Greuel M Nsc 2019.Pdf (8.792Mb)

Home , Crotaphopeltis

Foraging ecology of Naja nivea and Dispholidus typus

Janine Greuel

MSc degree

Supervisor: Dr Bryan Maritz

Submitted in fulfilment of the requirements for the degree of Master of Science

Department of Biodiversity and Conservation Biology

University of the Western Cape

Bellville, South Africa

November 2019

http://etd.uwc.ac.za/

Abstract

It is widely reported that snakes can be major predators of avian nests, but the use of a single avian prey type by competing species has rarely been examined. This study aimed to investigate predation of a single food resource by the sympatric snakes Naja nivea and Dispholidus typus. Specifically, I aimed to 1) identify factors influencing snake presence in sociable weaver colonies and 2) quantify snake predation and potential differences in the consumption of prey by the two competing snakes.

I used repeated visual surveys of sociable weavers to obtain presence-absence data of cape cobra and boomslang in sociable weaver colonies over an entire breeding season. I related the presence-absence data of the two snake species to spatially- and temporally-variable factors using principal component analyses (PCA) and multiple logistic regression analyses. The presence of snakes in sociable weaver colonies is primarily influenced by temporal factors, but spatial factors also play a role. The results indicate that cape cobra forage on warm and windy days, primarily in larger colonies, while greater breeding activity and lower wind speed results in increased nest visitation by boomslang.

Additionally, I examined snake predation in sociable weavers at a finer scale, specifically identifying differences in resource utilization by the two snake species by quantifying consumption rates. Although both species consumed eggs and chicks, my results showed a significant difference in predation events across the two snake species, with boomslang consuming more prey items than cape cobra. The snake species consumed similar numbers of chicks, but boomslang consumed significantly more eggs than did cape cobra. As a result, the mean age of prey consumed by cape cobra was significantly greater than that of boomslang, likely because boomslang arrive at colonies sooner than cape cobra after a new breeding attempt by the birds. My work has given novel insight into the predation by snakes in a sociable weaver system and identified significant differences in the foraging ecology of these two competing snakes.

Keywords: foraging ecology, movement activity, nest predation, interspecific competition, Philetairus socius, resource partitioning, visual surveys, principal component analysis, consumption rates, presence-absence

http://etd.uwc.ac.za/ii

Declaration

I declare that “Foraging ecology of Naja nivea and Dispholidus typus” is my own work, that it has not been submitted for any degree or examination at any university, and that all sources I have used or quoted have been indicated and acknowledged by complete references.

Full name: Janine Greuel Date: November 2019

Signature:

http://etd.uwc.ac.za/iii

Acknowledgements

First I would like to thank my supervisor Dr Bryan Maritz for providing me with the opportunity to conduct research on some of the most intriguing organisms I‘ve come across. Additionally, I greatly appreciate his guidance and knowledge that helped me to produce this thesis, and the understanding Bryan showed during the more challenging times that I faced during my MSc.

During this thesis I received student funding from the National Research Foundation of South Africa (NRF), which paid for my study fees and living costs. The project itself was funded by the NRF through a research grant provided to Bryan Maritz.

The project would not have been possible without the collaborations with the De Beers Mining Corporation and the E Oppenheimer & Son (PTY) Limited. I am grateful to them and Duncan MacFadyen for providing me the permission to conduct my research at the Benfontein Nature Reserve, which is one of their conservation areas. Furthermore, I would like to thank Corne Anderson for providing me with the shape file of Benfontein. Special thanks go to the wildlife managers Piet Oosthuizen and Corne Anderson for the opportunity to join them during game captures at Benfontein and other De Beers nature reserves and to learn more about wildlife conservation and management.

I thank Rita Covas and her research team for their collaboration with my project. I appreciate everyone that was involved collecting in the field work and providing data on the sociable weaver metapopulation of Benfontein. Special thanks goes to Thomas Pagnon and Hugo Pereira, the research managers, who were always on point with leading the team and helpful with my enquiries. Furthermore, I’m grateful to all the other researcher assistants and students that I have met and that helped during my field work: Juliette Linossier, Cécile Vansteenberghe, André Ferreira, Tanja Van de Ven, Liliana Silva and Alexandre Baduel.

Huge appreciation goes to the South African Weather Service for providing me with the required weather data for my study area. Without this data, my project would have been significantly smaller and I couldn’t have generated these results.

I would also like to thank Jody Michael Barends for his assistance in creating a map of my study area and Luvoyo Kani for providing modelled distribution maps for my two study species.

http://etd.uwc.ac.za/iv

I thank Tyrone James Ping, Bryan Maritz, Pierre De Jager, Deon Oosthuizen, Marcel Witberg, Barry Kleynhans, Gary Cusins, Marius Koekemoer, Ed Raubenheimer for their permission to use their photographs of cape cobra, boomslang and sociable weaver in my thesis.

Last but not least, I would like to thank my family in Germany for their encouragement and interest in my project. I’m especially grateful for their support during the difficult times when I needed them most.

http://etd.uwc.ac.za/v

List of Figures Figure 1.1: Locations of the studied sociable weaver (Philetairus socius) colonies in the southeast of the Benfontein Nature Reserve (28°52’S, 24°51’E), which is located on the border of the provinces of the Northern Cape and the Free State, South Africa. The block dots represent the study colonies with their corresponding colony ID. Image from Google Earth.

Figure 1.2: Distribution map of cape cobra (Naja nivea) considering historical sampling intensity and general habitat within southern Africa.

Figure 1.3: Colour variation and ontogenetic shift in colouration of the cape cobra (Naja nivea). See text for details of each image.

Figure 1.4: Distribution map of African boomslang (Dispholidus typus) considering historical sampling intensity and general habitat within southern Africa.

Figure 1.5: Ontogenetic and sexual dichromatism in boomslang (Dispholidus typus) showing colour variations between males and females, as well as juveniles and adults. See text for details of each image.

Figure 1.6: Distribution map of sociable weaver (Philetaius socius) considering historical sampling intensity and general habitat within southern Africa.

Figure 1.7: Sociable weaver (Philetairus socius) nest structure with a foraging cape cobra (Naja nivea) moving between chambers at the Benfontein Nature Reserve, Kimberley, NC, South Africa.

Figure 1.8: Morphology of the sociable weaver (Philetairus socius) showing adult birds with distinct dark throat patch markings (A, B), a nestling (17 days old) without throat markings (C) and hatchlings (D).

Figure 1.9: A sociable weaver (Philetairus socius) colony protected from snakes using clingwrap around the trunk of the host, camelthorn tree (Vachellia erioloba).

Figure 1.10: Colony check and tagging of sociable weaver (Philetairus socius) colonies at the Benfontein Nature Reserve. Yellow tags aid in the identification of all chambers in a colony (A) required for any checks conducted during daily field work (B).

http://etd.uwc.ac.za/vi

Figure 1.11: Cape cobra (Naja nivea) and boomslang (Dispholidus typus) in sociable weaver colonies. A – E were individuals found at the Benfontein Nature Reserve, while the green specimen (F) is from the Tswalu Game Reserve in the Kalahari.

Figure 2.1: Heat map matrix of Pearson’s correlation coefficients for multiple spatial and temporal variables.

Figure 2.2: Screeplots of the proportion of the spatial variability (A) and temporal variability (B) found in the data collected for the data set including all 13 sociable weaver (Philetairus socius) colonies at the Benfontein Nature Reserve. The total amount of variation for the spatial variables is explained by six spatial principal components with a distinct elbow visible after the inclusion of three PCs. The total amount of variation for the temporal variables is explained by seven temporal principal components.

Figure 3.1. Mass of sociable weaver offspring in relation to their age (in days). Day 0 refers to eggs and anything after day 1 refers to the age of chicks. Polynomial trendline shows an almost linear growth curve for nestlings.

Figure 3.2. Outcome of all eggs (n = 3129) produced across study colonies (n = 13) throughout the study period. The fate of offspring was categorized into i) offspring was predated by cape cobra (Naja nivea), ii) offspring was predated by boomslang (Dispholidus typus), iii) offspring failed but the cause was not known or iv) offspring fledged successfully.

Figure 3.3. Comparison of the percentage of the reproductive failure rates for individual colonies disregarding the fate (dark grey) and the proportion of such failures attributable to snakes (light grey). Colony number with an asterisk (*) indicates that a colony was protected from snakes. Numbers above bars indicate the total number of reproductive effort per colony over entire breeding period (i.e. total number of eggs laid).

Figure 3.4. Comparison of lost sociable weaver (Philetairus socius) offspring for A) all study colonies (n = 13) and B) all colonies predated by snakes at least once (n = 11) based on four predation assumptions. Failed sociable weaver breeding attempts (defined as number of eggs lost during the study period) were nA = 2607 and nB = 2401, respectively. The proportion of the total lost sociable weaver offspring are based on the following assumptions: i) lost to snakes observed at a colony, ii) predated by observed snakes and any events when four or more ( 4) breeding

http://etd.uwc.ac.za/vii chambers were empty, iii) predated by observed snakes and any events when two or more ( 2) breeding chambers were empty, iv) predated by observed snakes and any events when one or more ( 1) breeding chambers were empty.

Figure 3.5. Break down of the fate of all eggs laid across sociable weaver colonies (n = 11), predated at least once, during the study period. The stacked bars show a) the proportion of total offspring successfully fledged versus failed, b) the proportions of the total offspring failures being predated by snakes or not, and c) the proportions of consumed offspring by Naja nivea (n = 35) and Dispholidus typus (n = 19) of the total snake predation.

Figure 3.6. Histograms showing the distribution of the data for the number of chambers and prey items predated by Dispholidus typus (A, C) and Naja nivea (B, D). The numbers of prey items predated were binned.

Figure 3.7. Consumption of sociable weaver (Philetairus socius) offspring for 35 predation events by cape cobra (Naja nivea) and 19 events for boomslang (Dispholidus typus) showing the mean (± SE) depredation of chambers and prey items (i.e. eggs and chicks combined). The asterisk (*) indicates a significant statistical difference in predated chambers/prey items compared between the two snake species.

Figure 3.8. Consumption of sociable weaver (Philetairus socius) offspring for 35 predation events by cape cobra (Naja nivea) and 19 events for boomslang (Dispholidus typus) showing the mean (± SE) consumption of eggs and chicks. The asterisk (*) indicates a significant statistical difference in predated prey items compared between the two snake species.

Figure 3.9. Mean age of sociable weaver (Philetairus socius) offspring consumed by cape cobra (Naja nivea) and boomslang (Dispholidus typus) at Benfontein Nature Reserve. The asterisk (*) indicates a significant statistical difference of the mean age consumed by the two snake species.

Figure 3.10. Histograms of the age of sociable weaver chicks consumed by (A) cape cobra (Naja nivea) and (B) boomslang (Dispholidus typus) at Benfontein Nature Reserve. Sample sizes for predation events were n = 19 and n = 35, respectively.

http://etd.uwc.ac.za/viii

Figure 3.11. Time of arrival at sociable weaver colonies (Philetairus socius) after new breeding attempt by the communal breeders by cape cobra (Naja nivea) and boomslang (Dispholidus typus) at Benfontein Nature Reserve. The asterisk (*) indicates a significant statistical difference in the time of arrival between the two snake species.

Figure 3.12. Mass consumption measures for true predation events by cape cobra (n = 35) and boomslang (n = 19). Box and Whisker plot including outliers and means symbolised by dots and crosses, respectively.

http://etd.uwc.ac.za/ix

List of Tables Table 2.1: Summary of the number of observations and the snake occurrences for each study colony at the Benfontein Nature Reserve. The total values (Σ) display the data sets including all colonies (n = 13) and excluding protected colonies (n = 9).

Table 2.2: Comparison of three different methods to retain principal components explaining spatial and temporal variables for 13 Philetairus socius colonies.

Table 2.3: Eigenvalues, proportion of variance explained and cumulative proportion of variance explained for retained temporal and spatial variables for 13 Philetairus socius colonies.

Table 2.5: Eigenvectors/Factor loadings of retained principal components for temporal and spatial variables across 13 sociable weaver colonies collected at Benfontein Nature Reserve.

Table 2.7: Cumulative Proportion of variance for spatial and temporal principal components for the data set excluding protected sociable weaver (Philetairus socius) colonies.

Table 2.8: Results from multiple regression analysis for all retained temporal and spatial principal components for cape cobra (Naja nivea) for the data set excluding protected sociable weaver (Philetairus socius) colonies (n = 9). The asterix (*) indicates the significance of a variable (P < 0.05).

Table 2.9: Eigenvectors/Factor loadings of retained principal components for temporal and spatial variables across 9 sociable weaver colonies collected at Benfontein Nature Reserve.

Table 2.10: Results from multiple regression analysis for all retained temporal and spatial principal components for boomslang (Dispholidus typus) for the data set excluding protected

http://etd.uwc.ac.za/x sociable weaver (Philetairus socius) colonies (n = 9). The asterix (*) indicates the significance of a variable (P < 0.05).

Table 3.1: General information on the individual colonies. Protection is given in binary code with 1 equal to protected and 0 being non-protected.

http://etd.uwc.ac.za/xi

Table of Contents

Chapter 1: The Importance of predators ...... 1

1.1 Top-down control ...... 1 1.1.1 Case study of the Gray wolves at Yellowstone National Park ...... 2 1.1.2 Case study of the Brown tree snake of Guam ...... 4 1.1.3 Case study of the Burmese Python in the Everglades National Park ...... 6

1.2 Study system ...... 8

1.3 Measuring predation rates ...... 10

1.4 Approach ...... 11

1.5 Study site ...... 11

1.6 Study species ...... 13 1.6.1 Cape cobra ...... 13 1.6.2 Boomslang ...... 16 1.6.3 Sociable weaver ...... 18

1.7 Methods ...... 21 1.7.1 Terminology ...... 21 1.7.2 Study period ...... 23 1.7.3 Bird data ...... 23 1.7.4 Snake data ...... 26

1.8 Structure of thesis ...... 29

Chapter 2: Correlates of colony visitation and predation by the snakes Naja nivea and Dispholidus typus ...... 30

2.1 Introduction ...... 30 2.1.1 Influences on snake activity ...... 30 2.1.2 Nest predation ...... 31 2.1.3 Effect of climate change on avian reproduction ...... 32

2.2 Methods ...... 33 2.2.1 Field methods ...... 33

http://etd.uwc.ac.za/xii

2.2.2 Statistical analysis ...... 36

2.3 Results ...... 38 2.3.1 Temporal and spatial collinearity ...... 38 2.3.2 Data analysis including all 13 colonies ...... 39 2.3.3 Data analysis excluding protected colonies ...... 44

2.4 Discussion ...... 47 2.4.1 Influences on snake presence ...... 47 2.4.2 Temporal niche partitioning ...... 51 2.4.3 Sampling biases and limitations ...... 52

2.5 Conclusion ...... 54

Chapter 3: Predation rates and differential food resource utilization by two nest predators, Naja nivea and Dispholidus typus...... 55

3.1 Introduction ...... 55

3.2 Methods ...... 59 3.2.1 Field methods ...... 59 3.2.2 Statistical analysis ...... 60

3.3 Results ...... 64 3.3.1 Chambers and prey items ...... 68 3.3.2 Eggs and chicks ...... 70 3.3.3 Mean age of prey...... 71 3.3.4 Prey availability...... 72 3.3.5 Time of arrival ...... 72 3.3.6 Mass consumption ...... 73

3.4 Discussion ...... 74 3.4.1 Nest mortality ...... 75 3.4.2 Predation by cape cobra and boomslang ...... 76 3.4.3 Hypotheses for differential resource utilization ...... 78 3.4.4 Sampling biases and limitations ...... 80 3.4.5 Impact of snakes on prey ...... 81

3.5 Conclusion ...... 82

Chapter 4: Synthesis of the findings...... 83

http://etd.uwc.ac.za/xiii

4.1 Competing for a singular resource ...... 83 4.2 Recommendations for data collection ...... 84 4.3 Conservation planning ...... 85 4.4 Future work ...... 85

Literature cited ...... 87

http://etd.uwc.ac.za/xiv

Chapter 1: The Importance of predators

1.1 Top-down control

For decades, ecologists have debated about the driving forces shaping the amount of biomass within and among ecosystems (Hairston et al. 1960, White 1978, Leroux & Loreau 2015). A multitude of abiotic and biotic factors impact populations and communities of organisms (Hunter & Price 1992). The literature suggests two non-exclusive processes—top-down and bottom-up control—regulating population size of organisms (Leroux & Loreau 2015). According to the bottom-up perspective, resources (e.g., nitrogen and phosphorus) and primary producers control the biomass of all trophic levels in the food web from below (resource driven/limited; White 1978, Hunter & Price 1992, Gruner et al. 2008). The opposing view (top-down control) states that consumers such as predators and herbivores are the driving force in regulating the biomass of lower trophic levels from above (Legagneux et al. 2012, Leroux & Loreau 2015).

The green world hypothesis by Hairston et al. (1960) incorporates both a top-down and bottom- up approaches. The hypothesis states that predators, rather than food, regulate herbivore populations, resulting in the control of the plant biomass (i.e., top-down control). However, Hairston et al. (1960) also stated that producers, carnivores and decomposers are resource limited (e.g., bottom-up control). These concepts were described more precisely in the late 1970s and 1980s (White 1978, McQueen et al. 1986). Researchers argued about which forces apply to particular systems and as a result often classified systems as either top-down or bottom-up controlled (White 1978). McQueen et al. (1986) argued, however, that organisms at the bottom of the food chain are mostly bottom-up controlled, but organisms towards the top of the food chain are increasingly top-down controlled. After 1990, various studies of ecosystems confirmed that both consumers and resources shape the biomass pyramids (Hunter & Price 1992, Hassell et al. 1998, Polis 1999, Fath 2004, Borer et al. 2006, Gruner et al. 2008, Polishchuk et al. 2013, Whalen et al. 2013). Additionally, Hunter and Price (1992) stated that organisms are influenced by a variety of forces, which vary in strength within and among ecosystems.

http://etd.uwc.ac.za/1

The importance of predators maintaining ecosystem health has been widely recognized after trophic downgrading has been observed in all major biomes (Estes & Palmisano 1974, Paine 1980, Gruner 2004, Estes et al. 2011, Ripple & Beschta 2011). The removal of predatory species typically causes chain reactions down the food web potentially causing changes within and across communities and ecosystems. Estes et al. (2011) highlights that the loss of apex consumers not only impacts ecosystem biomass through direct changes in predation and herbivory, but also through changes in other processes such as disease, biogeochemical cycling and invasive species. Furthermore, other human induced changes such as pollution, climate change, habitat loss are amplified by trophic cascades (Estes et al. 2011). Measuring the impacts introduced or removed species can have on a system is a challenging exercise because the changes usually become visible only decades later and other factors can influence such changes synergistically (Fritts & Rodda 1998, Gurevitch & Padilla 2004, Rogers et al. 2017).

Natural experiments have shown that top-down predation by apex predators can be extremely important in the maintenance of ecosystem function. The removal of native apex predator species (e.g. gray wolves of Yellowstone) has resulted in trophic cascades as did the introduction of exotic apex predators. Invasion by non-indigenous species have shown that snakes are important apex predators and can cause devastating impacts on ecosystems. Unfortunately, our knowledge of top-down regulation by snakes is limited to examples of non-indigenous invasive species. However, the following case studies—Brown tree snake of Guam, and Burmese python in the Everglades National Park—prove that snakes can have important ecosystem wide impacts. Globally, studies investigating the impact by snakes on ecosystems are rare, especially for indigenous species. In southern Africa, hardly any knowledge exists on top-down control of prey species by snakes, neither in non-invasion systems nor for native species, because studies examining the role of snakes as apex predators do not exist. However, there are some candidate systems in which predation by snakes might be important, but the manner in which snakes interact with their prey is never the focus of studies.

1.1.1 Case study of the Gray wolves at Yellowstone National Park

The extirpation and subsequent reintroduction of the gray wolf (Canis lupus) in the Yellowstone National Park (YNP) is a good example of what effects the removal of an apex predator can have on an ecosystem. The gray wolf is a native carnivore in YNP which was extirpated from the park

http://etd.uwc.ac.za/2 during the 1920s (Ripple & Beschta 2011). With the disappearance of wolves, elk (Cervus elaphus) populations increased, their movement patterns changed, and their browsing behaviour was altered (Beschta 2005, Fortin et al. 2005). Because of the increased browsing pressure by elk, plant species such as aspen (Populus tremuloides), cottonwoods (Populus spp.), and willows (Salix spp.) experienced dramatic reductions in recruitment (Singer et al. 1994, Ripple et al. 2001, Beschta 2005). After seven decades of absence, YNP reintroduced wolves in 1995/1996 (Ripple & Beschta 2011). The reintroduction of wolves impacted plant communities by increasing plant recruitment through an increase in predation pressure on elk populations (Ripple & Beschta 2011). In the early 1990s, the elk population in the northern range of YNP was more than 15,000 and decreased to approximately 6,100 individuals by 2010 (Ripple & Beschta 2011). Recent research has revealed that this pattern may not be exclusively driven by wolves but rather involve more complex top-down effects involving humans (Wright et al. 2006, Eberhardt et al. 2007, Barber-Meyer et al. 2008, Eacker et al. 2016). However, humans and carnivorous predators affect different groups of the elk population. Hunters remove mostly females that are in their reproductive age, while natural predators predate on the young, weak and old individuals (Wright et al. 2006, Barber-Meyer et al. 2008, Eacker et al. 2016).

Because of the decline of elk numbers herbivory on species such as P. tremuloides dropped to <25%, resulting in increased recruitment and growth of young aspen (Ripple & Beschta 2011). By 2012, aspen densities had doubled (8,960 individuals/ha) since 1997/98—shortly after wolves had been reintroduced to complete Yellowstone’s predatory guild (Beschta et al. 2016). Also, the density (individuals/ha) of aspen larger than 1 m in height has increased drastically during that same time period (Beschta et al. 2016). Populus spp. and Salix spp. show a similar recovery success. In addition to plant recovery, the numbers of beaver (Castor canadensis) and bison (Bison bison) experienced a positive trend, likely because of increased food availability (Ripple & Beschta 2011, Smith & Tyres 2012).

The example of the Yellowstone wolves demonstrates how the removal of a predator species disrupted the top-down control of the lower trophic levels in the ecosystem. The following two examples show how top-down predation by invasive alien snake species can cause ecosystem disruptions and species extinctions (Rodda et al. 1999, Dove et al. 2011, Dorcas et al. 2012). Snakes’ ectothermic nature, the ability to fast for extended periods of time and thus eating fewer

http://etd.uwc.ac.za/3 prey, as well as opportunistic foraging behaviour are several reasons why it isn’t well understood how snakes affect food webs (Diller & Johnson 1998).

1.1.2 Case study of the Brown tree snake of Guam

The most well-studied snake species that has caused massive ecosystem changes is the Brown tree snake (Boiga irregularis) on Guam. The species is native to Australia, New Guinea, and the Solomon Islands (Savidge 1987). It is a large arboreal colubrid snake, which is mostly nocturnal (Rodda et al. 1999). The introduction to Guam allegedly occurred from New Guinea around 1950, but the pathway through which it arrived at Guam is not known (Rodda et al. 1992, Fritts & Rodda 1998). At the time of the snake’s arrival, the island had 18 native bird species (12 forest species, three seabirds, two wetland species and one reef-heron), three mammal species, and at least ten reptile species (Savidge 1987, Fritts & Rodda 1998).

Savidge (1987) observed the range retraction of native forest birds with an expanding distribution of the Brown tree snake on the island. Because of the snake’s habitat preferences, the forest avifauna was most threatened by its expanding range and increasing population size on the island. By the 1970s there were only two of the 12 forest birds and one of the three mammal species remaining—Guam rail (Gallirallus owstoni), island swiftlet Aerodramus bartschi (referred to as A. vanikorensis bartschi in Fritts & Rodda 1998), and the Mariana fruit bat Pteropus mariannus subsp. mariannus (Savidge 1987, Fritts & Rodda 1998, Cruz et al. 2008). The other bird and mammal species had been extirpated. Today, the Mariana fruit bat and the island or Mariana swiftlet are considered endangered (Allison et al. 2008, Cruz et al. 2008). The offspring of Mariana fruit bats are threatened by the snake resulting in reduced recruitment and thus a negative long-term prognosis for the species (Allison et al. 2008). The Guam rail, which was endemic to Guam, is now listed as extinct in the wild by the IUCN (BirdLife International 2016). Reintroduction programs to Guam should only be considered once the spread of B. irregularisis is controlled (BirdLife International 2016). Birds and bats have been extremely prone to the snake’s predation, while reptiles appear to be slightly less affected (Fritts & Rodda 1998).

Out of 12 indigenous reptiles, only two species (Emoia caeruleocauda, Lepidodactylus lugubris) are still common on Guam. Approximately six species have possibly survived in small numbers in unknown locations on the island while the rest such as Perochirus ateles and Gehyra oceanica

http://etd.uwc.ac.za/4 are likely extinct (Fritts & Rodda 1998). Introduced species such as the house gecko (Hemidactylus frenatus) and the Indonesian brown skink (Carlia fusca) have been found in stomach contents of B. irregularis (Savidge 1988). In addition to increased predation, introduced species such as the Indonesian brown skink out-compete native species (Rodda & Fritts 1992) and sustain a stable population of B. irregularis by serving as a new food source (Fritts & Rodda 1998), enabling the Brown tree snake to survive the loss of native prey populations. The snake’s nocturnal and arboreal behaviours and an ability to detect prey explains the success of this invasive predator (Savidge 1987). Guam’s indigenous species did not evolve with predators that exhibit such behaviours, explaining their proneness to snake predation.

