Direct Gradient Analysis Research Articles

Ecological Consideration for Restoration of the Degraded Urban River

Vegetation and riverine structures were compared among urban and several semi-natural rivers by applying direct gradient analysis and ordination method. Urban rivers showed different species composition from the semi-natural ones. Species composition of semi-natural rivers depended on the geographical positions, such as the upstream, midstream, and downstream and on micro-topographies, such as the waterside, flooding bed, and embankment. Semi-natural rivers showed gentle change in micro-topography, whereas urban one did not so. Our restoration plan to improve the ecological quality of the degraded urban river by imitating semi-natural river was prepared based on those data.

Predictive mapping of forest composition and structure with direct gradient analysis and nearest- neighbor imputation in coastal Oregon, U.S.A.

Spatially explicit information on the species composition and structure of forest vegetation is needed at broad spatial scales for natural resource policy analysis and ecological research. We present a method for predictive vegetation mapping that applies direct gradient analysis and nearest-neighbor imputation to ascribe detailed ground attributes of vegetation to each pixel in a digital landscape map. The gradient nearest neighbor method integrates vegetation measurements from regional grids of field plots, mapped environmental data, and Landsat Thematic Mapper (TM) imagery. In the Oregon coastal province, species gradients were most strongly associated with regional climate and geographic location, whereas variation in forest structure was best explained by Landsat TM variables. At the regional level, mapped predictions represented the range of variability in the sample data, and predicted area by vegetation type closely matched sample-based estimates. At the site level, mapped predictions maintained the covariance structure among multiple response variables. Prediction accuracy for tree species occurrence and several measures of vegetation structure and composition was good to moderate. Vegetation maps produced with the gradient nearest neighbor method are appropriately used for regional-level planning, policy analysis, and research, not to guide local management decisions.

A vegetation analysis of the pastoral landscapes of upland Wales, UK

Abstract. The upland moorlands of Wales are situated on the oceanic fringe of western Europe, and have experienced a long history of pastoral management. Recent vegetation data are analysed to assess the relative contribution of abiotic and anthropogenic factors to variation in habitat composition among the major upland ranges of this region. From a numerical analysis of plant community cover data, recorded from 65 sites covering 260 000 ha, a six‐cluster site classification emerged with striking biogeographical coherence. Direct gradient analysis and variance partitioning revealed strong correlation between vegetation composition and spatially‐structured climatic gradients, in particular temperature, rainfall and oceanicity; differences in bedrock geology appear to have a lesser role. The analysis also indicates a close correlation between habitat variation and anthropogenic parameters, especially grazing intensity, burning frequency, and sulphur and nitrogen deposition levels. At this regional scale, anthropogenic impacts appear to have accentuated, rather than obscured, vegetation patterns which are primarily determined by climate and other abiotic variables. The findings have considerable relevance for conservation planning and also for predictive studies on the consequences of climatic change for the biota of the uplands of southern Britain.

Niche differentiation and distribution of Carex curvula along a bioclimatic gradient in the southwestern Alps

Abstract. Carex curvula is a dominant sedge of European alpine tundra, exhibiting two morphological forms: C. curvula ssp. curvula (Ccc) and C. curvula ssp. rosae (Ccr). In this paper, we attempt to explore whether Ccc and Ccr are ecotypes or vicariant forms and whether between‐ and within‐regional distribution patterns can be explained by variations in the amount of available habitats and/or by changes in niche attributes. The study area covered three bioclimatically distinct regions of the southwestern Alps in which local abundances of Ccc and Ccr strongly differ.The realized niche of both forms was investigated by a direct gradient analysis performed on an extensive floristic and environmental data set. We found no evidence of niche overlap between Ccc and Ccr as their distribution curves differ strongly along disturbance, mesotopography and soil acidity gradients. We investigated the effect of region on the structure of local scale variables tables. Highly concordant patterns among regions were found, except that optimal habitat conditions for Ccc were infrequent in the southernmost regions. We compared the ecological behaviour of range centre and range margin populations of each form. We found a narrower ecological amplitude for Ccr in the northern part of its range, whereas for Ccc the niche breadth of range‐margin populations was not reduced compared to range centre populations. At its southern range limit, Ccc mostly occupies ecologically marginal habitats.We conclude that Ccc and Ccr represent ecotypes, not vicariant forms. Finally, we suggest that habitat availability, resulting from the interplay of regional‐scale and local‐scale variables, satisfactorily explains the distributional patterns of Carex curvula ecotypes in the southwestern Alps.

