Régis Céréghino

Applications of artificial neural networks for patterning and predicting aquatic insect species richness in running waters

Abstract Two artificial neural networks (ANNs), unsupervised and supervised learning algorithms, were applied to suggest practical approaches for the analysis of ecological data. Four major aquatic insect orders (Ephemeroptera, Plecoptera, Trichoptera, and Coleoptera, i.e. EPTC), and four environmental variables (elevation, stream order, distance from the source, and water temperature) were used to implement the models. The data were collected and measured at 155 sampling sites on streams of the Adour–Garonne drainage basin (South-western France). The modelling procedure was carried out following two steps. First, a self-organizing map (SOM), an unsupervised ANN, was applied to classify sampling sites using EPTC richness. Second, a backpropagation algorithm (BP), a supervised ANN, was applied to predict EPTC richness using a set of four environmental variables. The trained SOM classified sampling sites according to a gradient of EPTC richness, and the groups obtained corresponded to geographic regions of the drainage basin and characteristics of their environmental variables. The SOM showed its convenience to analyze relationships among sampling sites, biological attributes, and environmental variables. After accounting for the relationships in data sets, the BP used to predict the EPTC richness with a set of four environmental variables showed a high accuracy (r=0.91 and r=0.61 for training and test data sets respectively). The prediction of EPTC richness is thus a valuable tool to assess disturbances in given areas: by knowing what the EPTC richness should be, we can determine the degree to which disturbances have altered it. The results suggested that methodologies successively using two different neural networks are helpful to understand ecological data through ordination first, and then to predict target variables.

The ecology of European ponds: defining the characteristics of a neglected freshwater habitat

There is growing awareness in Europe of the importance of ponds, and increasing understanding of the contribution they make to aquatic biodiversity and catchment functions. Collectively, they support considerably more species, and specifically more scarce species, than other freshwater waterbody types. Ponds create links (or stepping stones) between existing aquatic habitats, but also provide ecosystem services such as nutrient interception, hydrological regulation, etc. In addition, ponds are powerful model systems for studies in ecology, evolutionary biology and conservation biology, and can be used as sentinel systems in the monitoring of global change. Ponds have begun to receive greater protection, particularly in the Mediterranean regions of Europe, as a result of the identification of Mediterranean temporary ponds as a priority in the EU Habitats Directive. Despite this, they remain excluded from the provisions of the Water Framework Directive, even though this is intended to ensure the good status of all waters. There is now a need to strengthen, develop and coordinate existing initiatives, and to build a common framework in order to establish a sound scientific and practical basis for pond conservation in Europe. The articles presented in this issue are intended to explore scientific problems to be solved in order to increase the understanding and the protection of ponds, to highlight those aspects of pond ecology that are relevant to freshwater science, and to bring out research areas which are likely to prove fruitful for further investigation.

Spatial analysis of stream invertebrates distribution in the Adour-Garonne drainage basin (France), using Kohonen self organizing maps

We analysed the regional distribution of 283 lotic macroinvertebrate species from four insect orders (Ephemeroptera, Plecoptera, Trichoptera, Coleoptera =EPTC) in the Adour-Garonne drainage basin (South–Western France, surface =116 000 km 2 ). The aim of this work was to provide a stream classification based on characteristic species assemblages. The faunistic data corresponded to the occurrence (presence or absence) of 283 species at 252 sampling sites. These data were computed with the Kohonen self organised map algorithm (SOM) (Kohonen, Self-Organizing Maps, volume30 of Springer Series in Information Sciences. Springer, Berlin, Heidelberg. (Second Extended Edition 1997)). This neural network algorithm has already been successfully used in ecology (Giraudel et al., Artificial neural networks, applications to ecology and evolution. Springer-Verlag, (in press); Chon et al., Ecol. Model., 90, 1996, 69–78) for communities patternizing. SOM enable visualisation of the complex species assemblage in a two-dimensional space, preserving the topology of the input data. Then, using the U-matrix method, it was possible to classify the data without prior knowledge. Four major EPTC regions were characterised within the

Biodiversity and distribution patterns of freshwater invertebrates in farm ponds of a south-western French agricultural landscape