Drastically declining populations and extinction rates of native species influence other species on lower trophic levels (Rogers et al. 2012, Rogers et al. 2017). With the extirpations of native avian fauna, Guam has become a spider-infested island in recent years. Rogers et al. (2012) compared spider web densities on Guam to three nearby islands with healthy native avian populations. During the wet season web densities on Guam were 40 times greater than on the other island and 2.3 times greater in the dry season (Rogers et al. 2012). Spider web size was up to 50% larger on Guam than on islands with insectivorous birds (Rogers et al. 2012). The local bird extinctions not only caused a lack of spider control, but also a reduction in competition, possibly further benefiting spider population growth (Rogers et al. 2012). Increased numbers of spiders might counter-act the loss of insectivorous avian predators, reducing the risk of insect pests (Rogers et al. 2012). However, it might also result in increased predation of pollinating insects impacting plant communities. There might be a large prey overlap between spiders and insectivorous birds, however there are likely to be prey species consumed less frequently or not at all by spiders. For example, plant eating larvae might experience lower predation, possibly resulting in increased plant damage.

While confirming the findings from small-scale experiments, the magnitude at which extirpations of avian apex predators impact systems might be significantly greater in large-scale natural experiments (Rogers et al. 2012). The loss of native insectivorous birds resulted in changes in avian prey populations, but also influenced Guam’s native flora. Many native floral species on Guam are dependent on birds for their recruitment (Mortensen et al. 2008). Mortensen et al. (2008) studied two bird-pollinated tree species, the mangrove tree (Bruguiera gymnorrhiza) and the forest tree (Erythrina variegata var. orientalis). Flower visitation by birds, seed set, and

http://etd.uwc.ac.za/5 germination were significantly lower on Guam than Saipan, where nearly no snakes are present (Mortensen et al. 2008). Other pollinators such as insects and lizards were only encountered occasionally but may be important in pollination of other plant species. Approximately 70% of Guam’s tree species depend on frugivores for seed dispersal (Rogers et al. 2017). The loss of the main pollinating avian species was long assumed to have negative long-term effects on plant recruitment, leading to floral extirpations and decreased forest diversity on Guam (Mortensen et al. 2008). Similarly, Rogers et al. (2017) showed that the lack of frugivores on Guam has had negative impacts on seed dispersal, germination and survival. Estimates of tree recruitment for the two studied tree species Psychotria mariana (61–72% ) and Premna serratifolia (87–92%) are lower on Guam than on neighbouring islands with healthy frugivore populations, respectively (Rogers et al. 2017). Therefore, the Brown tree snake not only affects prey populations, but also breaks mutualistic links between species and reduces floral recruitment throughout the entire ecosystem resulting in trophic cascades (Mortensen et al. 2008).

1.1.3 Case study of the Burmese Python in the Everglades National Park

The Burmese python (Python bivittatus), one of the largest snake species globally, is a constricting snake native to Southeast Asia (Barker & Barker 2008). This ambush predator is a habitat generalist, inhabiting terrestrial and aquatic habitats including forests, grassland, caves, burrows and even some deserts (Hart et al. 2015). Many of these habitats are typically associated with aquatic bodies such as wetlands, rivers, lakes and pools (Barker & Barker 2008, Dorcas et al. 2011, Hart et al. 2015, Sharma & Kandel 2015). In its native range this species has experienced dramatic declines, leading to a listing of vulnerable by the IUCN (Stuart et al. 2012). Two native populations are critically endangered due to reductions in population sizes by up to 80% (Stuart et al. 2012). Over the last ten years, the global Burmese python population has experienced a reduction of at least 30% (Stuart et al. 2012).

In the early 1980s, the Burmese python was introduced to Florida (Meshaka et al. 2000, Willson et al. 2011). It is uncertain how the Burmese python got to Florida, but the most likely pathway is via the pet trade (i.e., escaped or released pets; Willson et al. 2011). In 2000, the species was officially recognized as well-established in the southern parts of Florida (Meshaka et al. 2000), however more recent studies suggest an earlier establishment (Willson et al. 2011). Large numbers of Burmese python inhabit the Everglades National Park (ENP) (Dorcas et al. 2012).

http://etd.uwc.ac.za/6

The ENP represents an ecosystem that is made up of various types of wetlands resulting in favourable conditions in which Burmese pythons can thrive (Hart et al. 2015). Some of these wetland types are similar to those found in their natural habitat in Southeast Asia such as wet grasslands, mangroves, and tropical lowlands (Barker & Barker 2008, Dorcas et al. 2012). ENP is home to a vast variety of wildlife, especially avian and mammalian species. Several species found in ENP are of special concern such as little blue heron (Egretta caerulea), snowy egret (E. thula), while Key Largo woodrat (Neotoma floridana smalli) and wood stork (Mycteria americana) are considered endangered (Dove et al. 2011, Dorcas et al. 2012).

Dorcas et al. (2012) found that the sightings of once-common mammals—raccoons (Procyon lotor), Virginia opossums (Didelphis virginiana) and rabbits (Sylvilagus spp.)—have declined drastically since the Burmese python was accepted as established in 2000. However, it is likely that the species was established earlier than 2000 as sightings of Burmese pythons have occurred since the early 1980s (Meshaka et al. 2000). Large mammals have been recorded in the python’s diet including the white-tailed deer (Odocoileus virginianus) and bobcat (Lynx rufus; Snow et al. 2007, Reed & Rodda 2009, Rochford et al. 2010). Even larger reptiles such as the American alligator (Alligator mississippiensis) have been consumed by this giant constrictor (Reed & Rodda 2009, Rochford 2010). Many of the animals found in ENP did not evolve with a predator such as the Burmese python. Mostly nocturnal, this snake occasionally forages diurnally (Dove et al. 2011). As a result, native animals are highly susceptible to predation (Dove et al. 2011). Rodents have been sighted more frequently since the pythons’ proliferation, which may be the result of a reduction in other predators such as the bobcat (Maehr & Brady 1986, Dorcas et al. 2012). The observed declines in mammal populations may be a direct result of top-down predation by the Burmese python and due to increased competition for resources with other large predators. This case shows that snakes have great potential to influence terrestrial ecosystems. Invasive exotic species are more likely to affect a system negatively, while native species may ensure ecosystem functionality.

This study examines 1) the factors that influence the activity of snakes (Naja nivea and Dispholidus typus) and 2) the predation of snakes on an avian prey to test the hypotheses that influencing factors and predation differ among sympatric species despite feeding on an in- common prey (Philetairus socius).

http://etd.uwc.ac.za/7

1.2 Study system

The Kalahari system is a semi-arid sandy savanna in southern Africa, covering 2.5 million hectares of land making it the largest sand cover in the world (Thomas & Shaw 1993, Scholes et al. 2002). The system consists of the central Kalahari desert and the surrounding Kalahari basin. The Kalahari sands cover large parts of Botswana, Namibia, South Africa and extend into Zambia, Angola, Zimbabwe and the Democratic Republic of Congo (Mishra et al. 2015). The system experiences a precipitation gradient of 200–1000 mm of rain per annum from south to north (Scholes et al. 2002). The wet season, October/November to April, is typically warmer than the dry season, and temperatures are higher in the south than the north (Scholes et al. 2002). The system is characterized by its aridity, high radiation levels, sandy soils and a variety of flora and fauna (Dean et al. 1999, Bullock et al. 2011, Wasiolka & Blaum 2011). The vegetation of the Kalahari depends on rainfall and soil composition (which influences the soils ability to retain water) with a corresponding increase in numbers of woody plants along the rainfall gradient from south to north (Scholes et al. 2002, Mishra et al. 2015). Moving from south to north along this precipitation gradient, the tree composition is increasingly made up of members of the families Mimosaceae, Combretaceae and Caesalpinacea, respectively (Scholes et al. 2002). The part of the Kalahari basin that reaches into South Africa, is categorized as grassy shrubland and is dominated by Stipagrostis grasses, with shrubs (Rhigozum trichotomum and Vachellia haematoxylon) and scattered trees (Vachellia erioloba and Boscia albitrunca). Trees are often the only shaded cover available to animals, especially larger mammals like gemsbok (Oryx gazella). The Kalahari is home to a variety of mammals, birds, reptiles and amphibians capable of withstanding the harsh conditions (aridity and extreme seasonal temperatures; Haacke 1984, Bates 1991, Bullock et al. 2011, Muñoz 2013).

In South Africa, the cape cobra (Naja nivea) and boomslang (Dispholidus typus) are widespread and abundant species (Broadley 1983, Branch 1998). Despite their abundance, there are no studies on the relationships of these snakes with other organisms (such as predators or prey) nor to their physical surroundings. As a result, the potential role that these snakes play as predators is poorly understood. The distributions of the two species overlap in central South Africa (Marais 2014, Maritz & Alexander 2014) in an ecosystem where cape cobra and boomslang are known to prey heavily on sociable weavers (Philetairus socius; Spottiswoode 2007), an important

http://etd.uwc.ac.za/8 ecosystem engineer that builds large communal nests (i.e. colony). The impact of snakes on these birds can be dramatic with nest predation reaching up to 70% (Covas 2002).

Sociable weavers are small passerine birds endemic to southern Africa. A large proportion (78.9%) of their diet consists of a variety of insects; the harvester termite (Hodotermes mossambicus) being consumed most abundantly (48.9% of animal food; MacLean 1973a). A lesser part of their diet consists of seeds, mainly those of green grasses (MacLean 1973a). Such bird species are trophic-process linkers that impact invertebrate population through predation (Sekercioglu 2006), serving as controlling agents of invertebrate populations, behaviour and evolution (Bock et al. 1992). Also, by regulating invertebrate prey numbers, birds are capable of preventing plant damage and influencing primary production (Sekercioglu 2006). Records of seeds in the diet of P. socius indicate that they may play a role in seed dispersal as well. Moreover, sociable weavers are ecosystem engineers or non-trophic (e.g. physical) process linkers (Sekercioglu 2006, Rymer et al. 2014), meaning that they create and modify habitats (Sekercioglu 2006). Sociable weavers do so by building large colonial structures, using grass straws and twigs (MacLean 1973b). These colonial nests are used by a variety of species such as the African pygmy falcon (Polihierax semitorquatus), kalahari tree skink (Trachylepis spilogaster), barn owl (Tyto alba), giant eagle owl (Bubo lacteus), and red-headed finch (Amadina erythrocephala) (MacLean 1973c, Dean et al.1999, Covas et al. 2004a). For example, Rymer et al. (2014) suggested that skinks experience lower predation rates, thermal shelter and increased prey availability in trees with sociable weaver colonies. Ecological functions, such as the ones sociable weaver perform, are important in ecosystem maintenance (Lundberg & Moberg 2003). Hence, a change in sociable weaver dynamics could potentially alter ecosystems.

Sociable weaver colonies at the Benfontein Nature Reserve experience high predation rates by snakes (Covas & du Plessis, 2005), particularly cape cobra and boomslang. In reality, little is known regarding how and when snakes access sociable weaver colonies limiting our ability to understand the underlying dynamics of this top-down regulation. Benfontein Nature Reserve provides an excellent opportunity to examine this top-down predation of avian ecosystem engineers by native snake species. Specifically, the system offers an opportunity to examine factors such as predation rate, timing of predation by snakes and factors influencing snake movement.

http://etd.uwc.ac.za/9

1.3 Measuring predation rates

Studying snakes is challenging due to their unpredictable movement patterns, secretiveness and often low individual densities (Greene 1997, Diller & Johnson 1998). This had led to a poorer understanding of snake ecology compared to that of other tetrapod classes such as birds and mammals (Ray et al. 2012). Studies examining the effect of predation by snakes on its prey are still relatively rare, compared to other tetrapod classes. Most such studies are mainly on non- African snake species (Downes 2002, Sperry et al. 2008, Sperry et al. 2009).

Measuring the predation of one organism by another is a challenging task. Laboratory experiments are easier to control but the results often do not display the natural predation rates realistically (Puntila et al. 2012). Natural experiments on predation by snakes are often challenging for various reasons. In order to measure predation in the field, scientists have to detect the study species first, while in laboratory experiments captive bred individuals can be used. Previous studies on snake detectability suggest interspecific variation, with tendencies towards low detection probabilities overall (Kéry 2002, Harvey 2005, Durso et al. 2011, Steen et al. 2012). Snakes inhabit a range of habitats, of which many make detection of snakes a discouraging and difficult task (Fitch 1987, Durso et al. 2011). Moreover, snakes are cryptic and secretive species that avoid open space to minimize predation risk (Lima & Dill 1990, Webb & Whiting 2005). Their crypsis and unpredictable movement patterns make it challenging to locate and track them, even with radio tracking devices (Penner et al. 2008). Moreover, during field studies, predation events by snakes are seldom observed.

The diet of a species plays a fundamental role in its ecology (Leighton & Echeverri 2015). In order to understand species interactions with the environment, a comprehensive knowledge of the dietary composition of the species is a necessity (Leighton & Echeverri 2015). Our knowledge of feeding ecology of snakes is focused on North American, European and Australian species (Gregory 1984, Drobenkov 1995, Rodríguez-Robles et al. 1999, Rodríguez-Robles 2002, Filippi et al. 2005, de Queiroz & Rodríguez-Robles 2006, Glaudas et al. 2008, Webber et al. 2016), with research on African species being limited (Keogh et al. 2000, Luiselli 2003). However, empirical studies on African snake species have been published over the last two decades (i.e., Shine et al. 1996, Luiselli & Angelica 1998, Shine et al. 1998, Webb et al. 2000, Webb et al. 2001, Akani et al. 2003, Luiselli 2006a, Shine et al. 2007, Glaudas & Alexander

http://etd.uwc.ac.za/10

2017). Despite improving our knowledge of consumed prey, dietary studies from gut content analyses are limited by restricted access to museum specimens and general low feeding frequencies in snakes (Branch 1997, Layloo et al. 2017). Similarly, laboratory results may show how much snakes are capable of eating, but might not represent actual consumption rates in a natural environment.

The system examined in this study is the ideal solution to the problems regarding the detection of snakes and observing feeding events in the wild. The snakes visit the sociable weaver colonies to forage for the offspring of the birds. This results in high detectability of snakes within colonies and frequent observations of feeding events by snakes. The offspring can be counted efficiently, allowing the quantification of predation by snakes.

1.4 Approach

In my thesis, I used colonial nests of sociable weaver at Benfontein Nature Reserve to examine several aspects of snake predation on these birds. I aimed to identify the environmental predictors of snake use of sociable weaver colonies and to quantify general predation and detailed consumption rates of each snake species on the birds. I wished to close gaps in our knowledge of the basic biology of the snakes and to investigate the potential for snakes to control sociable weaver ecosystems through top down predation. This study examines the presence and absence of snakes in social weaver colonies to test the hypotheses that 1) the movement of Naja nivea and Dispholidus typus is influenced by different spatial and temporal variables and 2) nest predation by these two snake species differ.

1.5 Study site

The study was conducted on the Benfontein Nature Reserve (BNR), situated approximately 6 km south-east of Kimberley, in the Northern Cape Province and the bordering Free State Province, South Africa (Figure 1.1). The farm is owned by De Beers Group and is part of the Diamond Route conservation program emphasizing biodiversity conservation, education and sustainability. The reserve covers approximately 11,000 hectares and is made up of three biomes—the Nama Karoo, Grassland, and Savanna biomes converge in this area, resulting in a transitional zone of great species diversity (Figure 1.1). The three vegetation types that come together in the

http://etd.uwc.ac.za/11

Kimberley area are Thorn Bushveld, Eastern Mixed Nama Karoo and Dry Sandy Highveld Karoo (Mucina & Rutherford 2006). Components of each of those vegetation types are represented in the nature reserve. The study site is situated in the south-eastern part of the reserve, however, and consists primarily of open savanna dominated by camelthorn trees (Vachellia erioloba) and grasses of the genus Stipagrostis (Covas 2002). The nature reserve hosts a variety of wildlife, including the sociable weaver (Philetairus socius) with its large colonial nest structures and the associated pygmy falcon (Polihierax semitorquatus) are among the tourist attractions.

The study site experiences mainly summer rainfall (mean ± SD: 431 ± 127 mm; Covas et al. 2008), with mean minimum temperatures of 10.9 °C and mean maximum temperatures of 26.0 °C for the period between 1932–2009 (Kruger & Sekele 2013). In winter, temperatures range from -8 °C to 25 °C and in summer from 8 °C to 40 °C.

Figure 1.1: Locations of the studied sociable weaver (Philetairus socius) colonies in the south-east of the Benfontein Nature Reserve (28°52’S, 24°51’E) on the border of the

http://etd.uwc.ac.za/12 provinces of the Northern Cape and the Free State, South Africa. The black dots represent the study colonies with their corresponding colony ID. Image from Google Earth.

1.6 Study species

1.6.1 Cape cobra

Cape cobra (Naja nivea) are large diurnal elapid snakes, endemic to southern Africa including the southern parts of Namibia and Botswana as well as the southern, western, and central regions of South Africa (Figure 1.2; Broadley & Wüster 2004, Bates et al. 2014). The species is commonly found in arid karoo, succulent karoo, fynbos, grassland, and desert habitats (Phelps 2007, Shine et al. 2007, Bates et al. 2014). Mostly terrestrial, cape cobra occasionally ascend trees when foraging, but are typically observed basking or moving on the ground and taking refuge in burrows. Cape cobra are highly adaptable to modified habitats such as urban areas (Bates et al. 2014), where they are commonly found on private properties (gardens, dwellings; pers. obs.).

Figure 1.2: Distribution map of cape cobra (Naja nivea) considering historical sampling intensity and general habitat within southern Africa.

http://etd.uwc.ac.za/13

In comparison to its congeners the cape cobra is a medium sized cobra (Shine et al. 2007). They reach an average size (± SD) of 1058 ± 355 mm measured as snout-vent length (SVL; Layloo et al. 2017). There is no significant sexual dimorphism in SVL (Shine et al. 2007). However, Shine et al. (2007) demonstrated a difference in relative head and tail lengths, with males having longer heads and tails compared to females of the same SVL.

Figure 1.3: Colour variation and ontogenetic shift in colouration of the cape cobra (Naja nivea). See text for details of each image.

http://etd.uwc.ac.za/14

Based on museum collections, it appears that there is approximately a 2:1 sex ratio in cape cobra, with males being collected more frequently than females (Shine et al. 2007, Layloo et al. 2017). The skewed ratio observed in the studies is likely a result of higher male activity during the mating season. Cape cobra have smooth scales which produce a shiny appearance. Juveniles are usually yellow-brown with one distinct dark band on the throat that fades with age (Figure 1.3.G – H; Broadley & Wüster 2004). Adult cape cobra exhibit great colour variation, ranging from bright yellow through brown to black, with many individuals being speckled (Figure 1.3.A – F; Phelps 2007).

Empirical studies on the diet of cape cobra suggest that they consume a wide variety of prey items from various taxa, indicating a generalist feeding behaviour (Phelps 2007, Shine et al. 2007, Layloo et al. 2017). Shine et al. (2007) suggested that the diet of cape cobra consists of mammals, amphibians and reptiles. However, they did not record any birds in their analyses even though earlier studies note observations of cape cobra consuming eggs, chicks and adult birds. For example, Covas (2002) described the cape cobra as one of the main predators of sociable weavers, confirming the findings of MacLean (1973c). Both, MacLean (1973c) and Covas (2002) suggest severe impacts on the reproductive success of sociable weavers, with individuals often raiding the whole colony (MacLean 1973c, Covas 2002). Recently, Layloo et al. (2017) conducted a comprehensive quantitative study of the diet of cape cobra including data from literature, museum specimens and citizen science. Using three different sampling methods, they collected 101 feeding records (including multiple new records) and found the diet to consist of prey items including reptiles (42.6%), mammals (29.7%), amphibians (17.8%) and birds (9.9%). Reptiles, of which 81.4% of prey items were snakes, were the most abundant tetrapod class detected in their diet. Despite a generalist feeding behaviour, more than a third of the cape cobra diet consists of snakes (34.7% of all records), indicating a possible dietary preference. In addition to being ophiophagous, cape cobra are also cannibalistic (Maritz et al. 2019). Notably, all bird prey records in the study by Layloo et al. (2017) came from the literature and citizen science, highlighting the importance of including all available sources for dietary studies (Layloo et al. 2017).

http://etd.uwc.ac.za/15

1.6.2 Boomslang

Boomslang (Dispholidus typus) are colubrid snakes, endemic to sub-Saharan Africa reaching into Kenya in the North and Senegal in the West (Bates et al. 2014). Within South Africa boomslang are predominantly found in the southern, eastern and northern regions with some records from central South Africa (Figure 1.4; Bates et al. 2014). Individuals have arboreal habits excluding them from habitats such as the arid karoo but inhabit many others including fynbos, karoo scrub, arid and moist savanna, lowland forest and grassland (Bates et al. 2014).

Figure 1.4: Distribution map of African boomslang (Dispholidus typus) considering historical sampling intensity and general habitat within southern Africa.

Dispholidus typus are diurnal, mostly arboreal snakes with long, slender bodies and an average SVL (± SD) of 1133 ± 73 and 1137 ± 98 mm for adult males and females, respectively (Smith et al. 2019). Boomslang have keeled scales, producing a non-shiny appearance, which may aid in better camouflage in tree canopies. Boomslang are highly sexually dichromatic (Smith et al. 2019). Male colouration varies tremendously, ranging from yellow to green through to black, often with black-edged scales or occasionally with blue tints (Figure 1.5A – C; Branch & Jackson 2016, Smith et al. 2019). On rare occasions, other colours like a cream-orange can be

http://etd.uwc.ac.za/16 observed (male individual from Oudtshoorn, South Africa; Figure 1.5C). Typically, females have a light to olive brown dorsal colouration with a lighter ventral scale colouration (Figure 1.5D – E; Smith et al. 2019). Occasionally females exhibit male colourations (Marais 2014). Juveniles have grey brown dorsal scales and white to yellow bellies, often covered with small dark blotches (Figure 1.5F – G; Smith et al. 2019). Boomslang have large eyes in relation to their head size. Juveniles have bright emerald eyes, but become an olive green to brown colour at adulthood (Figure 1.5).

Until recently, dietary information for boomslang was mainly based on anecdotal notes, observations and field guides, suggesting that boomslang mostly prey on chameleons and other

http://etd.uwc.ac.za/17 vertebrates such as lizards, frogs, rodents, and bird eggs and nestlings (Broadley 1968, Branch 1998, Greene 1999, Covas 2002, Marais 2004, Pla et al. 2017). In contrast, Smith et al. (2019) showed that the diet of D. typus consists mainly of nestling birds (54 %) and chameleons (36%). The study confirmed that boomslang consume all tetrapod classes with the focus on tree- dwelling prey such as various species of birds and chameleons (Smith et al. 2019). Occasionally boomslang descend to the ground to forage for ground-dwelling prey such as rodents.

1.6.3 Sociable weaver

The sociable weaver (Philetairus socius) is a small passerine bird endemic to southern Africa, and occurring mainly in the Northern Cape of South Africa, and Namibia (Figure 1.6; MacLean 1973d, Mendelsohn & Anderson 1997). They are mostly found in arid savanna and dry woodland (Hockey et al. 2005). This species is highly colonial, building enormous dome-like structures in camelthorn trees, which can be inhabited by up to 500 individuals (Figure 1.7; Hockey et al. 2005). A colony consists of many chambers in which breeding pairs roost and breed and may be accompanied by so-called helpers (Covas 2002). Colonies of sociable weavers are common within camelthorn trees at the Benfontein Nature Reserve.

http://etd.uwc.ac.za/18

Figure 1.6: Distribution map of sociable weaver (Philetaius socius) considering historical sampling intensity and general habitat within southern Africa.

Adult birds weigh 26–32g (Covas 2002). In the field, males and females are morphologically indistinguishable. In adults, the feathers of chest and abdomen are beige in colour, with a black throat marking, and the wing and back feathers are different shades of brown to black (Figure 1.8.A–B). The head is light brown, slightly darker than the colouration of the chest. Juvenile colouration is not as defined as in adults and the throat marking only develops after the first moult indicating sexual maturity (Figure 1.8.C; MacLean 1973d). The legs of sociable weaver are grey-blue (Figure 1.8; MacLean 1973d). Hatchlings are mostly naked with three fine feathers on the head (Figure 1.8.D, pers. obs.).

Figure 1.7: Sociable weaver (Philetairus socius) nest structure with a foraging cape cobra (Naja nivea) moving between chambers at the Benfontein Nature Reserve, Kimberley, NC, South Africa. Arrow pointing to the snake.

http://etd.uwc.ac.za/19

On average, sociable weavers lay clutches of three eggs at one-day intervals, and breeding pairs often attempt to raise multiple clutches during a breeding season (Covas 2002). Breeding pairs often have ‘helpers’—typically offspring from previous successful clutches—which provide assistance with the feeding of new clutches (Covas & du Plessis 2005). The intervals between breeding attempts depends on the fate of the young and increases from clutch failure occurring during incubation, at the nestling stage, or for successfully fledged clutches (Covas 2002).

http://etd.uwc.ac.za/20

Sociable weaver populations appear to be regulated by predation of eggs and chicks, primarily by snakes (MacLean 1973c, Covas 2002). Both boomslang and cape cobra are frequently observed in colonies of P. socius during breeding season and represent the biggest threat to weaver reproductive success (Covas & du Plessis 2005, Spottiswoode 2007). These predators commonly raid colonies of their entire brood, occasionally missing only a few chambers (Spottiswoode 2007).

1.7 Methods

1.7.1 Terminology

My study relies on repeated observations of the sociable weaver nests using a standardised protocol that is used by Covas (2002). Below I define eleven terms that will aid in the understanding of my research protocol.

Colonies and chambers

For this study, I do not use the term ‘nest’ because it may be ambiguous or synonymous with the terms ‘colony’ and ‘chamber.’ Therefore, I use the term ‘colony’ to refer to the entire nest structure and ‘chamber’ for individual nests within the ‘colony.’

Breeding chamber

A chamber is considered a ‘breeding chamber’ when it contains offspring (eggs or chicks) at any point in time. A chamber becomes a non-breeding chamber once chicks fledge successfully or the offspring fails before fledging.

Clutches and broods

I use the term ‘clutch’ when I refer to offspring that are still eggs and the term ‘brood’ to refer to chicks.

