Jean-Christophe Foltête

A software tool dedicated to the modelling of landscape networks

Since landscape connectivity reflects a basic form of interaction between species and their environment, the modelling of landscape networks is currently an important issue for researchers in ecology and practitioners of landscape management alike. Graph-based modelling has recently been shown to be a powerful way of representing and analysing landscape networks. Graphab 1.0 is designed as a package integrating a complete set of connectivity analysis functions. The application can build graphs from a given landscape map by exploring several possibilities for link topology, types of distances and graph definitions. A wide range of connectivity metrics can be computed from these graphs at the global, component or local levels. By extrapolating patch-based metrics outside of the graph using a distance-dependent function, the relationship between the graph and any set of point data can be established in order to compare the connectivity properties of the landscape network and field observations of a given species. In conclusion, Graphab 1.0 provides a full set of coherent modelling functions for analysing and exploring landscape graphs with a single application.

Genetic structure of the cyclic fossorial water vole (Arvicola terrestris): landscape and demographic influences.

Genetic structure can be strongly affected by landscape features and variation through time and space of demographic parameters such as population size and migration rate. The fossorial water vole (Arvicola terrestris) is a cyclic species characterized by large demographic fluctuations over short periods of time. The outbreaks do not occur everywhere at the same time but spread as a wave at a regional scale. This leads to a pattern of large areas (i.e. some hundreds of km2), each with different vole abundances, at any given time. Here, we describe the abundance and genetic structures in populations of the fossorial water vole. We use the data to try to understand how landscape and demographic features act to shape the genetic structure. The spatial variability of vole abundance was assessed from surface indices, collected in spring 2002 (April) in eastern central France. Genetic variability was analysed using eight microsatellite loci at 23 localities sampled between October 2001 and April 2002. We found some congruence between abundance and genetic structures. At a regional scale, the genetic disruptions were associated with both sharp relief and transition between an area of low abundance and another of high abundance. At a local scale, we observed a variation of the isolation‐by‐distance pattern according to the abundance level of vole populations. From these results we suggest that the dispersal pattern in cyclic rodent populations varies throughout the demographic cycle.

Integrating graph-based connectivity metrics into species distribution models

Species distribution models (SDMs) are commonly used in ecology to map the probability of species occurrence on the basis of predictive factors describing the physical environment. We propose an improvement on SDMs by using graph methods to quantify landscape connectivity. After (1) mapping the habitat suitable for a given species, this approach consists in (2) building a landscape graph, (3) computing patch-based connectivity metrics, (4) extrapolating the values of those metrics to any point of space, and (5) integrating those connectivity metrics into a predictive model of presence. For a given species, this method can be used to interpret the significance of the metrics in the models in terms of population structure. The method is illustrated here by the construction of an SDM for the European tree frog in the region of Franche-Comté (France). The results show that the connectivity metrics improve the explanatory power of the SDM and emphasize the important role of the habitat network.

Coupling inter-patch movement models and landscape graph to assess functional connectivity

Landscape connectivity is a key process for the functioning and persistence of spatially-structured populations in fragmented landscapes. Butterflies are particularly sensitive to landscape change and are excellent model organisms to study landscape connectivity. Here, we infer functional connectivity from the assessment of the selection of different landscape elements in a highly fragmented landscape in the Île-de-France region (France). Firstly we measured the butterfly preferences of the Large White butterfly (Pieris brassicae) in different landscape elements using individual release experiments. Secondly, we used an inter-patch movement model based on butterfly choices to build the selection map of the landscape elements to moving butterflies. From this map, functional connectivity network of P. brassicae was modelled using landscape graph-based approach. In our study area, we identified nine components/groups of connected habitat patches, eight of them located in urbanized areas, whereas the last one covered the more rural areas. Eventually, we provided elements to validate the predictions of our model with independent experiments of mass release-recapture of butterflies. Our study shows (1) the efficiency of our inter-patch movement model based on species preferences in predicting complex ecological processes such as dispersal and (2) how inter-patch movement model results coupled to landscape graph can assess landscape functional connectivity at large spatial scales.

Deviations in Pedestrian Itineraries in Urban Areas: A Method to Assess the Role of Environmental Factors:

Walking has long been neglected in urban-mobility research, but it is now making its way into numerous studies using various approaches. Empirical data are often processed in well-known models of flow allocation to study the behaviour of pedestrians and to identify their preferences. However, these models assume that route choices are predetermined at the start of each trip and do not admit any possible intervening decision along these trips. We propose to overcome this limitation through a new method for the analysis of pedestrian behaviour. This method, which we call ‘deviation analysis’ consists of (1) identifying the intersections from which a pedestrian has chosen a route longer than the shortest path; (2) defining the segments of the network which diverge from each deviation; (3) testing the influence of the environmental variables of these segments on the choice of route by using a discrete choice model. Deviations are compared with the cases where pedestrians follow the shortest path (called ‘continuations’), which are assumed to be less strongly linked to environmental variables due to the ‘natural’ choice for minimising the distance travelled. This method is applied to a series of pedestrian trips recorded in the French city of Lille. Results show that the environmental variables used in this study contribute to explaining the route choices with more strength when the deviations involve a trip lengthening of at least 50 m. They also show that the influence of variables describing the visual aspect of urban landscape may influence the route choices and outline the positive role of the urban atmosphere linked to the commercial function of streets.

