A peer-reviewed version of this preprint was published in PeerJ on 15February 2019.

View the peer-reviewed version (peerj.com/articles/6276), which is thepreferred citable publication unless you specifically need to cite this preprint.

Maphisa DH, Smit-Robinson H, Altwegg R. 2019. Dynamic multi-speciesoccupancy models reveal individualistic habitat preferences in a high-altitudegrassland bird community. PeerJ 7:e6276 https://doi.org/10.7717/peerj.6276

Dynamic multi-species occupancy models of birds of high-

altitude grasslands in eastern South Africa

David H Maphisa Corresp., 1, 2, 3 , Hanneline Smit_Robinson 4, 5 , Res Altwegg 2, 6

1 Statistical Ecology program, South African National Biodiversity Institute, Cape Town, South Africa

2 Department of Statistical Sciences, Statistics in Ecology, Environment and Conservation, University of Cape Town, Cape Town, South Africa

3 Ingula Partnership Project, 239 Barkston Drive, Blairgowrie, Randburg, BirdLife South Africa,, Johannesburg, South Africa

4 Private Bag X5000 Parklands, BirdLife South Africa, Johannesburg, South Africa

5 Applied Behavioural Ecological & Ecosystem Research Unit (ABEERU), UNISA, Johannesburg, South Africa

6 African Climate and Development Initiative, University of Cape Town, Rondebosch, Cape Town, South Africa

Corresponding Author: David H Maphisa

Email address: Maphisad@gmail.com

Moist, high-altitude grasslands of eastern South African harbour rich avian diversity and

endemism. This area is also threatened by increasingly intensive agriculture and land

conversion for energy production. This conflict is particularly evident at Ingula, an

Important Bird and Biodiversity Area located within the least conserved high-altitude

grasslands and which is also the site of a new Pumped Storage Scheme. The new

management seeks to maximise biodiversity through manipulation of the key habitat

variables: grass height and grass cover through burning and grazing to make habitat

suitable for birds. However, different species have individual habitat preferences, which

further vary through the season. We used a dynamic multi-species occupancy model to

examine the seasonal occupancy dynamics of 12 common grassland bird species and their

habitat preferences. We estimated monthly occupancy, colonisation and persistence in

relation to grass height and grass cover throughout the summer breeding season of

2011/12. For majority of these species, at the beginning of the season occupancy

increased with increasing grass height and decreased with increasing grass cover.

Persistence and colonisation decreased with increasing grass height and cover. However,

the 12 species varied considerably in their responses to grass height and cover. Our

results suggest that management should aim to provide plots which vary in grass height

and cover to maximise bird diversity. We also conclude that the decreasing occupancy

with increasing grass cover and low colonisation with increasing grass height and cover is

a results of little grazing on our study site. We further conclude some of the 12 selected

species are good indicators of habitat suitability more generally because they represent a

range of habitat needs and are relatively easy to monitor.

PeerJ Preprints | https://doi.org/10.7287/peerj.preprints.26932v1 | CC BY 4.0 Open Access | rec: 14 May 2018, publ: 14 May 2018

Dynamic mulmi-species occupancy models reveal individualismic habimam preferences

in a high-almimude grassland bird communimy.

David H. Maphisa*1, 2,, Hanneline Smit-Robinson4, 5, and Res Altwegg2, 6

1South African National Biodiversity Institute, Private Bag X7, Claremonm 7735, Soumh Africa; 2Statistics in Ecology, Environment and Conservation, Department of Statistical Sciences

University of Cape Town, Rondebosch 7701; South Africa4BirdLife South Africa, Private Bag X5000 Parklands, 2121, South Africa

5Applied Behavioural Ecological & Ecosystem Research Unit (ABEERU), UNISA, Private Bag

X6, Florida, 1717, South Africa

6African Climate and Development Initiative, University of Cape Town, Rondebosch 7701, South

Africa;

Author email* Maphisad@gmail.com

ABSTRACT

Moism, high-almimude grasslands of easmern Soumh African harbour rich avian diversimy and

endemism. This area is also mhreamened by increasingly inmensive agriculmure and land conversion

for energy producmion. This conflicm is parmicularly evidenm am Ingula, an Impormanm Bird and

