SEARCH

SEARCH BY CITATION

Keywords:

  • Altitude;
  • diversity;
  • ferns;
  • local diversity;
  • pteridophytes;
  • regional diversity;
  • species–area relationship;
  • tropical mountains

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Biosketches
  10. Supporting Information

Aim  To calculate the degree to which differences between local and regional elevational species richness patterns can be accounted for by the effects of regional area.

Location  Five elevational transects in Costa Rica, Ecuador, La Réunion, Mexico and Tanzania.

Methods  We sampled ferns in standardized field plots and collated regional species lists based on herbarium and literature data. We then used the Arrhenius function cAz to correct regional species richness (S) for the effect of area (A) using three slightly different approaches, and compared the concordance of local and regional patterns prior to and after accounting for the effect of area on regional richness using linear regression analyses.

Results  We found a better concordance between local and regional elevational species richness after including the effect of area in the majority of cases. In several cases, local and regional patterns are very similar after accounting for area. In most of the cases, the maximum regional richness shifted to a higher elevation after accounting for area. Different approaches to correct for area resulted in qualitatively similar results.

Main conclusions  The differences between local and regional elevational richness patterns can at least partly be accounted for by area effects, suggesting that the underlying causes of elevational richness patterns might be the same at both spatial scales. Values used to account for the effect of area differ among the different study locations, showing that there is no generally applicable elevational species–area relationship.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Biosketches
  10. Supporting Information

Understanding the factors that drive gradients of species richness is one of the major challenges of ecological and biogeographical research. Regressing patterns of species richness against sets of variables that are presumed to be potentially responsible for these patterns is a common method, and comparing results from many studies enhances our insight into the nature of the distribution of species richness. The crucial problem is that comparing results from different studies requires comparable methods of data collection, especially concerning the basic data source. Studies documenting patterns of species richness can be divided into two main groups (Romdal & Grytnes, 2007). The first focuses on field inventories in a specific area (hereafter: local studies), whereas the second derives its data mainly from literature and scientific collections (hereafter: regional studies). These approaches differ substantially in geographical extent and data quality as well as in the factors that influence them (Fig. 1).

image

Figure 1.  Factors linking local and regional species richness to area. The true local and the true regional richness are the actual values of a given area. Because every sampling method has its own specific error and will not sample the complete richness (sampling completeness), it is impossible to measure this value exactly. Because of this, every method applied always results in an estimate of species richness (estimated local and regional richness). Area influences true richness in a number of ways, depending on the scale of the area. For the local area, the representativeness of small plots (which increases as plot size increases: passive sampling) and small-scale changes of habitats are the main factors influencing true local richness. At the regional scale, effects of biogeographical species turnover, increasing large-scale habitat diversity, and increasing carrying capacity have to be taken into account. True regional richness influences true local richness by providing the species pool from which true local richness is recruited. The abundance of species within this species pool increases with the size of the regional area and affects local richness by the higher dispersal probability into the local area (echo effect). We defined local and regional area at the spatial scales typically found in ecological studies, leaving a gap between them that is typically too large for plot sampling and too small for regional studies.

Download figure to PowerPoint

Neither of the two approaches can a priori be said to be superior to the other, because both have their own advantages and drawbacks. In general, local studies are costly, time-consuming and typically cover just a minute portion of the surface area, as well as only part of the regional biota, especially in species-rich environments (Whitmore et al., 1985). Thus, a constant sampling intensity in different environments may result in a systematic undersampling in species-rich communities compared with species-poor communities (Lawton et al., 1998). Properly conducted local studies have the advantage of a standardized sampling method, thus eliminating the influence of varying sampling intensity and sampling area on observed richness patterns. A further problem of local studies is that ongoing habitat destruction makes it increasingly difficult to find natural habitats for sampling.