A vegetation analysis of the pastoral landscapes of upland Wales, UK

Abstract The upland moorlands of Wales are situated on the oceanic fringe of western Europe, and have experienced a long history of pastoral management. Recent vegetation data are analysed to assess the relative contribution of abiotic and anthropogenic factors to variation in habitat composition among the major upland ranges of this region. From a numerical analysis of plant community cover data, recorded from 65 sites covering 260000 ha, a six-cluster site classification emerged with striking biogeographical coherence. Direct gradient analysis and variance partitioning revealed strong correlation between vegetation composition and spatially-structured climatic gradients, in particular temperature, rainfall and oceanicity; differences in bedrock geology appear to have a lesser role. The analysis also indicates a close correlation between habitat variation and anthropogenic parameters, especially grazing intensity, burning frequency, and sulphur and nitrogen deposition levels. At this regional scale, anth...

Niche differentiation and distribution of Carex curvula along a bioclimatic gradient in the southwestern Alps

Abstract Carex curvula is a dominant sedge of European alpine tundra, exhibiting two morphological forms: C. curvula ssp. curvula (Ccc) and C. curvula ssp. rosae (Ccr). In this paper, we attempt to explore whether Ccc and Ccr are ecotypes or vicariant forms and whether between- and within-regional distribution patterns can be explained by variations in the amount of available habitats and/or by changes in niche attributes. The study area covered three bioclimatically distinct regions of the southwestern Alps in which local abundances of Ccc and Ccr strongly differ. The realized niche of both forms was investigated by a direct gradient analysis performed on an extensive floristic and environmental data set. We found no evidence of niche overlap between Ccc and Ccr as their distribution curves differ strongly along disturbance, mesotopography and soil acidity gradients. We investigated the effect of region on the structure of local scale variables tables. Highly concordant patterns among regions were found,...

The impact of logging intensity on field-layer vegetation in Swedish boreal forests

The relationship between logging intensity and changes in ground cover vegetation was studied in 16 species and groups of species recorded at 10- or 11-year intervals in mature conifer-dominated forests. The 789 plots located in northern and central Sweden had been surveyed by the National Forest Inventory and the National Survey of Forest Soil and Vegetation. Thirty-seven percent of the plots had been subjected to a thinning or clear-cutting between the inventories. A principal components analysis showed that, of the variables considered, logging intensity had the highest explanatory power regarding change in ground cover vegetation between the inventories (the other variables were sum of temperatures, age of stand, timber volume, percentage Pinus sylvestris and site productivity). A multivariate direct gradient analysis technique (Redundancy analysis) showed that the logging intensity significantly affected the change in cover. This analysis also ranked the species in their responsiveness to logging. Epilobium angustifolium, narrow-leaved grasses and broad-leaved grasses, increased most with logging intensity. The response was not linear and only detectable at high logging intensities (>80%). In contrast, Vaccinium myrtillus seemed to decrease linearly with increased logging intensity. There was several years time-lag in the response to logging of E. angustifolium, V. myrtillus and narrow-leaved grasses. Several species and groups of species seemed unaffected by the logging. In sample plots unaffected by logging the cover of most species decreased.

An analysis of plant species distributions on the floodplain of the Okavango River, Namibia, with respect to impacts of possible water abstraction

The proposed abstraction of water from the Okavango River in Namibia could potentially result in significant changes in the vegetation of floodplains downstream of the abstraction point. Direct gradient analysis was used to determine the distribution of the most common floodplain plant species in relation to elevation above the water surface and distance from the channel, and therefore to the depth and duration of flooding. Species could be ranked and scored according to their distribution in relation to both elevation and distance from the channel, reflecting a gradient from species intolerant of flooding to those that withstand prolonged flooding. In general, as distance from the channel increased, the elevation at which each of the species occurred decreased, but there was a great deal of variation in the range of elevations over which each species occurred, primarily as a consequence of the irregular topography of these floodplains. Furthermore, several ruderal species were restricted to elevated sites in close proximity to the channel, occurring on recently formed point bars which are the product of fluvial processes. Disturbance in the form of sediment deposition on point bars is thus an important determinant of species distribution on floodplains of the Okavango River. Given these patterns of species distributions it is difficult to predict with accuracy the effect of abstraction on vegetation distribution. A preliminary analysis of possible changes in the hydrological regime after abstraction has also been attempted as part of this study. This was achieved by comparing mean stage heights before and after abstraction assuming a constant rate of abstraction of 3 cumecs, and based on historical daily records from Rundu and Mukwe. Uniform abstraction would have a greater impact at low flows than at medium and high flows. Should abstraction proceed, it would therefore be more appropriate to remove water during periods of high discharge. However, since the Okavango River in Namibia recharges local and possibly regional groundwater, its discharge declines downstream. Rates of water loss from the river thus need to be determined prior to any decision being made about whether or not abstraction should take place, or where to locate the point of abstraction.

