Jacques Lauga

Application of neural networks to modelling nonlinear relationships in ecology

Abstract Two predictive modelling principles are discussed: multiple regression (MR) and neural networks (NN). The MR principle of linear modelling often gives low performance when relationships between variables are nonlinear; this is often the case in ecology; some variables must therefore be transformed. Despite these manipulations, the results often remain disappointing: poor prediction, dependence of residuals on the variable to predict. On the other hand NN are nonlinear type models. They do not necessitate transformation of variables and can give better results. The application of these two techniques to a set of ecological data (study of the relationship between density of brown trout spawning sites (redds) and habitat characteristics), shows that NN are clearly more performant than MR (R2 = 0.96 vs R2 = 0.47 or R2 = 0.72 in raw variables or nonlinear transformed variables). With the calculation power now currently available, NN are easy to implement and can thus be recommended for modelling of a number ecological processes.

How do king penguins (Aptenodytes patagonicus apply the mathematical theory of information to communicate in windy conditions

In the king penguin (Aptenodytes patagonicus), both pair members alternate in incubating and rearing their chick. Mates can recognize each other among thousands of other birds in the hubbub of the colony using only acoustic signalling: the display call. Large penguin colonies are found on sub–Antarctic islands where strong winds blow throughout the year. We have shown by experiments under natural conditions that the level of background noise increases in windy conditions and thus leads to a diminution of the signal–to–noise ratio. Moreover the emergence level of the signal revealed by entropy calculation is statistically weaker in windy conditions. To achieve breeding success, birds must continue communicating in spite of the significant decrease in the total amount of information that can be transmitted in windy situations. For the first time, to our knowledge, we have shown that a bird species takes into account the constraints imposed by wind on their acoustic communication. In windy conditions, birds try to maintain the efficiency of communication by increasing both the number of calls emitted and the number of syllables per call. This result conforms with predictions from the mathematical theory of communication: increased redundancy in a signal improves the probability of receiving a message in a noisy channel.

Artificial neural networks as a classification method in the behavioural sciences

The classification and recognition of individual characteristics and behaviours constitute a preliminary step and is an important objective in the behavioural sciences. Current statistical methods do not always give satisfactory results. To improve performance in this area, we present a methodology based on one of the principles of artificial neural networks: the backpropagation gradient. After summarizing the theoretical construction of the model, we describe how to parameterize a neural network using the example of the individual recognition of vocalizations of four fallow deer (Dama dama). With 100% recognition and 90% prediction success, the results are very promising.

An experimental test of hypotheses explaining social segregation in dimorphic ungulates

In many sexually dimorphic ungulates, adults spend most of their time in single-sex groups outside the mating season. We tested two hypotheses proposed to explain this social segregation. The activity budget hypothesis predicts that females spend more time grazing than males, and that activity synchrony is higher within single-sex than mixed-sex groups. The social affinity hypothesis states that higher spatial proximity and aggregation within the sexes results from sexual differences in motivation to interact socially. Fifteen merino sheep, Ovis aries, 2 or more years old, of each sex were penned together in three 1-ha paddocks in succession during 7 winter weeks. Despite the high sexual dimorphism in body weight, both sexes were found together, and did not differ substantially in activity budget. Males grazed less than females but the difference was small. The degree of activity synchrony was similar within and between the sexes. When grazing, males walked more rapidly and tended to be at the front of the group more often than females. Ewes were rarely seen interacting with other sheep, in contrast to males, which interacted twice as often with females as with males. However, same-sex pairs of nearest neighbours were found more frequently than expected, whether or not the males were at the front of the group. These results indicate a social segregation at a small spatial scale, which can be explained by differences in speed of movement between males and females and/or higher social affinity between same-sex than opposite-sex peers. The lack of single-sex groups can be explained by the strong gregariousness of merino sheep, and possibly the young age of the adult males.

