Temporal Co-Occurrence in a Community of Subtropical Australian Birds

doi:10.21203/rs.3.rs-4639189/v1

Download PDF

Research Article

Temporal Co-Occurrence in a Community of Subtropical Australian Birds

https://doi.org/10.21203/rs.3.rs-4639189/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

I aimed to better understand the community ecology of sympatric birdlife in subtropical South East Queensland, Australia using patterns of temporal species co-occurrence and principles from network theory. In line with expectation, a hierarchical clustering analysis showed that sampling dates that were successive possesed similar patterns of observed bird species. Interestingly, there is an example of total observed bird biomass being almost identical (38 kg) on successive days in September, but comprising 36 species on the first day yet only 8 on the second, the latter date being dominated by biomass from a single species (Australian Pelican, Pelicanus conspicillatus). Ranking each species on its average pairwise correlation to the other 87 species in the set shows that, unsurprisingly, raptors (such as Whistling Kites, Haliastur splenurus) tend to be enriched at the extreme negative end of the list (hypergeometric enrichment statistic P = 0.00029) indicating their presence is inhibitory to other avian species. On the other hand, Silvereyes (Zosterops lateralis) possess the highest number of total connections, the highest radiality (or ‘network influence’) metric within the inferred co-occurrence network and have the second highest average positive correlation to all other bird species in the set (+0.32). Collectively, this suggests Z. lateralis can be seen as an indicator species whose presence indicates an enhanced likelihood of observing a diversity of other bird species. The use of the PCIT network reconstruction algorithm is demonstrated in ecology for the first time.

co-occurrence network

subtropical Australian birds

raptors

The observed pattern of life on earth – which species we observe, when and in what numbers - is dependent on a range of factors including prevailing habitat and the impact of species interactions. The sum total of observed species (the α diversity or species richness) and the individual occurrence data collectively add up to that environment’s particular tally of biodiversity. Building on our understanding of the causes and consequences of biodiversity is clearly important. Every species has intrinsic value that once lost can never be regained; pristine ecosystems are often beautiful places appreciated by human observers (Hooper, Chapin III et al. 2005); highly biodiverse ecosystems possess a complexity that is often thought to lend them relative stability (Odum 1953, Elton 1958), but see (May 1972) and (Ives and Carpenter 2007) for a different perspective on the complexity-stability relationship; many species (both known and unknown) represent tapped and untapped sources of scientific knowledge, medicines and consumable products (Hooper, Chapin III et al. 2005); and healthy, productive ecosystems contribute to tangible environmental features such as air quality and flood mitigation, among many other things (Kremen 2005).

Here, I was interested in investigating community ecology at Hiddenvale Research Station in the Lockyer Valley region of subtropical South East Queensland (https://hiddenvalewildlife.uq.edu.au/). The main focus of this reserve is to protect and promote native Australian ecosystems, engage in rewilding where appropriate and foster conservation of target species. Much of the reserve is composed of native bushland, with areas of open grassland and remnant and regrowth woodland. There is an ongoing effort to establish baseline species monitoring for a wide range of animal and plant groups. I elected to focus this analysis on birds as their combination of charisma, size, daylight activity and unique vocalisations means reliably detecting their presence is far less problematic than for many other smaller, more cryptic and less well studied taxa. This combination of species richness and occurrence signal reliability makes them a pragmatic choice. I used species co-occurrence as my main metric of interest as it reflects species co-existence, which in turn is fundamental to many aspects of community ecology.

Bird observation data were used to a) to reconstruct a temporal co-occurrence network and characterise its topology and b) identify bird species often (or rarely) observed in the presence of other bird species. I also assessed observed bird biomass as part of a strategy to estimate habitat carry capacity and generate insights into the likely flow of energy and material through the ecosystem. Using network theory adds considerable value to baseline biodiversity monitoring studies. After all, community ecology studies emphasise the need to study species and their interactions simultaneously (Delmas et al 2019), and network theory is an intuitive way of quantitating and visualising interactions (or ‘edges’) between entities (or ‘nodes’).