The effects of varied grazing management on epigeal spiders, harvestmen and pseudoscorpions of Nardus stricta grassland in upland Scotland

A hypothesis that epigeal arachnid assemblages benefit more from greater vegetation structure than botanical species composition in upland grasslands was tested. The test was carried out within a grazing experiment, initiated in 1991, to investigate vegetation dynamics in response to stocking with mixed livestock at varied rates. The experimental treatments comprised: no livestock, sheep only or sheep with cattle. Livestock treatments were grazed to maintain either 4.5 or 6.5 cm average sward heights between tussocks. Two replicates of each treatment were used and allocated to 10 plots across 22 ha of Nardus stricta-dominated grassland. The effects on epigeal arachnids (excluding acarines) of the botanical and structural differences of the grassland between treatments during April–October 1993 and 1994 were assessed. Epigeal arachnid species composition was estimated using continuous pitfall trapping and the densities of mainly money spiders (Araneae: Linyphiidae) were estimated from monthly suction sampling and visual counts of spider webs in micro-habitats. These data were later compared with stocking rate, botanical species composition and vegetation structure. Forty of the 84 sampled species occurred in all experimental treatments. There was a significant effect of treatment on the number of arachnid species in suction but not in pitfall samples. There was also a significant effect of treatment on the relative abundance of 26% of these arachnid species. For most species of spider, harvestmen and pseudoscorpion, abundance was greater in the ungrazed and taller, grazed swards although a few species were captured in greater numbers in the treatments with shorter swards. Botanical composition, mean vegetation height and grazing intensity accounted for 48.5–53.2% of the variability in the species composition/relative abundance of these arachnids, calculated by direct gradient analysis. Almost half of the species were randomly distributed across the experimental treatments and are recorded as widespread in upland heathland or grassland habitats and lowland grassland. More spider webs were counted during July–September 1993–1994, with greater numbers (dominated by the linyphiid species, Lepthyphantes mengii) counted in tall, ungrazed swards compared with taller grazed swards created by sheep alone or sheep with cattle. In the treatments with fewer webs, these were occupied by more linyphiid species. Suction sampling detected greater diversity of arachnids in the ungrazed N. stricta. This was related to increased plant litter below the leaf stratum where webs were counted. Vegetation structure and not botanical species composition within the N. stricta plant community determined arachnid species composition and abundance. Furthermore, no single grazing treatment supported the total number of arachnid species represented across the entire grazing experiment. It is concluded that varied grazing management, including some temporary ungrazed areas, is necessary to maintain the structural variability of grassland patches so as to maintain a spatial mosaic that favours the optimum arachnid fauna of upland grasslands.

Primary succession in post-mining landscapes of Lower Lusatia — chance or necessity

Based on a comprehensive 4-year survey the spontaneous vegetation of the Lusatian post-mining landscape is described. Thirteen vegetation types are distinguished and characterized with respect to characteristic species, dominant growth form, and frequency of occurrence. Indirect gradient analysis showed that there is a broad overlap of vegetation types, and that no successional trends can be inferred on the basis of floristic composition alone. The species composition was related to environmental variables by means of direct gradient analysis and canonical correspondence analysis. The multivariate approach showed that variables like pH, organic carbon, phosphate and water capacities have an obvious influence on species composition. However, in some cases it is not clear whether this is cause or consequence of the respective vegetation colonization. The age of the stands can be identified as an important variable, but cannot be distinguished from spatial autocorrelation effects. The univariate approach showed that no threshold values can be detected in important parameters, such as pH. Autocorrelation effects hamper the interpretation of the complex data set. All successional sequences observed or inferred so far are combined to a conceptual model of spontaneous vegetation development. According to the local circumstances, various types of initial vegetation are possible. Initial vegetation is subsequently followed by a large number of stages and types which form a complex network. Some of these stages are stabilized by a dense vegetation layer or competitive power of the dominant species. The direction of succession is influenced to varying portions by neighborhood effects, facilitation, disturbance, and other mechanisms. The duration of the stages and the transition probabilities cannot be exactly predicted in the present state of knowledge. Besides its scientific importance, primary succession is regarded as an adequate means of restoration leading to areas likewise interesting for nature conservation and considerate tourism.