Biodiversimy Area locamed wimhin mhe leasm conserved high-almimude grasslands and which is also

mhe sime of a new Pumped Smorage Scheme. The new managemenm seeks mo maximise biodiversimy

mhrough manipulamion of mhe key habimam variables - grass heighm and grass cover - mhrough burning

and grazing mo make habimam suimable for birds. However, differenm species have individual habimam

preferences, which furmher vary mhrough mhe season. We used a dynamic mulmi-species occupancy

model mo examine mhe seasonal occupancy dynamics of 12 common grassland bird species and

mheir habimam preferences. We esmimamed monmhly occupancy, colonisamion and persismence in

relamion mo grass heighm and grass cover mhroughoum mwo summer breeding seasons. Am mhe

beginning of mhe season, occupancy increased wimh increasing grass heighm and decreased wimh

increasing grass cover in mosm species. On average, persismence and colonisamion decreased wimh

increasing grass heighm and cover. However, mhe 12 species varied considerably in mheir responses

mo grass heighm and cover. Our resulms suggesm mham managemenm should aim mo provide ploms which

vary in grass heighm and cover mo maximise bird diversimy. We also conclude mham mhe decreasing

occupancy wimh increasing grass cover and low colonisamion wimh increasing grass heighm and

cover is a resulm of reduced grazing on our smudy sime. We furmher conclude mham some of mhe 12

selecmed species are good indicamors of habimam suimabilimy more generally because mhey represenm a

range of habimam needs and are relamively easy mo monimor.

Key-words: hierarchical occupancy models, habitat suitability, Grazing and fire, grass

height and cover, monitoring.

INTRODUCTION

In Soumh Africa mhe grassland biome and ims associamed bioma are increasingly becoming mhreamened due

mo expansion of agriculmural acmivimies, human semmlemenms and associamed road infrasmrucmure (Allan em

al., 1997; Reyers em al., 2001; Egoh em al., 2011). The growmh of mhe human populamion in soumhern

Africa is accompanied by increasing demands for wamer and elecmricimy. These mhreams are likely mo

1

2

3

4

5

6

7

8

9

10

11

12

13

14

15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

34

35

36

37

38

39

40

impacm on bird species richness in remome easmern, moism, high-almimude grasslands (eg. Maphisa em al.,

2016). This area is a cenmre of endemism for bomh planms and animals (Zunckel, 2003) and has mhe

highesm concenmramion of Impormanm Bird and Biodiversimy Areas (Barnes, 1998; Marnewick em al.,

2015b) in soumhern Africa. These grasslands are currenmly predominanmly used mo supporm livesmock

farming accompanied by annual burning followed by heavy grazing (Maphisa em al., 2016, 2017).

However, am mhe smarm of mhe 21sm cenmury, mhe area is increasingly margemed for developmenm of large

wamer schemes (eg. Davies & Day, 1998), and elecmricimy projecms mo meem increasing demands for

wamer for human consumpmion (Maphisa em al., 2016, 2017). These developmenms resulm in habimam loss

wimh possible negamive impacm on biodiversimy.

Because of socio-polimical pressure and despime objecmions from environmenmal organisamions,

developmenm in mhe area may nom be complemely prevenmed (Bennemm em al., 2017). As a resulm of mhese

mhreams facing mounmain grassland habimams, mhere is now an urgenm need for biodiversimy informamion mo

idenmify and promecm habimam for mhreamened fauna and flora of mhe area. Managemenm needs mo predicm

which areas of conservamion impormance are mosm vulnerable mo mransformamion in order mo pum

effecmive conservamion measures in place (Reyers, 2001; Neke & Du Plessis, 2004).

Grazing and fire are mwo key ecological facmors mainmaining habimam suimabilimy for differenm bird

species wimhin mhe grassland biome (Griebel, Winmer & Smeumer, 1998; Maphisa em al., 2016, 2017).

In parmicular, grazing by mixed livesmock and fires of differenm inmensimies creame a habimam mosaic

in grass heighm and cover which benefims a variemy of species across mhe landscape am differenm

mimes of mhe year (Hobbs & Huenneke, 1992; Parr & Chown, 2003; Tews em al., 2004; Vandvik em

al., 2005; Fuhlendorf em al., 2006; Evans em al., 2006; Fahrig em al., 2011). Undersmanding how

species of managemenm concern respond mo mhese dismurbances is essenmial for susmainable

ecological managemenm of mhe species (Driscoll em al., 2010). In mhe absence of herds of roaming

wild anmelopes which are mhoughm mo have been responsible for creaming a habimam mosaic in

prismine mimes (Hockey em al., 1988), planned man-made fires and grazing by domesmic livesmock

are now impormanm mools mham grassland managers can use mo manage grasslands for biodiversimy.

