Project Topics
Get Complete Project Material File(s) Now! »
The ecological niche concept in a competition framework
The modern concept of ecological niche was proposed by George Evelyn Hutchinson (1957) and it is defined as a quantitative n-dimensions hypervolume constructed by the range of environmental features that enable a species to maintain a viable population indefinitely (Blonder et al. 2014). Independent axes, that have a biological meaning for the species, characterize the dimensions of the niche (Maire et al. 2012). If we assume that interspecific competition does truly occur among co-existing species, the ecological niche can be divided into two categories: the fundamental niche is defined in absence of competition while the realized niche is characterized by the presence of competition among species (Hutchinson 1957, Maire et al. 2012).
Based on the principle of competitive exclusion, niche theory implicitly assumes that, in order to have a stable co-existence, niches of co-occurring species must differ (Hutchinson 1957; Chesson 1991) although a certain degree of similarity is permissible (May & Mac Arthur 1972; Pianka 1974). By measuring competitors’ niche overlap it is possible to assess the effects of density-dependent competition on the tolerable upper limit of niche overlap (Young 2004). Moreover, the conceptual niche framework allows testing if competition between sympatric species is happening with the use of species assemblages modeling.
The ecological niche concept is strongly related to habitat selection (Maire et al. 2012). The latter is considered as a density dependent process. In fact, when populations are at low-density levels, individuals can freely occupy the habitat that maximizes their fitness. At the opposite, when population density levels increase, the individual fitness decreases within the most favourable habitat, making adjacent and less favourable habitats providing the same fitness. If habitat suitability can vary in function of population densities (Morris 1988), then habitat selection depends not only on resources abundance but also on the density of the same and/or different species sharing the same area. In the latter case, which is the most commonly represented, the organization of the community can be based on shared or distinct preferences (Morris 1988). In the case of shared preferences, competition for resources may occur. It is therefore possible to use habitat distribution patterns to evaluate the role of interspecific competition.
The necessary link between ecosystem knowledge and conservation
We have seen above how the loss of biodiversity is becoming an urgent issue that needs to be addressed by the conservation and scientific world before it becomes irreversible. Effective conservation measures are therefore required to guarantee the persistence of ecosystems that might provide invaluable services for human well-being (Braat & de Groot 2012), in most cases still undiscovered (Wilson 1992).
However, successful conservation management cannot be implemented without well-developed knowledge on species biology and processes occurring among species sharing the same habitat and, eventually, the same preferences.
The discipline of conservation biology has already contributed to mitigating anthropogenic actions on biodiversity at different organization levels and it has helped to reveal underlying mechanisms inducing variation in populations’ demographic parameters (Primack & Miller-Rushing 2012). Thanks to this information, policy makers were able to act, most of the time under high urgency, setting up appropriate conservation programs to save endangered species from extinction.
Most of the conservation measurements put into place to preserve particular species deal only with the management of physical or habitat features without accounting for species interactions, which, most of the time, are unknown before the management action has been undertaken (Soulé et al. 2005). This kind of procedure is a risk for the ecosystem as ecological cascades may bring a better or a worst result than the expected one. For example, the eradication of invasive rats on North Island (Seychelles) in 2005 led to an unexpected decrease in the number of invertebrates, both on ground and leaves, despite the fact that rats were known to be feeding on invertebrates. This unexpected result was probably due to the trophic cascading effects of the removal of rats, which triggered a significant increase in land birds and lizards, and also in large invertebrates, all of which are feeding on small invertebrates and limited by rats (Galman 2011; Rocamora & Henriette, in press).
Therefore, when conservationists fail to understand the interactions occurring between species, conservation measurements might not achieve the desired results (Soulé et al. 2005). A solution to improve biodiversity maintenance can therefore be found in both practice and theory. Conservation evidence provide practical examples of what works in conservation (Sutherland et al. 2004a; Sutherland 2015) i.e. which actions have been already undertaken with successful results (see above). However, empirical studies on wild communities are strongly required to improve knowledge on mechanisms driving species coexistence (Morris 2003) and to track and predict changes on populations and communities when humans alter ecosystems properties.
In summary, theory and conservation should be closely related to effectively protect and manage important ecosystems and endangered or declining populations.
CASE STUDY: ESTIMATING SHEARWATER DENSITIES ON ARIDE ISLAND
We here present the application of the two-species model to the estimation of a mixed colony on Aride Island (Seychelles) with two breeding shearwaters: the wedge-tailed Puffinus pacificus and tropical shearwater P. bailloni. Aride Island Nature Reserve comprises 73 ha and is suspected to harbor the largest colony of Puffinus bailloni in the world (Del Hoyo et al. 1992, Skerrett et al. 2001). Both species breed in the same type of natural burrow on the hillsides of the island. While P. pacificus is a seasonal breeder found mainly from September to February, P. bailloni has no clear breeding season in Seychelles, and incubating birds may be found all year round (Skerrett et al. 2001).