D9 and D17

Chicks are referred to as ‘D9 chicks’ and ‘D17 chicks’ based on their age in days since hatching date. Therefore, ‘D9 chicks’ and ‘D17 chicks’ are 9 and 17 days old, respectively.

http://etd.uwc.ac.za/21

Full colony check

A full colony check is defined as the inspection of all chambers in a colony to detect new breeding activity (i.e. new clutches). During these checks missing clutches and broods are documented as well. Each colony undergoes these checks every third day in order to confidently calculate the initiation date of new clutches—sociable weavers lay eggs in 24-hour intervals.

Control check

A ‘control check’ is defined as a visitation of a specific breeding chamber in a colony for one of the following reasons: 1) to detect new eggs laid after a new clutch initiated, 2) to weigh eggs, 3) to confirm predicted hatching dates, 4) to process D9 or D17 chicks.

Figure 1.9: A sociable weaver (Philetairus socius) colony protected from snakes using clingwrap around the trunk of the host, camelthorn tree (Vachellia erioloba).

http://etd.uwc.ac.za/22

Protected and non-protected colonies

In this study, I differentiate between ‘protected colonies’ and ‘non-protected colonies’ to state whether anthropological measures were taken to exclude snakes from a colony. Colonies were protected from snakes using cling wrap covering approximately two meters of the trunk of the host tree (Figure 1.9). Fissures in the bark enable snakes to ascend trees using concertina movements (Mullin & Cooper 2002). The use of cling wrap removes this surface feature, hindering snakes from climbing trees, thus excluding snakes from predating on the offspring of sociable weavers.

1.7.2 Study period

The research was conducted from August 2016 until April 2017, which corresponds with the breeding season of sociable weavers (Covas 2002). I collected snake presence-absence data during three time periods representing early summer (23/9/2016 – 21/10/2016), peak summer (2/1/2017 – 29/1/2017) and late summer (3/4/2017 – 29/4/2017). I also assisted the sociable weaver project (SWP) team with their data collection. However, during the other times, the sociable weaver project continued the data collection to increase sample size.

1.7.3 Bird data

The research conducted on sociable weavers at BNR is a long-term project led by Rita Covas (FitzPatrick Institute of African Ornithology) and includes studies on population dynamics and biology, feeding and behavioural ecology, reproduction, and predation (Covas 2002, Covas et al. 2004a, Spottiswoode 2009, van Dijk et al. 2013, Lloyd 2016). The data collected for the sociable weaver project were important for my research as it provided valuable information on the breeding/offspring of sociable weavers. The bird data required for this study included total number of adult birds per colony, total number of chambers per colony, number of active chambers in each colony, and the number and mass of eggs and chicks throughout the breeding season.

In this study, I included data collected from 13 colonies of which 4 were protected from snakes. However, one of the protected colonies was poorly protected—due to the deeply furrowed bark of the trunk, snakes were able to climb the tree by squeezing their bodies through the spaces between the fissures and the cling wrap.

http://etd.uwc.ac.za/23

Colony captures

Annual colony captures were conducted to count and identify all individuals that utilized a colony. Before dawn, while the birds were still asleep, the research team set up mist nets on the ground surrounding the colony tree (Covas et al. 2004a). By creating noise, the birds woke up, tried to escape the colony and flew into the nets. The birds were removed from the nets, placed in separate linen bags, measured, and released again. Those captures provided information on the gains and losses of individuals (e.g. through migration of individuals between colonies) and the total number of birds. Previous studies measured colony size, however they used different approaches depending on the research question—absolute colony size (number of individuals), relative colony size (ranks) or colony volume (Spottiswoode 2009, van Dijk et al. 2013, Leighton & Echeverri 2014). For this study, I focused on the total number of individuals and the volume of colonies as measures of colony size.

Colony checks

The sociable weaver field work relevant to my project included daily visitations of colonies for full colony checks and control checks during the entire breeding season of sociable weavers. The total number of checks—full colony checks and control checks combined—conducted during the breeding season varied for colonies depending on colony size, breeding intensity and reproductive success.

Before conducting any checks, colonies were inspected for snakes to prevent field assistants from getting harmed, and to assess the loss of offspring to predation. In addition, the ground directly under colonies was inspected for any eggs or chicks that had fallen out. Afterwards, the research vehicle was parked directly under the colony. The vehicles used in this study were modified by placing a wooden board covering the whole rooftop for people to stand on when checking colonies (Figure 1.10B). When required, a step ladder was placed on the rooftop to reach higher chambers. The chambers were inspected using a small mirror with LED lights and an extendable handle for easy excess of the chambers. A torch was frequently used as an additional source of light.

Each colony underwent a full colony check every third day. The date and colony number (i.e. identification of the colony or individual sampling sites) was recorded for each full colony check. Yellow plastic tags, which were installed prior to this study, identified the chambers

http://etd.uwc.ac.za/24

(Figure 1.10A). However, newly built chambers were tagged immediately during the breeding season. Tags were necessary to eliminate mistaking chambers with one another. During these checks, all chambers were inspected for their reproductive content. New clutches were visited daily and first processed after the third egg was laid. Daily visitations continued, and new eggs were processed immediately. When the birds did not lay an egg within 24 hours, the clutch was complete. Accordingly, clutches of fewer than three eggs were processed earlier.

Eggs and chick processing

The eggs were carefully removed from the chambers and placed in small containers with cotton cushioning. The eggs were weighed and returned to their chamber of origin. The eggshell was very thin, hence eggs were handled with great caution. Processing of nine day old chicks included weight measurements, blood samples and tagging them with a numbered metal ring. The last processing of chicks occurs on the 17th day after hatching and includes weight and body measurements (wing and tarsus length).

http://etd.uwc.ac.za/25

1.7.4 Snake data Presence absence

Throughout the breeding season of sociable weavers, I conducted daily visual surveys at 13 colonies to observe Naja nivea and Dispholidus typus. The number of visited colonies varied according to breeding activity. During the peak of breeding season 10–13 colonies were visited for the collection of the bird data each day. Towards the end of the season, when breeding intensity decreased, bird data collection reduced to three to five colony visitations per day. Thus, I conducted additional snake checks to maintain a constant sampling effort throughout the season. Throughout the breeding season, at the beginning of each colony visit, colonies were checked for the presence of snakes before any chamber checks were conducted. A torch was used to illuminate each chamber to detect snakes. For each visitation, I recorded the date, colony number, the presence or absence of snakes, and the species and age of any observed snake.

Species identification

Each snake observed within a colony was identified to species level. Identification of snakes was easiest when individuals were moving between chambers, but more challenging when individuals were curled up inside a chamber (Figure 1.11). In a curled up position the head of snakes were often not or only partially visible. Cape cobra were typically visible on arrival, either moving between chambers or coiled up inside a chamber. As a result of their thicker bodies, cape cobra bodies always extended into the entrance tunnel. Due to a more slender body compared to those of cape cobra, boomslang were able to completely curl up inside a chamber.

The two study species of snakes have distinct morphological differences that are useful when only parts of the snake are visible. Such differences include the makeup of their dorsal scales and their body colouration. Dorsal scales of cape cobra are smooth leaving a shiny look, while boomslang have keeled dorsal scales resulting in a duller appearance. Even though both species have a wide range of colouration, only certain colours were observed within the study area. Cape cobra are typically yellow to golden brown in this area, whereas boomslang are usually grey- brown (pers. obs.; Figure 1.11.F is for explanatory purposes only). Other morphological characteristics such as head features (e.g. head size and shape, eye colouration and size) were used when available. The age of individuals (adult versus juvenile) was estimated based on the

http://etd.uwc.ac.za/26 size and colouration and distinct juvenile features. Video footage, sourced from the sociable weaver project, provided additional records of snake presence at colonies.

http://etd.uwc.ac.za/27

Colony measures

In addition to the data recorded on the sociable weaver offspring, the chambers, and the adults, I recorded information on several physical features related to the colony. The height, length, and width of each colony were measured to calculate colony volume as larger structures might be recognized by snakes more efficiently. The height was measured from the lowest point or bottom side of the colony to the top of the dome-like structure. Because of the irregular structures of colonies, the minimum and maximum height was recorded based on the lowest and highest chamber of a colony. The length of a colony was measured parallel to the supporting branch. Again, minimum and maximum values were recorded when the colonies were shaped irregularly. The colony width was measured perpendicular to the length. Furthermore, the distance between the ground to the lowest branch of host trees were recorded, as snakes use low hanging branches to ascend trees. Snakes avoid open spaces to reduce predation risk—nearby trees may serve them protective cover. Therefore the distance between the host tree to the nearest tree was recorded for each colony. The nearest tree was defined as woody plants with a DBH (diameter at breast height) greater than 10 cm. All measures were taken to the nearest 0.01 m using a 50 m measuring tape.

Weather conditions

Several weather related measures were required to detect factors predicting presence/absence of snakes. I obtained most of the weather data from the South African Weather Service (SAWS). The nearest weather station was at Kimberley Airport (station number 0290468A9) approximately 10 km from the study site. The obtained data included minimum and maximum temperature (°C), wind speed (m/s), humidity (%), air pressure (hPa) and precipitation (mm). However, rainfall was recorded directly on BNR using rain gauges installed by the sociable weaver research project. Minimum temperature reflects the lowest recorded temperature at night, while maximum temperature represents the highest record for the day. Wind speed, humidity and air pressure recordings reflect the morning conditions (at 8am) as sampling took place in the mornings.

http://etd.uwc.ac.za/28

1.8 Structure of thesis

The first chapter serves as an introduction to my thesis, describing the ecological context for wanting to study predation by snakes in natural systems. The chapter also describes some sections of the methods as they inform the overall project. However, some methods sections are repeated in the relevant chapters to ascertain a comprehensive understanding as the layout of the following two chapters is according to peer-reviewed literature standards. Using the data gathered, I set out to answer several questions arranged into the next two chapters.

In chapter two, I aim to identify temporal and spatial correlates of nest visitation and predation by cape cobra and boomslang. Therefore, I relate the presence-absence data of the two snake species to the spatially- and temporally-variable factors using principal component analyses (PCA) and multiple logistic regression analyses. PCA removed the auto-correlations among the variables, creating new representative components used in the multiple logistic regression analyses.

The third chapter examines snake predation in sociable weavers at a finer scale, specifically identifying differences in resource utilization by the two snake species by quantifying consumption rates. Therefore, data on snake presence and the breeding of sociable weaver were essential to answer questions on snake predation. It was especially important to know the exact content for each chamber for each colony at any point in time.

The fourth and final chapter synthesizes the results from chapter two and three and discusses how these findings aid in our current understanding of snakes as top-down controllers in ecosystems. Additionally, I summarize the biases and limitations of collecting and quantifying the presence/absence data and the consumption data. Furthermore, I give suggestions for future studies, specifically for a complete study design aiming to investigate if the differential resource utilization by cape cobra and boomslang are attributable to habitat specialization, interspecific competition or predator avoidance.

http://etd.uwc.ac.za/29

Chapter 2: Correlates of colony visitation and predation by the snakes Naja nivea and Dispholidus typus

2.1 Introduction

2.1.1 Influences on snake activity

Movement and activity patterns of snakes are influenced by a complex combination of environmental factors and are fundamental to understand their biology and ecology (Nathan et al. 2008, Eskew & Todd 2017). Snakes often display seasonal activity patterns that are linked to various biotic and abiotic factors including weather conditions, food availability and reproduction (Marques et al. 2011). Various species from tropical and temperate regions display seasonal activity patterns (Brown & Shine 2002, Eskew & Todd 2017). However, the associated predictors for seasonal activity can vary. For example, weather conditions influence activity of tropical snakes only modestly, while seasonal reproduction and food availability appear to be the dominant predictors (Brown et al. 2001, Brown & Shine 2002). Activity of temperate snakes on the other hand is strongly influenced by weather conditions (Eskew & Todd 2017). Because of the strong climate seasonality, snake activity peaks during warmer months (Gregory 1982). Climatic predictors vary between species and may include other factors such as humidity, wind speed and moonlight (Clarke et al. 1996, Sun et al. 2011, Eskew & Todd 2017). However, weather conditions are not the only factors influencing snake activity. During mating season, male individuals move significantly further distances than females in many species (Maritz & Alexander 2012), while female individuals move further during egg-laying migrations (Bonnet et al. 1999). Food availability can also influence predator foraging—the presence and absence, but also movement and activity of prey affect predatory movement (Brown et al. 2002, Brown & Shine 2002).

http://etd.uwc.ac.za/30

Most studies that focus on the correlates of activity patterns of snakes have largely focused on tropical snakes, and temperate snakes from North-America and Europe (Gregory 1982, Daltry et al. 1998, Sun et al. 2001, Brown et al. 2002, Brown & Shine 2002, Shine et al. 2002, Gregory 2009). In a southern African context, such studies are less common and only started to proliferate in more recent years (Alexander & Marshall 1998, Maritz & Alexander 2012, Glaudas & Alexander 2016, Fizzotti 2018). Movement studies are fundamental for various reasons— movement takes part in many aspects of an organism’s biology (Nathan et al. 2008). Ultimately, understanding movement and activity patterns improves sampling strategies, especially of elusive and infrequently moving animals such as snakes (Willson et al. 2011, Alexander & Maritz 2015, Eskew & Todd 2017). Furthermore, a comprehensive understanding of predator movement has important implications for prey populations and permits more conclusive interpretations of studies on prey mortality, especially in the context of nest predation (Lloyd 2016).

2.1.2 Nest predation

A significant proportion of avian mortality is caused by nest predation (Ricklefs 1969). Researchers have focused on identifying nest predators and found that snakes are often among the main predators (Thompson et al. 1999, Thompson & Burhans 2003, Thompson & Burhans 2004, Stake et al. 2004, Stake et al. 2005, Natusch et al. 2016). Thompson & Burhans (2004) found that predation studies using artificial nests might often be biased. Artificial nests appear to be predated mainly by mammals and birds, but not snakes. In comparison, a large proportion of the predators of real nests are snakes (65%), followed by birds and mammals (Thompson & Burhans 2004). Furthermore, nest predators may vary on seasonal and diel scales (Natusch et al. 2016). Thus, inferences based on results from artificial nest experiments should be made with caution. Weatherhead and Blouin-Demers (2004) pointed out the significance of ornithologists studying snakes to better understand nest predation. Unfortunately, studies on nest predation face limitations due to a lack of understanding of the predators themselves (Lloyd 2016).

Sociable weavers experience high rates of nest predation by snakes, specifically by cape cobra and boomslang (MacLean 1973c, Covas 2002). In areas where the distribution of the snakes overlap with those of sociable weaver, the snakes are thought to regulate the population dynamics of the colonial breeder (MacLean 1973c, Covas 2002, Lloyd 2016). Predation by

http://etd.uwc.ac.za/31 snakes causes up to 70% of nest mortality (Covas 2002). In areas where only one of the species occurs nest survival is potentially higher. However, the impact of individual snake species on the reproductive success of sociable weavers has not been quantified. In addition, a lack of information regarding of the influence of environmental conditions on snake movement hinders ornithological understanding of nest predation and variation thereof.

2.1.3 Effect of climate change on avian reproduction

Temperate regions are characterized by distinct summer and winter seasons. In these regions, the onset of avian reproduction is linked to temperature, with earlier clutch laying in warmer springs (Dunn 2004, Visser et al. 2009). However, in arid systems, rainfall appears to be the significant predictor for reproduction by birds (Zann et al. 1995, Bolger et al. 2005, Altwegg & Anderson 2009, Saunders et al. 2013). Food availability is correlated with both temperature and rainfall, but correlation strength with either one factor likely varies between systems (Dean & Milton 2001, Dunn 2004, Altwegg & Anderson 2009). In semiarid savannas in southern Africa, the reproductive onset and output of sociable weavers (Philetairus socius) is highly correlated with summer rainfall, which is a significant predictor of food availability (Covas et al. 2008, Altwegg et al. 2014). However, Mares et al. (2017) found that rainfall and temperature influence the reproduction of this communal breeder. Specifically, warmer springs result in earlier breeding onset and cooler summers can prolong the breeding season (Mares et al. 2017). A long-studied population of sociable weavers (Philetairus socius) in South Africa has experienced a decline in numbers potentially in response to climate change (Altwegg et al. 2014, Mares et al. 2017). Increasing temperatures and aridity result in lower food abundance, and ultimately influence reproductive output negatively (Covas et al. 2008, Altwegg et al. 2014, Lloyd 2016). Importantly, the potential impacts of climate change on sociable weaver reproduction combined with nest predation pressure by snakes, may accelerate population declines and potentially local extinctions (Lloyd 2016). Unfortunately, the environmental correlates of snake foraging behaviour remain poorly described, limiting our ability to predict such effects.

The aim of this chapter is to identify environmental correlates of sociable weaver colony visitation by snakes at Benfontein Nature Reserve. Specifically, I integrate more than 2000 sociable weaver colony observations in which I scored the presence or absence of cape cobra and boomslang respectively to understand patterns of colony visitation. Using logistic regression

http://etd.uwc.ac.za/32 analyses, I relate species-specific presence/absence patterns to a variety of abiotic (e.g. wind, temperature) and biotic factors (e.g. breeding activity, nest characteristics). I hypothesize that the presence of cape cobra and boomslang in sociable weaver colonies occurs non-randomly and (1) depends on a particular set of temporal and spatial variables, and (2) that those variables differ for the two snake species.

2.2 Methods

2.2.1 Field methods Presence/absence

Between 8/8/2016 and 19/4/2017, 2256 nest checks of sociable weaver colonies were conducted to detect snakes. The study, conducted at the Benfontein Nature Reserve (BNR), included 13 colonies. On arrival at a colony, before conducting any inspections of chambers, visual surveys were conducted to detect the presence or absence of any snake. Standing underneath the colony, I used a handheld torch to illuminate the inside of each chamber. Additionally, the tree canopy and the roof of the nest structure were examined for the presence of snakes. Individual snakes were identified based on physical features including their scales (smooth versus keeled), colouration and head features. All observed snakes were classified as adults based on their size and lack of juvenile markings (i.e. black banded throat of Naja nivea and overall colouration of Dispholidus typus.

The protocol for detecting snakes in sociable weaver colonies has been used in previous studies (Covas 2002, Covas et al. 2008, Spottiswoode 2007) and is the standard protocol for the research conducted at BNR. The protocol has succeeded in detecting snakes within colony chambers, but detectability of snakes in tree canopies, especially boomslang, is low. Observations on radio- tracked boomslang at Tswalu Kalahari Reserve have shown that individuals may remain undetected in tree canopies even when known to be present (B. Maritz pers. comm).

A variety of abiotic (e.g. wind, temperature) and biotic factors (e.g. breeding activity, nest characteristics) were collected. These factors were grouped into spatially- and temporally- variable factors. While spatially-variable factors remain constant through time, they vary from colony to colony. Conversely, temporally-variable factors change over time, but are likely to be

http://etd.uwc.ac.za/33 similar across the study site at any one time point. Hereafter, I will refer to them as spatial variables and temporal variables.

Spatial variables

For each colony, spatial variables were collected to relate them to the presence/absence of snakes and included the following: (1) number of adult birds, (2) number of chambers, (3) volume of the whole nest structure, (4) distance between lowest branch of host tree and the ground, and (5) distance between host tree to nearest tree. All measurements were taken to the nearest 0.01 m using a 50 m measuring tape.

(1) The number of adult birds per colony was quantified by annual colony captures, which are typically conducted at the beginning of a new breeding season by the sociable weaver research group, who shared their data with me. During such captures, all individuals of a colony were captured using mist nets set up around the colony (Covas et al. 2004a). The captured birds were counted and identified using the individual-specific colour- combination of plastic rings around their feet. (2) The SWP research group placed numbered, yellow plastic tags next to the entrance chambers to identify them. New chambers were tagged at the beginning of the breeding season. Chambers which collapsed or were newly built throughout the breeding season were tagged and documented. However, the change of the number of chambers throughout the season is small. Therefore, I used the number of chambers recorded in the beginning of the breeding season for all of my analyses. Again, these data were gathered by the sociable weaver team and shared with me. (3) Measurements of the length, width and height of each colony were recorded to estimate the volume of the structure. In cases of structures with irregular shapes, two measurements were taken (e.g. minimum and maximum length to get the average length). The length of a colony was measured parallel to the supporting branch and the width perpendicular to the length. The height of the structure was measured from the lower side of the colony (where chamber entrances are located) to the top of the dome-like structure. For height measurements the minimum and maximum height was recorded based on the lowest and highest chamber of the colony, as these often varied extensively.

http://etd.uwc.ac.za/34

(4) Measurements of the distance between the lowest branch and the ground were taken. Snakes are able to use low hanging branches to ascend trees. At the study site, camelthorn trees (Vachellia erioloba), hosting sociable weaver colonies, have different appearances or growths, resulting in some individual trees having very low hanging branches. (5) Distance from the trunk base of the tree hosting the colony to the trunk of the nearest tree was recorded. The nearest neighbouring tree was defined as a woody plant with a DBH (diameter at breast height) greater than 15 cm.

Temporal variables

The temporal variables represent (1) breeding related variables and (2) weather variables, which were obtained for every colony observation. Breeding variables included the numbers of eggs and chicks, and the number of active breeding chambers in each colony. I incorporated six weather variables in this study, namely minimum and maximum daily temperature, wind speed, precipitation, humidity and air pressure.

(1) The breeding variables were obtained through daily visitations to the colonies as performed by the sociable weaver team: full colony checks and control checks. Every third day, when colonies underwent a full colony check, all chambers were inspected for the presence of new clutches. Sociable weaver breeding pairs lay their eggs in 1-day intervals, thus this survey approach enabled estimation of the laying date of new clutches. Additionally, the examination of the remaining chambers identified the reproductive content or the absence of such. Hence, every three days, the exact number of eggs and chicks was updated, and any losses recorded, for every colony. Despite serving a different purpose, control checks often discovered the loss of eggs or chicks (broken eggs or dead chicks on the ground beneath the colony or snake in chambers) prior to full colony checks. In the case of dead chicks that were already ringed, the identification using the rings ascertained the chamber of origin. For broken eggs, the chamber of origin was usually identified by checking the chambers. These recordings aided in a more precise documentation of the chambers’ content at any point in time. The documentation of clutches and broods additionally enabled quantification of the number of active breeding chambers.

http://etd.uwc.ac.za/35

(2) Weather related measures were required to assess whether abiotic factors predict the presence/absence of snakes. I obtained most of the weather data from the South African Weather Service (SAWS). The nearest weather station, located at Kimberley Airport (station number 0290468A9), was approximately 10 km from the study site. The obtained data included minimum and maximum temperature (°C), wind speed (m/s), humidity (%), air pressure (hPa) and precipitation (mm). However, rainfall was recorded directly on BNR using rain gauges installed by the sociable weaver research project. Minimum temperature reflects the lowest recorded temperature at night, while maximum temperature represents the highest record for the day. Wind speed, humidity and air pressure recordings reflect the morning conditions (at 0800 h) as sampling took place in the mornings.

2.2.2 Statistical analysis

To identify any temporal and spatial correlates of sociable weaver nest visitation and predation by cape cobra and boomslang I used two different data sets—one set including all 13 study colonies and another set excluding the protected colonies. For both data sets, I performed Χ2 tests on the numbers of observed colonies to identify whether there were significant differences between the frequencies with each colony was observed. Additionally, I performed regression tests between the number of observed snakes (of each species) at individual colonies and the frequencies with which colonies were observed to test if they are related. The Χ2 tests were performed in SPSS and the regression tests were performed in Microsoft Excel. All other statistical analyses were performed in R (R Development Core Team, 2018).

Temporal and spatial data

For all variables I compared Pearson’s correlation coefficients to identify multicollinearity between variables, to inform whether a principal component analysis was required. I generated a heat map to visualize any existing correlations, including their corresponding correlation coefficients (r). I standardized the temporal and spatial data using the scale function in base R (R Development Core Team, 2018). This standardization keeps the relationships of the transformed variables equal to those of the untransformed variables.

http://etd.uwc.ac.za/36

Principal component analysis (PCA)

To meet the requirements of regression analyses, I ran principal component analyses (PCA) on the temporal and spatial data sets to remove the multicollinearity among the predictor variables. The PCA converted correlated variables into new uncorrelated multivariate scores, called principal components (PCs). Additionally, PC factor loadings, Kaiser’s criterion (eigenvalues) and the proportions of variance were quantified. Loading coefficients greater than or equal to 0.3 and less than or equal to -0.3 were considered significant.

Retaining components

Three different methods (screeplots, eigenvalues and cumulative percentage of variance) were used to decide which principal components to retain. Screeplots are graphs that show the variance for each PC, which typically results in an elbow-shaped figure. The screeplot method is based on the first pronounced break between components. All components before and at the point of break were retained. The next method uses the eigenvalues or Kaiser’s criterion to decide on the number of retained components. All PCs with eigenvalues greater than 1 were retained. In order to use the cumulative percentage of variance to retain PCs, one has to decide on a cut-off percentage. I retained the fewest number of PCs that explain at least 75% of the variation found in the data. The numbers of retained PCs from each method were compared to one another to identify any similarities or differences between the number of retained components.

Multiple logistic regression analysis

Two multiple logistic regression analyses were run on the retained principle components describing temporal and spatial variables. The first analysis examined the relationship between the presence/absence of cape cobra and the temporal and spatial PCs. The second analysis examined the same relationship for boomslang. The models tested which principal components significantly predict the presence or absence of the snakes. I performed a backwards model selection with all retained PCs dropping non-significant components for the final model. The best fit was given by the model with the lowest Akaike’s information criterion (AIC). To interpret the findings from the regression analyses, the eigenvectors of significant PCs were multiplied with the Z-value to obtain the correct direction of the examined relation.

http://etd.uwc.ac.za/37

2.3 Results

2.3.1 Temporal and spatial collinearity

Correlation analysis showed multicollinearity among both temporal and spatial variables (Figure 2.1). Some spatial variables, explaining nest characteristics, were highly correlated with each other (Figure 2.1).

Figure 2.1: Heat map matrix of Pearson’s correlation coefficients for multiple spatial and temporal variables.

