西双版纳热带季节雨林凋落叶分解与土壤动物群落：两种网孔分解袋的分解实验比较

 以西双版纳热带湿性季节沟谷雨林混合凋落叶作为分解基质,在不同位置季节雨林样地,采用不同网孔（ 2和0.15 mm）分解袋,开展大中型土壤动物对雨林凋落叶分解影响的实验,测定了不同网孔分解袋土壤动 物多样性、凋落叶分解速率和主要养分元素释放状况。结果显示：2 mm网孔分解袋土壤动物类群相对密度 年均值为2.67～2.83目•g－1凋落物干重,个体相对密度年均值为22.3～21.77个•g－1凋落物干重,显著 高于0.15 mm网孔分解袋的类群相对密度0.27～0.28目•g－1凋落物干重和个体相对密度2.88～2.77个•g－ 1凋落物干重（p<0.01）,并且0.15 mm网孔分解袋中极少量的动物个体主要为小型类群弹尾目和蜱螨目（ 原生动物、湿生土壤动物线虫不计）,由此我们视2 mm网孔分解袋凋落叶分解由绝大多数土壤动物和其它 土壤生物共同作用,而0.15 mm网孔分解袋基本排除了大中型土壤动物对袋内凋落叶分解的影响。2 mm网 孔分解袋凋落叶物质失重率（71%左右）、分解率指数（1.88～2.44）和主要养分元素释放率明显高于 0.15 mm分解袋（34%～35%,0.48～0.58）。通过比较两种不同网孔分解袋凋落叶失重率和元素释放率的 差异,显示出季节雨林大中型土壤动物群落对凋落叶物质损失的贡献率为年均值46%左右,并使凋落叶C/N 和C/P明显降低,而对不同元素释放率的影响不同,其中对N、S和Ca元素释放率的影响较大,而对K素释放 的影响作用最小。相关分析显示,2 mm网孔分解袋内土壤动物群落类群和个体的相对密度与凋落叶物质残 留率有较好的负相关关系,而群落香农多样性指数与凋落叶分解率指数表现出一定的正相关关系。     

Leaf size and inflorescence size may be allometrically related traits

Summary Corner's rules for plant form relate the degree of branching to branch diameter, and branch diameter to leaf or inflorescence size. We report the first interspecific test of these rules for inflorescence size and branch diameter. We derived a simple corollary of Corner's rules; since leaf size and inflorescence size are both correlated to branch thickness, they may be correlated to each other. This corollary holds for Leucadendron and Protea (Proteaceae), and in certain other taxa in the Asteraceae, Bruniaceae and Pinaceae which also have leaves and reproductive structures on the same shoot. For such taxa this implies that selection for aspects of floral display (inflorescence size, pollination type) may also be expressed at the level of leaf size and vice versa. This has implications for many aspects of botany and also points to the importance of the co-ordinating role of plant architecture for aspects of plant form.     

The effects of leaf size,leaf habit,and leaf form on leaf/stem relationships in plant twigs of temperate woody species

Question: Do thick‐twigged/large‐leaf species have an advantage in leaf display over their counterparts, and what are the effects of leaf habit and leaf form on the leaf‐stem relationship in plant twigs of temperature broadleaf woody species? Location: Gongga Mountain, southwest China. Methods: (1) We investigated stem cross‐sectional area and stem mass, leaf area and leaf/lamina mass of plant twigs (terminal branches of current‐year shoots) of 89 species belonging to 55 genera in 31 families. (2) Data were analyzed to determine leaf‐stem scaling relationships using both the Model type II regression method and the phylogenetically independent comparative (PIC) method. Results: (1) Significant, positive allometric relationships were found between twig cross‐sectional area and total leaf area supported by the twig, and between the cross‐sectional area and individual leaf area, suggesting that species with large leaves and thick twigs could support a disproportionately greater leaf area for a given twig cross‐sectional area. (2) However, the scaling relationships between twig stem mass and total leaf area and between stem mass and total lamina mass were approximately isometric, which indicates that the efficiency of deploying leaf area and lamina mass was independent of leaf size and twig size. The results of PIC were consistent with these correlations. (3) The evergreen species were usually smaller in total leaf area for a given twig stem investment in terms of both cross‐sectional area and stem mass, compared to deciduous species. Leaf mass per area (LMA) was negatively associated with the stem efficiency in deploying leaf area. (4) Compound leaf species could usually support a larger leaf area for a given twig stem mass and were usually larger in both leaf size and twig size than simple leaf species. Conclusions: Generally, thick‐twigged/large‐leaf species do not have an advantage over their counterparts in deploying photosynthetic compartments for a given twig stem investment. Leaf habit and leaf form types can modify leaf‐stem scaling relationships, possibly because of contrasting leaf properties. The leaf size‐twig size spectrum is related to the LMA‐leaf life span dimension of plant life history strategies.     