How does one go about reconstructing a species co-occurrence network from a data matrix of bird observations, where every column is a sampling date and every row a species? The approach taken here deployed the Partial Correlation and Information Theory (PCIT) algorithm of (Reverter and Chan 2008). This algorithm was invented for inferring co-expression networks of messenger RNA in the field of genomics. It is a soft thresholding method that exhaustively explores all trios (here there are 109,736 possible trios from the initial set of 88 species) to estimate the significance of the Pearson correlations between all 3,828 co-occurring bird species pairs. PCIT is uniquely powerful because it systematically assesses local data structure to help disentangle the truly correlated pairs of birds from the co-occurrences that are merely auto-correlated and therefore most likely spurious. It also makes no assumption about final network topology, producing a non-random network, such as one with a scale free connectivity distribution, only if it actually exists rather than by force. The resultant bird co-occurrence network was clustered, visualised, mapped and statistically analysed using the genomics software Cytoscape (Shannon, Markiel et al. 2003).

The structure – or topology - of the entire network can then be used as a baseline to help understand the ‘whole-of-system’ or holistic impact of subsequent interventions such as agriculture, fire and feral pests. Here, I hypothesised that 1) cumulative observed bird biomass would be similar during successive sampling dates reflecting the carrying capacity of the habitat at that time of year, even when the identity of the component species markedly differs 2) the temporal co-occurrence network would follow a scale-free connectivity distribution in line with most ‘real world’ networks, such as the structure of the internet, social networks and gene regulatory networks 3) that birds of prey would negatively co-occur with most other bird species and 4) that functionally related species of birds would positively co-occur in cohesive modules.

Site

Hiddenvale Research Station is a collaborative teaching and research venture between the philanthropic Turner Family Foundation and The University of Queensland located in the Lockyer Valley region of South East Queensland, Australia (latitude − 27.7287, longitude 152.467). It experiences a subhumid and subtropical climate with relatively long hot summers and relatively mild winters with occasional frosts. The site comprises a large tract of protected land but also incorporates a) some managed livestock grazing in target areas and b) some ecotourism activities, particularly mountain bike trails.

Bird occurrence data

Bird observation data from 2019 was downloaded from The Global Biodiversity Information Facility GBIF (https://www.gbif.org/ ) for the following dates: May 2nd (Autumn), July 9th, July 10th (mid-Winter), August 11th, August 12th (late Winter) and September 24th, 25th, 28th, 29th and 30th (Spring)). The opportunities afforded by online portals such as GBIF are discussed here (Petersen, Speed et al. 2021). The GBIF bird occurrence observations are made by interested and experienced members of the wider community. After cleaning the data such that only the most abundant observation for a given species was used in cases with multiple observers, a data matrix was produced with as many columns as dates (n = 10) and as many rows as observed bird species (n = 88) (Supplementary Data File 1). Of the 399 avian species documented in the Lockyer Valley region of South East Queensland, 88 were detected at HVRS in 2019 according to GBIF.

Bird variables

Two additional bird variables were assessed for deeper study, the continuous variable of species body mass (g) and the binary variable of whether or not a species is a raptor. Species biomass data (based on reported adult body mass) was derived from literature searches (Supplementary Data File 1). This phenotype was explored as part of a strategy to connect bird occurrence observations to habitat carrying capacity and the flow of energy and materials through the ecosystem. Secondly, there are 7 raptors in the data (Collared Sparrowhawks Accipiter cirrocephalus, Brown Goshawks Accipiter fasciatus, Gray Goshawks Accipiter novaehollandiae, Square Tailed Kites Lophorctinia isura, Whistling Kites Haliastur spehnurum, White Bellied Sea Eagles Icthyophaga leucogaster and Wedge Tailed Eagles Aquila audax) whose ecological position as avian predators is well documented. The latter categorisation means a particular subset of large, hypercarnivorous birds with hooked beaks and powerful talons is placed under explicit consideration as an ecological guild.