Tree protection was negatively correlated with measurements of colony size (volume of nest structure, number of adults, and number of chambers). The correlation coefficients revealed weaker correlations for temporal variables. Minimum temperature is positively correlated with maximum temperature and wind speed, but negatively correlated with pressure (Figure 2.1).

http://etd.uwc.ac.za/38

2.3.2 Data analysis including all 13 colonies

Among all 13 colonies, I gathered a total of 2256 presence/absence observations of snakes in sociable weaver colonies over the entire study period (Table 2.1). There was a significant difference in the frequencies with which each colony was observed (Χ2= 207.11, df = 11, P < 0.01; Table 2.1).

Table 2.1: Summary of the number of observations and the snake occurrences for each study colony at the Benfontein Nature Reserve. The total values (Σ) display the data sets including all colonies (n = 13) and excluding protected colonies (n = 9).

Naja nivea Dispholidus typus Colony ID Observations occurrences occurrences 2 149 4 0 7 193 3 3 8 187 2 6 11 182 4 6 20 179 7 0 21 144 0 1 27 178 0 1 31 182 5 0 32 183 3 2 37 195 9 0 38 134 1 0 43 153 0 0 71 197 0 0

Σ13 2256 38 19

Σ9 1549 31 18 The average number of observations per colony was 174. All individual snakes that were detected were classified as adults. Out of the 57 , 38 records were cape cobra and 19 records were boomslang (Table 2.1). I found that the number of cape cobra, boomslang and pooled snake occurrences were not related to the number of times colonies were observed (Regression tests: cape cobra – F1,11 = 1.99, P = 0.19; boomslang – F1,11 = 1.32, P = 0.27; pooled snake occurrences

– F1,11 = 4.19, P = 0.07).

http://etd.uwc.ac.za/39

The comparison of methods for retaining principal components showed that the number of retained PCs mostly differ (Table 2.2). The same number of PCs describing spatial variables were retained using either the Kaiser’s criterion or cumulative percentage of variance.

Table 2.2: Comparison of three different methods to retain principal components explaining spatial and temporal variables for 13 Philetairus socius colonies. Method Temporal PCs retained Spatial PCs retained Screeplot 2 3 Kaiser's criterion 3 2 Cumulative percentage of variance 4 2

The screeplot displaying spatial variability showed a pronounced break in the slope after the third spatial component (Figure 2.2A). However, two PCs already explain 81% of the total variation, well above the cut-off percentage (Table 2.3).

http://etd.uwc.ac.za/40

The screeplot method retained two temporal PCs (Figure 2.2B), explaining less than 50% of the total variation in the data (Table 2.3). The Kaiser’s criterion method retained three temporal PCs, explaining approximately 65% of data variation. The last method, cumulative percentage of variance, retained four temporal PCs, which together explained 77.9% of the variation found in the data (Table 2.3). Based on the low cumulative variation explained by two and three retained PCs, the results from the last method were used to retain components.

Table 2.3: Eigenvalues, proportion of variance explained and cumulative proportion of variance explained for retained temporal and spatial variables for 13 Philetairus socius colonies.

Proportion of variance Cumulative proportion of Eigenvalues explained variance explained

Temporal PCs PC1 2.09 0.299 0.299 PC2 1.33 0.189 0.489 PC3 1.12 0.160 0.648 PC4 0.91 0.131 0.779 Spatial PCs PC1 3.08 0.514 0.514 PC2 1.78 0.297 0.810 PC3 0.56 0.093 0.903

Cape cobra

Table 2.4 presents the results from the multiple logistic regression analysis for cape cobra when including all sociable weaver colonies. The presence or absence of cape cobra is significantly affected by the first two spatial PCs and first temporal PC (Table 2.4).

http://etd.uwc.ac.za/41

Table 2.4: Results from multiple regression analysis for all retained temporal and spatial principal components for cape cobra (Naja nivea) for the data set including all 13 sociable weaver (Philetairus socius) colonies. The asterix (*) indicates the significance of a variable (P < 0.05). Principal components Z P Standard Error Spatial PC1 -2.258 0.024* 0.111 Spatial PC2 -2.821 0.005* 0.221 Temporal PC1 2.709 0.007* 0.126 Temporal PC2 0.725 0.469 0.154 Temporal PC3 -0.208 0.835 0.160 Temporal PC4 0.196 0.844 0.173

Finally, spatial PC2 (29.7% of variance) is significantly driven by several variables including the distance to the nearest tree, the distance between the lowest branch and ground, and lastly the snake excluding protection of the tree (Table 2.5). After multiplying by their corresponding Z- values, these findings indicate that cape cobra are more likely to occupy sociable weaver colonies when weather conditions are warm and windy and, predominantly favour large, unprotected colonies with high tree canopies and nearby trees.

http://etd.uwc.ac.za/42

Table 2.5: Factor loadings of retained principal components for climatic and environmental variables across 13 sociable weaver colonies collected at Benfontein Nature Reserve. Temporal variables PC1 PC2 PC3 PC4 Max. temperature 0.309 -0.332 0.642 -0.180 Min. temperature 0.603 -0.198 0.068 0.059 Wind 0.460 0.202 -0.276 0.426 Humidity -0.005 -0.676 -0.136 -0.201 Air pressure -0.545 -0.162 0.063 -0.076 Rain 0.073 -0.405 -0.673 -0.094 Number of active nests 0.165 0.403 -0.178 -0.853 Spatial variables Number of adults -0.536 0.035 -0.044 - Volume of colony -0.540 -0.046 -0.210 - Number of chambers -0.561 0.063 -0.019 - Distance to nearest tree 0.079 0.598 -0.762 - Lowest branch -0.249 0.533 0.554 - Tree protected 0.194 0.592 0.259 -

Boomslang

The multiple logistic regression analysis identified only the fourth temporal PC as significant (Table 2.6). This component explains 13.1% of the overall variance and is driven by the number of active breeding chambers and wind, respectively (Table 2.5). Average wind speed for boomslang occurrences was 3.6 m/s. After multiplying by their corresponding Z-values, these findings indicate that boomslang are more likely to occupy weaver nests when it is less windy, and predominantly favour colonies with greater breeding activity. The results indicate that temperature does not meaningfully drive the foraging of boomslang within the range of temperatures recorded during the study period.

http://etd.uwc.ac.za/43

Table 2.6: Results from multiple regression analysis for all retained temporal and spatial principal components for boomslang (Dispholidus typus). The asterix (*) indicates the significance of a variable (P < 0.05). Principal components Z P Standard Error Spatial PC1 -0.832 0.405 0.154 Spatial PC2 -1.116 0.264 0.264 Temporal PC1 0.938 0.349 0.188 Temporal PC2 -0.112 0.911 0.206 Temporal PC3 -0.714 0.475 0.168 Temporal PC4 -2.611 0.009* 0.250

2.3.3 Data analysis excluding protected colonies

Among all non-protected colonies (n = 9) a total of 1549 colony observations were recorded (Table 2.1). There was a significant difference between the total observations for each colony (Χ2 = 192.1, df = 7, P < 0.01; Table 2.1). Excluding protected colonies reduced the sample size for cape cobra from 38 to 31 individuals, and for boomslang from 19 to 18 individuals. Colony 43 and 71 were not visited by any snakes, while Colony 27 was visited by one boomslang and Colony 20 was visited by cape cobra on seven occasions. Again, I found that the number of cape cobra and boomslang occurrences were not related to the number of times colonies were observed (Regression tests: cape cobra – F1,7 = 3.97, P = 0.08; boomslang – F1,7 = 1.71, P = 0.23). However, the number of the pooled snake occurrences per colony were related to the number of times colonies were observed (Regression test: pooled snake occurrences – F1,7 = 16.04, P < 0.05).

For consistency, I used the same method to retain principal components explaining spatial and temporal variables as in the previous data set for all colonies, thus excluding the reporting of further screeplots. The cumulative percentage of at least 75% was reached when retaining the first two spatial PCs and the first four temporal PCs (Table 2.7).

http://etd.uwc.ac.za/44

Table 2.7: Cumulative proportion of variance for spatial and temporal principal components for the data set excluding protected sociable weaver (Philetairus socius) colonies. Cumulative Proportion PC1 PC2 PC3 PC4 PC5 PC6 PC7 of Variance Spatial data 0.648 0.839 0.974 0.995 1.000 - - Temporal data 0.296 0.483 0.642 0.776 0.880 0.957 1.000

Cape cobra

When excluding protected sociable weaver colonies, the presence or absence of cape cobra is only significantly affected by the first temporal PC (Table 2.8). Temporal PC1 explains 29.6% of the total variance found in the temporal data.

Table 2.8: Results from multiple regression analyses for all retained temporal and spatial principal components for cape cobra (Naja nivea) for the data set excluding protected sociable weaver (Philetairus socius) colonies (n = 9). The asterix (*) indicates the significance of a variable (P < 0.05).

Principal components Z P Standard Error

Spatial PC1 -0.733 0.465 0.122

Spatial PC2 1.016 0.310 0.212

Temporal PC1 2.342 0.019* 0.139

Temporal PC2 0.871 0.384 0.171

Temporal PC3 -0.246 0.805 0.182

Temporal PC4 -0.150 0.881 0. 205

The contributions by variables are the same as for the previous data set: minimum temperature, air pressure, wind and, maximum temperature, respectively (Table 2.9). The spatial PCs, which are significant for the analyses with all colonies, are not significant when removing the protected colonies from the analysis. After multiplying the factor loadings by their corresponding Z-values,

http://etd.uwc.ac.za/45 these findings indicate that cape cobra are more likely to occupy sociable weaver colonies when weather conditions are warm and windy.

Table 2.9: Factor loadings of retained principal components for temporal and spatial variables across 9 sociable weaver colonies collected at Benfontein Nature Reserve. Temporal variables PC1 PC2 PC3 PC4 Max. temperature 0.311 -0.328 0.646 -0.185 Min. temperature 0.609 -0.180 0.067 -0.072 Wind 0.465 0.218 -0.278 0.371 Humidity 0.004 -0.677 -0.139 -0.215 Air pressure -0.544 -0.186 0.680 -0.067 Rain 0.077 -0.425 -0.659 -0.032 Number of active nests 0.119 0.372 -0.207 -0.878 Spatial variables Number of adults -0.523 0.053 - - Volume of colony -0.516 0.163 - - Number of chambers -0.549 0.078 - - Distance to nearest tree -0.149 -0.982 - - Lowest branch -0.370 -0.024 - -

Boomslang The results for the analysis using the data set excluding protected colonies are the generally the same as for the analysis that included all colonies. The multiple logistic regression analysis identified only the fourth temporal PC as significant (Table 2.10). This component explains 13.4% of the overall variance and is highly contributed to by the number of active breeding chambers, and wind, but at lower strength (Table 2.9). After multiplying by their corresponding Z-values, these findings indicate that boomslang predominantly favour colonies with high breeding activity are more likely to occupy weaver nests when it is less windy.

http://etd.uwc.ac.za/46

Table 2.10: Results from multiple regression analysis for all retained temporal and spatial principal components for boomslang (Dispholidus typus) for the data set excluding protected sociable weaver (Philetairus socius) colonies (n = 9). The asterix (*) indicates the significance of a variable (P < 0.05).

Principal components Z P Standard Error Spatial PC1 0.445 0.657 0.177 Spatial PC2 -0.851 0.395 0.246 Temporal PC1 0.438 0.661 0.196 Temporal PC2 -0.178 0.859 0.214 Temporal PC3 -1.139 0.255 0.179 Temporal PC4 -3.172 0.002* 0.282

2.4 Discussion

2.4.1 Influences on snake presence

Snake presence in sociable weaver colonies varies temporally and spatially, correlating with environmental variables. Moreover, the correlates of snake detection in colonies and environmental variables are different for cape cobra and boomslang. Both species appear to be predominately impacted by temporally variable factors.

A combination of principal components, including spatial and temporal variables, predict the presence of cape cobra in sociable weaver colonies. The results indicate that cape cobra primarily forage on warm and windy days and possibly favour larger colonies. Cape cobra presence in colonies increased with higher minimum and maximum temperature, and greater wind speed, but low air pressure. In this study, the average minimum temperature for active cape cobra was 15.8 °C. However, cape cobra were absent when minimum temperatures dropped below 9 ºC, indicating a minimum temperature threshold for movement. Active on warmer days, minimum temperature appears to be the more significant predictor for cape cobra presence than maximum temperature. The first analysis including data from all colonies identified three significant nest variables (number of adults, colony volume, number of chambers), which were highly correlated and generally described colony size. Mostly living in terrestrial and

http://etd.uwc.ac.za/47 subterranean habitats, cape cobra might detect larger colonies more readily due to the physical size and greater avian activity around the colony’s perimeter (Spottiswoode 2007). Furthermore, nest predation by cape cobra increases for colonies with nearby trees and lower branches. Cape cobra may favour colonies with greater surrounding tree coverage to avoid open spaces where they are more vulnerable to predation (e.g. predatory birds and mongooses). Additionally, trees with low branches might provide them with extra pathways to ascend a tree besides using the stem.

The datasets produced similar findings—factors influencing boomslang presence in colonies are limited to two variables. The results showed that greater breeding activity and lower wind speed results in increased nest visitation by boomslang. Arboreal snakes might experience higher rates of cooling or desiccation at high wind velocity (Baeyens & Rountree 1983, Sun et al. 2001). Also, boomslang might even avoid cape cobra as they forage during windy conditions. Eskew & Todd (2017) suggested that foraging success and predation risk might explain predicting factors for snake activity. Temperature was found to be non-significant for boomslang activity. However the data showed that boomslang foraged in colonies at similar temperatures as cape cobra. Analyzing the presence-absence data for boomslang showed that the cumulative proportion of variance was the most appropriate method for retaining principal components.

The differences of environmental variables predicting the presence or absence of cape cobra and boomslang in sociable weaver colonies might arise due to several reasons. Visitations of colonies by cape cobra are strongly impacted by climatic conditions while boomslang visitation are predominantly predicted by breeding activity. These differences are potentially due to major differences in the biology and ecology of the two snake species (e.g. habitat use, diet, sensory organs, functional morphology; Lillywhite & Henderson 1993, Bates et al. 2014, Lillywhite 2014, Marais 2014, Layloo et al. 2017, Smith et al. 2019). Despite distributional and vegetational overlaps, the preferred habitats of the two species differ substantially—cape cobra are mostly terrestrial while boomslang is a primarily arboreal species (Bates et al. 2014, Marais 2014, Smith et al. 2019). The arboreal lifestyle of boomslang might cause the earlier detection of breeding activity (i.e. when most chambers are filled with eggs) than cape cobra. Mullin & Cooper (1998) found that foraging gray rat snakes use visual stimulus to detect the movement of provisioning adult birds around nests. Boomslang might detect the breeding activity more readily through visual senses as a result of living in the same habitat as the prey. During incubation

http://etd.uwc.ac.za/48 period, the visual stimulus is greater in the direct vicinity of the nests as adult birds spend more time incubating than on foraging. Arboreal snakes usually have well developed eyes, potentially aiding in the detection of their prey (Lillywhite & Henderson 1993). Once nest provisioning increases after hatching, cape cobra start targeting nests. The differences in habitat use reflect the differences in the diet of the two snakes. Individuals of both consume all tetrapod classes, but the proportions at which these classes are consumed by the two snakes differ considerably. Cape cobra feed mostly on mammals and reptiles, while boomslang prey mostly on birds and reptiles (Layloo et al. 2017, Smith et al. 2019). Additionally, the different determinants of snake presence might be influenced by predator avoidance by boomslang. Cape cobra are known to predate heavily on other snakes (Layloo et al. 2017), including boomslang (Maritz et al. 2019). Therefore, boomslang might maximise their net benefits, by visiting colonies earlier to reduce exposure to predation by cape cobra and to increase foraging success as most chambers are actively breeding. However, further analyses of other factors are required to produce a better understanding of this activity pattern.

Spottiswoode (2009) showed that nest predation increases with colony size. Similarly, my results showed that cape cobra presence is related to larger colonies. However, only the analysis including all 13 colonies identified colony size as a predicting variable of cobra presence. Therefore, using colony size as a variable has less statistical power to predict cape cobra presence in sociable weaver colonies than using climatic variables such as minimum temperature and wind. The presence of boomslang increases with increasing breeding activity, which indicates that colony size is a potential predictor for presence as well because larger colonies usually have greater numbers of adult individuals than smaller colonies producing greater breeding activity. The metallic starling (Aplonis metallica), also a communal nesting passerine, has a variety of predators including mammals, birds, reptiles and amphibians (Natusch et al. 2016). Snakes and amphibians are the main predators of starling eggs and nestlings and primarily congregate under colonies during the nesting stage (Natusch et al. 2016). Predator presence is significantly greater at larger communal nests of metallic starlings (Natusch et al. 2016). Nest predators such as snakes may detect larger colonies more readily than smaller ones.

Reasons for increased snake presence in larger colonies might involve easier detection of the nests, greater avian activity around the nest’s perimeter and greater food resource to be exploited by predators in order to maximize their energy intake. Previous studies on nest predation suggest

http://etd.uwc.ac.za/49 that predators detect avian nests through increased parental activity (Skutch 1949, Martin et al. 2000). Sociable weavers forage in flocks resulting in greater activity, which might make them more conspicuous to cape cobra than individually foraging birds (Spottiswoode 2007). In tropical forests, snakes use olfactory cues to locate communal bird nests (Natusch et al. 2017). Parental activity and the strength of the scent present at the colony likely increases with colony size as the amount of breeding increases as well. Thus, increased scent and parental foraging after the onset of breeding are possible cues for nest predators to start their search for colonies, specifically cape cobra. Martin et al. (2000) showed that nest site is a significant factor predicting nest predation. Similarly, Sperry et al. (2008) suggest that greater predation rates of golden-cheeked warblers (Dendroica chrysoparia) compared to those of black-capped vireos (Vireo atricapillais) are likely the consequence of Texas rat snakes (Elaphe obsoleta) frequenting warbler habitats more than those of vireos. This indicates that certain colonies may experience more predation due to its location within the landscape. My results showed that the location of a colony in terms of distance to nearest tree predicted cobra presence negatively. The same pattern was observed for branch height (i.e. higher branches decreased snake visitation).

Other studies of climatic influences on snake activity show similar results. For example, activity pattern of pit vipers (Gloydius shedaoensis), a species from north-eastern China, is influenced by spatial (different parts of a path) and temporal (season and weather conditions) factors (Sun et al. 2001). The study found that the presence of pit vipers was significantly affected by temperature, wind speed and humidity. More snakes were observed on hot days, with lower wind speed and humidity. Specifically smaller snakes (juveniles) are less active on windy days, indicating that snake activity varies depending on body size (Sun et al. 2001). Eskew & Todd (2017) found that daily activity of temperate snakes is positively and negatively correlated with temperature and rainfall, respectively. Possible explanations for these factors included physiological constraints, and trade-offs between foraging success and predation risk (Eskew & Todd 2017). Studies have suggested that snakes are more active on warmer days (Sun et al. 2001, Sperry at al. 2008). Sperry et al. (2008) reported also that the activity pattern of Texas rat snakes (Pantherophis obsoletus) is correlated stronger with minimum daily temperatures than maximum daily temperatures. These studies and the finding from this study show that variables affecting snake activity differ. Such differences may differ between and among species, likely as a result of

http://etd.uwc.ac.za/50 differences in their natural histories (e.g. body size, foraging mode, habitat; Sun et al. 2001, Eskew & Todd 2017).

2.4.2 Temporal niche partitioning

On a broad scale, cape cobra and boomslang are less likely to experience high interspecific competition in areas where they co-occur because 1) they largely inhabit different types of microhabitats (terrestrial versus arboreal) and 2) the primarily consumed prey types differ extensively. However, in areas of syntopy with sociable weavers, these snakes heavily depredate the same food resource, raising the question how they are able to coexist. Competing for the same resource as the ophiophagous cape cobra, involves an inherent predation risk for boomslang. My finding of the differences in temporal factors influencing the activity of these two snakes, indicate that they may partition the singular food resource on the temporal scale axis. Temporal niche partitioning (or conditional differentiation) occurs when competing species are able to coexist based on varying environmental conditions (Schoener 1974, Angert et al. 2009). Based on the results here, the presence of snakes in sociable weaver colonies is influenced by varying temporal factors—boomslang presence depends on breeding activity and cape cobra presence depends on warm minimum temperature and windy conditions. Additionally, opposed to cape cobra, I found that boomslang forage in less windy conditions. Research on the mountain lizard (Iberolacerta aurelioi) showed that the thermoregulation of these reptiles is negatively affected by increased windspeed, possibly reducing daily activity (Ortega et al. 2016). With increasing wind speed, temperatures in arboreal sites are lower than terrestrial sites, relating to differences in body temperatures (Sun et al. 2002). Sun et al. (2001) found that wind speed affects pit-viper (Gloydius shedaoensis) numbers, specifically juveniles as they may be affected more than larger individuals. By virtue of a smaller SVL relative to body mass, cape cobra have a lower surface-area-to-volume ratio (SA:V) and therefore greater thermal inertia than boomslang. With a lower SA:V, the effect of windspeed on body temperature would be less for cape cobra.This suggests that boomslang forage in less windy conditions because arboreal habitats are cooler than terrestrial habitats (Sun et al. 2002). Therefore, cape cobra forage on windy days to counter the high temperatures of the soil.

Differences in activity patterns of other organisms has been linked to temporal niche partitioning (Albrecht & Gotelli 2001, Rouag et al. 2007, Sivan et al. 2013). The two sympatric lizards

http://etd.uwc.ac.za/51

Psammodromus algirus and Acanthodactylus erythrurus share the same food resource, but partition it based on differences in daily activity (i.e. they forage at different diel times). This temporal difference reduces direct competition and potentially aids in the coexistence of two species with greatly overlapping habitats and diets (Rouag et al. 2007). Additionally, a recent study on two desert snakes in Israel investigated coexistence between Saharan sand viper (Cerastes vipera) and the Crowned leafnose snake (Lytorhynchus diadema; Sivan et al. 2013). Dietary selection differs for these snakes, but both ingest a large proportion of a specific food item, the Nidua fringe-fingered lizard (Acanthodactylus scutellatus). Additionally, both snakes are nocturnal and active during spring and summer, but their seasonal and daily activities peak at different times, suggesting that temporal niche differentiation contributes to the coexistence of these two desert snakes (Sivan et al. 2013).

As a result of different environmental conditions favouring the presence or activity of a specific species, cape cobra and boomslang might reduce interspecific competition. The competitive interactions between the two predators contribute to the structuring of the natural communities. For sociable weavers, the temporal niche partitioning between these snakes means that different environmental conditions favour nest predation by specific predators.

2.4.3 Sampling biases and limitations

Temporally variable factors are inherently temporally autocorrelated with potential influences on my findings. Extended periods of warmth might be needed to trigger foraging behaviour, and so the actual measure of climatic variables on a single day might provide little information. The analyses of this study are strongly linked to fine scale temporal variability due to the selection of variables, daily sampling and restriction of sampling to a single breeding season of sociable weavers. On account of my working with the sociable weaver project, I opted for a more fine scale temporal approach. However, seasonal effects might be an important factor as well. Snakes potentially forage more in warm weather—not because it is warm, but because it is summer time when birds breed. The findings of this study are directly related to when the birds are breeding. My approach makes it difficult to make inferences on a broad temporal scale. However, snake activity within sociable weaver colonies is expected to decrease during winter when the birds are usually not breeding. Including presence-absence data for snakes during the winter months over

http://etd.uwc.ac.za/52 consecutive years will allow for more broad scale inferences and increase the sample size of observed snakes, which was especially low for boomslang.

The protection of colonies is an anthropogenic measure taken to eliminate nest predation by snakes. Such human induced interferences usually result in a sampling bias, which had implications for the analyses of both predicting spatial and temporal determinants of snake presence in colonies. I therefore performed the analyses of the presence-absence data of snakes including and excluding the protected colonies to identify any biases of the results. The results from these two analyses showed that the variables explaining colony size as predicting variable for cape cobra presence were significant when including all colonies but non-significant when excluding protected colonies. The removal of protected colonies from the analyses reduced the sample size severely. When predated, protected colonies should be included in analyses as these observations provide valuable information. However, the other variables identified as significant were the same for the analyses of the different data sets, which I interpret as statistically more powerful than the variables explaining colony size.

The frequencies at which colonies were observed differed throughout the breeding season. The number of observations was greater for larger colonies than for smaller ones, relating to more breeding activity. Thus data collection for the Sociable Weaver Project occurred more frequently at larger colonies. At a species level, the regression tests showed that the different numbers of times at which colonies were observed did not bias the number of cape cobra or boomslang observations. This held true for both analyses including and excluding protected colonies. However, the inclusion of protected colonies in the analysis for all snake occurrences affected the results. Colonies observed at higher frequencies did result in higher numbers of observed snakes at non-protected colonies.

The duration of colony sampling varied extensively with colony size and breeding intensity. As a result snake checks were conducted at varying times, which resulted in a different range of conditions at which colonies were observed. Snake presence in colonies is potentially influenced by these different conditions. Snakes visiting colonies for a short time period may be missed due to late arrivals at colonies. For future studies I recommend that a stricter protocol be in place ensuring similar times and conditions at which colonies are visited, reducing potential biases.

http://etd.uwc.ac.za/53

2.5 Conclusion

Cape cobra and boomslang activity in sociable weaver colonies appears to be influenced by different variables, potentially resulting in temporal resource partitioning. Both snakes are impacted by temporal variables, but these variables differ for cape cobra and boomslang. Additionally, cape cobra activity is influenced by spatial variables such as colony size, but likely as a secondary predicting variable. Climatic conditions have stronger statistical power in predicting cape cobra presence in sociable weaver colonies. Boomslang activity is strongly linked the breeding intensity of sociable weavers. These differences in snake activity indicate that certain conditions make predation by cape cobra more likely, whereas other conditions make predation by boomslang more likely.

