The Influence of Island Area on Ecosystem Properties

See allHide authors and affiliations

Science  29 Aug 1997:
Vol. 277, Issue 5330, pp. 1296-1299
DOI: 10.1126/science.277.5330.1296


Island area is frequently a major determinant of the species composition of biological communities; community structure, in turn, often has important effects on ecosystem-level properties. Fifty islands of varying area were selected in an archipelago in the northern Swedish boreal forest zone, in which larger islands burn more frequently than smaller ones through wildfire arising from lightning strike, thus inducing a significant relationship between island area and plant species composition. This relationship was found to be a major factor in determining several ecosystem-level properties of these islands, including standing biomass, plant litter decomposition, nitrogen mineralization, terrestrial carbon partitioning, humus accumulation, and plant nitrogen acquisition.

Gradients of island area have frequently been used to help understand the factors responsible for structuring ecological communities (1), and it is apparent that the area of islands is important in regulating the occurrence and abundance of component species (2) as well as their interactions (3). There is an increasing awareness that individual species effects in communities are important determinants of ecosystem-level properties and, consequently, of functioning of the ecosystem (4). Therefore, it is expected that islands with different areas and thus different species compositions would contain different ecosystem-level attributes (5). However, there have been few attempts at using island-area gradients for evaluating processes and factors that operate at the ecosystem-level of resolution (5).

We selected an island archipelago in the northern boreal forest zone of Sweden, located within two adjacent lakes—Lake Hornavan and Lake Uddjaure (65°55′–66°09′N; 17°43′–17°55′E). We chose 50 islands, ranging in area from 0.02 to 15.0 ha (and with these areas being distributed lognormally) which were formed on morainic deposits created by the retreat of land ice 9000 years ago. Our island system is ideal for testing hypotheses relating to island area effects because all the islands are of the same age and origin, and because they have all been subjected to comparatively minor human interference (6). The main disturbance regime on the islands is wildfire by lightning strike; lightning (and therefore wildfire) appears to strike larger islands more frequently than smaller ones, presumably because they have a larger area to intercept. This is reflected in both the vegetation composition (7) and fire history data (8) collected from each of these islands (Table1). The vegetation data demonstrated that larger islands are dominated by earlier-successional plant species which dominate in the presence of regular wildfire, for example, Pinus sylvestris and Vaccinium myrtillus, whereas smaller islands show a greater abundance of late successional species which occur in the prolonged absence of fire, for example Picea abies and Empetrum hermaphroditum (9). Further, evaluation of fire history with both fire-scar data and14C dating of charcoal particles in humus profiles reveals that the larger islands, in general, have burned much more recently than the smaller ones and have a greater fire frequency (Table 1).

Table 1

Vegetation and fire history characteristics in relation to island area. Values presented are means ± SE.

View this table:

Those late successional plant species that dominate on the smaller islands typically produce foliage and litter of poor quality with high levels of secondary metabolites, principally phenolics (10), and these compounds have the potential to reduce the ability of the soil microflora to decompose substrates and mineralize nutrients (11). Measurements of soil chemical and biological properties (12) were consistent with this—we found higher concentrations of water-soluble phenolics, reduced microbial biomass, and reduced microbial activity in the humus of the smaller islands (Fig. 1). This retardation of microbial activity was also reflected in a reduction of the rates of decomposition and N mineralization of V. myrtillus litter placed on the smaller islands (13). Further, the N concentration of both the humus and the added litter was highest on the smaller islands, suggesting that on these islands organic N becomes bound in protein-phenolic complexes, which are notoriously resistant to microbial attack (14). The inhibition of soil biotic processes on the smaller islands probably contributes to the substantial accumulation of humus that occurs on them (Fig. 1); the smallest islands contain up to 10 times more humus per unit area than do the largest ones. Therefore, our results demonstrate that island area is critical in regulating key ecological processes (and therefore ecosystem function) and that this is most likely attributable to plant species effects.