Other non-raptorial carnivorous (e.g. Laughing Kookaburras, Dacelo novaeguineae and Australian Pelicans Pelicanus conspicillatus), primarily piscivorous (e.g. White faced Herons Egretta novaehollandiae and Sacred Kingfishers Todiramphus sanctus) and insectivorous (e.g. Australian magpies Gymnorhina tibicen) bird species that therefore occupy a similar trophic level (i.e. that of secondary or tertiary consumers) to that of the raptors have not been included in this particular designation.

Statistical properties of the species occurrence data matrix

Basic descriptive statistics were computed within Graphpad Prism software version 10.2.2.

Presence / absence matrix

The occurrence matrix (where each cell represents the number of individuals of a given species that were observed) was converted into a binary species presence / absence matrix within Microsoft Excel version 16 using the “=IF” function (Supplementary Data File 2).

Observed bird species biomass

The cumulative observed biomass for each species across the 10 dates and for all species in a given date was computed within Microsoft Excel version 16. The cumulative biomass for all species on each of the 10 dates was represented as a bar chart using Sigmaplot version 15.0.

Hierarchical clustering

In order to estimate the relationships among the 10 dates based on shared patterns of bird observations, a hierarchical clustering approach was used. This produces a dendrogram of relationships, awarding the shortest branch lengths to those relationships that are strong and positive, and longer branch lengths to those that are either a) weak and positive or b) negative. The transposed data matrix was imported into Permut Matrix software version 1.9.3 (Caraux and Pinloche 2005), normalised on columns and subject to hierarchical clustering on both rows (dates) and columns (bird species). Euclidean Distance was used as the Dissimilarity metric and McQuitty’s method for the Linkage Rule. The resultant clustered matrix is displayed as a heatmap on a continuous red to green colour spectrum, where bright red denotes a large number of occurrences and bright green the converse.

Co-occurrence network reconstruction

In this network reconstruction all 88 observed bird species were included in the source file; there was no a priori pre-screen for associations deemed more likely based on ecological expectation. PCIT (Reverter and Chan 2008) was used for reconstruction. PCIT is a soft thresholding method that systematically explores local data structure to disentangle truly correlated pairs from those that are merely auto-correlated and therefore likely spurious.

This algorithm was originally developed with a genomics application in mind (mRNA co-expression) but the concepts of mRNA co-expression and species co-occurrence are analogous. Indeed, PCIT can be considered a universal network reconstruction method for any situation where pairwise correlations between entities can be reliably computed. This approach is uniquely powerful and hypothesis-free as it makes no assumption about final network structure. For example, a scale-free connectivity distribution common to many ‘real world’ networks will only emerge if it really exists, not by force or assumption. The partial correlations of all trios are computed, in this case there are (88*87*86)/6 or 109,736 bird species trios under consideration. The correlations are computed across the 10 time points, hence this is a temporal (rather than spatial) co-occurrence network.

The output of the PCIT algorithm is a 3 column file with as many rows as there are pairwise comparisons (88*87/2 or 3,828 in this case), with the first column being the first bird species of the pair, the second column being the second bird species of the pair and the 3rd column being either the Pearson correlation (if deemed significant by PCIT) or a ‘0’ if deemed not significant by PCIT. After filtering to remove all the non-significant pairs and adding an extra column denoting whether the significant Pearson correlations were either positive or negative, this file was imported into Cytoscape software version 3.8.2 as the network file (Shannon, Markiel et al. 2003).