This study highlights the value of studying the influences on snake movement. From a herpetological perspective, this research covers gaps in our knowledge of the natural history of snakes, specifically environmental factors that influence their movement activity. Movement plays a fundamental role in all parts of an organism’s life, such as foraging, reproduction and migration. Although this study focused on movement for foraging, the findings likely apply to the snakes’ general movement activity. Identifying environmental factors predicting snake movement is important for future studies in terms of sampling snakes. Sampling of snakes has always been considered challenging because of their unpredictable movement patterns (Penner et al. 2008). Identifying activity predicting factors improves sampling technique and success, and ultimately detectability (Willson et al. 2011).

From an ornithological perspective, understanding how environmental conditions impact different nest predators, helps to make inferences on nest predation risk of birds (Weatherhead & Blouin-Demers 2004). The finding of this study shows that different factors influence different predators. While one predator favours warm and windy conditions for foraging, another prefers generally less windy conditions. Therefore, the predation risk for the sociable weavers is complex because one set of conditions can make predations from one predator more likely but changing conditions might simply make another predator more likely.

http://etd.uwc.ac.za/54

Chapter 3: Predation rates and differential food resource utilization by two nest predators, Naja nivea and Dispholidus typus

3.1 Introduction

Predator-prey relationships, which are two-way interactions, drive population dynamics of organisms thereby shaping ecological communities (Jaksic & Simonetti 1987, Codron et al. 2017). Organisms aim to maximize energy intake and minimize energy expenditure, which is related to foraging and predation risk (Pettorelli et al. 2011). The effect of predators on prey populations involves direct and indirect costs. Predators directly affect prey populations by consuming individuals and thus increasing prey mortality rates (Pettorelli et al. 2011). In order to decrease predation risk, prey species allocate more energy into predator avoidance, increasing their energy expenditure thereby indirectly effecting prey. At the same time, foraging activity and efficiency and as a result, energy intake decreases. Furthermore, the reproductive output of prey may reduce due to reduced mating opportunity. Thus, predators directly and indirectly affect the fitness of their prey.

Comprehensive knowledge of the predator's diet improves our understanding of predator-prey relationships. However, quantifying the diet of snakes is challenging for several reasons, including infrequent feeding events by snakes and rare observations of such, and limited accessibility of museum specimens (Greene 1997, Branch 1998, Durso et al. 2011, Glaudas et al. 2017, Layloo et al. 2017). These challenges lead to a limited understanding of an organism's natural history, highlighting the importance of dietary studies (Litvaitis 2000). However, knowing which prey species are consumed by a predator is not enough to understand the effects the predator has on its prey (Litvaitis 2000). To describe predator-prey relationships, one has to quantify rates of consumption by the organisms (Litvaitis 2000). Quantifying such interactions is challenging as well for similar reasons (e.g. secretiveness of snakes, few feeding events, low

http://etd.uwc.ac.za/55 detectability). Additionally, prior information on the prey species that are consumed by a particular snake species is required. Studies that have quantified consumption rates focused on one or two prey species, while snakes often have a wide range of prey.

Few studies have attempted to quantify prey consumption rates by snake species in the wild, despite the fact that snakes are globally distributed predators. Imler (1945) found that bull snakes (Pituophis catenifer sayi) raid 42% of duck nests and Fitch (1948) estimated that Western rattlesnake (Crotalus oreganus) predate 34% of the offspring of California ground squirrels (Otospermophilus beecheyi). Diller and Johnson (1988) estimated that Prairie rattlesnakes (Crotalus viridis) remove approximately 14% of juveniles of Townsend's ground squirrel (Urocitellus townsendii) and up to 11% of juvenile mountain cottontail rabbits (Sylvilagus nuttallii). They also estimated predation rates of the same prey species by gopher snakes (Pituophis catenifer) to be 4% and 22-43%, respectively. Moreover, 81% of the biomass consumed by rattlesnakes comprised ground squirrels, and 60% of the ingested biomass of gopher snakes consisted of similar amounts of ground squirrels and cottontail rabbits.

Several methods are used to quantify the diet of snakes, including dissection of gut contents, palpation of life individuals, direct observations, videography, examination of faecal samples using undigested fragments or DNA analysis. A more recent approach used by scientists is citizen science, which has been valuable not only adding new records of prey species, but also displaying a fairly wide range of the diet of snake species (Layloo et al. 2017, Smith et al. 2019). While these different methods provide valuable data on the prey consumed by snakes, they all have limitations and are often biased. For example, Glaudas et al. (2017) compared data received from the dissection of gut contents of museum specimens with that of videography of wild puff adders (Bitis arietans). They found that diet breadth was narrower when measured using museum specimens compared to when using fixed videography. Additionally, on average, the prey items found in museum specimens were more than three times larger than those observed in video footage. Therefore, smaller prey species are often undetected in studies using museum specimens only and do not provide reliable information on predation and consumption rates by snakes.

Studies on Naja nivea and Dispholidus typus have shown that their diets are diverse, including prey from all tetrapod classes (Layloo et al. 2017, Smith et al. 2019). However, the main consumed prey classes differ for the two species. Both snake species have a relatively low

http://etd.uwc.ac.za/56 standardized Levin’s measure of niche breadth using family-level prey frequencies (cape cobra–

BA = 0.29; boomslang–BA = 0.13; Layloo et al. 2017, Smith et al. 2019). However, Layloo et al. (2017) suggested that cape cobra are generalist predators because of the diversity of prey consumed and the absence of patterns of broad-scale dietary variation. Their data indicated that they consume any adequately sized prey of tetrapod classes when encountered. Interestingly, more than a third of the cape cobra diet consists of other snakes (35%; Layloo et al. 2017). Boomslang have a narrower dietary niche breadth, mostly predating on birds and reptiles (Smith et al. 2019). Close to 90% of reptiles were members of the family Chamaeleonidae (Smith et al. 2019). Amphibians and mammals only make up 8% of their diet (Smith et al. 2019). Smith et al. (2019) found that 36% and 54% of the prey items were chameleons and nestling birds, respectively, and suggest that boomslang diet is restricted to prey with limited mobility as a result of habitat specialization. Both species show broad-scale dietary similarities and some species specific overlaps, such as the sociable weaver (Philetairus socius).

In systems where the distribution of those snakes overlap with that of sociable weavers, cape cobra and boomslang are known to predate on the passerine’s brood. Research on sociable weavers has also identified honey badger (Mellivora capensis) and pygmy falcon (Polihierax semitorquatus) as nest predators of sociable weavers (MacLean 1973c). However, these two predators have a relatively low impact on sociable weaver reproductive success compared to that of the two snake species (MacLean 1973c, Covas et al. 2004a). Snakes typically raid whole colonies during a single foraging bout. In areas of syntopy of predator and prey, birds (particularly nestlings and eggs) likely make up a substantial proportion of the diet of these snakes, specifically during the breeding season of the birds (Covas 2002). Usually, snakes are found in the colonial nest structures shortly after the first summer rain when sociable weavers start breeding. Breeding success of sociable weavers is variable as it depends on rainfall and predation (Covas 2002, Lloyd 2016). Covas (2002) quantified that predation was the fate for over 70% of the offspring. Cape cobra and boomslang have been identified as the primary nest predators (MacLean 1973c, Covas 2002, Spottiswoode 2007, Covas et al. 2008). However, Lloyd (2016) identified that there is a gap in our understanding of how and when these snakes forage in sociable weaver colonies. Specifically, how snake movement is influenced by environmental variables (e.g. weather or the physical nest structure of sociable weavers) and whether there are any differences in predation by different snake species.

http://etd.uwc.ac.za/57

Interspecific competition occurs when two species use the same resource in an ecosystem, for example a food or space. Competition can lead to the exclusion of one species (“Competitive exclusion principle”)—when two sympatric species occupy the same ecological niche, one species exhibiting an advantage in resource use, will eventually exclude the other (Hardin 1960). Species competing for the same resource can only coexist if they use the shared resource differently, also referred to as “niche differentiation” (Kraft et al. 2008). There is a range of ways in which a resource can be utilized differently. For example, a shared resource can be used at different times, at different life stages of prey and predator or at different spatial levels (Kronfeld-Schor & Dayan 2003, Tran et al. 2014). Broad dietary differences exist between cape cobra and boomslang, but both are known to predate on the brood of sociable weavers. Boomslang are likely more reliant on the breeding of these birds than cape cobra, however, because cape cobra have a wide variety of prey especially non-avian prey species (Layloo et al. 2017). In addition to competing for a shared food resource, there is an inherent risk for boomslang to forage in sociable weaver colonies because cape cobra prey on other snakes, including boomslang (Layloo et al. 2017, Maritz et al. 2019). This raises the question of whether there are differences in how these snakes use the shared resource in order to co-exist.

This chapter aims to understand differences in the use of a singular food resource by potentially competing snake species and the impact snake predation has on sociable weaver reproduction. The concurrent research on the sociable weavers provided an opportunity to investigate nest predation by snakes and the potential competition between snakes within this system. Therefore, I quantified nest predation of the sociable weavers and the time of arrival at colonies, after a new communal breeding attempt, for both cape cobra (Naja nivea) and boomslang (Dispholidus typus) over an entire breeding season. I tested the null hypotheses that (i) nest predation by cape cobra and boomslang would be similar in terms of depredated chambers and prey items, (ii) the frequency with which chicks and eggs were predated by each snake species would be similar, (iii) that the mean age of prey consumed by each species of snake would be similar, (iv) that the two species of snakes arrive at colonies at similar times following the onset of egg laying, and (v) that the log10 transformation of total mass intake would be similar for both species. In addition, I compared the total loss of offspring attributable to snake predation for this season to previous findings on snake predation in this system.

http://etd.uwc.ac.za/58

3.2 Methods

3.2.1 Field methods

Throughout the breeding season of sociable weavers, daily colony visitations for full colony checks and control nest checks were conducted. The aim of these checks was to collect data on the presence and absence of snakes and several bird breeding related variables. The breeding data included the number of eggs and chicks present in each chamber of a colony, and the masses of eggs, and chicks at different ages (9 and 17 days old).

Snake presence/absence

For each colony visitation the tree canopy, the top of the nest structure and all chambers were inspected for snakes. A torch illuminated the inside of a chamber to increase visibility and detection of snakes, especially those positioned deep within a chamber. In the presence of a snake, the colony was not further inspected for eggs and chicks, but visited daily until the snake had left. Any eggs or chicks missing from the colony following the observation of a snake were inferred to have been consumed by that individual snake. In the case of two snakes present in a colony at the same time, empty chambers near the chamber the snake was observed in were inferred to have been predated by that individual snake.

Breeding variables

Sociable weavers usually lay eggs at 1-day intervals (MacLean 1973e). After detection of a new clutch, the laying date of eggs and the clutch size were recorded throughout the breeding season. The mass of eggs was taken to the nearest 0.01 gram using a digital scale after the third egg was laid. Eggs laid after that, were weighed directly after detecting them. Therefore, daily visitation of a new clutch stopped when no more eggs were laid.

Incubation period for sociable weaver eggs is between 13 to 14 days (MacLean 1973e). Thus, chambers with eggs were visited daily near the estimated hatching date, and visits continued until all eggs hatched and actual hatching dates were recorded. In order to estimate growth rate of chicks, the mass of chicks was recorded nine and seventeen days (D9 and D17) after hatching. After removing chicks from a chamber, they were kept in cotton bags for the duration of the measuring process. Individuals were placed upside down into a 35 mm film canister to prevent the birds from escaping and weighed to the nearest 0.01 g. Afterwards the chicks were placed

http://etd.uwc.ac.za/59 back into their chamber. Chicks fledge between 21–24 days after hatching (MacLean 1973e). However, previous observations showed that nestlings may fledge prematurely when disturbed after 17 days from hatching. Therefore, after D17 the chambers were not checked again to avoid premature fledging.

On site arrival, the ground beneath the colony was checked for any broken eggs or dead chicks. Eggs on the ground were usually considered as ejected by either the parents or other individuals of a colony. Parent birds eject unfertilized or dead eggs, and chicks that are sick or too weak, or when they cannot provide enough food for all the chicks. In this case, the smallest chick is typically ejected. It may also happen that siblings or other individuals of a colony ejected chicks. However, determination of the exact cause of death was often not possible. Snakes typically consume the whole content of a chamber (MacLean 1973c, Spottiswoode 2007, Lloyd 2016, Greuel pers. observation). Therefore, chambers with only partially missing clutches or broods, were not considered as predated by snakes but rather as ejections (when eggs or chicks are found on the ground) or as death unknown. On occasion however, snakes dropped chicks while trying to consume them. The confined space within chambers made it sometimes difficult for snakes to consume chicks close to fledging age (pers. observation).

For the quantification of consumption rates, I only used events in which snakes were present and consumed the content of at least one chamber. Therefore, the presence of snake protection (cling wrap around stem) is irrelevant for my statistical analyses.

3.2.2 Statistical analysis

I include observations from protected colonies into all analyses (except one), because observations of feeding events are generally low for snakes (Greene 1997). I do not expect any biases by including the data of snake consumption records in protected colonies in my analyses, because once in a colony snakes likely eat as much as possible or available. The potential impact of the clingwrap hindering them from descending the tree with full stomachs is probably insignificant, as they can digest the consumed prey in the tree or colony (and leave afterwards) or they potentially fall to the ground on purpose when descending the trunk (Lillywhite 2014). However, including the data from protected colonies (especially colonies that were never predated) potentially results in a bias of nest predation across the landscape. Therefore, I perform analyses using two different data sets to show how strong the results are biased.

http://etd.uwc.ac.za/60

I used the nest check data, which included the number as well as the age of eggs and chicks present in any chamber in a given colony at any point in time, and the mass of eggs and chicks at D9 and D17, to test the various consumption hypotheses. Comparing statistics of predation between cape cobra and boomslang were limited to true predation events (when more than one chamber or prey item was predated by a snake).

In order to estimate predation rates at my study site, I first had to infer how many chambers were raided by snakes in a single visit on average. Cape cobra and boomslang were the only snake species found within colonies during this study and are known as the main predators of sociable weaver offspring (MacLean 1973c, Covas 2002). Even though other snakes might go into the colonies, observations are very seldom (Covas, pers. comm); therefore other snake species are likely to have only a marginal effect on the predation rate. Hence I only used the data for those two species to make inferences on general snake predation. I compared the mean number of chambers predated by cape cobra and boomslang. In order to compare the results from this study to those of previous studies, I calculated predation rates under the same assumption as Covas (2002). Under further assumptions I assessed snake predation rates based on: 1) using only predation events when a snake was encountered at a colony (true predation events), 2) using true predation events plus all observations in which four or more breeding chambers were empty without encountering a snake, 3) using true predation events plus all encounters when two or more adjacent breeding chambers were empty without encountering a snake (Covas 2002) and 4) using observed predation events plus all encounters when one or more breeding chambers were empty without encountering a snake. The first assumption displays the minimum predation sociable weaver experience by snakes, while the last assumption displays the maximum snake predation rate (likely a slight overestimate due to different fates such as ejections or starvation). The second assumption is based on previous predation estimates (Covas 2002, Covas et al. 2008) and the third assumption is based on the calculated average number of breeding chambers predated by snakes in this study. In order to calculate these predation rates I first had to quantify the total number of reproductive attempts (defined as the total number of eggs laid) and the total reproductive loss (defined as all eggs and nestlings lost despite their fate) for the entire study period. Using the total reproductive attempts, I described the predation rates as the percentage of offspring that is lost to snakes, followed by the proportion of the total reproductive loss attributed to snake predation.

http://etd.uwc.ac.za/61

Next, I compared the average number of prey items (eggs and chicks combined) consumed by each snake species to test whether one of the two species has a greater impact on the sociable weavers numerically than the other. I tested the data on the number of prey items predated per event for normality using a histogram and tested the hypothesis that the consumption of prey items by cape cobra and boomslang do not differ by using a non-parametric test (Wilcoxon test). Subsequently I quantified the number of eggs and chicks consumed per predation event and grouped them by species. I used a Shapiro-Wilk test, accounting for small sample size, to check the data for normality. Additionally, I used histograms to compare the results from the normality test, and found that some of the data identified as normal by the Shapiro-Wilk test were not normally distributed. Further on, the data on the number of eggs and chicks consumed by each snake species were treated as non-parametric. For each study species the average number of eggs and chicks consumed was calculated and compared between species. Wilcoxon rank sum tests tested for any significant differences in the consumption of eggs and chicks between and within groups (i.e. whether species A consumes a food item more than species B versus a species consuming food item 1 more than food item 2).

Additionally, I compared the mean age of prey consumed by the two snake species. The time from egg laying to fledging ranges from 34 to 38 days (incubation period 13 – 14 days and nesting period 21 – 24 days; MacLean 1973e). Using the data from this study, an average time from egg laying to fledging of 35 days was calculated. In order to calculate the average age of the prey consumed by the two study species, I first calculated the mean age of the prey consumed for the individual predation events by identifying the exact age for each consumed prey. The data were tested for normality using Shapiro-Wilk test and for equality of variances using the F- statistic. Using a two-sample t-test, I compared the mean age of consumed prey between cape cobra and boomslang to identify differences in their food choice. In addition, I compared the availability of eggs and chicks at the different consumption events. I quantified the proportions of eggs and chicks present before each snake event, and the proportions of eggs and chicks consumed. The proportions of consumed prey (eggs and chicks) were compared between cape cobra and boomslang with a Fisher’s exact test for count data.

In order to identify whether one snake species arrives earlier at the food resource than the other species, I assessed the number of days a snake was detected after the start of a colonial breeding attempt, which was defined as the date of the first egg laid of such attempt. The birds laid their

http://etd.uwc.ac.za/62 eggs mostly synchronously, but on some instances old breeding chambers (i.e. chambers from a previous attempt that had succeeded) were still active when a new breeding attempt started. In such cases the date of the first egg laid from the new breeding attempt was used as those made up the bulk of all active breeding chambers. The mean number of days since breeding commenced was compared between the two species using a two-sample t-test for unequal variances.

40 y = 1.7728x + 0.9376 R² = 0.9661 35

20 Mass (g) Mass

0 0 2 4 6 8 10 12 14 16 18 Age of offspring (days)

Figure 3.1. Mass of sociable weaver offspring in relation to their age (in days). Day 0 refers to eggs and anything after day 1 refers to the age of chicks. A polynomial trendline shows an almost linear growth curve for nestlings.

Lastly, I assessed the total mass consumed by snakes in the wild by using estimates of the mass of eggs and chicks at the time of consumption. The data on mass of eggs and chicks (D9 and D17) were graphed using a scatter plot and a polynomial trend line was fitted. The mass of eggs stays constant during incubation. However, after hatching the chicks weight increases daily.

http://etd.uwc.ac.za/63

Therefore, I identified the age of each consumed chick. The polynomial function was then used to calculate the weight of chicks for the days when no mass was recorded (Figure 3.1). The averages of mass of eggs, and chicks at D9 and D17 were calculated using the collected data on egg and chick weight. For each feeding event, the frequencies of prey items consumed for each age group (i.e. 1-35 days old) were quantified and multiplied by the corresponding mass. The mass of all eggs and chicks consumed per event was summed and then average mass for all cape cobra and boomslang was calculated. The data were tested for normality and log-transformed because they were not normally distributed. I tested whether cape cobra and boomslang consume significantly different masses of food using an independent two-sample t-test.

3.3 Results

I gathered 57 presence records for snakes in sociable weaver colonies. However, three observed cape cobra did not consume any available offspring. Therefore, the number of predation events was 54, including 35 records for cape cobra and 19 for boomslang. These snakes were the only predators observed in sociable weaver colonies during the study period. In addition, all individuals were classified as adults.

17% 14%

10%

59%

Naja nivea Dispholidus typus unknown fate fledged

http://etd.uwc.ac.za/64

Regardless of the cause of death, the overall reproductive failure among 3129 eggs laid across all 13 colonies amounted to 83.3%, with 16.7% of eggs eventually producing successfully fledged chicks. For 59% of the failed offspring the cause of failure could not be verified, while at least 14% were lost to cape cobra and at least 10% were lost to boomslang (Figure 3.2). This means that snakes were conservatively responsible for at least 24% of all failures, but probably more.

The individual failure rates for each colony varied from 47.6% to 96.4% (Figure 3.3). Two of the three lowest failure rates were colonies with snake protection (Figure 3.3). The highest snake- related loss in relation to individual failure rates was in colony 8 (Figure 3.3), which was visited by mostly boomslang. This colony was the largest in terms of number of adults and colony volume (Table 3.2).

83 100.0 416 404 437 221 375 267 41

80.0 296 184 261 61 60.0 82

40.0 Lost offspring offspring Lost (%) 20.0

0.0 2 7 8 11 20* 21 27* 31 32 37 38 43* 71* Colony Number Total lost offspring Proportion of lost offspring predated by snakes

http://etd.uwc.ac.za/65

The snake protected colonies number 43 and 71 didn’t experience any known predation from snakes and produced 26.2% of all successful offspring (n = 522). Following Covas (2002), snakes consumed 60.8% of the total produced offspring (Figure 3.4A). After excluding the non- predated colonies, the percentage of snake predated offspring increased to 70,6% (Figure 3.4B).

Table 3.1. General information on the individual colonies. Protection is given in binary code with 1 equal to protected and 0 being non-protected.

Colony Number of Colony Number of Protection Snake adults volume (m3) chambers occurrences

2 16 5.27 12 0 4

7 66 5.19 30 0 6

8 134 13.56 79 0 8

11 65 5.75 40 0 10

20 48 6.60 22 0 7

21 13 2.93 11 0 1

27 32 4.59 22 1 1

31 63 5.44 43 0 5

32 68 8.85 48 0 5

37 83 10.49 73 0 9

38 23 3.36 10 0 1

43 26 3.38 14 1 0

71 76 5.55 35 1 0

Figure 3.4 shows that the exclusion of non-predated colonies resulted in an increase (7 – 21%) of the proportion of lost offspring. Assuming that one or more missing chambers indicates a snake predation event, snake inferred loss of offspring across the landscape reached a maximum of 89.1% (Figure 3.4B).

http://etd.uwc.ac.za/66

100 A 100 B

80 80

60 60

40 40

20 20 Proportion of lost offspring (%) lost of Proportion offspring 0 (%) lost of Proportion offspring 0 snakes ≥4 ≥2 ≥1 snakes ≥4 ≥2 ≥1 Assumptions for snake predation Assumptions for snake predation Figure 3.4. Comparison of lost sociable weaver (Philetairus socius) offspring for A) all study colonies (n = 13) and B) all colonies predated by snakes at least once (n = 11). Failed sociable weaver breeding attempts (defined as number of eggs lost during the study period)

were nA = 2607 and nB = 2401, respectively. The proportion of the total lost sociable weaver offspring are based on the following criteria: i) lost to snakes observed at a colony, ii) predated by observed snakes and any events when four or more ( 4) breeding chambers were empty, iii) predated by observed snakes and any events when two or more ( 2) breeding chambers were empty, iv) predated by observed snakes and any events when one or more ( 1) breeding chambers were empty.

Across the entire study period, between one and 10 snakes were sighted at each of the 11 predated colonies (Table 3.1). Across these colonies, there was a total reproductive effort of 2786 eggs. Of these eggs, approximately 26.8% were preyed upon by observed snakes (Figure 3.5). However, the overall offspring loss was 86.2%, of which 31.2% was attributed to snake predation (Figure 3.5). The nest success decreased to 13.8%.

http://etd.uwc.ac.za/67

n = 2786 n = 2401 n = 748 100

20 Proportion of offspring (%) offspring of Proportion 0 Total offspring Total failures Total snake predation

Figure 3.5. The fate of all eggs laid across sociable weaver colonies (n = 11), predated at least once, during the study period. The stacked bars show a) the proportion of total offspring successfully fledged versus failed, b) the proportions of the total offspring failures being predated by snakes or not, and c) the proportions of consumed offspring by Naja nivea (dark grey; n = 35) and Dispholidus typus (light grey; n = 19) of the total snake predation.

3.3.1 Chambers and prey items

Across the landscape, the absolute consumption of sociable weaver offspring is higher for cape cobra (n = 423) than boomslang (n = 325). Histograms showed that the data for the number of chambers and prey items predated by cape cobra were not normally distributed (Figure 3.6).

Wilcoxon sum rank tests showed that there was a significant difference between ranks given for breeding chambers predated (W = 215.5, P = 0.03). The mean (± SD) number of chambers predated by cape cobra was 4.03 ± 2.76 chambers compared to 5.79 ± 2.88 chambers predated by boomslang (Figure 3.7). Ranks given for the number of prey items predated by each snake species differed significantly (W = 223 , P < 0.05; Figure 3.7). The mean number of prey items predated by cape cobra was 12.09 ± 7.48 prey items compared to 17.11 ± 9.33 prey items predated by boomslang.

http://etd.uwc.ac.za/68

6 10 Naja nivea Naja nivea Dispholidus typus Dispholidus typus 8 4 y c y n 6 c e n u e q u e q r e F r 4 F 2

0 0 0 1 2 3 4 5 6 7 8 9 10 11 5 10 15 20 25 30 35 Number of chambers depredated Number of prey items depredated

Figure 3.6. Histograms showing the distribution of the data for the number of chambers and prey items depredated by Dispholidus typus and Naja nivea. The data for depredated prey items were binned.

20 Naja nivea ＊ Dispholidus typus

＊

5 Number of items predated of Number

0 Chambers Prey items

http://etd.uwc.ac.za/69

The asterisk (*) indicates a significant statistical difference in predated chambers/prey items compared between the two snake species.

3.3.2 Eggs and chicks

Shapiro-Wilk normality tests on the number of eggs and chicks consumed by each species showed that all data subsets were not normally distributed, except for the data for the number of eggs consumed by boomslang. However, the histograms showed that all data, including those describing the number of chicks consumed by boomslang, do not display normal distributions. Thus the Wilcoxon rank sum test for non-parametric data was used to compare the two study groups. The results from the Wilcoxon rank sum test showed a significant difference (two-sided, W = 193.5, P = 0.01) between ranks given for the number of eggs consumed by cape cobra and boomslang (Figure 3.8). A one-sided Wilcoxon rank test showed that cape cobra consume significantly fewer eggs than do boomslang (W = 193.5, P < 0.01). On average (± SD), boomslang consumed 12.95 ± 9.65 eggs whereas cape cobra only consumed 6.46 ± 6.61 eggs (Figure 3.8). Boomslang fed significantly more frequently on eggs than on chicks (Wilcoxon sum rank test: W = 284.5, P = 0.0012).