Figure 1

Humus chemical and biological characteristics in relation to island area. (A through F) Chemical and biological properties are shown for humus samples collected from each island. Concentrations of phenolics are expressed as microgram of gallic acid equivalent per gram of humus. (G through J) Properties are shown for leaf litter of V. myrtillus placed in litter bags and left to decompose for 1 year. (K and L) Total mass of humus and humus N was determined on an areal basis.

Carbon partitioning is clearly regulated by island area. With increasing island size, there is a distinct trend of an increasing proportion of organic C that is bound in living organisms, especially trees (Fig. 2). Because smaller islands contain much higher terrestrial C levels on an areal basis than do larger ones, our data indicates that wildfire is of critical importance (either directly or indirectly) in reversing C lock-up in boreal forest ecosystems. This finding suggests that deliberate anthropogenic supression of fires in boreal forests over the past century has the potential to lead to retardation of soil biological processes and substantial terrestrial C sequestration (15), which is likely to be of global significance given the role of boreal forests in the global C cycle (16).

Figure 2

Proportion of terrestrial C present in humus and plant pools in relation to island area. Total C values are mean ± SD.

There is evidence that N partitioning is also likely to be affected by island area. Although the humus N content was negatively related to island area, the N concentrations of leaves of E. hermaphroditum, shoots of the moss Pleurozium schreberi, and the soil microbial biomass were all positively related to island area (Fig. 3). This is indicative of enhanced acquisition of N by these organisms, and means that the N present in the humus of the smaller islands is clearly less available to living organisms than that of the larger islands. This again supports the concept that higher levels of phenolics on the smaller islands contribute to reduced N availability. The lower plant tissue N content of plants on smaller islands would be expected to result in subsequently produced plant litter with a lower C:N ratio, and thus, a reduced rate of decomposition (17), ultimately resulting in further organic matter accumulation.

Figure 3

Total N concentration of selected organisms in relation to island area. The relationship of humus N concentration to island area is included for comparative purposes. The N concentration of leaves from V. myrtillus and V. vitis-idaea did not show a statistically significant relationship with island area.

Island area studies have potential for investigation of relationships between species diversity and ecosystem-level properties (5). In our study, plant species diversity was highest on the smallest islands (Table 1), meaning that ecosystem process rates were lowest on those islands with the greatest diversity. This finding is in direct contrast to other studies which have shown elevated process rates in more diverse communities (18). However, we found that the plant biomass of the larger islands was dominated by a single early successional competitive tree species, P. sylvestris, which has comparatively favorable litter quality but which contributes to a reduced species diversity. The greater ecological stresses on the smaller islands (due to higher phenolics, lower pH, and reduced N availability) presumably prevented dominance by a single competitive species, resulting in a greater diversity (19), but with a greater abundance of those plant species with traits likely to contribute to retarding ecosystem-level processes. Other possible explanations for the discrepancy between our results and earlier investigations (18) are that we considered the long-term effects of plant litter and belowground interactions on ecosystem properties, rather than the live-plant effects and productivity aspects that have characterized other studies (20); that possible artifact associated with those studies claiming positive effects of species diversity on ecosystem processes has led to incorrect conclusions (21); and that long-term feedbacks between plant species composition and ecosystem processes occurred in our study. Our results point to the utility of island archipelago studies in investigating relationships between biodiversity and ecosystem function, and reveal that the nature of such relationships are ultimately dependent on the ecological attributes of each of the plant species present (22).

Our study has demonstrated, through the assessment of a range of islands with differing plant communities, that species effects can be of critical importance in determining ecosystem properties, and we suggest that island archipelago studies provide unique, and largely unrealized, opportunities to test hypotheses relating to species effects in ecosystems. In using islands to address ecosystem-level questions, an obvious extension is to address relevant issues through experimentation rather than through the correlation-based approaches that have characterized island biogeography research to date. Ultimately, island-area effects are far more wide-reaching than simply influencing the structure of ecological communities; they are also important determinants of how the ecosystem itself functions.

  • * To whom correspondence should be addressed. E-mail: wardled{at}

  • Present address: Landcare Research, Post Office Box 69, Lincoln 8152, New Zealand.


View Abstract

Navigate This Article