Within Cytoscape the resultant network was clustered using the ‘Organic Layout’ clustering algorithm. This centralises nodes that are ‘influentially connected’, that is they allow the network to be spanned in a relatively small number of steps. It also produces cohesive ‘modules’ of groups of nodes that are highly inter-connected. Thus, functional assemblages of interacting species (whether positively or negatively related) would be expected to appear as modules within the network. Network analysis was then used to compute the connectivity distribution of the whole network and also a set of topological statistics, including ‘radiality’ or network influence, for each individual node present in the network. The network was visually mapped such that node size reflects bird mass, node colour denotes whether or not a bird is a raptor and edge colour reflects whether that particular pairwise co-occurrence is negative (red) or positive (green) (Supplementary Data File 3).

Average pairwise correlation

For each bird species its average pairwise correlation to all other 87 bird species was computed. After sorting on this average the extent of enrichment of Birds of Prey at the extreme negative end of the list was estimated using the = hypergeom.dist function within Excel. For 10 exemplar bird species (the 5 species bearing on average the most positive co-occurrences and conversely those 5 bearing the most negative) the pairwise correlations distributions were represented as box plots using Sigmaplot,

Statistical properties of the species occurrence data matrix

The basic descriptive statistics of the occurrence data matrix are to be found in Supplementary Data File 4.

Presence / absence data matrix

The α diversity, or species richness, ranged from 8 to 50 across the 10 sampling dates (Fig. 1 Panel A). On September 29th and 30th, two dates where ~ 38 kg of birds are observed in total (Fig. 1 Panel B), the α diversity is 5-fold different. On the 29th the biomass of the 39 contributing avian species are somewhat evenly represented whereas on the 30th the biomass of only 8 species are detected, and biomass is almost entirely dominated by Australian Pelicans (Fig. 2).

Observed bird species biomass

The highest cumulative bird biomass across all species is observed in the Austral Autumn, the lowest in Winter, with the various Spring sampling values intermediate (Fig. 1 Panel B).

Hierarchical clustering

Successive sampling dates (July 9 and 10, September 24 and 25, September 28 and 29, August 11 and 12) are awarded the shortest pairwise branch lengths in the dendrogram indicating high similarity in the overall pattern of observed birds (Fig. 3). The only Autumn sampling date (May 2) is awarded its own branch in the tree indicating a somewhat unique composition of observed species. Season, therefore, has a substantial (and generally repeatable) influence on the observed pattern of bird species at this location.

Patterns of Co-occurrence

Two different examples of significant pairwise species co-occurrence, one significantly positive (Little Friarbirds with Torresian Crows) and one significantly negative (Whistling Kites with Superb Fairywrens Malurus cyaneus), are shown in Fig. 4.

After sorting on average pairwise co-occurrence the birds of prey are significantly enriched at the extreme negative end of the list when viewed collectively as a functional group (Hypergeometric P value = 0.00029). The distributions in pairwise co-occurrence for the extreme 10 birds are shown in Fig. 5. Those birds with extreme negative average co-occurrences possess much tighter overall distributions than their positive counterparts.

A set of 636 co-occurrences were deemed significant by PCIT. Out of the total possible pairwise comparisons of 3,828, this represents a percentage of 17%. The overall distribution of pairwise co-occurrences is shown in Fig. 6. The mass of the data is not centred on 0, but rather skewed towards the negative, indicating that in general the presence of one bird species is at the expense of others. On the other hand, when focussing on the extremes of the distribution there are many more very strong positives than there are strong negatives.

The co-occurrence network (Fig. 7) comprises all 88 birds connected by 636 edges. Among other things this means no bird species was entirely absent from the network as every species possessed at least one significant co-occurrence to another species.

In terms of overall topology, the network has a tendency towards a bimodal rather than scale-free connectivity distribution (Supplementary Fig. 1). There is a very cohesive module comprising numerous positive co-occurrences dominated by birds of very small mass such as Silvereyes, White-browed Scrub Wrens Sericornis frontalis, Little Friarbirds and Red Backed Fairy Wrens. 6 of the 7 birds of prey are found on the periphery of the network, the one exception being the Wedge Tailed Eagle, Aquila audax.