Size structure of benthic freshwater fish communities in relation to environmental gradients

The benthic fish communities of 161 Swedish lakes, sampled with multi‐mesh gillnets, were examined by direct gradient analyses with ten environmental factors. A geographic gradient, correlated with climatic factors and productivity, was most important for ordination of species optima. By contrast, the distribution of biomass in different size‐classes and trophic groups, was influenced more directly by local factors such as water quality (pH) and lake morphometry (area and maximum depth). The communities are not only structured at the species level, but also by size‐related allocation of resources, within and between coexisting species. This national survey confirmed patterns observed in previous local field studies, as well as experiments designed for testing mechanisms acting on size‐structured populations. However, it highlighted the need for more extensive studies on if and how large‐scale environmental variation affects the cues for size‐related ontogenetic niche shifts in facultative piscivores, because species such as perch Perca fluviatilis, brown trout Salmo trutta, and Arctic charr Salvelinus alpinus often dominate the benthic fish community.

Restoration of an oak forest in east-central Missouri: Early effects of prescribed burning on woody vegetation

We investigated early effects of prescribed burning on the structure and species composition of woody plants in Cuivre River State Park (2750 ha), Lincoln County, Missouri, USA, where a program of prescribed burning in forest understory was instituted to return the vegetation to conditions similar to those found at the time of European settlement. Prescribed burning in forest was initiated in 1989 and approximately 17% of the park is now burned every other year. We compared woody vegetation in 34 0.05 ha plots in burned sections of the park and in 34 0.05 ha plots in unburned sections of the park. Burned areas had experienced two to four burns at the time vegetation was sampled in 1996. Burned areas had significantly fewer small trees, fewer live woody shrubs, and less foliage in the understory (1.5–10 m) than did unburned areas. Small individuals of several species (e.g., Ulmus americana) were less abundant on burned areas and, overall, similarity of sample plots (based on basal areas of small trees) was significantly greater within burned areas than between burned and unburned areas. There were no differences in overstory structure or species composition that could be attributed to burning. Direct gradient analyses indicated that environmental factors (e.g., slope, elevation, aspect) accounted for some of the variation in species composition among plots but also indicated that burned and unburned plots did not differ consistently with respect to environmental factors. Thus, observed differences in vegetation structure and composition between burned and unburned areas of the park are due to the burning program and not to pre-existing differences in environmental factors (e.g., topography). Results to date reflect relatively short-term (<10 years) responses to burning. It is expected that changes in canopy composition and structure will occur over time.

Restoring Savanna Using Fire: Impact on the Breeding Bird Community

AbstractRestoration of many terrestrial plant communities involves the reintroduction of fire. However, there have been few studies of the effects of fire on the avifauna during the restoration process. To study the effects of oak savanna restoration on avian communities, breeding birds were censused and the vegetation structure documented in seven experimental burn units (8–18 ha) that had experienced different frequencies of controlled burns during the past 31 years (0–26 burns). Data were analyzed with both direct and indirect gradient analyses using multivariate techniques. The results showed that, as savanna restoration proceeded, there was a general decline in predominantly insectivorous species, particularly those that feed in the upper canopy region (leaves and air space), and a general increase in omnivorous species, particularly those that feed on the ground and in the lower canopy. Insectivorous bark gleaners (woodpeckers) also increased during restoration and were correlated with the increase in standing dead trees resulting from the fires. Overall, savanna restoration resulted in increases in the abundance of many open country bird species, including many species that have been declining in central and eastern North America, including red‐headed woodpecker, Baltimore oriole, eastern kingbird, vesper sparrow, field sparrow, lark sparrow, brown thrasher, American goldfinch, and brown‐headed cowbird. The shifts in species and guilds were correlated with changes in burn frequency and the macro vegetation structure—tree and shrub density, leaf area index, and relative proportion of standing dead trees. The findings show that savanna restoration can increase bird diversity and provide important habitat for uncommon or declining bird species. These birds are most likely attracted to one or more of the distinctive habitat features of the restored savanna environments, including scattered mature trees, standing dead trees and snags, and presence of both shrubby and grassland vegetation. The findings also suggest that restoration ecologists and wildlife biologists will need to work together to achieve desired goals, since different types of savanna restoration efforts may produce different effects on the breeding bird community.