从水力结构的角度解释生长温度对烟草叶片形状和大小的影响（英文）

叶片形状和大小在不同的生长温度下变化非常大,但少有从水力结构的角度解释其变化原因的研究。本研究测定了生长于两个不同温度下(24℃/18℃昼/夜;32℃/26℃昼/夜)的烟草叶片的解剖结构,导水率,叶片长宽比和叶面积。生长在24℃/18℃下的烟草叶片与生长在32℃/18℃的叶片相比,更狭窄,并有更小的叶柄导管直径,更低的叶脉密度和导水率。然而,在不同的生长温度下,烟草叶面积并没有显著差异。叶片导水率与叶脉密度呈正相关,但与叶片长宽比呈负相关。结果表明在不同的生长温度下叶片解剖结构和叶片导水率可能对于改变叶宽比起着重要作用。     

Habitat structure, trophic structure and ecosystem function: interactive effects in a bromeliad–insect community

Although previous studies have shown that ecosystem functions are affected by either trophic structure or habitat structure, there has been little consideration of their combined effects. Such interactions may be particularly important in systems where habitat and trophic structure covary. I use the aquatic insects in bromeliads to examine the combined effects of trophic structure and habitat structure on a key ecosystem function: detrital processing. In Costa Rican bromeliads, trophic structure naturally covaries with both habitat complexity and habitat size, precluding any observational analysis of interactions between factors. I therefore designed mesocosms that allowed each factor to be manipulated separately. Increases in mesocosm complexity reduced predator (damselfly larva) efficiency, resulting in high detritivore abundances, indirectly increasing detrital processing rates. However, increased complexity also directly reduced the per capita foraging efficiency of the detritivores. Over short time periods, these trends effectively cancelled each other out in terms of detrital processing. Over longer time periods, more complex patterns emerged. Increases in mesocosm size also reduced both predator efficiency and detritivore efficiency, leading to no net effect on detrital processing. In many systems, ecosystem functions may be impacted by strong interactions between trophic structure and habitat structure, cautioning against examining either effect in isolation.     

Leaf trait co‐variation,response and effect in a chronosequence

Question: Is there any generality in terms of leaf trait correlations and the multiple role of leaf traits (response to and/or effect on) during secondary succession? Location: A secondary successional sere was sampled at four different ages since abandonment from several years to nearly 150 years on the Loess Plateau of northwestern China. Method: Specific leaf area (SLA), leaf mass per area (LMA), leaf nitrogen (N_mass, N_area), leaf phosphorus (P_mass, P_area) and leaf dry matter content (LDMC) were measured for all species recorded in the successional sere. Above‐ground net primary productivity (ANPP) and specific rate of litter mass loss (SRLML) were measured as surrogates for ecosystem properties. Soil total carbon (C) and nitrogen (N) were measured in each stage. Leaf traits were related to ecosystem properties and soil nutrient gradients, respectively. Results: LMA is correlated with N_area and P_area' and negatively with N_mass. Correlation between N_area and P_area was higher than between N_mass and P_mass. At the community level, field age, community hierarchy and their interaction explain 64.4 ‐ 93.5% of the variation in leaf traits. At the species level, field age explains 22.4 ‐ 45.5% of the variation in leaf traits (excl. P_area) while plant functional group has a significant effect only for N_mass. LDMC is correlated with ANPP and negatively with SRLML; P_mass is correlated with SRLML. Conclusions: Mean values of LMA, N_mass and N_area are close to the worldwide means, suggesting that large‐scale climate has a profound effect on leaf mass and leaf nitrogen allocation, while environmental gradients represented by succession have little influence on leaf‐trait values. Correlations between leaf traits, such as LMA‐N_area, LMA‐P_area and LMA‐N_mass shown in previous studies, are confirmed here. Although none of the leaf traits is proved to be both a response trait and an effect trait independent of time scale and community hierarchy, mass‐based leaf N is likely a sensitive response trait to soil C and N gradients. In addition, LDMC can be a marker for ANPP and SRLML, while mass‐based leaf P can be a marker for SRLML.     