We assessed the importance for biodiversity of man-made farm ponds in an agricultural landscape in SW France lacking natural wetlands. The ponds were originally created to provide a variety of societal services (irrigation, visual amenity, water for cattle, etc.). We also assessed the environmental factors influencing invertebrate assemblages in these ponds. Only 18 invertebrate taxa out of 114 taxa occurring in the study area were common to ponds and rivers indicating that the contribution of farm ponds to freshwater biodiversity was potentially high. A Self-Organizing Map (SOM, neural network) was used to classify 36 farm ponds in terms of the 52 invertebrate families and genera they supported, and to specify the influence of environmental variables related to land-use and to pond characteristics on the assemblage patterns. The SOM trained with taxa occurrences showed five clusters of ponds, most taxa occurring only in 1–2 clusters of ponds. Abandoned ponds tended to support higher numbers of taxa, probably because they were allowed to undergo a natural succession. Nevertheless, abandoned ponds were also amongst the largest, so that it remained difficult to separate the effects of pond size and abandonment, although both factors were likely to interact to favour higher taxon richness. The invertebrate communities in the ponds appeared to be influenced mainly by widely acting environmental factors (e.g. area, regionalization of assemblages) with little evidence that pond use (e.g. cattle watering, amenity) generally influenced assemblage composition. Our results support the idea that agricultural landscapes containing man-made ponds make a significant contribution to freshwater biodiversity indicating that protection of farm ponds from threats such as in-filling and pollution can make a positive contribution to the maintenance of aquatic biodiversity. This added value for biodiversity should be considered when calculating the economic costs and benefits of constructing water bodies for human activities.

Spatial patterns of macroinvertebrate functional feeding groups in streams in relation to physical variables and land-cover in Southwestern France

Artificial neural networks were used to quantify the distribution of macroinvertebrate functional feeding groups (FFGs) in relation to physical variables and to land-cover in the Adour–Garonne stream system (SW France; 116,000xa0km2). The relative abundances of 5 FFGs were calculated from macroinvertebrate data recorded at 165 sampling sites. Each site was characterized using 5 physical variables (elevation, stream order, stream width, distance from the source, slope) and 3 land-cover variables (% forested, % urban areas, % agricultural areas). The sites were first classified using the Self-Organizing Map algorithm (SOM), according to the physical and land-cover variables. Two major clusters of sites corresponded to anthropogenically modified and natural areas, respectively. Anthropogenically modified areas were clearly divided into agricultural and urban landscapes. Each major cluster was divided into 3–4 subsets of sites according to a topographic gradient of physical variables. To examine the variability of the communities, FFG proportions at the 165 sites were examined on the SOM trained with physical and land-cover variables. When the riverine landscape was natural, FFG patterns responded to the upstream–downstream gradient in physical variables. When the landscape was altered by agriculture or urbanization, the effects of land-cover on FFGs overcame the influence of the physical variables. The categorization of the landscape into forested, agricultural, and urban areas was relevant to detect changes in FFG patterns. In light of increasing development along riparian zones, the use of SOMs to detect responses of FFGs to landscape alterations at regional scales exemplifies an effective technique for assessing river health based on ecological indicator groups.

Influence of intermittent hydropeaking on the longitudinal zonation patterns of benthic invertebrates in a mountain stream

The longitudinal zonation patterns of invertebrate species were studied in the River Oriege (Pyrenees, France) from 920 to 819 m a.s.l. At 912 m a.s.l., the river receives hypolimnetic water diverted from a nearby reservoir lake, and the natural flow may be enhanced several times a day from 1 to 11 m 3 s -1 in summer and winter, and from 5 to 15 m 3 s -1 during spring spates. During hydropeaking, the water was cooled in summer and slightly warmed in winter, but this was attenuated 3500 m downstream from the plant. Invertebrate densities were recorded in July (end of spates) and October (low flow period) at ten sampling sites : a reference site upstream of the hydrostations discharge point (site 1), and nine sites each 400m below the outlet (sites 2-10). The longitudinal distribution of invertebrate taxa was studied using Factorial Analyses, taxa and sites were clustered using Analysis of Dynamic Cluster. The hydroelectric facility did not clearly modify the qualitative composition of the benthic fauna, but clearly affected the longitudinal zonation of several populations. The low abundance of several species below the outlet reflected the impact of both hydropeaking and zonation. These species were high mountain species, the density of which decreases towards downstream sites in the Pyrenees, and low mountain species, the density of which naturally decreases towards upstream sites. Some taxa disappeared in the regulated section, so their distribution in the River Oriege was primarily influenced by hydropeaking. Finally, some taxa preferentially lived in the lower section of the river where the impact of peaking flows was strongly attenuated, so their longitudinal distribution was first governed by their natural longitudinal zonation. Under this kind of river regulation (natural discharge and temperature except during periods of power generation, intermittent hydropeaking from a separate reservoir) modifications of the thermal regime had a minor effect on the population dynamics, unlike hydraulic disturbances which strongly influenced the structural attributes of benthic communities.

Farm ponds make a contribution to the biodiversity of aquatic insects in a French agricultural landscape

Manmade ecosystems provide a variety of resources that have strong economic values. We assessed the importance of 37 farm ponds for the biodiversity of Odonata in an agricultural landscape lacking natural wetlands in southwestern France. Farm ponds captured 40% of the regional species pool, including both common and rare species. The species assemblages were not correlated with pond use (e.g., cattle watering, duck farming, etc.) or to landscape variable. Species richness was correlated with pond area, suggesting that community diversity was primarily driven by autoecological processes. Farm ponds thus made a positive contribution to the maintenance of aquatic biodiversity. This added value for biodiversity should be considered when calculating the costs and benefits of constructing water bodies for human activities.