Vegetation along hydrologic and edaphic gradients in a North Carolina coastal plain creek bottom and implications for restoration

We described the vegetation of two alluvial swamp forest stands along Durham Creek in Beaufort County, North Carolina, USA in relation to elevation, hydrologic, and edaphic gradients. Over 3,000 sur- veyed elevations of individual plant microsites were used in conjunction with 26 years of stream gage data to examine individual species responses to annual and growing season flooding frequencies. Direct gradient analyses combined with plot ordinations derived from detrended correspondence analysis and canonical correspondence analysis suggested that differences in vegetation between the stands were primarily the result of variations in elevation, growing season flooding frequency, percent base saturation, exchangeable acidity, and soil physical properties. Although the stands were less than 4.5 km apart and without significant inter- mediate tributaries, growing season flooding frequency and duration were magnified in the lowest elevations of the downstream stand. An elevation difference of as little as 10 cm resulted in a 20% difference in the frequency of surface flooding during the growing season. Species distributions were significantly correlated with depth to mottling (r 2 5 0.75), flooding frequency (r 2 52 0.57), elevation (r 2 5 0.70), and several soil chemical properties. The two stands had very similar annual surface flooding regimes, but subtle differences in growing season flooding frequency, soil characteristics, and disturbance history have apparently resulted in dissimilar plant community composition and structure. These results suggest that the lack of quantitative data on vegetation-environment interactions occurring at the microtopographic scale (10 21 m) in alluvial swamp forests makes precise prediction, planning, or design of created or restored wetland composition and function a formidable challenge.

Effect of AM symbiosis exclusion on grassland community composition

Experimental data on grassland community response to exclusion of arbuscular mycorrhizal (AM) symbiosis using fungicide benomyl and to phosphate addition were evaluated after four years of continued manipulation using multivariate direct gradient analysis. While the community responded only to phosphate application in the third year (phosphate application suppressing forbs in favour of graminoids), the data from the fourth season indicate relative suppression of some forbs (includingPlantago lanceolata, Cerastium holosteoides) with both phosphate and fungicide applications. Positive response of graminoids to phosphate addition seems to interact with fungicide application, while an important community constituent —Achillea millefolium — seems to profit primarily from fungicide applications. The effect of the fungicide on the AM symbiosis was confirmed on roots ofPlantago lanceolata. The direct evaluation of frequency of AM fungal structures in plant roots also revealed a diverse response of different plant species to phosphate application. Nevertheless, in all species (except non-mycorrhizalLuzula campestris), phosphate decreased the frequency of arbuscules and increased the frequency of non-infected root segments.

The influence of coniferous canopies on understorey vegetation and soils in mountain forests of the northern Calcareous Alps

Abstract. Compositional and edaphic gradients were studied in montane forests of the Bavarian Alps (Germany), in which natural mixed deciduous‐coniferous tree layers have been altered by past management in favour of Picea abies. Data on species composition and ecological factors were collected in a stratified random sample of 84 quadrats comprising a gradient from pure Picea to pure Fagus sylvatica stands. Data about the understorey composition were subjected to indirect (DCA) and direct gradient analysis (RDA) with the proportion of Picea in the canopy as a constraining variable. Three principal components of a matrix containing seven descriptors of mineral soil, relief and tree layer cover were included as covariables describing the variability of primary ecological factors. Gradients of organic topsoil morphology and chemistry were extracted correspondingly. Responses of individual species, species group and topsoil attributes were studied by simple and partial correlation analysis.Mosses were significantly more abundant and diverse under Picea stands. Few graminoid and herb species were partially associated with Picea, and total understorey richness and cover did not differ systematically by stand type. No relationship between tree layer and understorey diversity was detected at the studied scale. Juvenile Fagus sylvatica was the only woody species significantly less abundant under Picea. In the topsoil lower base saturation, lower pH and larger C/N‐ratios in the litter layer were partially attributable to the proportion of Picea, only for base saturation a relationship was detected in greater soil depth also. The frequency of broad humus form types did not differ by tree species, nor was overall depth of organic forest floor attributable to canopy composition.