There is also a tendency for other large birds (namely Australian Pelicans Pelecanus conspicillatus, Australasian Darters Anhinga novaehollandiae, Pied Cormorants Phalocrocorax varius, Little Black Cormorants Phalocrocorax sulcirostris and Royal Spoonbills Platalea regia) to be found on the periphery. Birds with the highest radiality (or network ‘influence’ score) can be found in Table 1.

A temporal co-occurrence network has been reconstructed for a subtropical group of 88 sympatric Australian birds at a wildlife refuge in South East Queensland. This network is the first of its kind in ecology to make use of the PCIT network inference algorithm. Species occurrence records are a cornerstone of community ecology and biodiversity assessment. When taken in conjunction with species biomass, occurrence data can generate insights into the flow of energy and biological material through ecosystems.

The main findings are that a) the distribution of all co-occurrences is skewed to the negative rather than being centred on 0, indicating the presence of one bird species tends to exclude another broadly in line with a competitive exclusion model b) total observed bird biomass is similar on successive sampling dates (e.g. ~38 kg on the 29^th and 30^th September) indicating the habitat has an upper limit on the birdlife it can support, although α diversity and the component species may differ enormously from one day to the next c) raptors tend to possess many negative co-occurrence relationship with other avian species indicating that their presence is strongly inhibitory d) a cohesive ‘module’ or cluster of positively co-occurring diminutive bird species is clearly detected and e) the network as a whole deviates from the expectation of a scale-free connectivity distribution for unknown reasons.

The co-occurrence output represents an abstraction of bird community ecology via an estimate of temporal co-existence. It can be contrasted with, say, an empirically derived food web which is a ‘real’ network anchored on bona fide documented patterns of consumption (and therefore of empirically proven energy and material transfer). Here, the fact we are using an abstraction means some of the co-occurrence observations may be incidental, in which case their ecological meaning is less clear. Nevertheless, the partial correlation approach taken does represent a step towards minimising the detection of spurious interactions arising from auto-correlation.

In taking this abstract approach one can create new hypotheses regarding bona fide species interactions such as those driven by niche competition and predator-prey relationships. Moreover, 88 bird species is a large enough assemblage to make a co-occurrence network that has a complex yet meaningful topology. That is, it can be subject to a formal analysis of its constituent modules and one can quantitative features of individual nodes which may reveal new biology. The co-occurrence network approach adds value to baseline species monitoring activities as it explicitly quantitates species interactions from the basic isolated ‘species-level’ observations that have been made.

With this in mind, having a quantitative understanding of bird species co-occurrence will a) help assess the impact of future interventions such as duration and intensity of fire, livestock agriculture and the behaviour and populations of feral pests and b) can be used as a model for how one might quantitate complex interactions in other taxa for whom reasonably large and reliable data sets also exist.

Other authors have previously used the basic concept of co-occurrence in the context of bird ecology. For example, patterns of bird species co-occurrence have been estimated for avian assemblages in the Black Forest of Germany (Basile, Asbeck et al. 2021), Mediterranean wetlands (Fois, Cuena-Lombrana et al. 2022), Alpine environments (Garcia-Navas, Sattler et al. 2021) and Spanish Steppes (Barrero, Ovaskainen et al. 2023). However, none of these examples report a bird species co-occurrence network whose detailed topology has been clearly visualised, formally analysed using principles from network theory and submitted in a format that can be interrogated by other members of the scientific community. Cytoscape software is a freely available community resource and the .cys file has been made available as part of this publication such that interested readers can visualise and navigate the network themselves.