Sensitivity of aquatic insect species richness to disturbance in the Adour-Garonne stream system (France)

Abstract Ephemeroptera, Plecoptera, Trichoptera and Coleoptera (EPTC) were sampled at 153 sites distributed among 47 rivers from the Adour–Garonne stream system (south-western France). Geographic differences in abiotic conditions (elevation, slope, distance from source, stream order, maximum water temperature) were assessed by statistical analysis of data from unstressed sites, and three clusters (or sub-regions) were thus derived from a Principal Component Analysis (PCA). In terms of number of species at unstressed sites, these three sub-regions (i.e. “brooks”, “streams” and “large rivers”) grouped sampling sites having significantly comparable richness, whereas between-regions differences in species richness distributions were significantly different. Therefore, species richness to disturbance relationships should be assessed differently according to the considered area. For each sub-region, we have determined five richness classes, using box-plots of species richness distribution among disturbed and undisturbed sampling sites. The results provided by the EPTC species richness and the Indice Biologique Global Normalise (IBGN, a normalised index used for the biological surveillance of French streams, with a family level of taxonomic resolution) were significantly different at the whole-drainage-basin scale but had different responses for streams and large rivers. By using the species level of taxonomic resolution, the EPTC richness was more sensitive to slight disturbances in mountain rivers (streams), i.e. where invertebrate diversity and number of species per insect families were the highest. If the species richness of a restricted number of taxonomic groups is considered an integrative descriptor of the ecosystem health, then the method has the interest to involve little computation, and to be easily transposable to any large stream system.

Influence of hydropeaking on the distribution and larval development of the Plecoptera from a mountain stream

The downstream distribution and larval development of the Plecoptera from a Pyrenean stream were studied upstream and downstream of a hydroelectric power plant with intermittent hypolimnetic releases. During power generation, flow and temperature were the two main environmental factors modified. The downstream distribution of the various taxa reflected both the impact of hydropeaking and the natural zonation. The lowest densities and biomasses were estimated at 700 m downstream from the plant, underlining the prominent role of hydropeaking. Above the plant, Plecoptera larvae showed a constant drift due to accidental dislodgement, and behavioural drift was mainly nocturnal. Below the plant, the flushing action of peaking flows added to this a catastrophic drift, which was even higher since the amplitude between natural flow and peak flow was elevated. The life cycle patterns and the growth rates of the five dominating species (Siphonoperla torrentium, Isoperla acicularis, Perla grandis, Amphinemura sulcicollis, Protonemura beatensis) were similar at the various sampling sites, and only slight differences in hatching and/or emergence periods were recorded. The Plecoptera of the Oriege complete their larval development during periods where artificial thermal fluctuations are low and have a minor effect on the populations: from autumn to spring—i.e. when the temperatures remain rather low and during snow melt flood (e.g. S. torrentium, I. acicularis, A. sulcicollis)–or in summer when hypolimnetic releases increase the daily thermal fluctuations but reduce the mean diel temperatures (e.g. P. beatensis). Therefore, under this kind of river regulation, the plecopteran population structure and distribution is mainly governed by hydraulic disturbances.

Are algae relevant to the detritus-based food web in tank-bromeliads?

We assessed the occurrence of algae in five species of tank-bromeliads found in contrasting environmental sites in a Neotropical, primary rainforest around the Nouragues Research Station, French Guiana. The distributions of both algal abundance and biomass were examined based on physical parameters, the morphological characteristics of bromeliad species and with regard to the structure of other aquatic microbial communities held in the tanks. Algae were retrieved in all of the bromeliad species with mean densities ranging from ∼102 to 104 cells/mL. Their biomass was positively correlated to light exposure and bacterial biomass. Algae represented a tiny component of the detrital food web in shaded bromeliads but accounted for up to 30 percent of the living microbial carbon in the tanks of Catopsis berteroniana, located in a highly exposed area. Thus, while nutrient supplies are believed to originate from wind-borne particles and trapped insects (i.e., allochtonous organic matter), our results indicate that primary producers (i.e., autochtonous organic matter) are present in this insectivorous bromeliad. Using a 24-h incubation of size-fractionated and manipulated samples from this plant, we evaluated the impact of mosquito foraging on algae, other microorganisms and rotifers. The prey assemblages were greatly altered by the predation of mosquito larvae. Grazing losses indicated that the dominant algal taxon, Bumilleriopsis sp., like protozoa and rotifers, is a significant part of the diet of mosquito larvae. We conclude that algae are a relevant functional community of the aquatic food web in C. berteroniana and might form the basis of a complementary non-detrital food web.