蒙古栎群落叶型的分析

1　引　　言叶的大小即叶型是群落的重要外貌特征之一 ,与群落的生产率有关 ;叶子的形态与气候有密切的关系 ,一个扩展着的叶片所能达到的最大程度 ,受温度和湿度有效性的影响 ;大的叶片经常地出现于热带温暖而潮湿的气候中 ,而小的叶片则是十分干燥和寒冷地区植物的特征[1] .对于不同的群落的叶型分析 ,前人已做了一些研究 ,如对常绿阔叶林和季雨林的群落学研究[2～ 4 ] 及北方落叶林的群落学研究[5,6] .另一个与气候相关联的叶子特征是叶子的叶缘 ,有关方面的研究资料比较少 .据研究 ,在一个植物区系的双子叶乔木中全缘叶植物种的百分率 ,…     

Effects of pond size and consequent predator density on two species of tadpoles

Experimental ponds were used as a model system of habitat patches to study the effect of habitat size on the relative growth performance of tadpoles of Bufo americanus and Pseudacris triseriata, and on colonization by predatory insects. Three pond depths and surface areas were habitat size treatments in a replicated, factorial experiment. Tadpoles of both species were astablished together at a single density and ponds were left open to natural colonization by aquatic insects. Pond area had a significant effect on the multivariate response of P. triseriata larval period, survival, and metamorphic mass. P. triseriata survived better relative to B. americanus in larger ponds. However, increasing pond area led to greater incidence of predacious beetle larvae (Dytiscus, Coleoptera: Dytiscidae). Dytiscus larvae had a significant negative effect on the survival of P. triseriata and led to reduced P. triseriata survival relative to B. americanus in colonized ponds. The results suggest that habitat size can influence community structure by altering the distribution of predation among habitat patches.     

Leaf dynamics and shoot phenology of eleven warm-temperate evergreen broad-leaved trees near their northern limit in central Japan

Dynamic features of shoot phenology including leaf emergence and leaf fall, and leaf life span for eleven evergreen broad-leaved tree species were investigated in a warm-temperate rain forest in Mount Kiyosumi, central Japan. All species had periodic leaf emergence or flushing pattern, and were classified into two types; single and multiple flush and only one species, Eurya japonica, represented the latter type and the rest had single flush in spring. The single flush type can further be subdivided into two groups according to their duration of shoot growth; short and long flush. Seasonal patterns of leaf fall were categorized into four; unimodal, bimodal, broad unimodal, and multimodal type though they were not fixed pattern. The leaf emergence and leaf fall patterns were correlated for the eleven species, and five phenological types were categorized. Four of them were the single flush types, i.e., short flush of leaf emergence with unimodal leaffall (SSU) type of Castanopsis sieboldii and Quercus salicina, short flush with bimodal leaf fall (SSB) type of Quercus acuta, Machilus thunbergii, Neolitsea sericea, and Cinnamomum japonicum, long flush with bimodal leaffall (SLB) type of Myrsine seguinii, and long flush with broad unimodal leaffall (SLR) type of Symplocos prunifolia, Cleyera japonica, and Illicium anisatum. The multiple flush type is only one species, Eurya japonica, and it had multimodal leaffall pattern (MM type). The phenological pattern varied in relation to leaf life span, leaf size, and tree habit. Leaf life span ranged from 1.1 to 5.8 yr. The short flush species or SSU and SSB types were all canopy or subcanopy trees, and the former had short and the latter had long leaf life spans. The long flush species were all microphyllous small trees, and SLB type had a relatively long leaf life span in understory, SLR type had a long leaf life span in understory or in open habitat and/or forest gap as a pioneer tree. MM type had a long leaf life span and colonizing species in open habitat but they can survive in understory as well. The phenological attributes of evergreen trees were well corresponded to the ecological guild of the tree in both forest structure and successional stage, and were also constrained by phylogenetic groups.     