16 Naja nivea

* 12

4 Number of predated items items of predated Number 0 Eggs Chicks Figure 3.8. Consumption of sociable weaver (Philetairus socius) offspring for 35 predation events by cape cobra (Naja nivea) and 19 events for boomslang (Dispholidus typus) showing the mean (± SE) consumption of eggs and chicks. The asterisk (*) indicates a significant statistical difference in predated prey items compared between the two snake species.

http://etd.uwc.ac.za/70

Boomslang ate five times as many eggs than chicks, with a recorded maximum of 29 eggs eaten during a single predation event. Wilcoxon rank sum test showed that there was no significant difference between ranks given for the number of chicks consumed by cape cobra and boomslang (W = 384.5, P = 0.35). The mean (± SD) number of chicks consumed by cape cobra was 5.63 ± 5.53 compared to 3.95 ± 3.84 chicks consumed by boomslang (Figure 3.8). There was no significant difference in ranks given for eggs and chicks consumed by cape cobra (W = 612.5, P = 0.99; Figure 3.8).

3.3.3 Mean age of prey

The data on mean age of consumed prey were normally distributed for both study species and have equal variances (F34,18 = 1.50, P = 0.37). There was an overlap of the mean age of the prey that is consumed by the two snakes (Figure 3.9). However, the mean age of consumed prey differs by more than 10 days. The mean (± SD) age of prey consumed by cape cobra (18.17 ±

7.82 days) was significantly greater than that for boomslang (12.97 ± 6.39 days; one-tailed, t52 = 2.48, P = 0.01). The most frequently consumed chicks by cape cobra and boomslang were 14 and 2 days old, respectively (Figure 3.10).

n = 35 *

n = 19

http://etd.uwc.ac.za/71

30 30 A B 25 25 s s k k

c 20 20 c i i h h c c

f 15 f 15 o o

r r e e b 10 b 10 m m u u N 5 N 5

0 0 1 2 3 4 5 6 7 8 9 10 1112 1314151617 1 2 3 4 5 6 7 8 9 10111213 1415 1617 Age of chicks (days) Age of chicks (days)

3.3.4 Prey availability

At snake occurrences in colonies, the availability of eggs and chicks differed for the two study species. For cape cobra presences, the availability of eggs and chicks was similar (53.47% and 46.53% respectively). For boomslang presences, the availability of eggs and chicks differed substantially (68.17% and 31.83% respectively). There was no significant difference in the proportions of available prey (eggs and chicks combined) consumed by the two snake species (Fisher’s exact score = 0.16; P > 0.05). The proportions of available eggs consumed by cape cobra and boomslang differed (Fisher’s exact score = 0.03, P < 0.05), while there was no significant difference in the proportions of available chicks consumed by the two snake species (Fisher’s exact score = 0.27; P > 0.05).

3.3.5 Time of arrival

The sample variances, for the time (in days) cape cobra and boomslang arrive at a colony after a new communal breeding attempt has started, do not differ significantly (F34,18 = 1.76, P = 0.21). Quantifying the time each study species requires to find sociable weaver nests after breeding attempts during the entire season revealed a statistically significant difference (two-tailed, t52 = 3.06, P < 0.01). Among all snake occurrences for which I documented actual predation events (n

http://etd.uwc.ac.za/72

= 54), the mean (± SD) number of days a snake occurred in a colony after attempting a new breeding cycle, was significantly greater for cape cobra (n = 35, 21.97 ± 10.10 days) than for boomslang (n = 19, 13.84 ± 7.62 days; one-tailed, t52 = 3.06, P < 0.01; Figure 3.11).

n = 35 * n = 19

3.3.6 Mass consumption

The mass for eggs varies much less than those of chicks at the age of day nine or seventeen (Figure 3.1). Quantifying the mass consumed by cape cobra and boomslang revealed that they consume similar amounts of food (Figure 3.12). The data were not normally distributed, therefore for further statistical analyses the data were standardized using log10 transformation for size data. Comparing the variation of transformed data on the mass consumed by each species showed equal variances (F34, 18 = 1.81, P = 0.18). Independent t-test showed no significant difference in the mean (± SD) mass consumed by cape cobra (117.24 ± 89.51) and boomslang

(90.02 ± 80.30; t52 = 0.70, P = 0.49) across the entire sampling period. However, on an

http://etd.uwc.ac.za/73 individual level, any one boomslang may consume the same mass as any one cape cobra (e.g. cape cobra individual 17 and boomslang individual 14 both consumed 90 grams of mass).

3.4 Discussion

At Benfontein Nature Reserve, predation rates of sociable weaver offspring by snakes (76.8%) were similar to those in previous studies (Covas 2002, Covas et al. 2008). The results indicate differences in food resource utilisation by cape cobra and boomslang, potentially impacting sociable weaver reproduction at different stages. Cape cobra arrive at colonies later after a new breeding attempt by the birds, typically after egg hatching, than do boomslang, removing fewer but older offspring. Boomslang on the other hand arrive just before egg hatching, primarily preying on eggs. Despite these differences in consumption, the two species of snake consume similar biomass.

http://etd.uwc.ac.za/74

3.4.1 Nest mortality

Both snake species have been known to heavily predate on sociable weaver offspring during breeding seasons (MacLean 1973c, Covas 2002, Spottiswoode 2009). Previous studies estimated predation by snakes to cause over 70% of nest failure in sociable weavers (Covas 2002, Covas et al. 2008). However, no specification was made to the individual snake species. I quantified that snakes remove at least 25% of the biomass of sociable weaver offspring. However, that only included the observed predation events and is thus an underestimate of the actual loss. Covas (2002) and Covas et al. (2008) assumed that when two or more adjacent chambers were empty, they had been predated by snakes. Using the same assumption, predation rates by snakes in my study were similar to those estimated in previous years (Covas 2002, Covas et al. 2008). However, inclusion of the protected, non-predated colonies into predation analyses biased the results towards a lower percentage of snake predation across the landscape. Therefore, I excluded them from further analyses and focused on the colonies where a predation event was witnessed at least once, whether they were protected or not. The sociable weaver metapopulation at BNR does not experience brood predation by honey badgers. Even though pygmy falcons occur at BNR, it is not unusual for these predators to be absent (i.e. not detected) during some breeding seasons (Covas et al. 2004b). Pygmy falcons impact the breeding success of sociable weavers significantly less than snakes do (Covas 2002). During this study, no pygmy falcons were detected in the study area and there was no evidence (e.g. pink feaces around entrance of chambers or suddenly reduced numbers of adult sociable weavers; Covas et al. 2004b) of roosting individuals in any of the study colonies. It is possible that predation events by pygmy falcons were undetected, however pygmy falcons typically predate on chicks from a single chamber at a time (Covas et al. 2004b, Spiby 2014). Therefore, the impact of pygmy falcons on the offspring of sociable weavers can be considered as negligible to this study and that up to 100% of the predated weaver offspring are predated by snakes.

There were cases when snakes did not predate on any offspring or when only one brood was predated, indicating that snakes did not always find broods or may have been disturbed during foraging. Accordingly, the maximum predation rate by snakes was estimated to be over 80%, including all missing chambers (excluding missing offspring when individuals of same brood survived until a later stage), which is higher than previous estimates for sociable weaver nest

http://etd.uwc.ac.za/75 predation by snakes. The estimated predation rate using the mean number of chambers predated by snakes as an assumption (48%), likely provides a reliable minimum reference point.

For this breeding season, the nest success across all colonies was 16.7%. The protected colonies positively affected nest success resulting in approximately 12%, which is lower compared to other recent studies on nest success (Liebezeit & Goerge 2002, Arlt et al. 2013, Kellett & Alisauskas 2019). A study on colonial geese showed that nest survival probability varies annually (Kellett & Alisauskas 2019). Applying techniques from avian ecology studies, Byer et al. (2018) estimated similar nest survival for two species of turtles experiencing high nest predation. Snake predation in this study was similar to previous breeding seasons, it is likely that nest success was similar too. Some annual variation in nest success is however possible. Nest success varied among colonies depending on snake protection status and number of snake visitations. In this study, the larger colonies were visited more frequently and had the lowest nest success (<10%).

3.4.2 Predation by cape cobra and boomslang

My comparison of the consumption rates provided insight into the individual impact of the two snake species. Across the landscape, cape cobra were responsible for more than 56% of snake predated offspring, but the sample size for cape cobra was twice as high as that for boomslang. Mean prey consumption was higher for boomslang than for cape cobra due to the consumption of high quantities of eggs. Therefore, numerically, boomslang have a greater impact on sociable weaver reproduction than do cape cobra. In terms of total number of offspring predated in the ecosystem, the results indicate that cape cobra have a greater impact on sociable weaver brood.

Cape cobra arrive at colonies approximately one week later than boomslang, when a large proportion of the eggs has already hatched. The cape cobra’s later arrival is likely the result of the species’ natural habitats and foraging ecology. Cape cobra have a wide variety of prey from all tetrapod classes, primarily reptiles and mammals (Layloo et al. 2017). The cape cobra’s primary prey types are ground-dwelling organisms and forage primarily above ground and in burrows. However, they are also known to climb trees in search of food, but they spend only a small fraction of their time in trees compared to arboreal snakes (B. Maritz pers. comm.). Therefore, compared to boomslang, cape cobra may arrive at colonies later and feed on older

http://etd.uwc.ac.za/76 prey, because they are less likely to be in the close vicinity of the colonies to detect new breeding attempts by the sociable weavers.

As a result of the cape cobra’s late arrival at colonies, significantly fewer eggs are available, therefore they consume fewer eggs than boomslang, but similar numbers of chicks. The mean age of their prey is 18 days from egg laying, indicating that the chicks they consume are significantly older than those consumed by boomslang. The most frequent age group of chicks consumed by cape cobra was 14 day old chicks, weighing approximately 26 g each. Despite fewer prey eaten, cape cobra consume similar masses per feeding event as boomslang. However, similar numbers of predated eggs and chicks indicate that cape cobra do not have any preferences for either food item. These predators are also known to scavenge on dead prey such as roadkill (Phelps 2006).

Boomslang are responsible for 44% of the snake related nest mortality, which is slightly less than that for cape cobra (probably attributable to lower sampling size for boomslang). After a new breeding attempt by sociable weavers, boomslang usually arrive at colonies before egg hatching, which is significantly earlier than cape cobra arrival. Preying on both eggs and chicks, boomslang consume primarily younger prey compared to cape cobra. Specifically, boomslang consume higher quantities of eggs than chicks, with chicks generally being younger than those depredated by cape cobra. Despite consuming small prey, boomslang consume a large amount in terms of mass.

The early arrival of boomslang in colonies of sociable weavers is likely the result of their general presence in trees. Arboreal snakes spend most of their time in vegetation above the ground and prey mostly on organisms that are related to such environments (Lillywhite & Henderson 1993). Like other arboreal species, boomslang exploit a wide variety of prey including all tetrapod classes, primarily however on reptiles and birds including their eggs (Lillywhite & Henderson 1993, Smith et al. 2019). Arboreal snakes typically feed on small prey (Lillywhite & Henderson 1993), with diurnal active foragers preying primarily on quiescent prey such as bird eggs (Henderson 1982). This study found the same pattern of small inactive prey being favoured by boomslang. This pattern may be explained by the overall morphology of arboreal snakes. They have slenderer bodies and narrower heads compared to terrestrial snakes such as cape cobra (Lillywhite & Henderson 1993). However, it could also be argued that they consume primarily

http://etd.uwc.ac.za/77 eggs as a result of greater egg availability at the early arrival at colonies. This is supported by the fact that boomslang are capable of consuming larger prey than observed in this study (Smith et al. 2019). However, the confined physical environment of the individuals chambers within sociable weaver colonies likely make it more challenging to consume larger prey.

Layloo et al. (2017) hypothesized that in systems where sociable weaver, cape cobra and boomslang occur together, the snakes are competing for the same food resource. The results from this study indicate that in areas of distributional overlap, interspecific competition between cape cobra and boomslang for sociable weaver offspring does exist. Both consume eggs and chicks— the latter at similar frequencies—but cape cobra consume older, ultimately larger prey than do boomslang. For that reason, the biomass consumed by the two snake species does not differ despite boomslang consuming more prey items overall. The consumption of younger prey, primarily eggs, by boomslang is reflected in the earlier arrival of boomslang at colonies after a new breeding attempt by the sociable weavers.

These differences are likely the result of ecological dissimilarities of the two species of snake, such as their habitats (terrestrial versus arboreal). Luiselli et al. (1998) found significant differences between terrestrial and arboreal snake guilds based on their diet compositions. Boomslang are mostly found in trees, explaining why birds (specifically their brood) and chameleons are the most abundant prey types. Cape cobra are mainly terrestrial, spending most of their time in burrows or cruising on the ground, thus foraging mostly on ground-dwelling organisms. Occasionally, cape cobra explore and feed on birds in trees and shrubs. Savidge’s (1988) suggested that within the consumption of avian prey Boiga irregularis displays a food preference with 65% of the total avian prey items being eggs. A possible reason for choosing eggs over chicks might be that eggs are easier to consume than mobile chicks, especially older chicks. Observations of snakes failing to feed on 17 day old chicks showed that consumption of larger chicks is more challenging likely due to increased mobility and size of chicks, constraints due to prey shape (e.g. chicks expand their wings, thus being more challenging to consume in a small space such as a chamber), and confined space within chambers.

3.4.3 Hypotheses for differential resource utilization

The results indicate that differential food resource utilization by the two nest predators is taking place. As discussed above, their different habitats might be the reason for the difference in

http://etd.uwc.ac.za/78 arrival time at colonies and forms the first hypothesis, possibly explaining the difference in resource use. Besides this, there are two other hypotheses that might explain this finding. In light of the ecology of the two species, I will discuss the following three hypotheses explaining the observed differential resource utilization: i) boomslang detect resource earlier due to their arboreal habits, ii) boomslang consume smaller prey for greater locomotion, and iii) they avoid cape cobra to reduce risk of predation.

Firstly, previous studies have shown that birds make up a large proportion of a variety of arboreal snakes (Savidge 1988, Luiselli et al. 1998, Hockey et al. 2005, Smith 2016). Snakes are more likely to feed on bird eggs if they feed on the corresponding adult animals, which may be explained by the habitat of the predator (de Queiroz & Rodríguez-Robles 2006). Arboreal snakes are more likely to encounter and feed on birds than terrestrial snakes, ultimately consuming more eggs than terrestrial snakes (de Queiroz & Rodríguez-Robles 2006). Previously, Shine (1983) reviewed the diets of snakes revealing that birds including their eggs are more important in the diet of arboreal snakes than of terrestrial snakes. The same pattern is evident in this study—the arboreal boomslang consumed more eggs than the terrestrial cape cobra. The arboreal habits of boomslang may allow them to detect breeding colonies before cape cobra, simply because of sharing the same habitat as their prey. Previous studies suggest that predation rates are higher for prey that shares the same space as the predator (Sperry et al. 2008). Skutch (1949) hypothesized that nest predation is higher at nests with greater parental activity. More recent findings confirm that nest predation increases with parental activity (Martin et al. 2000). Gray rat snakes (Pantherophis spiloides) use visual cues (detection of adult birds) to locate trees with active nests (Mullin & Cooper 1998). The initial increase in parental activity after the first breeding pulse of the season may be the key trigger for cape cobra to actively start foraging for sociable weaver colonies.

Previous studies found that locomotion of terrestrial and arboreal snakes is negatively affected by increased meal size (Garland & Arnold 1983, Ford & Shuttlesworth 1986, Crotty & Jayne 2015). This effect appears to be greater for arboreal snakes (Crotty & Jayne 2015). Based on a review of movement of arboreal animals, Crotty & Jayne (2015) suggest that the likelihood of slipping and falling increases with an increase in speed. Their research found that consumption of larger meals resulted in decreased speed, decreased balance and thus increased chance of slipping and falling. Therefore, arboreal snakes are likely to decrease their speed after feeding on

http://etd.uwc.ac.za/79 large prey to decrease the chances of falling when cruising through tree canopies. Despite a decrease in speed, the change in body shape and increased mass, localized at a specific body part, may still cause increased falling due to loss of balance (Crotty & Jayne 2015). Thus, a preference for consuming smaller prey such as eggs may be beneficial for arboreal snakes to maintain greatest possible locomotion and ultimately reducing predation risk.

The risk of being eaten increases with the increased exposure while foraging for food (Sih 1980). Lastly, predation risk and foraging behaviour is often described as a trade-off (Sih 1980, McNamara & Houston 1987, Abrams 1992). This study dealt with two snake predators utilising the same food resource. However, Cape cobra are known to frequently feed on other snakes including boomslang (Layloo et al. 2017). Most snake species consumed by cape cobra are terrestrial, however recent records from citizen science show that boomslang are predated as well (Layloo et al. 2017, Maritz et al. 2017). The lack of records of cape cobra predating on boomslang within trees may simply be a lack of observation or it might mean that boomslang evolved a way that allows them to efficiently avoid cape cobra when they are found in arboreal habitats. Where their distributions overlap with that of sociable weavers their pathways are known to at least occasionally cross when foraging for the offspring of the weavers. Boomslang arrive at colonies when most of the prey are still eggs. In terms of biomass, older prey such as chicks should be a more nutritiously valued prey item, but boomslang feed predominantly on eggs rather than chicks. Therefore, the choice to forage earlier and feed on less nutritious food suggests that they might avoid cape cobra and thus lower the risk of being predated. During the study, a cape cobra and a boomslang were in the same colony at the same time. However, no ophiophagic event occurred, possibly because the cape cobra was curled up in a chamber, and either didn’t notice the boomslang or wasn’t interested as it had already fed. A combination of the hypotheses may likely be the most plausible explanation as to why boomslang and cape cobra utilise the same food resource differently.

3.4.4 Sampling biases and limitations

The sampling was restricted to a single breeding season and the total number of sighted snakes was low and varied for the two snake species. Information on the abundance of the two snake species in this ecosystem lacks. Thus, further studies are required for broad-scale inferences on

http://etd.uwc.ac.za/80 the exact impact on sociable weaver reproduction by these snakes. Differences in sample size might be biased due to differences in detectability of the two snake species.

The probability of detecting an organism varies extensively—snakes typically have low detectabilities (Kéry 2002, Harvey 2005, Durso et al. 2011, Steen et al. 2012). Snakes inhabit a range of habitats some of which make it more challenging to detect snakes (Fitch 1987, Durso et al. 2011). The use of camouflage allows snakes to blend in with their surroundings to a degree that observations become close to impossible (Penner et al. 2008). Within trees, cape cobra are more conspicuous than boomslang, which blend in with their arboreal habitats magnificently. This may have led to the difference in sample sizes for the two species. It is a challenging task to prevent losing sight of already detected snakes, especially when seeking refuge (Firtch 1987, pers. obs.). Even with radio tracking devices, locating and tracking snakes is highly challenging and sometimes even impossible due to their cryptic and unpredictable movement patterns (Penner et al. 2008, B. Maritz pers. comm.).

Protecting colonies from snakes using cling wrap is an anthropogenic interference causing a bias towards lower predation by snakes, ultimately misrepresenting nest predation by snakes across the landscape. However, by excluding the predation events from protected colonies I would dismiss valuable information on prey consumption by those snakes. Including all protected colonies likely causes a larger bias than by only including protected colonies that were predated at least once.

3.4.5 Impact of snakes on prey

Sociable weavers experience high rates of predation by cape cobra and boomslang. Even though boomslang may be less abundant, the findings indicate that they have a similar impact on the sociable weavers as cape cobra. The one nest predator removes primarily eggs, while the other consumes fewer but often older chicks. The reproductive investment by the birds is lower for eggs than that for older chicks, as less time and energy was spend on parental activities such as foraging. The loss of an egg is therefore less severe than that of a chick. However, the high frequencies at which boomslang remove eggs may balance out this difference, resulting in a similar impact on sociable weavers. Basically, different predators are likely to raid nests at different times due to ecological differences, impacting the success of reproduction in different ways.

http://etd.uwc.ac.za/81

Similar to Lloyd (2016), my results also showed that colonies with snake protection have lower relative reproductive failure than unprotected colonies. Lloyd (2016) found that females from protected colonies are twice as successful in producing fledglings compared to females from unprotected colonies. In the light of climate change, the predation pressure by snakes may be problematic for some populations of sociable weavers. Climate change has resulted in increasing aridity in areas where sociable weavers occur. Because the onset of breeding of sociable weavers depends on rainfall, increasing aridity is likely going to cause declining populations, such as the metapopulation is experiencing at Benfontein Nature Reserve (Altwegg et al. 2014). Lloyd (2016) stated that the predation by snakes increases the pressure on these declining populations.

The findings of this study form the basis for new testable hypotheses, highlighting the importance for ornithologists and herpetologists alike to focus their research on basic ecology and natural history studies. Quantifying the abundance of cape cobra and boomslang in this area would allow us to answer whether cape cobra are more abundant than boomslang and make more detailed inferences on the individual impact these snakes have on sociable weavers. Also, the discussed hypotheses make up for interesting research on the underlying mechanisms that allow these snakes to exploit the same resource. In light of climate change, long-term research of these two nest predators within sociable weaver systems might reveal how significant their impact may be, especially across the landscape.

3.5 Conclusion

In quantifying the consumption of sociable weaver offspring by snakes, I showed that there are subtle differences in the utilization of a singular food resource by cape cobra and boomslang. Previous findings on nest predation of sociable weavers were similar. The analyses that identified the differences in consumption, ultimately differences in resource utilization, were based on measures per snake visit. Also, I’m confident that an appropriate sample size was achieved for answering the research questions. The findings from this study led to new hypotheses potentially explaining the underlying mechanisms that result in this differential resource utilization.

http://etd.uwc.ac.za/82

Chapter 4: Synthesis of the findings

4.1 Competing for a singular resource

The main aim of this thesis was to investigate predation of a singular food resource by the sympatric snakes (Naja nivea and Dispholidus typus). Several research questions were answered to assess nest predation by native snake species. Chapter two focused on defining environmental predictors for the presence of both snake species in sociable weaver colonies and predicting differences in snake activity based on varying environmental factors. Chapter three involved quantifying nest predation, defining differences in prey consumption and use of a singular food resource by the two snake species. The findings in chapter two indicate that activity of cape cobra and boomslang is predicted by different spatially-varying and temporally-varying factors, which means that they are more likely to be observed in colonies at different times, suggesting coexistence through temporal resource partitioning. Cape cobra presence is predicted by weather conditions and colony size—they forage in larger colonies on warm days, with high winds speeds and reduced air pressure—but climatic conditions appear to have stronger predictive power of cobra presence. Boomslang predate more on colonies with intense breeding activity on days with less wind. The findings in chapter three indicate that cape cobra and boomslang are utilizing the same food resource differently by predating at different times and consuming prey at different life stages. Boomslang arrive at colonies when most birds are incubating their eggs, primarily consuming eggs and young chicks. Cape cobra arrive at colonies after hatching, when mostly chicks are present, resulting in the consumption of older chicks. Overall, the snakes consume similar amounts of biomass. Predation rates of sociable weaver offspring by snakes in this study were similar to previous findings by Covas et al. (2008).

Despite having more than minor impacts in their respective ecosystems, it was difficult to quantify their impact in the past. Particularly resulting from infrequent detections ascribed to their elusiveness and low consumption rates as feeding events are rarely observed. During recent years however, this general consensus is being overruled by the herpetological community focusing more on the role of snakes as important predatory species in ecosystems (Nowak et al.

http://etd.uwc.ac.za/83

2008, Campbell III et al. 2012, Steen et al. 2014, Willson & Winne 2015, Willson 2017). The findings from this thesis support the idea that snakes have the potential to regulate prey populations. My results indicate that native snakes are capable of regulating prey populations with high densities. Nowak et al. (2008) focused on viperids, which are typical ambush-foragers with low energy requirements, and suggested that vipers are more likely to regulate low prey densities. This thesis however, paid attention to snakes exhibiting a different foraging technique (actively searching for prey) characterized by higher energy requirements (Secor & Nagy 1994, Secor 1995). Cape cobra and boomslang often raid whole sociable weaver colonies and are responsible for the majority of nest mortality. Without the regulation by snakes, sociable weavers would likely reproduce more fledglings as seen from higher relative reproductive success in protected colonies. Thus, snakes with either foraging strategy have the potential to control population dynamics of ecosystems. However, the severity of impact is expected to be higher for active foraging snakes than for ambush predators with low energy requirements (such as vipers; Nowak et al. 2008).

4.2 Recommendations for data collection

I recommend that an additional sampling season is conducted including data collection during the winter months, allowing inferences on broad-scale temporal variation. For future sampling designs, the protocol for daily data collection should be set out more strictly. This means that instead of joining the SWP research team for their data collection, colonies should be visited independently for snake presence check. Independent sampling would ensure similar sampling times because snake checks only take a few minutes, while bird data sampling can take up to an hour for a single colony. Additionally, I suggest snake checks in the morning, early and late afternoon to increase sample size and make more detailed inferences on time of day when snakes visit colonies. Furthermore, instead of using the climatic data from the nearest weather station, a pocket weather station should be used. Such a device would allow site-specific documentation of climatic data, independent of time differences in sampling. The potential bias in sample size for the two snake species will be resolved only by conducting population studies at BNR, which will also allow for more detailed quantification of the individual species’ impact on the sociable weaver reproduction.

http://etd.uwc.ac.za/84

4.3 Conservation planning

From an ornithological conservation perspective, gaining information about spatial and temporal influences on snake activity is important for the sociable weaver metapopulation at BNR to implement appropriate measures to protect the birds from snakes. Previous studies stated that the population size is declining with increasing temperature and aridity and that nest predation by snakes amplifies the pressure on this population (Altwegg et al. 2014, Lloyd 2016). Snakes cause dramatic declines and even local extinctions of prey species as shown by the brown tree snake of Guam or the Burmese python in Florida (Savidge 1987, Fritts & Rodda 1998, Cruz et al. 2008, Dove et al. 2011, Dorcas et al. 2012). Even though these are cases of exotic invasive snake species, they show that snakes have great potential to cause dramatic damage to the functionality of ecosystems. Secretive and seldom observed, snakes are widespread and often the most abundant and/or top predators in ecosystems (Nowak et al. 2008). Already regulating many prey populations, native snake species may have the potential to cause extirpations of prey species in combination with other pressures induced on their prey (e.g. habitat loss, diseases, climate change).