A method of independent time and frequency decomposition of bioacoustic signals: inter-individual recognition in four species of penguins

The aim of this paper is to introduce a method for analyzing acoustic signals capable of assessing the potential for individual coding information. Signals are analysed both in the time domain (rhythm of emission of the song independent of its frequency content) and in the spectral domain (spectral content of the song independent of the rhythm of emission). The method is then applied to a comparative study of four penguin species, where the problem posed by inter-individual recognition differs from species to species. A direct relationship was shown between the potential of individual coding and the difficulty in partner identification.

Local enhancement and seed choice in the juvenile canary, Serinus canarius

The influence of the presence and feeding behaviour of the father on the behaviour of young canaries was examined. The naive offspring had access simultaneously to canary seed or oat seeds in the same place as did the parent, who either did or did not have access to seeds (canary seeds, oats or hemp). In the absence of an adult, the juveniles took, husked and ate few seeds. The presence of the father increased the picking up, husking and ingestion of seeds by the juveniles, who also tended to feed in the same place as the parent (local enhancement). The choice of seed was also influenced by the parents behaviour as the young birds picked up more oat seeds when these seeds were available to the father. This was not observed for canary seeds, which tended to be selected spontaneously by the young birds. The amounts of canary seeds and oats husked and ingested by the juveniles did not only depend on seed choice. Manipulation of canary seed by the father facilitated its husking by the juvenile, whereas manipulation of oats by the father had no facilitatory effect on the use of this seed by the young birds. Overall, the effects of social facilitation tended to favour ingestion of seeds spontaneously selected by the offspring.

Acoustic communication in a king penguin colony: importance of bird location within the colony and of the body position of the listener

Abstract This study focuses on the propagation of the display call used for individual recognition between penguin partners. Transmission of acoustic information in a noisy environment such as a king penguin colony is very difficult. Conditions of propagation were examined for two areas of the colony (hatching and courtship areas) and reception at three heights from the ground (10, 45 and 90 cm). Signal modification was assessed in terms of attenuation of signal, amplitude modulation and spectral content. The acoustic recognition of the mate, which is decisive for breeding success, took place in the hatching area where signal degradation was less great for all parameters studied. When the receiver was located 10 cm above the ground, degradation of the signal was much more pronounced than when it was located 45 or 90 cm above ground. The characteristic incubating attitude of the king penguin (i.e. standing with the head at 45 cm above the ground) enables it to receive the main part of the signal in spite of the noisy environment.

Analysis of forestry impacts and biodiversity in two Pyrenean forests through a comparison of moth communities (Lepidoptera, Heterocera)

We have compared the biodiversity variations in moth communities between unmanaged forests and commercial forests in a mountainous environment (Pyrenees France). The aim was to evaluate the impact of forestry activities on moth diversity. The data collected from the insects were analysed with a Bayesian specific similarity index (noted SSP index) and by statistical biodiversity indexes comparison. It was seen that diversity and richness were decreased in the plantation compared to the unmanaged forest. Interestingly, the composition of the communities of moths was shown to be not only related to the presence/absence of host plants, but also to be in relation with changes in the differences in forest architecture (i.e. the relative coverage by the different vegetation levels). However, the moth community in the commercial forest still has a high capacity to converge with the moth community present in more natural forests. We report here a list of 177 species of moths, providing information on the distribution of some uncommon species, poorly studied as yet in the Pyrenees.

A soil-plant-water model with a case study in a forested catchment

Abstract A dynamic model of water utilization by vegetation is given here; the model covers a period of 1 year, and the spatial unit chosen is a forested catchment area. The transpiration process is shown to differ from evaporation over a water surface as soon as a water deficit develops, corresponding to the closure of stomata and a sharp reduction in transpiration, although the evaporative power of the air may still be strong. The mathematical model is first discussed with reference to hydrological processes over the watershed, followed by the addition of state equations which describe the living part of the watershed and its water consumption processes.