVI. Pendergast, J. R., R. M. Quinn, J. H. Lawton, B. C. Eversham, and D. W. Gibbons. 1993. Rare species, the coincidence of diversity hotspots and conservation strategies. Nature 365:335–337. Pressey, R. L., and M. Bedward. 1991. Mapping the environment at different scales: benefits and costs for nature conservation. Pages 7– 13 in C. R. Margules and M. P. Austin 1991, editors. Nature conservation: cost effective biological surveys and data analysis. CSIRO, Australia. Pressey, R. L., and A. O. Nicholls. 1989a. Efficiency in conservation evaluation: scoring versus iterative approaches. Biological Conservation 50:199–218. Pressey, R. L., and A. O. Nicholls. 1989b. Application of a numerical algorithm to the selection of reserves in semi-arid New South Wales. Biological Conservation 50:263–278. Pressey, R. L., C. J. Humphries, C. R. Margules, R. I. Vane-Wright, and P. H. Williams. 1993. Beyond opportunism: key principles for systematic reserve selection. TREE 8:124–128.
Efficiency of Bird and Insect Conservation
111
Purdie, R. W., and R. Blick. 1986. Selection of a conservation reserve network in the Mulga biogeographic region of south-western Queensland, Australia. Biological Conservation 38:369–384. Rebelo, A. G., and W. R. Siegfried. 1992. Where should nature reserves be located in the Cape floristic region, South Africa? Models for the spatial configuration of a reserve network aimed at maximizing the protection of floral diversity. Conservation Biology 6:243–252. Saalas, U. 1917. Die Fichtenkäfer Finlands I. Annales Academiae Scientiarium Fennicae, ser. A 8:1–547. Seppanen, E. 1970. Suomen suurperhostoukkien ravintokasvit. WSOY, Helsinki. Simberloff, D. S., and L. C. Abele. 1976. Island biogeography theory and conservation practice. Science 191:285–286. Simberloff, D., and L. C. Abele. 1982. Refuge design and island biogeography theory: effects of fragmentation. American Naturalist 120: 41–50. Solhøy, T. 1976. Terrestrial gastropods (Mollusca, Gastropoda: Basommatophora and Stylommatophora). Pages 23–45 in Fauna of the Hardangervidda, no. 10. Zoological Museum, Bergen. Stokland, J. N. 1994a. Avian species diversity in relation to tree composition and productivity in mature boreal forests. Paper I in Biological diversity and conservation strategies in Scandinavian boreal forests. Ph.D. thesis. University of Oslo, Oslo. Stokland, J. N. 1994b. Altitude and latitude gradients in species richness of beetles (Coleoptera) in northern Europe. Paper III in Biological diversity and conservation strategies in Scandinavian boreal forests. Ph.D. thesis. University of Oslo, Oslo. Stokland, J. N. 1995. Artsmangfold og virkesproduksjon i sydøst-norske naturskoger. Aktuelt fra Skogforsk 13:1–16. Sætersdal, M., and H. J. B. Birks. 1993. Assessing the representativeness of nature reserves using multivariate analysis: vascular plants and breeding birds in deciduous forests, Western Norway. Biological Conservation 65:121–132. Sætersdal, M., J. M. Line, and H. J. B. Birks. 1993. How to maximize biological diversity in nature reserve selection: vascular plants and breeding birds in deciduous woodlands, Western Norway. Biological Conservation 66:131–138. Tomter, S. M., editor. Statistics of forest conditions and resources in Norway. Norwegian Institute of Land Inventory, Ås. Tveite, B., and H. Braastad. 1981. Bonitering av gran, furu og bjørk. Norsk Skogbruk 1981 (4):17–22. Underhill, L. G. 1994. Optimal and suboptimal reserve selection algorithms. Biological Conservation 70:85–87. Väisanen, R. A., O. Järvinen, and P. Rauhala. 1986. How are extensive human-caused habitat alterations expressed on the scale of local bird populations in boreal forests? Ornis Scandinavica 17:282–292. Väisänen, R., and K. Heliövaara. 1994. Hot spots of insect diversity in northern Europe. Annales Zoologici Fennici 31:71–81. Whitcomb, R. F., J. F. Lynch, P. A. Opler, and C. S. Robbins. 1976. Island biogeography and conservation: strategy and limitation. Science 193:1030–1032.
Conservation Biology Volume 11, No. 1, February 1997