Regional studies, on the other hand, are advantageous because richness patterns can be studied without costly field surveys by data mining based on extensive prior studies, which often include areas that can no longer be sampled because of habitat destruction. Regional studies cover larger areas and larger fractions of the total biota because they combine data from numerous collecting efforts. Collectors, however, usually prefer easily accessible areas that look particularly interesting, so field collections tend to be spatially uneven (Nelson et al., 1990; Soria-Auza & Kessler, 2008). They might also focus on rare and interesting species, such that common species are under-represented. Actual species records are always patchy and it is contentious whether species distributions should be interpolated between records. Along elevational gradients, for example, some authors have suggested that species should be considered as present in a given elevational belt if they have been recorded at both higher and lower elevations (Williams et al., 1996; Lees et al., 1999). This may, however, cause artificially elevated species numbers at mid-elevations, because such interpolated data are overproportionally added to mid-elevations as opposed to edges of the gradient (Grytnes & Vetaas, 2002). Moreover, unusual distribution patterns (e.g. bimodal) may be masked (Hemp, 2002).

Patterns of species richness along elevational gradients have received considerable attention in the last decade, and have become firmly established as a complementary, replicable alternative to the traditional studies of latitudinal gradients (Rahbek, 1995, 2005; Lomolino, 2001). At both the local and regional level, elevational richness patterns across a wide range of taxa commonly show a more or less hump-shaped pattern, with maximum richness at some intermediate point of the gradient (Rahbek, 1995, 2005; Grytnes & McCain, 2007). Such patterns have been documented for mammals (McCain, 2005, 2007), birds (Herzog et al., 2005; McCain, 2009), reptiles (Fu et al., 2006), moths (Brehm et al., 2007), flowering plants (Hemp, 2001; Moser et al., 2005; Oommen & Shanker, 2005), vascular epiphytes (Küper et al., 2004; Krömer et al., 2005; Cardelús et al., 2006; Hemp, 2011), ferns (Hemp, 2001; Bhattarai et al., 2004; Carpenter, 2005; Kluge et al., 2006) and bryophytes (Grau et al., 2007). In addition, some studies have revealed monotonic declines of species richness from lowlands to high elevations (e.g. Araceae: Kessler, 2002; Acebey & Krömer, 2008), or roughly constant values from lowlands to mid-elevations, followed by a marked decline (Rahbek, 1995, 2005; Kessler, 2001). The causes determining elevational patterns of species richness are still being debated, but include the combined effects of surface area, geographical constraints, climate and ecosystem productivity, evolutionary and historical processes, and population-level processes such as source–sink effects (Grytnes & McCain, 2007; McCain, 2009). The degree to which these causes may differently influence local and regional patterns remains unclear.

The studies listed above include both local and regional approaches. Biases of both types of studies may be expected to result in different perceptions of patterns of species richness (Kessler et al., 2009). It is still unknown if local and regional studies differ in any systematic way, and hence it is unclear whether they can be combined, for example in meta-analyses of elevational richness patterns (McCain, 2009). In a recent comparison of elevational richness patterns recovered from local and regional studies, Kessler et al. (2009) found minor, but distinct, deviations between the two approaches. This might be a result of statistical noise, but there may also be a systematic deviation between the two types of datasets.

Kessler et al. (2009) suggested that regional area may lead to deviations between local and regional richness, as area is well known to influence patterns of species richness through a variety of mechanisms at various scales (Rosenzweig, 1995). At a regional scale, larger areas are well known to support more species, both because they maintain viable populations of more species and because they typically include higher habitat diversity (MacArthur & Wilson, 1967; Drakare et al., 2006). As land surface area typically declines with increasing elevation (Körner, 2000; Lomolino, 2001), regional species richness may be expected to peak at lower elevations compared with local richness. For this reason, a number of studies have corrected the regional number of species along elevational gradients for area effects (e.g. Rahbek, 1997; Sanders, 2002; Bachman et al., 2004).