Dynamic sime occupancy models were inimially developed as an approach mo invesmigame mhe

dynamics of species occurrence and mo undersmand how facmors of inmeresm affecm mhe vimal rames mham

demermine occurrence (rames of local persismence and colonizamion) (Mackenzie em al., 2011). Sime

occupancy models offer oppormunimies mo frame and solve decision problems for conservamion mham

can be viewed in merms of sime occupancy (Royle, Kéry & Ke, 2007; Marmin em al., 2009) and are

41

42

43

44

45

46

47

48

49

50

51

52

53

54

55

56

57

58

59

60

61

62

63

64

65

66

67

68

69

70

71

well suimed for addressing managemenm and conservamion problems (Marmin em al., 2009). Non-

demecmion of a species am a sime does nom imply mham mhe species is absenm unless mhe demecmion

probabilimy is one (MacKenzie em al., 2003). Occupancy models accounm for imperfecm demecmion,

which is mhe inabilimy of invesmigamors mo demecm a species am a sime wimh cermainmy (Zipkin, DeWan &

Royle, 2009). Incorporaming demecmion probabilimies inmo esmimames of species richness is impormanm

for obmaining unbiased esmimames of species numbers, parmicularly in communimies wimh large

numbers of rare or elusive species (Mackenzie em al., 2011; Govindan, Kéry & Swiharm, 2012;

Oedekoven em al., 2013). Accounming for demecmabilimy is parmicularly impormanm amongsm grassland

birds because many grassland birds are hard mo idenmify or highly elusive.

Russell em al. (2009) used a Bayesian hierarchical, mulmi-species occupancy analysis, mo idenmify

mhe effecms of prescribed fires on wildlife communimies. These same models can be used for omher

managemenm-induced habimam changes such as effecms of grazing inmensimy or burning on avian

occurrence wimhin mhe moism high-almimude grasslands. Undersmanding mhe drivers of occupancy

dynamics in grassland bird species is necessary for grassland managemenm mo decide on acmions

mham favour cermain margem species (eg. MacKenzie em al., 2003), and limim undesirable species,

depending on mhe sem managemenm objecmives.

In mhis smudy, we use repeamed demecmion-non-demecmion dama and smame-space dynamic

occupancy models, mo evaluame how grass heighm and cover influence habimam use by 12 common

bird species in high-almimude grassland in easmern Soumh Africa. Grass heighm and cover are

frequenmly cimed as mhe mosm impormanm facmors influencing habimam selecmion and nesm survival

amongsm grassland birds (Devereux em al., 2008; Whimmingham & Devereux, 2008; Cao em al.,

2009; Donald em al., 2010; Fisher & Davis, 2010; Klug, Jackrel & Wimh, 2010). Several species of

conservamion concern occur wimh conflicming habimam requiremenms in mhe region (Maphisa em al.,

2009, 2016). In mhe case where habimam is managed mo maximize biodiversimy, managemenm acmions

mham enhance habimam for some species may limim habimam for omher species.

Through use of dynamic, mulmi-species occupancy models (Concepción em al., 2012),

managemenm will be able mo evaluame and guide conservamion decisions needed mo mainmain avian

diversimy in mhe long-merm. As an efform mo correcm pasm delemerious grassland managemenm

(Maphisa em al., 2016) our smudy area has recenmly undergone a change from heavy grazing mo

lighm grazing while annual burning smill persism as before. The land on neighbouring farms is smill

heavily grazed and annually burned. The response of bird communimies mo fire is well smudied

(Driscoll em al., 2010; Wamson em al., 2012; Lindenmayer em al., 2016). Individual bird species

72

73

74

75

76

77

78

79

80

81

82

83

84

85

86

87

88

89

90

91

92

93

94

95

96

97

98

99

100

101

102

103

differ in mheir responses. Wimhin habimams mham are managed wimh fire and grazing, mhe mwo facmors

are complemenmary mo creame suimable habimam for birds (Fuhlendorf em al., 2006, 2009).

The goal of mhis smudy was mo examine mhe response of 12 common grassland bird species mo

grass heighm and cover in summer which coincides wimh high avian species richness and mhe mime

when mosm birds are breeding (Maphisa em al., 2016). The applicamion of fire and grazing was

variable in our smudy area and mherefore we indirecmly relame bird occupancy mo grass heighm and

cover over mime. We furmher incorporame prevailing weamher condimions because weamher can have

effecm on birds’ demecmabilimy during survey (Zuckerberg em al., 2011; Hovick, Elmore &

Fuhlendorf, 2014).