Snake populations are declining on a global scale (Reading et al. 2010). The causes for such declines are often not known. Despite population studies forming a vital part in conservation and management plans by detecting changes such as population declines (Dorcas & Willson 2009), few studies on African snake population dynamics exist (Luiselli et al. 2000, Luiselli 2006, Maritz & Alexander 2012). The lack of population studies for most snake species is alarming with respect to the diversity present in Africa (Maritz & Alexander 2012). In order to identify areas of conservation need population studies should be prioritized in the herpetological science community. Thus, estimating and monitoring snake population dynamics is the first step towards the implementation of appropriate and efficient conservation management plans for both predator and prey species. By doing so, the primary goal of conservation biology—maintaining biodiversity by protecting species and habitats from extinction caused by human interference or others (Soulé 1985)—can be achieved.

4.4 Future work

On a species level, the impact a predator has on its prey depends on the abundance of a species, which often varies on a spatial scale and has important implications for the ecology and

http://etd.uwc.ac.za/85 conservation of species (Brown et al. 1995). A new approach to study secretive species is occupancy modelling—an approach to estimate abundance (or rather a surrogate of abundance) using presence-absence data and incorporating detectability (Durso et al. 2011). By incorporating the identified environmental influences on snake activity, this study can aid in planning a robust sampling design to quantify occupancy of cape cobra and boomslang (Willson et al. 2011). Such population studies provide the opportunity to monitor snake population dynamics, unfold predator-prey relationships and model the effect of environmental changes on the population size of the two snake species and the corresponding changes in populations size of sociable weaver.

http://etd.uwc.ac.za/86

Literature cited

AKANI, G.C., E.A. ENIANG, I.J. EKPO, F.M. ANGELICI & L. LUISELLI. 2003. Food habits of the snake Psammophis phillipsi from the continuous rain-forest region of southern Nigeria (West Africa). Journal of Herpetology 37(1): 208–211.

ALBRECHT, M. & N.J. GOTELLI. 2001. Spatial and temporal niche partitioning in grassland ants. Oecologia 126(1): 134–141.

ALEXANDER, G.J. & B. MARITZ. 2015. Sampling interval affects the estimation of movement parameters in four species of African snakes. Journal of Zoology 297(4): 309–318.

ALLISON, A., F. BONACCORSO, K. HELGEN, & R. JAMES. 2008. Pteropus mariannus. The IUCN Red List of Threatened Species 2008: e.T18737A8516291. http://dx.doi.org/10.2305/IUCN.UK.2008.RLTS.T18737A8516291.en. Downloaded on 06 November 2018.

ALTWEGG, R. & M.D. ANDERSON. 2009. Rainfall in arid zones: possible effects of climate change on the population ecology of blue cranes. Functional Ecology 23(5): 1014–1021.

ALTWEGG, R., C. DOUTRELANT, M.D. ANDERSON, C.N. SPOTTISWOODE, & R. COVAS. 2014. Climate, social factors and research disturbance influence population dynamics in a declining sociable weaver metapopulation. Oecologia 174(2): 413–425.

ANDERSON, M.D. 2004. Aardwolf adaptations: a review: animals. Transactions of the Royal Society of South Africa 59(2): 99–104.

ANGERT, A.L., T.E. HUXMAN, P. CHESSON & D.L. VENABLE. 2009. Functional tradeoffs determine species coexistence via the storage effect. Proceedings of the National Academy of Sciences 106(28): 11641–11645.

ARLT, D., M. LOW & T. PÄRT. 2013. Effect of geolocators on migration and subsequent breeding performance of a long-distance passerine migrant. PLOS ONE 8(12): 82316.

BAEYENS, D.A. & R.L. ROUNTREE. 1983. A comparative study of evaporative water loss and epidermal permeability in an arboreal snake, Opheodrys aestivus, and a semi-aquatic snake,

http://etd.uwc.ac.za/87

Nerodia rhombifera. Comparative Biochemistry and Physiology Part A: Physiology 76(2): 301– 304.

BARBER‐MEYER, S.M., L.D. MECH & P.J. WHITE. 2008. Elk calf survival and mortality following wolf restoration to Yellowstone National Park. Wildlife Monographs 169(1): 1–30.

BARKER, D.G. & T.M. BARKER. 2008. The distribution of the Burmese python, Python molurus bivittatus. Bulletin of the Chicago Herpetological Society 43(3): 33–38.

BATES, M.F. (1991. New distribution records for amphibians and reptiles from the Cape Province, South Africa. The Journal of the Herpetological Association of Africa 39(1): 14–18.

BESCHTA, R.L. 2005. Reduced cottonwood recruitment following extirpation of wolves in Yellowstone's northern range. Ecology 86(2): 391–403.

BIRDLIFE INTERNATIONAL. 2016. Hypotaenidia owstoni. The IUCN Red List of Threatened Species 2016: e.T22692441A93353974. http://dx.doi.org/10.2305/IUCN.UK.2016- 3.RLTS.T22692441A93353974.en. Downloaded on 06 November 2018.

BOCK, C.E., J.H. BOCK & M.C. GRANT. 1992. Effects of bird predation on grasshopper densities in an Arizona grassland. Ecology 73(5): 1706–1717

BOLGER, D.T., M.A. PATTEN & D.C. BOSTOCK. 2005. Avian reproductive failure in response to an extreme climatic event. Oecologia 142(3): 398–406.

BONNET, X., G. NAULLEAU & R. SHINE. 1999. The dangers of leaving home: dispersal and mortality in snakes. Biological Conservation 89(1): 39–50.

BORER, E.T., B.S. HALPERN & E.W. SEABLOOM. 2006. Asymmetry in community regulation: effects of predators and productivity. Ecology 87(11): 2813–2820.

BRANCH, W.R. 1998. Field guide to snakes and other reptiles of southern Africa. Struik Nature, Century City, South Africa. BRANCH, W.R. & A. JACKSON. 2016. Dispholidus typus (Smith, 1928): Melanistic colouration. African Herp News 63: 9–11.

BROADLEY, D.G. 1968. The venomous snakes of Central and South Africa. In: BÜCHERL, W., E.E. BUCKLEY & V. DEULOFEU. 1968. Venomous animals and their venoms: Venomous vertebrates. Elsevier 425.

http://etd.uwc.ac.za/88

BROADLEY, D.G. & W. WÜSTER. 2004. A review of the southern African ‘non‐spitting’ cobras (Serpentes: Elapidae: Naja). African Journal of Herpetology 53(2): 101–122.

BROWN, J.H., D.W. MEHLMAN & G.C. STEVENS. 1995. Spatial variation in abundance. Ecology 76(7): 2028–2043.

BROWN, G.P., R. SHINE & T. MADSEN. 2002. Responses of three sympatric snake species to tropical seasonality in northern Australia. Journal of Tropical Ecology 18(4): 549–568.

BULLOCK, K.L., G. MALAN & M.D. PRETORIUS. 2011. Mammal and bird road mortalities on the Upington to Twee Rivieren main road in the southern Kalahari, South Africa. African Zoology 46(1): 60–71.

BYER, N.W., B.N. REID, R.A. SEIGEL & M.Z. PEERY. 2018. APPLYING LESSONS FROM AVIAN ECOLOGY TO HERPETOLOGICAL RESEARCH: TECHNIQUES FOR ANALYZING NEST SURVIVAL DUE TO PREDATION. Herpetological Conservation and Biology 13(3): 517–532.

CAMPBELL, III, E.W., A.A.Y. ADAMS, S.J. CONVERSE, T.H. FRITTS & G.H. RODDA. 2012. Do predators control prey species abundance? An experimental test with brown treesnakes on Guam. Ecology 93(5): 1194–1203.

CARPENTER S.R., J.J. COLE, J.F. KITCHELL & M.L. PACE. 2013. Trophic Cascades in Lakes: Lessons and Prospects. In: TERBORGH, J. & J.A. ESTES. 2013. Trophic cascades: predators, prey, and the changing dynamics of nature. Island Press 55–69.

CLARKE, J.A., J.T. CHOPKO & S.P. MACKESSY. 1996. The effect of moonlight on activity patterns of adult and juvenile prairie rattlesnakes (Crotalus viridis viridis). Journal of Herpetology 192–197.

CODRON, J., J. BOTHA-BRINK, D. CODRON, A.K. HUTTENLOCKER & K.D. ANGIELCZYK. 2017. Predator–prey interactions amongst Permo‐Triassic terrestrial vertebrates as a deterministic factor influencing faunal collapse and turnover. Journal of Evolutionary Biology 30(1): 40–54.

COVAS, R. 2002. Life-history evolution and cooperative breeding in the sociable weaver. Unpubl. PhD thesis, University of Cape Town, South Africa.

http://etd.uwc.ac.za/89

COVAS, R., C.R. BROWN, M.D. ANDERSON & M.B. BROWN. 2002. Stabilizing selection on body mass in the sociable weaver Philetairus socius. Proceedings of the Royal Society of London B: Biological Sciences 269(1503): 1905–1909.

COVAS, R., C.R. BROWN, M.D. ANDERSON & M.B. BROWN. 2004a. Juvenile and adult survival in the sociable weaver (Philetairus socius), a southern-temperate colonial cooperative breeder in Africa. The Auk: Ornithological Advances 121(4): 1199–1207.

COVAS, R., O. HUYSER & C. DOUTRELANT. 2004b. Pygmy Falcon predation of nestlings of their obligate host, the Sociable Weaver. Ostrich-Journal of African Ornithology 75(4): 325–326.

COVAS, R. & M.A. DUPLESSIS. 2005. The effect of helpers on artificially increased brood size in sociable weavers (Philetairus socius). Behavioral Ecology and Sociobiology 57(6): 631– 636.

COVAS, R., M.A. DU PLESSIS & C. DOUTRELANT. 2008. Helpers in colonial cooperatively breeding sociable weavers Philetairus socius contribute to buffer the effects of adverse breeding conditions. Behavioral Ecology and Sociobiology 63(1): 103–112.

COVAS, R., O. HUYSER & C. DOUTRELANT. 2004. Pygmy Falcon predation of nestlings of their obligate host, the sociable weaver. Ostrich-Journal of African Ornithology 75(4): 325–326.

CROTTY, T.L. & B.C. JAYNE. 2015. Trade-offs between eating and moving: what happens to the locomotion of slender arboreal snakes when they eat big prey? Biological Journal of the Linnean Society 114(2): 446–458.

CRUZ, J.B., S.R. KREMER, G. MARTIN, L.L. WILLIAMS & V.A. CAMACHO. 2008. Relative Abundance and Distribution of Mariana Swiftlets (Aves: Apodidae) in the Northern Mariana Islands. Pacific Science 62(2): 233–246.

DALTRY, J.C., T. ROSS, R.S. THORPE & W. WÜSTER. 1998. Evidence that humidity influences snake activity patterns: a field study of the Malayan pit viper Calloselasma rhodostoma. Ecography 21(1): 25–34.

DEAN, W.R.J., S.J. MILTON & F. JELTSCH. 1999. Large trees, fertile islands, and birds in arid savanna. Journal of Arid Environments 41(1): 61–78.

http://etd.uwc.ac.za/90

DEAN, W.R.J. & S.J. MILTON. 2001. Responses of birds to rainfall and seed abundance in the southern Karoo, South Africa. Journal of Arid Environments 47(1): 101–121.

DEGREGORIO, B.A., P.J. WEATHERHEAD & J.H. SPERRY. 2016. Ecology and predation behaviour of Corn snakes (Pantherophis guttatus) on avian nests. Herpetological Conservation and Biology 11(1): 150–159.

DE VRIES, J.L., C.W.W. PIRK, P.W. BATEMAN, E.Z. CAMERON & F. DALERUM. 2011. Extension of the diet of an extreme foraging specialist, the aardwolf (Proteles cristata). African Zoology 46(1): 194–196.

DE QUEIROZ, A. & J.A. RODRIGUEZ-ROBLES. 2006. Historical contingency and animal diets: the origins of egg eating in snakes. The American Naturalist 167(5): 684–694.

DILLER, L.V. & D.R. JOHNSON. 1988. Food habits, consumption rates, and predation rates of western rattlesnakes and gopher snakes in southwestern Idaho. Herpetologica 44(2): 228–233.

DORCAS, M.E. & J.D. WILLSON. 2009. Innovative methods for studies of snake ecology and conservation. In: Mullin, S.J. & R.A. Seigel. 2011. Snakes: ecology and conservation. Cornell University Press. 5–37.

DORCAS, M.E., J.D. WILLSON, R.N. REED, R.W. SNOW, M.R. ROCHFORD, M.A. MILLER, W.E. MESHAKA, P.T. ANDREADIS, F.J. MAZZOTTI, C.M. ROMAGOSA & K.M. HART. 2012. Severe mammal declines coincide with proliferation of invasive Burmese pythons in Everglades National Park. Proceedings of the National Academy of Sciences 109(7): 2418– 2422.

DOVE, C.J., R.W. SNOW, M.R. ROCHFORD, F.J. MAZZOTTI. 2011. Birds consumed by the invasive Burmese python (Python molurus bivittatus) in Everglades National Park, Florida, USA. The Wilson Journal of Ornithology 123: 126–131.

DROBENKOV S.M. 1995. Comparative analysis of the feeding of sympatric snakes Vipera berus (L), Natrix natrix (L), and Coronella austriaca (Laur). Russian Journal of Ecology 26(3): 197–201.

DUNN P. 2004 Breeding dates and reproductive performance. Advances in Ecological Research 35: 69–87.

http://etd.uwc.ac.za/91

DOWNES, S.J. 2002. Size-dependent predation by snakes: selective foraging or differential prey vulnerability?. Behavioral Ecology 13(4): 551–560.

DURSO, A.M., J.D WILLSON & C.T. WINNE. 2011. Needles in haystacks: estimating detection probability and occupancy of rare and cryptic snakes. Biological Conservation 144(5): 1508–1515.

EACKER, D.R., M. HEBBLEWHITE, K.M. PROFFITT, B.S. JIMENEZ, M.S. MITCHELL & H.S. ROBINSON. 2016. Annual elk calf survival in a multiple carnivore system. The Journal of Wildlife Management 80(8): 1345–1359.

EBERHARDT, L.L., P.J. WHITE, R.A. GARROTT & D.B. HOUSTON. 2007. A seventy‐year history of trends in Yellowstone's northern elk herd. The Journal of Wildlife Management 71(2): 594–602.

ESKEW, E.A. & B.D. TODD. 2017. Too Cold, Too Wet, Too Bright, or Just Right? Environmental Predictors of Snake Movement and Activity. Copeia 105(3): 584–591.

ESTES, J.A. & J.F. PALMISANO. 1974. Sea otters: their role in structuring nearshore communities. Science 185(4156): 1058–1060.

ESTES, J.A., J. TERBORGH, J.S. BRASHARES, M.E. POWER, J. BERGER, W.J. BOND, S.R. CARPENTER, T.E. ESSINGTON, R.D. HOLT, J.B.C. JACKSON, R.J. MARQUIS, L. OKSANEN, T. OKSANEN, R.T. PAINE, E.K. PIKITCH, W.J. RIPPLE, S.A. SANDIN, M. SCHEFFER, T.W. SCHOENER, J.B. SHURIN, A.R.E. SINCLAIR, M.E.SOULÉ, R. VIRTANEN & D.A. WARDLE. 2011. Trophic downgrading of planet Earth. Science 333(6040): 301–306.

FATH, B.D. 2004. Distributed control in ecological networks. Ecological Modelling 179(2): 235–245.

FILIPPI, E., L. RUGIERO, M. CAPULA, D. CAPIZZI & L. LUISELLI. 2005. Comparative food habits and body size of five populations of Elaphe quatuorlineata: the effects of habitat variation, and the consequences of intersexual body size dimorphism on diet divergence. Copeia 2005(3): 517–525.

FITCH, H.S. 1948. Ecology of the California ground squirrel on grazing lands. American Midland Naturalist 39: 513–596.

http://etd.uwc.ac.za/92

FITCH, H.S., 1963. Natural history of the black rat snake (Elaphe o. obsoleta) in Kansas. Copeia 1963(4): 649–658.

FITCH, H.S. & R.R. FLEET. 1970. Natural history of the milk snake (Lampropeltis triangulum) in northeastern Kansas. Herpetologica 26(4): 387–396.

FITCH, H.S. 1987. Collecting and Life-history techniques. In Snakes: ecology and evolutionary biology. Macmillan, New York, United States of America. 143–164.

FORTIN, D., H.L. BEYER, M.S. BOYCE, D.W. SMITH, T. DUCHESNE & J.S. MAO. 2005. Wolves influence elk movements: behavior shapes a trophic cascade in Yellowstone National Park. Ecology 86(5): 1320–1330.

FREIRIA, F.M., J.C. BRITO & M.L. AVIA. 2006. Ophiophagy and cannibalism in Vipera latastei Boscá, 1878 (Reptilia, Viperidae). Herpetological bulletin 96: 26 pp.

FRITTS, T.H. & G.H. RODDA. 1998. The role of introduced species in the degradation of island ecosystems: a case history of Guam. Annual review of Ecology and Systematics 29(1): 113–140.

FIZZOTTI, B. 2018. The effects of roads on Puff Adder (Britis arietans) movement. Unpubl. MSc thesis. University of the Witwatersrand, South Africa.

GLAUDAS, X. & G.J. ALEXANDER. 2017. Food supplementation affects the foraging ecology of a low-energy, ambush-foraging snake. Behavioral Ecology and Sociobiology 71(1): 5 pp.

GLAUDAS, X., KEARNEY, T.C. & ALEXANDER, G.J. 2017. Museum Specimens Bias Measures of Snake Diet: A Case Study Using the Ambush-Foraging Puff Adder (Bitisarietans). Herpetologica 73(2): 121 ̶128.

GLAUDAS, X., T. JEZKOVA & J.A. RODRÍGUEZ-ROBLES. 2008. Feeding ecology of the Great Basin rattlesnake (Crotalus lutosus, Viperidae). Canadian Journal of Zoology 86(7): 723– 734.

GREENE, H.W. 1997. Snakes: the evolution of mystery in nature. University of California Press. Berkley.

GREENE, H.W. 1997. Snakes: The evolution of mystery in nature. University of California Press, Berkeley.

GREENE, H.W. 1999. Nattern der Neuen Welt. Schlangen. Birkhäuser. Basel. pp. 191–206.

http://etd.uwc.ac.za/93

GREGORY, P.T. 1984. Habitat, diet, and composition of assemblages of garter snakes (Thamnophis) at eight sites on Vancouver Island. Canadian Journal of Zoology 62(10): 2013– 2022.

GRUNER, D.S. 2004. Attenuation of top‐down and bottom‐up forces in a complex terrestrial community. Ecology 85(11): 3010–3022.

GRUNER, D.S., J.E. SMITH, E.W. SEABLOOM, S.A. SANDIN, J.T. NGAI, H. HILLEBRAND, W.S. HARPOLE, J.J. ELSER, E.E. CLELAND, M.E. BRACKEN, & E.T. BORER. 2008. A cross‐system synthesis of consumer and nutrient resource control on producer biomass. Ecology Letters 11(7): 740–755.

GUREVITCH, J. & D.K. PADILLA. 2004. Are invasive species a major cause of extinctions?. Trends in Ecology & Evolution 19(9): 470–474.

HAACKE, W.D. 1984. The herpetology of the southern Kalahari domain. Koedoe 27(2): 171– 186.

HAIRSTON, N.G., F.E. SMITH & L.B. SLOBODKIN. 1960. Community structure, population control, and competition. The American Naturalist 94(879): 421–425.

HARDIN, G. 1960. The competitive exclusion principle. Science 131(3409): 1292–1297.

HART, K.M., M.S. CHERKISS, B.J. SMITH, F.J. MAZZOTTI, I. FUJISAKI, R.W. SNOW & M.E. DORCAS. 2015. Home range, habitat use, and movement patterns of non-native Burmese pythons in Everglades National Park, Florida, USA. Animal Biotelemetry 3(1): 8.

HARVEY, D.S. 2005. Detectability of a large-bodied snake (Sistrurus c. catenatus) by time- constrained searching. Herpetological Review 36(4): 413.

HASSELL, M.P., M.J. CRAWLEY, H.C.J. GODFRAY & J.H. LAWTON. 1998. Top-down versus bottom-up and the Ruritanian bean bug. Proceedings of the National Academy of Sciences 95(18): 10661–10664.

HENDERSON, R.W. 1982. Trophic relationships and foraging strategies of some New World tree snakes (Leptophis, Oxybelis, Uromacer). Amphibia-Reptilia 3(1): 71–80.

HOCKEY, P.A.R., W.R.J. DEAN & P.G. RYAN. 2005. Roberts birds of southern Africa. The Trustees of the John Voelcker Bird Book Fund, Cape Town.

http://etd.uwc.ac.za/94

HUNTER M.D. & P.W. PRICE. 1992. Playing chutes and ladders: heterogeneity and the relative roles of bottom‐up and top‐down forces in natural communities. Ecology 73(3): 724–732.

IMLER, R.H. 1945. Bullsnakes and their control on a Nebraska wildlife refuge. The Journal of Wildlife Management 9(4): 265–273.

JAKSIC, F.M. & J.A. SIMONETTI. 1987. Predator/prey relationships among terrestrial vertebrates: an exhaustive review of studies conducted in southern South America. Revista Chilena de Historia Natural 60: 221–244.

KELLETT, D.K. & R.T. ALISAUSKAS. 2019. Mayfield estimates versus apparent nest success in colonial geese. The Journal of Wildlife Management 83(4): 954–962.

KEOGH, J.S., W.R. BRANCH & R. SHINE. 2000. Feeding ecology, reproduction and sexual dimorphism in the colubrid snake Crotaphopeltis hotamboeia in southern Africa. African Journal of Herpetology 49(2): 129–137.

KÉRY, M. 2002. Inferring the absence of a species: a case study of snakes. Journal of Wildlife Management. 66(2): 330–338.

KOK, O.B. & J.A.J. NEL. 1992. Diet of the bat-eared fox in the Orange Free State and northern Cape Province. South African Journal of Wildlife Research 22(2): 36–39.

KRAFT, N.J., R. VALENCIA & D.D. ACKERLY. 2008. Functional traits and niche-based tree community assembly in an Amazonian forest. Science 322(5901): 580–582.

KRONFELD-SCHOR, N. & T. DAYAN. 2003. Partitioning of time as an ecological resource. Annual Review of Ecology, Evolution, and Systematics 34(1): 153–181.

KRUGER, A.C. & S.S. SEKELE. 2013. Trends in extreme temperature indices in South Africa: 1962–2009. International Journal of Climatology 33(3): 661–676.

LAYLOO, I., C. SMITH & B. MARITZ. 2017. Diet and feeding in the cape cobra, Naja nivea. African Journal of Herpetology 66(2): 147–153.

LEECH, S.M. & M.L. LEONARD. 1997. Begging and the risk of predation in nestling birds. Behavioral Ecology 8(6): 644–646.

LEGAGNEUX, P., G. GAUTHIER, D. BERTEAUX, J. BÊTY, M.C. CADIEUX, F. BILODEAU, L. BOLDUC, A. MCKINNON, J.F. TARROUX, L. THERRIEN, L.

http://etd.uwc.ac.za/95

MORISSETTE & C.J. KREBS. 2012. Disentangling trophic relationships in a High Arctic tundra ecosystem through food web modeling. Ecology 93(7): 1707–1716.

LEIGHTON, G.M. & S. ECHEVERRI. 2014. Non-linear influence of nest size on thermal buffering of sociable weaver nests and the maintenance of cooperative nest construction. Avian Biology Research 7(4): 255–260.

LEIGHTON, G.M. & S. ECHEVERRI. 2015. Stable isotope signatures reveal cryptic differences in diet of sociable weavers. Avian Biology Research 8(2): 104–108.

LEROUX, S.J. & M. LOREAU. 2015. Theoretical perspectives on bottom-up and top-down interactions across ecosystems. Trophic Ecology 3–28.

LIEBEZEIT, J.R. & T.L. GEORGE. 2002. Nest predators, nest-site selection, and nesting success of the Dusky Flycatcher in a managed ponderosa pine forest. The Condor 104(3): 507– 517.

LILLYWHITE, H.B. 2014. Perceiving the snake's world: Structure and function of sense organs. In: How snakes work: structure, function, and behavior of the world's snakes. Oxford University Press 163–179.

LILLYWHITE, H.B. & R.W. HENDERSON. 1993. Behavioral and functional ecology of arboreal snakes. In: SEIGEL, R.A. & J.T. COLLINS. Snakes–Ecology and Behaviour.

LIMA, S.L. & L.M. DILL. 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology 68(4): 619–640.

LITVAITIS, J.A. 2000. Investigating food habits of terrestrial vertebrates. In: Research techniques in animal ecology: controversies and consequences 165–190.

LLOYD, K. 2016. An experimental approach to assess the role of nest predation in the population dynamics of the sociable weaver (Philetairus socius). Unpubl. MSc thesis, FitzPatrick Institute of African Ornithology, University of Cape Town, South Africa.