Plant growth-form alters the relationship between foliar morphology and species shade-tolerance ranking in temperate woody taxa

Variation in leaf size (area per leaf) and leaf dry weight per area (LWA) in relation to species shade- and drought-tolerance, characterised by Ellenberg's light (ELD) and water demand (EWD) values, respectively, were examined in 60 temperate woody taxa at constant relative irradiance. LWA was independent of plant size, but leaf size increased with total plant height at constant ELD. Canopy position also affected leaf morphology: leaves from the upper crown third had higher LWA and were larger than leaves from the lower third. Leaf size and LWA were negatively correlated, and leaf size decreased and LWA increased with decreasing species shade-tolerance. Mean LWA was similar for trees and shrubs, but trees had larger leaves than shrubs. Furthermore, all relationships were altered by plant growth-form: none of the qualitative tendencies was significant for trees. This implies the considerably lower plasticity of foliar parameters in trees than those in shrubs. Accordingly, shade-tolerance of trees, having relatively constant leaf structure, may be most affected by the variability in biomass partitioning and crown geometry which influence foliage distribution and spacing and finally determine canopy light absorptance. Alteration of leaf form and investment pattern for construction of unit foliar surface area which change the efficiency of light interception per unit biomass investment in leaves, is a competitive strategy inherent to shrubs. EWD as well as wood anatomy did not control LWA and leaf size, though there was a trend of ring-porous tree species to be more shade-tolerant than diffuse-porous trees. Since ring-porous species are more vulnerable to cavitation than diffuse-porous species, they may be constrained to environments where irradiances and consequently evaporative demand is lower.     

南疆地区巴仁杏果园土壤矿质元素与叶片营养水平分析

为探讨果园土壤矿质元素与叶片营养水平的相互关系,于巴仁杏盛果期采集南疆阿克陶县巴仁杏果园的壤土和沙土土样,以及植株叶片,测定土壤与叶片矿质元素质量分数,结果表明,该地区2种类型土壤巴仁杏果园叶片全氮质量分数均偏低,沙土果园叶片镁质量分数偏低。不同土壤类型果园叶片全氮、钾、钙、镁质量分数差异显著。该地区果园土壤养分整体较低,其中,土壤有机质、氮、磷、锌、锰、铜质量分数均偏低,此外不同矿质元素在不同土层的分布规律因土壤类型不同而各有差异。2种土壤类型果园的土壤有机质质量分数与叶片磷质量分数均呈正相关,叶片其他矿质元素质量分数与土壤有机质质量分数的相关性与土壤类型有关;土壤矿质元素与叶片相应矿物质元素的相关性因土壤质地不同而存在较大差异。因此,在巴仁杏果园的土肥管理中应增施有机质及氮、磷、锌、锰、铜、硼等矿质元素。此外,对不同土壤类型果园的管理应根据果园土壤类型进行针对性施肥。     

Plant diversity patterns in semi-natural grasslands along an elevational gradient in southern Norway

Leaf demography and growth of six common, co-occurring woody plant species that varied in stature (tree vs. shrub) and leaf texture (sclerophyllous, coriaceous, malacophyllous) were examined in a subtropical savanna parkland in southern Texas, USA. We tested the hypotheses that, (a) leaves of plants with evergreen canopies would have longer life spans than those of deciduous species; (b) supplementation of soil moisture would decrease leaf life span in both evergreen and deciduous species; (c) species responses to increased soil moisture availability would be inversely related to leaf longevity; and (d) deciduous growth forms would exhibit a greater growth response to increased soil moisture availability than their evergreen counterparts.A variety of seasonal leaf habits (evergreen, winter-deciduous and summer-deciduous canopies) and leaf life spans (median = 66 to 283 days) were represented by the targeted species, but there was no clear relationship between seasonal leaf habit and leaf longevity. Among species with evergreen canopies, median leaf longevity ranged from short (Zanthoxylum fagara = 116 days; Condalia hookeri = 158 days) to long (Berberis trifoliolata = 283 days) but did not exceed 1 yr. In fact, leaf longevity in evergreen shrubs was often comparable to, or shorter than, that of species with deciduous canopies (Ziziphus obtusifolia = 66 days; Diospyros texana = 119 days; Prosopis glandulosa = 207 days). Augmentation of surface soil moisture had no detectable effect on median leaf life span in any species and there was no clear relationship between leaf longevity and species growth responses to irrigation. Contrary to expectations, species with evergreen canopies responded to irrigation by producing more leaf biomass, longer shoots and more leaf cohorts/year than did deciduous species.Species differences in the annual cycle of leaf initiation, leaf longevity and canopy development, combined with contrasts in root distributions and a highly variable climate, may allow for spatial and temporal partitioning of resources and hence, woody species coexistence and diversity in this system. However, the lack of expected relationships between leaf longevity, leaf habit and plant responses to resource enhancement suggests that structure-function relationships and functional groupings developed in strongly seasonal environments cannot be applied with confidence to these subtropical savannas and thorn woodlands.     