At the local scale, larger sampling areas typically include more species because they more completely sample the regional species pool (‘passive-sampling hypothesis’, Connor & McCoy, 1979). However, even if sampling area is held constant, regional area will influence local species indirectly through the increase of the regional species pool (‘echo-effect’, sensuRosenzweig & Ziv, 1999; see also Brehm et al., 2003; Herzog et al., 2005). This indirect effect of regional area on local species richness has been documented in a meta-analysis including 71 published local elevational transect datasets (Romdal & Grytnes, 2007). This analysis, however, was based only on correlation analyses between local richness and regional area, without considering possible co-variances of other factors. For example, if local species richness is influenced by climatic factors that systematically change with elevation (e.g. temperature), and regional area also changes systematically with elevation, then a spurious correlation between local richness and regional area may result. To unravel any causal relationships between local richness and regional area, it is thus necessary to consider the intermediate link between them, that is, regional richness.

In this study we therefore extend the approach of Romdal & Grytnes (2007) by taking regional richness into account. Because the proposed indirect effect of regional area on local richness is mediated through regional richness, consideration of the latter can help to discern the causality of these relationships more clearly.

To test this assumption, we compiled data from both local and regional studies of ferns and lycophytes (hereafter collectively termed simply ferns) along five elevational gradients (Pichincha in Ecuador, Braulio Carrillo in Costa Rica, Los Tuxtlas in Mexico, Kilimanjaro in Tanzania, and La Réunion). Our aim was to assess the difference in local and regional patterns, and to calculate the degree to which these differences can be accounted for by the effects of area.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Biosketches
  10. Supporting Information

Local elevational species richness data were derived from five field studies using the same consistent sampling method (except for the Kilimanjaro transect, where the plot area was 1000 m2) (Table 1 and Appendix S1 in the Supporting Information). Sample plots of 400 m2 each, and of square shape, were placed in natural forest, that is, avoiding secondary vegetation and special habitats such as gaps, ravines or ridges, in order to keep vegetation structure as homogeneous as possible (Kessler & Bach, 1999; Kessler, 2001). Local species richness is given by the mean number of species in all plots sampled in a given elevational band. Regional richness data were generated from published databases and species lists (Table 1). Elevational extremes for each species in the study regions were interpolated to generate distribution data of species for each elevational belt under the assumption that species have continuous distributions (Grytnes & Vetaas, 2002). We defined regional area as the land surface covering the respective elevational belt where the regional richness data were obtained. For example, the regional area for the La Réunion transect consisted of the surface area of the whole island, because only specimens collected on this island were used to calculate the regional species richness. Surface area was calculated for Los Tuxtlas, Kilimanjaro, La Réunion and Pichincha using Spatial Analyst in ArcView 3.2 (ESRI, Redlands, CA, USA). For the Costa Rica transect, we estimated the proportion of total area found at intervals of 100 m in a 30-km-wide strip in Braulio Carrillo National Park and on Cerro de la Muerte (local transect area) by counting grid cells on topographical maps (1:50,000) between the contour lines (Kluge et al., 2006).

Table 1.   Data type and sources of the five local and regional datasets of elevational fern and lycophyte species richness. For the regional data, in some cases richness estimates were corrected by interpolation of elevational species ranges between the recorded elevational maxima. The number of sample plots per elevational belt is provided in Appendix S1.
TransectScaleMethodSourceData correction
PichinchaLocal22 plots of 20 × 20 mM. Kessler & M. Lehnert, unpublished dataNo
RegionalCountry listJørgensen & León-Yánez (1999)
Costa RicaLocal156 plots of 20 × 20 mKluge et al. (2006)Interpolation
RegionalRegional floraMoran & Riba (1995)
Los TuxtlasLocal42 plots of 20 × 20 mT. Krömer & A. Acebey, unpublished dataNo
RegionalCountry floraMickel & Smith (2004)
Kilimanjaro (south side)Local379 plots of 1000 m2Hemp (2001)Interpolation
RegionalRegional listHemp (2002)
La RéunionLocal29 plots of 20 × 20 mM. Kessler, unpublished dataNo
RegionalHerbarium recordsM. Kessler, unpublished data

Our assessment of the mismatch between local and regional richness patterns by incorporating the effect of regional area was based on the Arrhenius (1921) equation,

  • image(1)

where S is the number of species, c is the number of species in the smallest sampling area, A is area (in arbitrary units), and z is a constant describing the slope of the species–area relationship in the log–log space.