To carry out the census here analyzed, we followed the playback census protocol described in (Betts 1998). The census plots were sampled in November 2011, February 2012 and May 2012 (only once per plot), in order to ensure good representation of P. bailloni. For the seasonal P. pacificus, only the November and February months are considered. We surveyed 19 circular plots of 100 m2 randomly selected in 1996 for a previous survey (Betts 1998). All surveys occurred at night between 20:00h and 23:00h once every census month. In each plot, we noted all potentially suitable burrows and inspected them visually for the presence of a nesting bird of either species. We played recorded male-female duet calls for both species (Rocamora et al. 2000) in the case of unknown content, and for the observed species when the bird was visible. We played the call at the opening of the burrow and noted whether the bird responded by the end of the recordings (1:24 min for P. bailloni and, 1.58 for P. pacificus).
We used the two-species model specified above to estimate the average densities of shearwaters on the island. In order to account for plot-variation in occupancy and produce estimates of average island densities, we modelled species-specific occupancy probabilities ψ as a random effect. That is, rather than estimating plot-specific occupancy probabilities, given plot i and species S the probabilities of occurrence followed a logit-normal distribution: !,! ~ ( ! », ! »).
We evaluated the importance of heterogeneity in detection probabilities by comparing models that differed on whether the detection probabilities (pVA, pVB, pVA, pVB) were fixed or randomly (logit-normally) varying across plots. As a measure of model performance we used the Deviance Information Criterion, DIC (Spiegelhalter et al. 2002; Supplementary material Table ESM2.1). Parameters were estimated using a Bayesian framework, as it performed better in the simulations. We specified the following uninformative parameter priors for the random variables: ~ (0,10) ~ (0,10).
When probabilities were set as constant across plots, we used a uniform prior ranging from 0 to 1. We ran three independent MCMC chains with 10 000 iterations each, a burnin of 5000 and thinning of every 10 samples. Convergence was considered achieved when the Gelman-Rubin statistic (Gelman and Rubin 1992) for all parameters was lower than 1.1.
Calculating the number of pairs per sampled plot
To obtain an unbiased number of pairs per species and per plot, we need to account for the content of invisible burrows, detection rate and breeding failure, i.e. treating i) the burrows that could not be visually inspected and for which no response was obtained after playback, and ii) the currently empty burrows, that may have contained a pair earlier in the season (e.g. premature failures) or not. We used two different methods: in the first method, census data were integrated within a Bayesian statistical framework (see CHAPTER 2). This method accounts for the imperfect detection based on response rate to playback in order to calculate the probability of occupancy. This probability is then used to estimate the number of burrows with unknown content (content not visible and no response to the playback) that are occupied by one or the other species. However, while accounting for imperfect detection, this method does not account for breeding failures. Therefore in our case, the resulting estimates correspond to a minimum population size for the 2014 census.
To account for breeding failure in both species we used a second method that considers the data collected during the three breeding seasons. To estimate which proportion of empty burrows results from breeding failure, we took into account time (since total number of breeding failure will increase with season) and location in the island (as spatial variation in breeding density of the two species might occur). We assumed that the plots surveyed at the beginning of the breeding season (November) were at the maximum occupancy rate. Data were available in November (all years combined) for 73 plots and were used to map (using inverse distance weighting interpolation) the relative proportions of wedge-tailed, tropical shearwaters and empty burrows for the whole island, at a 1 ha resolution (see Appendix C). Then, we used the following procedure to calculate the number of pairs per plot accounting for imperfect detection, breeding failures and coexistence of the two species (i.e., an unknown burrow cannot be occupied simultaneously by the two species). The logic used in the calculation is described in details in Appendix C (which discusses also the assumptions used), a summary is provided below. The total number of burrows is obtained for tropical (TS) and wedge-tailed shearwater (WS) using the following formula (see Figure 3.2 for explanations on the various counts, N1 to N20): TOT TS = N1+N4+ N14+N7+N16 (EQN 1).
Species habitat and spatial models
The count data were modelled using a negative binomial distribution to account for over-dispersion (Richards 2008). The candidate habitat variables were checked for collinearity. The highest Pearson correlation obtained was 0.5, thus lower than the limit of 0.7 (Dormann et al. 2013), indicating that all variables could be used in the same model. Habitat variables were selected using the Akaike’s Information Criterion (Appendix D) (Zurr et al. 2013). A step-wise selection procedure starting from the full model (i.e. including all the habitat variables) was applied to identify the best model. A choice among interactions and quadratic variables was first done separately and the significant ones were added to the full model (Zurr et al. 2013). This variables selection step was performed using only the 2014 data. It was therefore simplified by eliminating the presence of replicates (in 2012-2013) while maximizing the number of data, thus avoiding using a random factor for the plot identity.