Behavioral constraints for the spread of the eastern mosquitofish, Gambusia holbrooki (Poeciliidae)

Eastern mosquitofish (Gambusia holbrooki) are native to the southeastern United States but notoriously invasive elsewhere, and are aggressive predators in ecosystems they inhabit. Information on dispersal behavior is needed to better understand mosquitofish spread upon introduction and potential means to mitigate that spread. We experimentally tested the effects of shallow water depths (3–24 mm) and obstacles (leaf litter) on mosquitofish dispersal behavior, plus a range of conditions relevant to field situations. Mosquitofish dispersed significantly faster in deeper water (p < 0.001) but some dispersed in only 3 mm water depth (i.e., one-half average body depth). Wetland and upland leaf litter at natural densities strongly interfered with mosquitofish dispersal behavior. Based on our results, introduced mosquitofish spread rapidly given unimpeded dispersal corridors (e.g., mowed ditches), and may do so at rates >800 m/day. Also, consistent lack of sexual dimorphism in dispersal behavior indicates that mosquitofish spread is not strongly dependent on female poeciliid reproductive biology. Our results support designation of mosquitofish as highly invasive and suggest that barriers to mosquitofish spread must obstruct dispersal pathways as shallow as 3 mm depth.     

The process of population increase and patterns of resource utilization of two spider mites,Oligonychus ununguis (Jacobi) andPanonychus citri (McGregor), under experimental conditions (Acari: Tetranychidae)

Summary Pattern of population growth and characteristics of habitat utilization and of migration by two species of spider mites were studied under experimental conditions. The population growth ofOligonychus ununguis (Jacobi) on a chestnut occurred only on a single mite-release leaf over a long period, and few individuals moved away. Most of the 2nd progeny generation females of this species emigrated from the mite-release leaf as well as the sapling by means of ballooning threads. During this growth period, population density on the mite-release leaf levelled off, whereas that on the sapling increased. In contrast, the foundress ofPanonychus citri (McGregor) on citrus actively moved over several neighbouring leaves, and until the 2nd progeny generation females emerged, individuals were distributed over all the sapling leaves by means of walking. Emigration from the sapling was not observed until the 2nd progeny females emerged, and after that the mites emigrated by means of ballooning threads. The population density ofP. citri on the sapling levelled off and was rather decreased on the mite-release leaf at the time of mite emigration. Comparing the changing pattern of the relative degree of aggregation (m ^*/m) measured in two different units between these two species, the pattern ofm ^*/m in 1 cm² on the mite release leaf inO. ununguis resembled that of the unit of leaves on the sapling inP. citri. This result as well as behavioural observations indicate that migration ofO. ununguis is the movement from leaf to leaf and that ofP. citri from sapling to sapling. It is, therefore, concluded that the boundary of the microhabitat is a single leaf forO. ununguis but sapling or foliage forP. citri. This work was presented in Annual Meeting of Jap. Soc. Appl. Ent. Zool., 1981 in Okayama.     

Influence of fine-scale habitat structure on nest-site occupancy,laying date and clutch size in Blue Tits Cyanistes caeruleus