In our case, we set = observed regional species richness, and = theoretical (area-corrected) regional richness at an arbitrary minimum area common to all elevational belts. This resulted in the formula:

  • image(2)

where the subscript ‘R’ indicates regional species richness and the subscript ‘cor’ indicates area-corrected values. To calculate SRcor, this equation was then modified to

  • image(3)

The main challenge of applying this formula was to obtain realistic z-values in the Arrehnius function. These values can be derived empirically from the slope of the species–area relationship in log–log space. Deriving the z-value by using the slope of a species accumulation curve is difficult in this case, because our small dataset shows a high degree of variance. In addition, z-values can vary with spatial scale (Crawley & Harral, 2001), and basing our estimates on just one single derived z-value might be misleading. Thus we used two additional approaches to estimate the z-value and validate the results of the empirically derived z-values. The second approach consisted of a nonlinear model (SR SLAz) to estimate the z-value that will give the best concordance between regional species richness (SR) and local species richness (SL) patterns. The calculated z-values from the nonlinear model were then used to correct for the effect of area, and the resulting patterns of corrected regional richness were then compared with local richness using linear regression analyses. Because empirical z-values typically fall between 0.2 and 0.4 and cannot be higher than 1 or lower than 0 (MacArthur & Wilson, 1967; Rosenzweig, 1995; Crawley & Harral, 2001), we used a range from 0 to 1 for z in the model. Our third approach used linear regression analyses between local and regional richness. We checked the values of the coefficient of determination (R2) given by the linear regression between local and regional richness for all z-values ranging from 0 to 1 in steps of 0.01. To test if a linear regression is valid to describe the relationship between local and regional richness, we calculated Akaike’s information criterion (AIC; Wagenmakers, 2004) using a linear and a quadratic nonlinear model. Because the number of observations divided by number of parameters sometimes falls below 40 (for the small datasets: Pichincha, La Réunion, Los Tuxtlas), we used the bias-corrected form, AICc (Burnham & Anderson, 2002).

Prior to the analyses, we transformed species numbers and area at each elevational band relative to a maximum of 100 to eliminate the effects of different measuring units as well as to make the patterns graphically more easily comparable. Trend lines within the figures were fitted with distance-weighted least-squares smoothing, using Xact 7.2 (SciLab, Hamburg, Germany).

Finally, we compared the elevational patterns of local richness with uncorrected and area-corrected regional species richness through linear regression analyses. All calculations were carried out in spss 11.5 and R (R Development Core Team, 2008).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Biosketches
  10. Supporting Information

Values of the coefficient of determination (hereafter: regression values) of the linear regression analyses between local and uncorrected regional species richness ranged from 0.24 (Kilimanjaro) to 0.84 (La Réunion) (Table 2). On four transects (Costa Rica, Kilimanjaro, La Réunion, Los Tuxtlas), uncorrected regional species richness peaked at a lower elevation than did local richness, whereas along the Pichincha transect, peaks roughly coincided in elevation (Fig. 2).

Table 2.   Empirically derived z-values, z-values estimated with the nonlinear model (nlm), and z-values that resulted in the highest concordance between local and regional richness. Linear regressions were calculated for local richness versus uncorrected regional richness, as well as against regional richness corrected with the empirically derived z-values, the derived z-value from the nonlinear model (nlm), and the z-value calculated with the linear model that resulted in the highest regression values between local and regional richness (highest R2).
Transectz-valuesR2-values
EmpiricalnlmHighest R2Regional richness corrected for area with z-values derived by:
UncorrectedEmpiricalnlmHighest R2
  1. ***< 0.001; **< 0.01; *< 0.05.