In terms of overall topology there are two striking observations. Firstly, that there is a cohesive module of positively co-occurring species dominated by diminutively sized non-threatening birds such as Red Backed Fairy Wrens, Little Friarbirds, Silvereyes, White Browed Scrub Wrens and Willie Wagtails Rhipidura leucophrys. And secondly, that Birds of Prey (Gray Goshawks, Brown Goshawks, Collared Sparrowhawks, White Bellied Sea Eagles and Square Tailed Kites) and indeed a number of other non-raptorial large sized birds (such as Australasian Darters, Australian Pelicans, Pied Cormorants, Royal Spoonbills, Little Black Cormorants, Laughing Kookaburras, Glossy Black Cockatoos Calyptorhynchus lathami and Channel Billed Cuckoos Scythrops novaehollandiae), tend to be located on the periphery of the network. This indicates large (and sometimes aggressive) birds either possess a small number of connections and / or the connections they do have do not allow the network to be easily traversed.

Viewing the pattern of all pairwise correlations, whether deemed significant by PCIT or not, reinforces the view that Birds of Prey tends to possess negative co-occurrences to the other 87 bird species with whom they are being compared. This is consistent with their presence causing other birds to either flee the environment completely or to seek refuge such that they cannot be detected by an observer. Further to this, the distribution of all pairwise correlations (based on all 88 bird species, not just raptors) is skewed to the negative, indicating that in general the presence of a given avian species tends to inhibit the presence of other species.

This latter observation is consistent with niche competition and the concept that any habitat possesses a carrying capacity that represents the upper limit for a given ecological guild. This resonates with the arguments of (Diamond 1975) regarding forbidden species combinations of birds in the Bismark archipelago and also the observations of (Barrero, Ovaskainen et al. 2023) where the bird assemblage in a natural Steppe in central Spain is configured around the Eurasian skylark which possesses principally negative competitively driven associations with many of the coexistent species.

The season with the highest cumulative observed bird biomass is Autumn and the lowest is Winter, with Spring sampling dates intermediate. These differences presumably relate to some combination of seasonally based bird migration patterns and also the restricted availability of energetic resources in the form of nuts, seeds, fruit and other important dietary components that may vary on a seasonal basis. The hierarchical clustering analysis indicates that Autumn has the most unique assemblage of observed birds, and that when sampling does occur on consecutive days it is very likely to lead to similar patterns of bird observations.

In this subtropical, native woodland habitat at Hiddenvale, a reliable indicator species for the presence of a wide diversity of other birds looks to be the Silvereye. This species is not only present in the cohesive module of positively co-occurring species within the network but also possesses the highest total number of significant connections, the highest radiality (or network influence) and the second most positive average pairwise co-occurrence. In Australia, Silvereyes are widely distributed small, omnivorous passerine birds. They breed in September to December (Austral Spring to Summer) and in late summer, South East Queensland populations migrate further north. These co-occurrence findings suggests a hypothesis that deliberately creating habitat conducive to populations of Silvereyes may represent a rational means of promoting a high biodiversity of numerous other bird species at Hiddenvale reserve.

The various analytic approaches used here (hierarchical clustering, network inference and hypergeometric enrichment) capture numerous ecological relationships (such as the predation threat created by the presence of Raptors both as individual birds and also as a collective functional group) that would likely be considered ‘real’ to ecologists but that in many cases would not pass significance thresholds by conventional statistical measures. This highlights the power of these more ‘holistic’ methods to reliably detect cryptic, ecologically relevant signals in observation data that undoubtedly possesses some inherent noise.

NJH declares that no funds, grants or other support was received during the preparation of this manuscript. NJH declares that he has no financial interests. NJH conceived of this study, performed the analyses and composed the manuscript.

Funding declaration

No funding was received for this research.

Author Contribution

NJH conceived of the research, performed the analyses, interpreted the data and composed the manuscript.

Acknowledgement

The author acknowledges Megan Brady, Peter Murray, Greg Baxter and Annabel Smith for stimulating conversations.

Data Availability

Data is provided within the manuscript and supplementary information files.