LUISELLI, L. 2003. Do snakes exhibit shifts in feeding ecology associated with the presence or absence of potential competitors? A case study from tropical Africa. Canadian Journal of Zoology 81(2): 228–236.

http://etd.uwc.ac.za/96

LUISELLI, L. 2006a. Broad geographic, taxonomic and ecological patterns of interpopulation variation in the dietary habits of snakes. Web Ecology 6(1): 2–16.

LUISELLI, L. 2006b. Site occupancy and density of sympatric Gaboon viper (Bitis gabonica) and Nose–horned viper (Bitis nasicornis). Journal of Tropical Ecology 22: 555–564.

LUISELLI, L., G.C. AKANI & D. CAPIZZI. 1998. Food resource partitioning of a community of snakes in a swamp rainforest of south-eastern Nigeria. Journal of Zoology 246(2): 125–133.

LUISELLI, L., F.M. ANGELICI, & G.C. AKANI. 2000. Large elapids and arboreality: The ecology of Jameson’s green mamba (Dendroaspis jamesoni) in an Afro–tropical forested region. Contributions to Zoology 69:1–11.

LUISELLI, L. & ANGELICI, F.M. 1998. Sexual size dimorphism and natural history traits are correlated with intersexual dietary divergence in royal pythons (Python regius) from the rainforests of southeastern Nigeria. Italian Journal of Zoology 65(2): 183–185.

LUNDBERG, J. & F. MOBERG. 2003. Mobile link organisms and ecosystem functioning: implications for ecosystem resilience and management. Ecosystems 6(1): 87–98.

MACLEAN, G.L. 1973a. The sociable weaver, part 5: food, feeding and general behaviour. Ostrich 44(3–4): 254–261.

MACLEAN, G.L. 1973b. The sociable weaver, part 2: nest architecture and social organization. Ostrich 44(3-4): 191–218.

MACLEAN, G.L. 1973c. The sociable weaver, part 4: predators, parasites and symbionts. Ostrich 44(3–4): 241–253.

MACLEAN, G.L. 1973d. The sociable weaver, part 1: description, distribution, dispersion and populations. Ostrich 44(3-4): 176-190.

MACLEAN, G.L. 1973e. The sociable weaver, part 3: breeding biology and moult. Ostrich 44(3- 4): 219–240.

MAEHR, D.S. & J.R. BRADY. 1986. Food habits of bobcats in Florida. Journal of Mammalogy 67(1): 133–138.

MARAIS, J. 2004. A complete guide to the snakes of southern Africa. Struik Nature, Century City, South Africa.

http://etd.uwc.ac.za/97

MARAIS, J. 2014. Dispholidus typus (A. Smith, 1828). In: BATES, M.F., W.R. BRANCH, A.M. BAUER, M. BURGER, J. MARAIS, G.J. ALEXANDER & M.S. DEVILLIERS. 2014. Atlas and red list of the reptiles of South Africa, Lesotho and Swaziland. South African National Biodiversity Institute 415.

MARES, R., C. DOUTRELANT, M. PAQUET, C.N. SPOTTISWOODE & R. COVAS. 2017. Breeding decisions and output are correlated with both temperature and rainfall in an arid-region passerine, the sociable weaver. Royal Society Open Science 4(9): 170835.

MARITZ, B. & G.J. ALEXANDER. 2012. Population density and survival estimates of the African viperid, Bitis schneideri. Herpetologica 68(2): 195–202.

MARITZ, B., G.J. ALEXANDER & R.A. MARITZ. 2019. The underappreciated extent of cannibalism and ophiophagy in African cobras. Ecology 100(2): e02522

MARITZ, B., L. VERBUGT, A. COETZER, T. PING, R.A. MARITZ, G. NICHOLAU, N. PARRY & J. BOYCE. 2017 Facebook group: Predation records—reptiles and frogs. https://web.facebook.com/groups/888525291183325/ [Accessed 21 July 2018].

MARITZ, B. & G.J. ALEXANDER. 2014. Naja nivea (Linnaeus, 1758). In: BATES, M.F., W.R. BRANCH, A.M. BAUER, M. BURGER, J. MARAIS, G.J. ALEXANDER & M.S. DEVILLIERS. 2014. Atlas and red list of the reptiles of South Africa, Lesotho and Swaziland. South African National Biodiversity Institute 406.

MARTIN, T.E., P.R. MARTIN, C.R. OLSON, B.J. HEIDINGER & J.J. FONTAINE. 2000. Parental care and clutch sizes in North and South American birds. Science 287(5457): 1482– 1485.

MARTIN, T.E., J. SCOTT & C. MENGE. 2000. Nest predation increases with parental activity: separating nest site and parental activity effects. Proceedings of the Royal Society of London. Biological Sciences 267(1459): 2287–2293.

MCQUEEN, D.J., J.R. POST & E.L. MILLS. 1986. Trophic relationships in freshwater pelagic ecosystems. Canadian Journal of Fisheries and Aquatic Sciences 43(8): 1571–1581.

MENDELSOHN, J.M. & M.D. ANDERSON. 1997. Sociable weaver Philetairus socius. In: HARRISON, J.A., D.G. ALLAN, L.G. UNDERHILL, M. HERREMANS, A.J. TREE, V.

http://etd.uwc.ac.za/98

PARKER & C.J. BROWN. The atlas of Southern African birds. Vol. 2: Passerines. BirdLife South Africa 534–535.

MESHAKA JR, W.F. LOFTUS, & T.M. STEINER. 2000. An annotated inventory of the herpetofauna of Everglades National Park, Florida. Florida Scientist 63: 84–103.

MILLER, D.G., C.T. IVEY & J.D. SHEDD. 2009. Support for the microenvironment hypothesis for adaptive value of gall induction in the California gall wasp, Andricus quercuscalifornicus. Entomologia Experimentalis et Applicate 132(2): 126–133.

MISHRA, N., K. CREWS, J. MILLER & T. MEYER. 2015. Mapping vegetation morphology types in southern Africa savanna using MODIS time-series metrics: a case study of central Kalahari, Botswana. Land 4(1): 197–215.

MORTENSEN, H.S., Y.L. DUPONT & J.M. OLESEN. 2008. A snake in paradise: disturbance of plant reproduction following extirpation of bird flower-visitors on Guam. Biological Conservation 141(8): 2146–2154.

MUCINA, L. & M.C. RUTHERFORD. 2006. The vegetation of South Africa, Lesoto and Swaziland. In Strelitzia 19. South African National Biodiversity Institute, Pretoria. Memoirs of the Botanical Survey of South Africa.

MULLIN, S.J. & R.J. COOPER. 1998. The foraging ecology of the gray rat snake (Elaphe obsolete spiloides)—visual stimuli facilitate location of arboreal prey. The American Midland Naturalist 140(2): 397–401.

MULLIN, S.J. & R.J. COOPER. 2002. Barking up the wrong tree: climbing performance of rat snakes and its implications for depredation of avian nests. Canadian Journal of Zoology 80(4): 591–595.

MUÑOZ, J. 2013. Mammal densities in the Kalahari, Botswana–impact of seasons and land use. MSc thesis. Uppsala University, Sweden.

NATHAN, R., W.M. GETZ, E. REVILLA, M. HOLYOAK, R. KADMON, D. SALTZ & P.E. SMOUSE. 2008. A movement ecology paradigm for unifying organismal movement research. Proceedings of the National Academy of Sciences 105(49): 19052–19059.

http://etd.uwc.ac.za/99

NATUSCH, D.J., J.A. LYONS, G. BROWN AND R. SHINE. 2016. Communally nesting migratory birds create ecological hot-spots in tropical Australia. PLOS ONE 11(10): p.e0162651.

NATUSCH, D.J., J.A. LYONS & R. SHINE. 2017. How do predators and scavengers locate resource hotspots within a tropical forest?. Australian Ecology 42(6): 742–749.

NOWAK, E.M., T.C. THEIMER & G.W. SCHUETT. 2008. Functional and numerical responses of predators: where do vipers fit in the traditional paradigms?. Biological Reviews 83(4): 601– 620.

PAINE, R.T. 1980. Food webs: linkage, interaction strength and community infrastructure. Journal of Animal Ecology 49(3): 667–685.

PAULAY, G. 1994. Biodiversity on oceanic islands: its origin and extinction. American Zoologist 34(1): 134–144.

PENNER, J., C. FRUTEAU, F. RANGE & M. RODEL, M. 2008. Finding a needle in a haystack: new methods of locating and working with rhinoceros vipers (Bitis rhinoceros). Herpetological Review 39(3): 310.

PETTORELLI, N., T. COULSON, S. DURANT & J.-M. GAILLARD. 2011. Predation, individual variability and vertebrate population dynamics. Oecologia 167: 305–314.

PHELPS, T. 2007. Observations of the cape cobra, Naja nivea (Serpentes: Elapidae) in the DeHoop Nature Reserve, Western Cape Province, South Africa. Herpetological Bulletin 99: 29– 35.

PLA, D., L. SANZ, G. WHITELEY, S.C. WAGSTAFF, R.A. HARRISON, N.R. CASEWELL & J.J. CALVETEJ. 2017. What killed Karl Patterson Schmidt? Combined venom gland transcriptomic, venomic and antivenomic analysis of the South African green tree snake (the boomslang), Dispholidus typus. Biochimica et Biophysica Acta (BBA)-General Subjects 1861(4): 814–823.

POLIS, G.A. 1999. Why are parts of the world green? Multiple factors control productivity and the distribution of biomass. Oikos 86(1): 3–15.

http://etd.uwc.ac.za/100

POLISHCHUK, L.V., J. VIJVERBERG, D.A. VORONOV & W.M. MOOIJ. 2013. How to measure top–down vs bottom–up effects: a new population metric and its calibration on Daphnia. Oikos 122(8): 1177–1186.

PUNTILA, R.I., C.W. MARTIN & J.F. VALENTINE. 2012. Measuring predation with a new design of submersible chronographic timer. Bulletin of Marine Science 88(4): 1115–1122.

RAY, J.M., C.E. MONTGOMERY, H.K. MAHON, A.H. SAVITZKY & K.R. LIPS. 2012. Goo- eaters: diets of the neotropical snakes Dipsas and Sibon in Central Panama. Copeia 2012(2): 197–202.

READING, C.J., L.M. LUISELLI, G.C. AKANI, X. BONNET, G. AMORI, J.M. BALLOUARD, E. FILIPPI, G. NAULLEAU, D. PEARSON & L. RUGIERO. 2010. Are snake populations in widespread decline? Biology Letters 6(6): 777–780.

REASER, J.K., L.A. MEYERSON, Q. CRONK, M.A.J. DE POORTER, L.G. ELDREGE, E. GREEN, M. KAIRO, P. LATASI, R.N. MACK, J. MAUREMOOTOO, D. O'DOWD, W. ORAPA, S. SASTROUTOMO, A. SAUNDERS, C. SHINE, S. THRAINSSON & L. VAIUTU. 2007. Ecological and socioeconomic impacts of invasive alien species in island ecosystems. Environmental Conservation 34(2): 98–111.

RICKLEFS, R.E. 1969. An analysis of nesting mortality in birds. Smithsonian Contributions to Zoology 9: 1–48.

RIPPLE, W.J., E.J. LARSEN, R.A. RENKIN & D.W. SMITH. 2001. Trophic cascades among wolves, elk and aspen on Yellowstone National Park’s northern range. Biological Conservation 102(3): 227–234.

RIPPLE, W.J. & R.L. BESCHTA. 2011. Trophic cascades in Yellowstone: the first 15 years after wolf reintroduction. Biological Conservation 145(1): 205–213.

RODDA, G.H., T.H. FRITTS & P.J. CONRY. 1992. Origin and population growth of the Brown tree snake, Boiga irregularis, on Guam. Pacific Science 46(1): 46–57.

RODDA, G.H. & T.H. FRITTS. 1992. The impact of the introduction of the colubrid snake Boiga irregularis on Guam's lizards. Journal of Herpetology 26(2): 166–174.

http://etd.uwc.ac.za/101

RODDA, G.H., T.H. FRITTS, M.J. MCCOID & E.W. CAMPBELL, III. 1999. An Overview of the Biology of the Brown Treesnake* (Boiga irregularis), a Costly Introduced Pest on Pacific Islands. USDA National Wildlife Research Center – Staff Publications. 659.

RODRÍGUEZ-ROBLES, J.A., C.J. BELL & H.W. GREENE. 1999. Gape size and evolution of diet in snakes: feeding ecology of erycine boas. Journal of Zoology 248(1): 49–58.

RODRÍGUEZ-ROBLES, J.A., 2002. Feeding ecology of North American gopher snakes (Pituophis catenifer, Colubridae). Biological Journal of the Linnean Society 77(2): 165–183.

ROGERS, H., J.H.R. LAMBERS, R. MILLER & J.J. TEWKSBURY. 2012. ‘Natural experiment’ demonstrates top-down control of spiders by birds on a landscape level. PLOS ONE 7(9): e43446.

ROGERS, H.S., E.R. BUHLE, J. HILLERISLAMBERS, E.C. FRICKE, R.H. MILLER & J.J. TEWKSBURY. 2017. Effects of an invasive predator cascade to plants via mutualism disruption. Nature Communications 8: 14557.

ROUAG, R., H. DJILALI, H. GUERAICHE & L. LUISELLI. 2007. Resource partitioning patterns between two sympatric lizard species from Algeria. Journal of Arid Environments 69(1): 158–168.

RYMER, T.L., R.L. THOMSON & M.J. WHITING. 2014. At home with the birds: Kalahari tree skinks associate with sociable weaver nests despite African pygmy falcon presence. Austral Ecology 39(7): 839–847.

SAKAI, A.K., F.W. ALLENDORF, J.S. HOLT, D.M. LODGE, J. MOLOFSKY, K.A. WITH, S. BAUGHMAN, R.J. CABIN, J.E. COHEN, N.C. ELLSTRAND & D.E. MCCAULEY. 2001. The population biology of invasive species. Annual Review of Ecology and Systematic 32(1): 305– 332.

SAUNDERS, D.A., B.A. WINTLE, P.R. MAWSON & R. DAWSON. 2013. Egg-laying and rainfall synchrony in an endangered bird species: Implications for conservation in a changing climate. Biological Conservation 161: 1–9.

SAVIDGE, J.A. 1987. Extinction of an island forest avifauna by an introduced snake. Ecology 68(3): 660–668.

http://etd.uwc.ac.za/102

SAVIDGE, J.A. 1988. Food habits of Boiga irregularis, an introduced predator on Guam. Journal of Herpetology 275–282.

SCHOENER, T.W. 1974. Resource partitioning in ecological communities. Science 185(4145): 27–39.

SCHOLES, R.J., P.R. DOWTY, K. CAYLOR, D.A.B. PARSONS, P.G.H. FROST & H.H. SHUGART. 2002. Trends in savanna structure and composition along an aridity gradient in the Kalahari. Journal of Vegetation Science 13(3): 419–428.

SECOR, S.M. & K.A. NAGY. 1994. Bioenergetic correlates of foraging mode for the snakes Crotalus cerastes and Masticophis flagellum. Ecology 75(6): 1600–1614.

SECOR, S.M. 1995. Ecological aspects of foraging mode for the snakes Crotalus cerastes and Masticophis flagellum. Herpetological Monographs 9: 169–186.

SEKERCIOGLU, C.H. 2006. Increasing awareness of avian ecological function. Trends in Ecology & Evolution 21(8): 464–471.

SHARMA, B.K. & R.C. KANDEL. 2015. Assessment of python (Python bivittatus Kuhl.) habitats in Bardiya National Park, Nepal. Ecoprint: An International Journal of Ecology 22: 85– 90.

SHINE, R. 1983. Arboreality in snakes: ecology of the Australian elapid genus Hoplocephalus. Copeia 1983: 198–205

SHINE, R., P.S. HARLOW, W.R. BRANCH & J.K. WEBB. 1996. Life on the lowest branch: sexual dimorphism, diet, and reproductive biology of an African twig snake, Thelotornis capensis (Serpentes, Colubridae). Copeia 1996: 290–299.

SHINE, R., W.R. BRANCH, P.S. HARLOW & J.K. WEBB. 1998. Reproductive biology and food habits of horned adders, Bitis caudalis (Viperidae), from southern Africa. Copeia 1998: 391–401.

SHINE, R., L.X. SUN, M. KEARNEY & M. FITZGERALD. 2002. Thermal correlates of foraging-site selection by Chinese pit-vipers (Gloydius shedaoensis, Viperidae). Journal of Thermal Biology 27(5): 405–412.

http://etd.uwc.ac.za/103

SHINE, R., W.R. BRANCH, J.K. WEBB, P.S. HARLOW, T. SHINE, & J.S. KEOGH. 2007. Ecology of cobras from southern Africa. Journal of Zoology 272(2): 183–193.

SIH, A. 1980. Optimal behavior: can foragers balance two conflicting demands? Science 210(4473): 1041–1043.

SINGER, F.J., L.C. MARK & R.C. CATES. 1994. Ungulate herbivory of willows on Yellowstone's northern winter range. Journal of Range Management 47(6): 435–443.

SIVAN, J., M. KAM, S. HADAD, A.A. DEGEN, I. ROZENBOIM & A. ROSENSTRAUCH. 2013. Temporal activity and dietary selection in two coexisting desert snakes, the Saharan sand viper (Cerastes vipera) and the crowned leafnose (Lytorhynchus diadema). Zoology 116(2): 113–117.

SKUTCH, A.F. 1949. Do tropical birds rear as many young as they can nourish? Ibis. 91: 430– 455.

SMITH, C.C.D., I. LAYLOO, R.A. MARITZ & B. MARITZ. 2019. Sexual dichromatism does not translate into sex-based difference in morphology or diet for the African boomslang. Journal of Zoology.

SMITH, D.W. & D.B. TYRES. 2012. The history and current status and distribution of beavers in Yellowstone National Park. Northwest Science 86(4): 276–288.

SNOW, R.W., M.L. BRIEN, M.S. CHERKISS, L. WILKINS & F.J. MAZZOTTI. 2007. Dietary habits of the Burmese python, Python molurus bivittatus, in Everglades National Park, Florida. Herpetological Bulletin 101: 5.

SOULÉ, M.E. 1985. What is conservation biology?. BioScience 35(11): 727–734.

SPERRY, J.H., R.G. PEAK, D.A. CIMPRICH & P.J. WEATHERHEAD. 2008. Snake activity affects seasonal variation in nest predation risk for birds. Journal of Avian Biology 39(4): 379– 383.

SPERRY, J.H., D.A. CIMPRICH, R.G. PEAK & P.J. WEATHERHEAD. 2009. Is nest predation on two endangered bird species higher in habitats preferred by snakes?. Ecoscience 16(1): 111– 118.

http://etd.uwc.ac.za/104

SPIBY, J. 2014. Not so friendly neighbours: Pygmy Falcon eating Sociable Weaver nestling. Biodiversity Observations 357–360.

SPOTTISWOODE, C.N. 2007. Phenotypic sorting in morphology and reproductive investment among sociable weaver colonies. Oecologia 154(3): 589–600.

SPOTTISWOODE, C.N. 2009. Fine‐scale life‐history variation in sociable weavers in relation to colony size. Journal of Animal Ecology 78(3): 504–512.

STAKE, M.M., J. FAABORG & F.R. THOMPSON, III. 2004. Video identification of predators at Golden-cheeked Warbler nests. Journal of Field Ornithology 75(4): 337–344.

STAKE, M.M., F.R. THOMPSON, III, J. FAABORG & D.E. BURHANS. 2005. Patterns of snake predation at songbird nests in Missouri and Texas. Journal of Herpetology 39(2): 215–222

STEEN, D.A., C. GUYER & L.L. SMITH. 2012. A case study of relative abundance in snakes. In: MCDIARMID,, R.W., M.S. FOSTER, C. GUYER, N. CHERNOFF & J.W. GIBBONS. 2012. Reptile biodiversity: standard methods for inventory and monitoring. University of California Press, Berkeley, California, USA. 287–294.

STEEN, D.A., C.J. MCCLURE, J.C. BROCK, D.C. RUDOLPH, J.B. PIERCE, J.R. LEE, W.J. HUMPHRIES, B.B. GREGORY, W.B. SUTTON, L.L. SMITH & D.L. BAXLEY. 2012. Landscape-level influences of terrestrial snake occupancy within the southeastern United States. Ecological Applications 22(4): 1084–1097.

STEEN, D.A., C.J. MCCLURE, W.B. SUTTON, D.C. RUDOLPH, J.B. PIERCE, J.R. LEE, L.L. SMITH, B.B. GREGORY, D.L. BAXLEY, D.J. STEVENSON & C. GUYER. 2014. Copperheads are common when kingsnakes are not: relationships between the abundances of a predator and one of their prey. Herpetologica 70(1): 69–76.

STUART, B., T.Q. NGUYEN, N. THY, L. GRISMER, T. CHAN-ARD, D. ISKANDAR, E. GOLYNSKY & M.W.N. LAU. 2012. Python bivittatus. The IUCN Red List of Threatened Species 2012: e.T193451A2237271. http://dx.doi.org/10.2305/IUCN.UK.2012- 1.RLTS.T193451A2237271.en. Downloaded on 30 August 2016.

SUN, L.X., R. SHINE, Z. DEBI, & T. ZHENGREN. 2001. Biotic and abiotic influences on activity patterns of insular pit-vipers (Gloydius shedaoensis, Viperidae) from north-eastern China. Biological Conservation 97(3): 387–398.

http://etd.uwc.ac.za/105

THOMAS, D.S.G. & P.A. SHAW. 1993. The evolution and characteristics of the Kalahari, southern Africa. Journal of Arid Environments 25(1): 97–108.

TOFT, C.A. 1985. Resource partitioning in amphibians and reptiles. Copeia 1985: 1–21.

TOLEDO, L.F., A.S. ABE & D.V. ANDRADE. 2003. Temperature and meal size effects on the postprandial metabolism and energetics in a boid snake. Physiological and Biochemical Zoology 76: 240–246.

TRAN, M.V., M. O’GRADY, J. COLBORN, K. VAN NESS & R.W. HILL. 2014. Aggression and food resource competition between sympatric hermit crab species. PLOS ONE 9(3): e91823.

VAN DIJK, R.E., J.C. KADEN, A. ARGÜELLES‐TICÓ, L.M. BELTRAN, M PAQUET, R. COVAS, C. DOUTRELANT & B.J. HATCHWELL. 2013. The thermoregulatory benefits of the communal nest of sociable weavers Philetairus socius are spatially structured within nests. Journal of Avian Biology 44(2): 102–110.

VAN DIJK, R. E., J.C. KADEN, A. ARGÜELLES‐TICÓ, D.A. DAWSON, T. BURKE & B.J. HATCHWELL. 2014. Cooperative investment in public goods is kin directed in communal nests of social birds. Ecology Letters 17(9): 1141–1148.

VISSER, M.E., L.J.M. HOLLEMAN & S.P. CARO. 2009 Temperature has a causal effect on avian timing of reproduction. Proceedings of the Royal Society B: Biological Sciences 276: 2323–2331.

WASIOLKA, B. & N. BLAUM. 2011. Comparing biodiversity between protected savanna and adjacent non-protected farmland in the southern Kalahari. Journal of Arid Environments 75(9), 836-841.

WEATHERHEAD, P.J. & G. BLOUIN-DEMERS. 2004. Understanding avian nest predation: why ornithologists should study snakes. Journal of Avian Biology 35(3): 185–190.

WEBB, J.K., R. SHINE, W.R. BRANCH & P.S. HARLOW. 2000. Life-history strategies in basal snakes: reproduction and dietary habits of the African thread snake Leptotyphlops scutifrons (Serpentes: Leptotyphlopidae). Journal of Zoology 250(3): 321–327.

WEBB, J.K., W.R. BRANCH & R. SHINE. 2001. Dietary habits and reproductive biology of typhlopid snakes from southern Africa. Journal of Herpetology 35(4): 558–567.

http://etd.uwc.ac.za/106

WEBB, J.K. & M.J. WHITING. 2005. Why don't small snakes bask? Juvenile broad‐headed snakes trade thermal benefits for safety. Oikos 110(3): 515–522.

WEBBER, M.M., T. JEZKOVA & J.A. RODRÍGUEZ-ROBLES. 2016. Feeding ecology of sidewinder rattlesnakes, Crotalus cerastes (Viperidae). Herpetologica 72(4): 324–330.

WHALEN, M.A., J.E. DUFFY & J.B. GRACE. 2013. Temporal shifts in top-down vs. bottom- up control of epiphytic algae in a seagrass ecosystem. Ecology 94(2): 510–520.

WHITE, T.C.R. 1978. The importance of a relative shortage of food in animal ecology. Oecologia 33(1): 71–86.

WILCOVE, D.S., D. ROTHSTEIN, J. DUBOW, A. PHILLIPS & E. LOSOS. 1998. Quantifying threats to imperiled species in the United States. BioScience 48(8): 607–615.

WILLSON, J.D. 2017. Indirect effects of invasive Burmese pythons on ecosystems in southern Florida. Journal of Applied Ecology 54(4): 1251–1258.

WILLSON, J.D., M.E. DORCAS & R.W. SNOW. 2011. Identifying plausible scenarios for the establishment of invasive Burmese pythons (Python molurus) in southern Florida. Biological Invasions 13(7): 1493–1504

WILLSON, J.D. & C.T. WINNE. 2016. Evaluating the functional importance of secretive species: A case study of aquatic snake predators in isolated wetlands. Journal of Zoology 298(4): 266–273.

WILLSON, J.D., C.T. WINNE & B.D. TODD. 2011. Ecological and methodological factors affecting detectability and population estimation in elusive species. The Journal of Wildlife Management 75(1): 36–45.

ZAIDAN, F. & S.J. BEAUPRE. 2003. Effects of body mass, meal size, fast length, and temperature on specific dynamic action in the timber rattlesnake (Crotalus horridus). Physiological and Biochemical Zoology 76: 447–458.

ZANN, R.A., S.R. MORTON, K.R. JONES & N.T. BURLEY. 1995. The timing of breeding by zebra finches in relation to rainfall in central Australia. Emu 95(3): 208–222.

http://etd.uwc.ac.za/107