Ecological classification of Nigerian mangroves using soil nutrient gradient analysis

Direct gradient analysis was used to relate the structure of mangrove communities to soil nutrient gradients. The predominant cations in the alluvial soils were magnesium and calcium, the values ranging from 8.6±0.9 to 24.6±2.0 me per 100 g. Organic carbon was high in the soils, ranging from 3.5% to 10.4%. All soil nutrients varied seasonally, in response to wet and dry periods of the climate. Correlation of species with soil nutrients indicated the nutrient status of the soils to be best defined in terms of calcium for the A stratum (species > 3 m tall), calcium/postassium for the B stratum (species 1–3 m tall) and magnesium for the C stratum (species < 1 m tall). Ecological group classification shows that Avicennia africanaand Nypa fruticansoccurred almost exclusively at the highest calcium values of 20.8 and 17.6 me per 100 g, respectively. Acrostichum aureumand Sesuvium portulacastrumare insensitive to the magnesium gradient. The analysis revealed an overlapping range of occurrences for most of the species although with varying ecological optima along the gradient.

Keeping things in order: multivariate direct gradient analysis of a strongly fluctuating benthic community

Two spatial surveys of the macrobenthos of an estuarine intertidal area, the Oosterschelde in the southwestern part of The Netherlands, were conducted in different years. In the period between the surveys changes in the hydrodynamics of the estuary took place as a result of the construction of a storm-surge barrier. The works reduced the tidal volume in the basin and as a consequence the altitude of the tidal flats decreased by approximately 0.1 to 0.2 m. Considerable differences in species densities were found between the two surveys, but these large differences could not be explained by changes in the `spatial' environmental variables (such as altitude) that were taken into account in a direct gradient analysis. Unknown and `only-to-time-related' (e.g. weather-related) factors had considerably influenced species abundance. This implies that the impact of the works on the macrobenthos could not be evaluated on the basis of observed species densities in the before- and after-the-works surveys. Generally, it implies that even if the levels of future `spatial' environmental variables are known, direct gradient analysis based on a single survey cannot be used for prediction of future densities in absolute terms in such strongly fluctuating communities. Yet, the gradient analyses for the two years revealed rather similar estimates of the major environmental gradients. The same was true for the parameters that described the relative change in species densities in response to a change in the environmental gradient levels. Hence the most appropriate way of evaluating the changes that have occurred in the benthic community as a result of the works was in terms of the observed changes in the `spatial' environmental gradient levels. The levels of the first environmental gradient, which was strongly related to the sediment structure, did not change between years. Levels of the second gradient, which could be interpreted as a tidal zone gradient, decreased. This suggests that the works particularly affected taxa that are strongly related to this second gradient, such as Corophium living in the upper tidal areas and Nephthys living in the lower tidal zone.

Data diving with cross‐validation: an investigation of broad‐scale gradients in Swedish weed communities

Summary 1 Multivariate analysis of complex data sets is plagued by problems of subjectivity and of finding statistically valid ways to test a large number of plausible hypotheses. We show how patterns in the data can be identified (data diving) as well as rigorously tested statistically by subdividing the data set. 2 We analysed data on weed biomass and environmental variables from more than 2000 plots in cereal and oil‐seed crops in Sweden during 1970–94. Half the data set was used in an exploratory phase while the other half was used in a subsequent confirmatory phase. 3 The exploratory analyses included multivariate statistics [detrended correspondence analysis (DCA) and canonical correspondence analysis (CCA)] with various options and combinations of variables, and led to the formation of hypotheses that were then tested. 4 We tested the hypotheses in a sequential analysis with CCA and Monte Carlo permutation tests: after establishing the influence of one set of environmental variables, this set was covaried out in subsequent analyses. In this way the influence of (i) season of sowing of the crop; (ii) geographical region; (iii) soil type; (iv) crop species; and (v) temporal trends was tested. The four latter were tested separately for spring‐ and autumn‐sown crops. 5 The sowing season of the crop had an overwhelming influence on the weed flora, and many weed species, both annual and perennial, showed strong associations with either autumn or spring. There were significant differences in weed flora composition between the geographical regions and soil types as well as between crop species. There were significant temporal trends only in the weed flora of autumn‐sown crops. 6 This study provides a protocol that combines exploratory ‘data diving’ with strict hypothesis testing using direct gradient analysis methods such as CCA. Such two‐phase analysis should improve the way complex data are analysed and patterns are interpreted.

Principal Curves: A New Technique for Indirect and Direct Gradient Analysis

Principal Curves: A New Technique for Indirect and Direct Gradient Analysis