Once habitat variables were selected, they were implemented in a spatio-temporal explicit model to estimate the population abundance and distribution of the two species. The model included the habitat variables as fixed effect and a random spatio-temporal term in order to account for spatio-temporal dependences. Model parameters were estimated using the Integrated Nested Laplace Approximation (INLA, see Rue et al., 2009). INLA allows estimating complex Bayesian hierarchical models in low computation times. (Carson and Flemming 2014; Muñoz et al. 2013; Rue et al. 2009). For continuous spatial modelling, the stochastic partial differential equation approach (SPDE) is preconized. It consists in using another approximation making the estimation of Gaussian random fields (a random field correspond to a random variable having a spatial structure) of the Matérn class much easier (Lindgren et al. 2011). Such approach requires defining a constrained Delaunay triangulation, which was done in such a way to keep the distance between the nodes lower than 10 m (see Appendix E). As the data set is composed of three distinct years of sampling, a temporal term was also added to the spatial structure. The temporal term chosen was an autoregressive order 1 structure (Blangiardo et al. 2013; Cameletti et al. 2013), which can be interpreted as the correlation of the random field at a given location between year t and t-1. The three years of sampling (2012, 2013, 2014) were used as time steps and the data collected in May 2012 and 2013 were removed in order to have a more homogeneous time period within each time step (from November to March). INLA is available in the R software (R core team, 2014) with the package R-INLA (www.r-inla.org). Predictions were made on a grid of resolution 10 x 10 m covering the whole island except the rocky coastal areas since shearwaters could not dig burrows in these areas, and were, indeed, totally absent.
Prediction maps and total abundance
For the spatio-temporal modeling, we retained values obtained within the second method, since our goal was to predict breeding numbers irrespective of breeding failure. Results for the wedge-tailed and tropical shearwater are shown in Figure 3.3 (standard errors maps are provided in Appendix F). As distribution for the two species over the three year time period was comparable, we decided to show only the map for 2014, this being the season with a larger sample size. The distribution of breeding pairs across the island was more homogeneous for tropical than for wedge-tailed shearwater (Figure 3.3), since breeding pairs of the former were present also on flat areas while the latter was mainly found on the hill or on steep areas (see also maps in Castle and Mileto, 1991 indicating presence/absence of the species). The temporal correlation between the three surveys for both tropical and wedge-tailed were very high: 0.98 and 0.97 respectively, suggesting highly similar distribution of the breeding pairs among the three years, which could indicate that the birds (not necessarily the same ones) choose to breed in the same areas during different breeding seasons. The spatial autocorrelation slightly differed between the two species, but variance stabilized at rather short distances: 196 m for the wedge-tailed and 88 m for the tropical.
The total abundance during the seasons 2012, 2013 and 2014 were comparable, with 14,977 (97.5% confidence interval 11,914-19,715), 14,848 (12,340-18,969) and 14,297 (11,891-18,152) pairs of wedge-tailed and 23,142 (19,119-29,634), 22,660 (18,738-28,523) and 22,610 (18,861-28,518) breeding pairs of tropical shearwaters respectively. In tropical shearwater, these values must be considered as lower bound values, since we could not totally account for breeding failure in this species as the survey did not cover the whole breeding season (being a year-round breeder) therefore the maximum occupancy rate (intercept) cannot be identified. Furthermore, the breeding season for this species only extends to 9 months and multiple pairs occupy the same burrow within one year. Applying a 4/3 multiplication coefficient (accounting for 9 months out of 12) suggest that 30,000 breeding pairs must be a more reliable mean estimate for the tropical shearwater on Aride Island, twice the numbers of wedge-tailed 15 000 pairs).
Table of contents :
CHAPTER 1
1.1 GENERAL INTRODUCTION
1.2 GENERAL METHODS
CHAPTER 2 ANALYSIS OF PLAYBACK CENSUS TO ESTIMATE THE DENSITY OF CAVITY-DWELLING BIRDS
2.1 INTRODUCTION
2.2 MATERIALS AND METHODS
2.3 RESULTS
2.4 DISCUSSION
TABLES
FIGURES
EXTERNAL SUPPLEMENTARY MATERIALS (ESMS)
CHAPTER 3 ASSESSING POPULATION SIZE IN NOCTURNAL DWELLING SEABIRDS ACCOUNTING FOR DETECTION PROBABILITY AND HABITAT PREFERENCES: A SEQUENTIAL APPROACH
3.1 INTRODUCTION
3.2 MATERIAL AND METHODS
3.3 RESULTS
3.4 DISCUSSION
TABLES
APPENDIXES
CHAPTER 4 COMPARATIVE FORAGING DISTRIBUTION AND ECOLOGY SUGGESTS INTERSPECIFIC COMPETITION BETWEEN TWO SYMPATRIC SHEARWATERS FROM THE SEYCHELLES
4.1 INTRODUCTION
4.2 MATERIALS AND METHODS
4.3 RESULTS
4.4 DISCUSSION
TABLES
FIGURES
EXTERNAL SUPPLEMENTARY MATERIALS (ESMS)
CHAPTER 5 MOVEMENT PATTERNS AND HABITAT SELECTION OF WEDGE-TAILED SHEARWATERS (PUFFINUS PACIFICUS) BREEDING AT ARIDE ISLAND, SEYCHELLES
5.1 INTRODUCTION
5.2 MATERIALS AND METHODS
5.3 RESULTS
5.4 DISCUSSION
CHAPTER 6 GENERAL DISCUSSION
APPENDIX I BREEDING SUCCESS ANALYSIS OF TROPICAL SHEARWATER AND WEDGE-TAILED SHEARWATER POPULATIONS ON ARIDE ISLAND.
BIBLIOGRAPHY .