Kilimanjaro0.630.530.69***0.24**0.58***0.55**0.59***
Costa Rica0.260.11*0.27***0.60***0.79***0.72**0.79***
Los Tuxtlas−0.050.210.06**0.36*0.300.53*0.75**
La Réunion0.160.210.36***0.84***0.88***0.88**0.89***
Pichincha0.270.260.19**0.79**0.89**0.89**0.89**
image

Figure 2.  Uncorrected and area-corrected patterns of regional (black circles, continuous lines) and local (open circles, dashed lines) fern and lycophyte species richness along all five elevational transects. To account for the effect of area, the following z-values were used: Costa Rica = 0.27; Kilimanjaro = 0.69; Pichincha = 0.19; La Réunion = 0.36; Los Tuxtlas = 0.6. Trend lines were fitted with distance-weighted least-squares smoothing, using Xact 7.2 (SciLab, Hamburg, Germany). Note that species numbers for each elevation were transformed to a relative maximum of 100 prior to analysis to make elevational trends visually comparable.

Download figure to PowerPoint

After accounting for the effect of area, regional richness peaks shifted upwards (Fig. 2), resulting in higher regression values with local richness along all transects (Table 2). Using the empirically derived z-values, regression values increased to 0.58 (Kilimanjaro), 0.79 (Costa Rica), 0.88 (La Réunion) and 0.89 (Pichincha) and decreased for Los Tuxtlas.

When we used the z-values obtained by estimation of the z-value using the nonlinear model, regression values increased to 0.53 (Los Tuxtlas), 0.88 (Pichincha and La Réunion), 0.72 (Costa Rica) and 0.55 (Kilimanjaro).

The test for all possible z-values from 0 to 1 showed higher regression values between local and regional richness for small z-values (< 0.45) for Kilimanjaro, Costa Rica, Pichincha and La Réunion. For Los Tuxtlas, higher regression values between local and regional richness were obtained at z-values < 0.26. When we used the z-values that resulted in the highest concordance of local and regional richness, regression values were, unsurprisingly, higher. In one case (Los Tuxtlas) the regression values were considerably higher than those obtained with the empirically derived z-value and the z-values estimated by the nonlinear model, whereas in the other cases they reached similar values. Interestingly, derived z-values of all three approaches were often in good concordance (standard deviation: ± 0.1). Thus, at Kilimanjaro, the empirically estimated z-value was 0.63, while the highest regression value was obtained with a z-value of 0.69, and the nonlinear model predicted a z-value of 0.53. Similar deviations were found at all other sites. The AICc values for the linear model were lower than those for the non-linear model for all sites when testing the relationship between local and regional richness. Differences of the AICc values ranged from 8.4 in Los Tuxtlas to 1.45 at Kilimanjaro.

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Biosketches
  10. Supporting Information

Our results support the assumption that area is partly responsible for the observed differences between local and regional species richness patterns along elevational gradients. Often, local and regional species richness patterns along such transects are already in good concordance (Herzog et al., 2005; Kessler et al., 2009), as best shown in our dataset by the Pinchinca and La Réunion transects (Fig. 2). However, both in our study and more generally, regional richness patterns tend to show peaks at lower elevations compared with local species richness (Kessler et al., 2009). By statistically accounting for the effect of varying surface area along the study gradients on the estimates of regional species richness, as previously applied for example by Rahbek (1997), Sanders (2002) and Bachman et al. (2004), we improved the concordance between local and regional richness patterns. In several cases, especially in Costa Rica, La Réunion and Kilimanjaro, local and regional patterns were very similar after accounting for area (Fig. 2). In all cases, the maximum regional richness shifted to a higher elevation after accounting for area, so that regional area seems to be the main factor responsible for these deviations in maximum species richness. This strongly implies that, despite the indirect character of the area effect on local richness (Rosenzweig & Ziv, 1999; Romdal & Grytnes, 2007), regional richness is influenced to a greater degree by area than is local richness. This difference is intuitively appealing, considering that local richness is clearly spatially restricted and depends on the size of the local sampling units. Thus, richness in smaller units may be primarily limited by the number of individuals and hence species that fit into small areas (Romdal & Grytnes, 2007). Indeed, Romdal & Grytnes (2007) found that the indirect area effect increased with increasing sample size in their meta-analysis. As we used only two sizes of sample plots (400 m2 on most transects, 1000 m2 on Kilimanjaro), we are unable to conduct a similar analysis. It would be worthwhile, however, to apply our approach to transect datasets where local richness has been sampled at different scales, such as the Norwegian dataset of Grytnes (2003).