"Birds of Lockyer Valley local government area." from https://wetlandinfo.des.qld.gov.au/wetlands/facts-maps/wildlife/?AreaID=lga-lockyer-valley&Kingdom=animals&Class=birds.
Barrero, A, O Ovaskainen, J Traba and J Gomez-Catasus (2023). Co-occurrence patterns in a steppe bird community: insights into the role of dominance and competition. Oikos(5): e09780.
Basile, M, T Asbeck, JM Cordeiro Pereira, G Mikusinski and I Storch (2021). Species co-occurrence and management intensity modulate habitat preferences of forest birds. BMC Biol 19(1): 210.
Caraux, G and S Pinloche (2005). PermutMatrix: a graphical environment to arrange gene expression profiles in optimal linear order. Bioinformatics 21(7): 1280-1281.
Diamond, JM (1975). Assemblies of species communities. Ecology and evolution of communities. ML Cody and JM Diamond. Cambridge, Massachusetts, Harvard University Press.
Elton, CS (1958). Ecology of invasions by animals and plants. London, Chapman and Hall.
Fois, M, A Cuena-Lombrana, C Zucca, N S. and G Bacchetta (2022). Investigating plant-bird co-occurrence patterns in Mediterranean wetlands: can they reveal signals of Ecosystem connectivity. Diversity 14.
Garcia-Navas, V, T Sattler, H Schmid and A Ozgul (2021). Bird species co-occurrence patterns in an alpine environment supports the stress-gradient hypothesis. Oikos 130(11): 1905-1918.
Hooper, DU, FS Chapin III, JJ Ewel, A Hector, P Inchausti and DA Wardle (2005). Effects of biodiversity on ecosystem functioning: a consensus of current knowledge. Ecological Monographs 75(1): 3-35.
Ives, AR and SR Carpenter (2007). Stability and diversity of ecosystems. Science 317(5834): 58-62.
Kremen, C (2005). Managing ecosystem services: what do we need to know about their ecology? Ecol Lett 8(5): 468-479.
May, RM (1972). Will a large complex system be stable? Nature 238(5364): 413-414.
Odum, EP (1953). Fundamentals of Ecology. Philadelphia, Saunders.
Petersen, TK, JDM Speed, V Grotan and G Austrheim (2021). Species data for understanding biodiversity dynamics: the what, where and when of species occurrence data collection. Ecological Solutions and Evidence 2: e12048.
Reverter, A and EK Chan (2008). Combining partial correlation and an information theory approach to the reversed engineering of gene co-expression networks. Bioinformatics 24(21): 2491-2497.
Shannon, P, A Markiel, O Ozier, NS Baliga, JT Wang, D Ramage, N Amin, B Schwikowski and T Ideker (2003). Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 13(11): 2498-2504.

Table 1 The 10 bird species with the highest radiality (network ‘influence’) scores as computed by Cytoscape software (data not shown). Radiality is somewhat independent of the total number of connections indicated in the third column of the table, that is a particularly ‘well connected’ bird may or may not possess as many total connections as a less ‘well connected’ bird.

Species	Body mass (g)	Number of significant connections (by PCIT)
Silvereye	10	33
Golden Whistler	30	28
Little Pied Cormorant	860	30
Pied Butcherbird	120	19
Superb Fairywren	11	21
Torresian Crow	565	31
Pied Currawong	300	15
Pacific Black Duck	1070	26
Varied Sitella	13	25
Noisy Miner	75	25

No competing interests reported.

Download PDF

Version 1

posted

You are reading this latest preprint version

Temporal Co-Occurrence in a Community of Subtropical Australian Birds

Status:

Version 1

Abstract

Figures

Introduction

Materials and Methods

Site

Bird occurrence data

Bird variables

Statistical properties of the species occurrence data matrix

Presence / absence matrix

Observed bird species biomass

Hierarchical clustering

Co-occurrence network reconstruction

Average pairwise correlation

Results

Statistical properties of the species occurrence data matrix

Presence / absence data matrix

Observed bird species biomass

Hierarchical clustering

Patterns of Co-occurrence

Discussion

Declarations

Author Contribution

Acknowledgement

Data Availability

References

Table

Additional Declarations

Supplementary Files

Status:

Version 1