Most birds have specific habitat requirements for breeding. The vegetation structure surrounding nest-sites is an important component of habitat quality, and can have large effects on avian breeding performance. We studied 13 years of Blue Tit Cyanistes caeruleus population data to determine whether characteristics of vegetation structure predict site occupancy, laying date and number of eggs laid. Measurements of vegetation structure included the density of English Oak Quercus robur, European Beech Fagus sylvatica, and other deciduous, coniferous and non-coniferous evergreen trees, within a 20-m radius of nest-boxes used for breeding. Trees were further sub-divided into specific classes of trunk circumferences to determine the densities for different maturity levels. Based on Principal Component Analysis (PCA), we reduced the total number of 17 measured vegetation variables to 7 main categories, which we used for further analyses. We found that the occupancy rate of sites and the number of eggs laid correlated positively with the proportion of deciduous trees and negatively with the density of coniferous trees. Laying of the first egg was advanced with a greater proportion of deciduous trees. Among deciduous trees, the English Oak appeared to be most important, as a higher density of more mature English Oak trees was associated with more frequent nest-box occupancy, a larger number of eggs laid, and an earlier laying start. Furthermore, laying started earlier and more eggs were laid in nest-boxes with higher occupancy rates. Together, these findings highlight the role of deciduous trees, particularly more mature English Oak, as important predictors of high-quality preferred habitat. These results aid in defining habitat quality and will facilitate future studies on the importance of environmental quality for breeding performance.     

青岛崂山茶区土壤生境因子对应分析

通过对崂山茶区土壤生境因子与不同茶园进行对应分析研究,发现可以按照崂山茶区土壤特征将不同的茶园分成三类:富磷区、富有机质区、老茶园区.从而为崂山茶区不同种类茶树种植以及如何改良土壤促进茶树生长提供依据.     

Resource tracking in aphids: programmed reproductive strategies anticipate seasonal trends in habitat quality

Most aphids show cyclical parthenogenesis, have short generation times and complete several generations each season. Two hypotheses have been proposed to account for the adaptive seasonal trends observed in the reproductive strategies of aphids. Firstly, individuals of each generation modify their reproductive strategy in direct response to the conditions they experience during their development. Secondly, the reproductive strategies of the different generations are to a large extent programmed and anticipate seasonal trends in habitat quality. These hypotheses were tested by rearing individuals of three generations of the host-alternating willow-carrot aphid, Cavariella aegopodii, on both willow and carrot. This revealed that the way this aphid allocates resources to gonads and lipid reserves is independent of an aphid's weight and the host plant on which it is reared. In addition each generation shows a specific relationship between offspring size and adult size, which tends to keep the absolute investment in individual offspring relatively constant from generation to generation, inspite of the big differences in adult size between generations. That is, through programmed allometric engineering aphids anticipate the predictable seasonal trends in habitat quality and so more closely track their resources, investing relatively more in gonads when food quality is high and relatively more in lipoidal reserves when food quality is poor. Received: 25 August 1997 / Accepted: 19 October 1997     

Connectivity compensates for low habitat quality and small patch size in the butterfly <Emphasis Type="Italic">Cupido minimus</Emphasis>

Habitat size, habitat isolation and habitat quality are regarded as the main determinants of butterfly occurrence in fragmented landscapes. To analyze the relationship between the occurrence of the butterfly Cupido minimus and these factors, patch occupancy of the immature stages in patches of its host plant Anthyllis vulneraria was studied in the nature reserve Hohe Wann in Bavaria (Germany). In 2001 and 2002, 82 A. vulneraria patches were surveyed for the presence of C. minimus larvae. The occurrence was largely affected by the size of the food plant patches. In a habitat model that uses multiple logistic regression, the type of management and habitat connectivity are further determinants of species distribution. Internal and temporal validation demonstrate the stability and robustness of the developed habitat models. Additionally, it was proved that the colonization rate of C. minimus was significantly influenced by the distance to the next occupied Anthyllis patch. Concerning long-term survival of (meta-) populations in fragmented landscapes, the results show that lower habitat quality may be compensated by higher connectivity between host plant patches. Electronic supplementary material The online version of this article (doi:) contains supplementary material, which is available to authorized users.     

Characteristics of “Number of Veins” to Estimate Leaf Maturity in Pteris mutilata (Pteridaceae)

Pteris mutilata , as has been established in ferns of cool temperate regions. NV expressed developmental stages better than any other leaf size parameters, such as blade length, blade width, stipe length, and total length (blade+stipe length). The leaf shape became more oblong and/or slender after the plant matured, which could be measured by two shape parameters, (blade width)/(total length) and (blade width)/(blade length). Principal component analysis using all the morphological parameters showed that NV is categorized into size parameters, although NV has been considered to differ somehow from the other size parameters. Thus NV represents one of the size parameters that is the most appropriate to estimate leaf maturation. Received 6 July 2000/ Accepted in revised form 22 August 2000