Although we used the species–area relationship to account for the area effect on regional area, this does not imply that we consider that we cannot account for the effect of regional area on local richness estimates. Our aim was to assess the degree to which the different effects of regional area on local and regional richness estimates lead to the observed differences in elevational richness patterns. We could just as well have conducted the analysis the other way around, accounting for area on local richness to match regional richness. Indeed, we did such an analysis, but because it produced qualitatively identical results, the results are not reported here.

Accounting for area effects did not result in perfect concordance between local and regional richness, and concordance was higher at three localities (Kilimanjaro, Costa Rica, Los Tuxtlas) than at the other two. Especially for the three transects with R2 values > 0.8, the remaining variance may simply reflect sampling noise. In other cases, however, especially where local and regional patterns remain distinctly different despite accounting for area and where regression values remain relatively low, additional factors may play a role. Another potential reason for the continuing mismatch between local and regional richness estimates despite taking regional area into account is that, in mountains, ground surface area is also influenced by the inclination of the mountain slopes and the roughness of the terrain, so that steeper and more dissected mountains have a larger total area (Vetaas & Grytnes, 2002; Rahbek et al., 2007). We did not attempt to model this effect for two reasons. First, estimates of the three-dimensional surface area depend on the spatial resolution of the digital elevation models, making it difficult to choose a ‘correct’ one. Second, plants still grow upright on mountain slopes, so that a higher three-dimensional surface area does not necessarily lead to proportional changes in the number of plant individuals. At the same time, very steep slopes may not be inhabited at all by most plants, despite their large surface area. One remaining reason for the continuing mismatch between local and regional richness patterns could be the different data corrections that we used for the regional richness (Table 1). These data corrections, however, only accounted for minor differences in the magnitude of statistical noise between the patterns, and therefore were not taken into account.

Along three of our study transects (Kilimanjaro, Los Tuxtlas, La Réunion), regional area estimates accounted for the entire mountain or mountain range. On the remaining transects (Costa Rica, Pichincha), regional area was delimited in a more arbitrary way to include the ecologically homogeneous regions about 100–200 km north and south of the study transects. This approach was also influenced by the availability of regional distributional data, which are often given only for large political entities, for example the country Costa Rica, forcing us to accept these entities as regions. This certainly affected our estimates of both regional area and regional species richness, but we do not consider that this fundamentally influenced our analyses because both parameters were evaluated for the same region. Had we chosen a larger regional area, then regional richness estimates would also have increased. While the absolute values of the relationships between local and regional richness would then also change, the qualitative relationships between local and regional richness estimates would be maintained.

The best concordance of local and regional richness after accounting for area effects was achieved with different z-values on each of the five study transects. This underlines the observation that z-values are not a given constant but rather are dynamic and hard to estimate (Crawley & Harral, 2001). Our three approaches for estimating z-values arrived at qualitatively similar but quantitatively different results. The derived z-values using the three approaches in many cases differed to some degree from each other. However, all of the three approaches have their strengths and weaknesses. Empirically derived z-values are the only ones that can be directly derived from the data and might therefore be the best choice. However, for small datasets with only a few data points, there is a relatively high probability that the derived z-value might be inaccurate. This might be the case for Los Tuxtlas, where the empirical z-value is actually negative.