Localising missing plants in squared-grid patterns of discontinuous crops from remotely sensed imagery

The purpose of this work is to localise and characterise missing plants on very high resolution (VHR) aerial images of agricultural parcels, in the case of discontinuous crops like wine and olive tree, which are planted according to a squared-grid pattern. It aims to establish an assisted, image processing system for remote sensed images, allowing the inventory the missing or withering plants, and the monitoring of their evolution during time. The global approach considers the planted parcel as a topological graph of vertices, whose reciprocal location conforms to a set of geometrical rules about orientation and length. The proposed system initiates the graph from the original image; then it adds missing vertices and refines its knowledge of the spatial pattern on an iterative basis. Quality indicators are assigned at each added vertex, and several stopping criteria are estimated for each iteration, permitting an automated use of the algorithm. Test cases have been conducted on two data sets of three parcels each: olive groves and goblet vineyards. The results are compared to validation data. They show an efficient reconstruction of the geometry and satisfactory omission-commission errors; they allow drawing up a typology of the major errors, and propose calibration parameters based on a sensitivity analysis. The main improvements include essentially the preprocessing, filtering step of the initial image. The process is being used for Languedocian vineyards (France), and may be potentially usable for other problematic with the same kind of spatial patterns.

A multi-species approach for assessing the impact of land-cover changes on landscape connectivity

ContextLand-cover changes (LCCs) could impact wildlife populations through gains or losses of natural habitats and changes in the landscape mosaic. To assess such impacts, we need to focus on landscape connectivity from a diachronic perspective.ObjectivesWe propose a method for assessing the impact of LCCs on landscape connectivity through a multi-species approach based on graph theory. To do this, we combine two approaches devised to spatialize the variation of multi-species connectivity and to quantify the importance of types of LCCs for single-species connectivity by highlighting the possible contradictory effects.MethodsWe begin with a list of landscape species and create virtual species with similar ecological requirements. We model the ecological network of these virtual species at two dates and compute the variation of a local and global connectivity metric to assess the impacts of the LCCs on their dispersal capacities.ResultsThe spatial variation of multi-species connectivity showed that local impacts range from −6.4% to +3.2%. The assessment of the impacts of types of LCCs showed a variation in global connectivity ranging from −45.1% for open-area reptiles to +170.2% for natural open-area birds with low-dispersion capacities.ConclusionsThis generic approach can be reproduced in a large variety of spatial contexts by adapting the selection of the initial species. The proposed method could inform and guide conservation actions and landscape management strategies so as to enhance or maintain connectivity for species at a landscape scale.

Does regional landscape connectivity influence the location of roe deer roadkill hotspots

Linear infrastructures have both direct and indirect effects on ecosystems. Wildlife-vehicle collisions are the most visible direct effect and can result in severe human injuries. In Europe, the roe deer population is growing and roe deer roadkills are becoming more common. Roe deer movements depend on landscape features and regional-scale connectivity. Here, we investigate the influence of the landscape network on the location of roe deer roadkill hotspots. In order to localize new potential hotspots along the national roads of Franche-Comté (Eastern France), we first show that roe deer roadkills are not distributed randomly and we identify hotspots. Then, we explain roe deer hotspot locations using a predictive model by combining landscape composition variables, road-related properties, and graph-based connectivity metrics. We test three centrality metrics at three dispersal distances and we assess the relative contribution of the connectivity metrics to the best model. Finally, in order to define high-risk sections, we find the probability that reduces the costs of misclassification that the model produces. We validate the model with a new set of roe deer roadkills.

Using landscape graphs to delineate ecologically functional areas

ContextLandscape graphs are widely used to model connectivity and to support decision-making in conservation planning. Compartmentalization methods applied to such graphs aim to define clusters of highly interconnected patches. Recent studies show that compartmentalization based on modularity is suitable, but it applies to non-weighted graphs whereas most landscape graphs involve weighted nodes and links.ObjectivesWe propose to adapt modularity computation to weighted landscape graphs and to validate the relevance of the resulting compartments using demographic or genetic data about the patches.MethodsA weighted adjacency matrix was designed to express potential fluxes, associating patch capacities and inter-patch distances. Eight weighting scenarios were compared. The statistical evaluation of each compartmentalization was based on Wilks’ Lambda. These methods were performed on a grassland network where patches are documented by annual densities of water voles in the Jura massif (France).ResultsThe scenarios in which patch capacity is assigned a small weight led to the more relevant results, giving high modularity values and low Wilks’ Lambda values. When considering a fixed number of compartments, we found a significant negative correlation between these two criteria. Comparison showed that compartments are ecologically more valid than graph components.ConclusionsThe method proposed is suitable for designing ecologically functional areas from weighted landscape graphs. Maximum modularity values can serve as a guide for setting the parameters of the adjacency matrix.

Integrated GIS software for computing landscape visibility metrics: XXXX

As an important component of the quality of the living environment, landscape is increasingly addressed in terms of its visual dimension. In contrast to the point‐like character of in situ observations and photographic analyses, the modeling of landscape visibility from digital data has the advantage of scanning geographical space in a systematic way. However, the tools currently available for visibility modeling are limited to the mapping of viewsheds. They require complementary operations for a complete landscape assessment, including direct and easy computation of landscape metrics. Furthermore, none of those tools integrates recent technical advances to better characterize the visible landscape by tangential vision (i.e., from ground level as opposed to vision by viewshed from above). Starting from this, PixScape software proposes to integrate a large set of functions for modeling landscape visibility while remaining interfaced with GIS software. This software can be used to perform a complete landscape assessment by computing a wide range of original landscape metrics. It performs tangential analysis in addition to viewshed analysis, which can produce more realistic outcomes. Because landscape visibility analysis over large areas implies significant computation time, the software also integrates a multi‐resolution process intended to speed up calculations while also taking into account the cognitive abilities of human vision.