Calculating regression values with all possible z-values shows that better concordance of local and regional richness can be achieved when using z-values < 0.45 for most localities (Fig. 3). z-values in this range have been also reported in previous studies (MacArthur & Wilson, 1967; Rosenzweig, 1995; Crawley & Harral, 2001), while higher z-values, which in our study lead to lower concordance between local and regional richness, are rarely documented. This approach helps in finding the z-value that results in the highest concordance between local and regional richness, but it may not result in empirically realistic values.

image

Figure 3. z-values ranging from 0 to 1 and the resulting regression values between local and regional richness patterns of ferns and lycophytes when the respective z-values are used to correct for area. The continuous bold lines show the R2 values for 100 steps of z-values (0, 0.01, 0.02, etc.). The dotted lines indicate the R2 values of the uncorrected richness patterns. The grey area indicates where regression values are highly significant ( 0.001).

Download figure to PowerPoint

The values estimated with the nonlinear model take into account that local–regional richness relationships can be nonlinear. On the other hand, the AICc values calculated to test for a linear versus a nonlinear relationship between local and regional richness show that a linear model is probably more appropriate. However, the good qualitative concordance of the z-values for all models (with some deviations for Los Tuxtlas) and the qualitatively similar results in most of the cases, irrespective of the approach used to select the z-values, support the conclusion that differences between local and regional elevational richness patterns can partly be accounted for by the effect of area on regional richness.

At present, we are unable to explain the causes of the different local versus regional patterns. They could be the result of uneven or incomplete regional sampling. They might, however, also reflect the fact that local sampling was usually restricted to forest habitats whereas regional species lists included species from a wider range of habitats, although these habitats typically have fewer fern species (Hemp, 2001; Kessler, 2001). Finally, these differences may reflect spatially varying impacts of factors that influence elevational richness patterns, such as temperature and humidity (Bhattarai et al., 2004; Krömer et al., 2005; Kluge et al., 2006), as well as energy availability and ecosystem productivity (Hawkins et al., 2003; Currie et al., 2004), historical and evolutionary processes (Ricklefs, 2004; Wiens & Donoghue, 2004; Smith et al., 2007), source–sink effects (Grytnes et al., 2008; Kessler, 2009), or even human impact (Nogués-Bravo et al., 2008; Marini et al., 2010).

In conclusion, regional surface area seems to be largely responsible for the differences between local and regional species richness patterns along elevational gradients. This has important implications for studying and understanding elevational diversity patterns. First, from an analytical point of view, we suggest that a direct comparison of local and regional patterns is feasible once the latter have been corrected for area effects. This may greatly increase the number of studies that can be combined in meta-analyses (e.g. McCain, 2009). More fundamentally, if area is indeed the only factor important for the difference between local and regional elevational richness patterns, then the similarity between the two patterns may be taken as an indication that both are reflections of one general pattern driven by a common suite of underlying processes, rather than by different processes at different spatial scales.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Biosketches
  10. Supporting Information

We thank all the people who supported our fieldwork; there are far too many to acknowledge individually here but without them this study would not have been possible. Rodrigo Wilber Soria-Auza conducted the GIS calculation of area. Fieldwork was partly funded by the Deutsche Forschungsgemeinschaft DFG (grants to M.K., A.H.), the Deutscher Akademischer Austauschdienst DAAD (to J.K., S.K.H.) and the Universidad Nacional Autónoma de México (to T.K.).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Biosketches
  10. Supporting Information

Biosketches

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Biosketches
  10. Supporting Information

Dirk N. Karger is a PhD student at the Department of Systematic Botany at the University of Zurich, Switzerland. His current research focuses on patterns of diversity among ferns on islands of different sizes in the Philippines and Indonesia.

The authors share a strong research focus on tropical montane forests, including ferns. Together, they are interested in understanding how tropical mountain biodiversity has evolved and is maintained, and how it can be sustainably managed.

Author contributions: D.K., J.K. and M.K. conceived the ideas; J.K., T.K., A.H., M.L. and M.K. collected the data; D.K., J.K. and M.K. analysed the data; D.K., J.K. and M.K. led the writing.

Editor: Ole Vetaas

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Biosketches
  10. Supporting Information

Appendix S1 Number of sampled plots per location.

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JBI_2468_sm_apps1.doc370KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.