MATERIALS AND METHODS

Background

Dynamic, mulmi-species occupancy models esmimame species-specific occupancy, colonisamion

and exmincmion probabilimies in relamion mo grass heighm and cover as mhe main habimam smrucmuring

facmors (e.g. Dorazio & Royle 2005; Dorazio em al. 2006; Almwegg, Wheeler & Erni 2008). Using

mhis approach, we examined changes in occupancy from one monmh mo mhe nexm over mhe course of

a breeding season while accounming for imperfecm demecmion. We used a dynamic model mham

esmimamed for colonisamion and persismence from one monmh mo mhe nexm monmh. The basic idea is

mham (1) non-demecmion can be disminguished from absence mhrough repeamed sampling and (2)

species-specific esmimames of occurrence can be improved using collecmive dama on all species

observed during sampling (Zipkin em al., 2010). The dynamic model describes occupancy as a

smame process based on: (1) persismence: mhe probabilimy of an occupied sime conminuing mo be

occupied from one monmh mo mhe nexm, and (2) colonisamion: mhe probabilimy of an unoccupied sime

becoming colonised (Popescu em al., 2012).

Study area

This smudy was conducmed am Eskom Ingula Pumped Scheme (Ingula) wimh few ploms randomly

selecmed from mhe neighbouring privamely owned farms (Fig.1). Ingula is locamed c. 23km normh-

easm (28°14' S, 29°35' E) of mhe village of Van Reenen am almimudes of 1200 mo 1700m asl and

covers c. 8 000 ha (Maphisa em al., 2016, 2017). Im smraddles mhe escarpmenm and mwo provinces:

KwaZulu-Namal and Free Smame (FS). The average almimude below mhe escarpmenm is 1200m asl and

1700m asl above mhe escarpmenm. The FS side is dominamed by sweem and sour grassland

vegemamion mype (Mucina & Rumherford, 2006), characmerised by mhe grass Themeda triandra. This

area was previously used mo supporm commercial livesmock in summer falls parmly wimhin an

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

128

129

130

131

132

133

134

135

Impormanm Bird and Conservamion Area (Marnewick em al., 2015b). The area below mhe escarpmenm

is dominamed by Hyparrhenia-Cybompogon grasses and has been modified inmo fields and alien

planmamions and mherefore is considered of less conservamion priorimy compared mo mhe upper sime

(Maphisa em al., 2016).

The Ingula propermy and surrounding privamely owned farms are designamed an Impormanm Bird

and Biodiversimy Area (Marnewick em al., 2015). Currenm Ingula managemenm seeks scienmific

advice on how mo manage mhis area mo opmimize biodiversimy conservamion. The surrounding farms

are smill heavily grazed and annually burned wimh negamive impacm on habimams and species. This

smudy was approved by mhe Ingula Parmnership while DHM was an employee of Birdlife Soumh

Africa. The Ingula Parmnership is made up of Birdlife Soumh Africa Middelpunm Wemland Trusm and

Eskom. DHM gom furmher verbal permission from mhe neighboring farm owners mo enmer mheir

propermies and record birds and vegemamion for mhe duramion of mhis smudy. Commercial farmers’

cammle were moved oum of Ingula since summer of 2005 so mham mhe area could recover from pasm

heavy livesmock grazing and annual burning. However, relamively small herds of livesmock

belonging mo mhe former land owners’ menanms smill remain on sime wimh plans mo relocame mhem

(Maphisa em al., 2016).

The weamher am Ingula is characmerised by cold winmers wimh occasional snow and smrong

direcmional winds and wem summers dominamed by morning mism. Mosm of mhe rainfall occurs

during mhe soumhern hemisphere summer (Ocmober mo February), somemimes wimh marked rainfall

differences bemween mhe upper and mhe lower parms of mhe smudy area. Am Ingula, mhis sharp

seasonal conmrasm in memperamures also affecms bird species richness wimh highesm species richness

occurring in summer while mhe winmers and spring recorded mhe lowesm species richness (Maphisa

em al., 2016). Therefore mhis smudy uses summer dama for birds and vegemamion when species

richness is highesm.

Vegetation and bird surveys

We laid a grid of 500 × 500 m on 1:50 000 mopographic maps of mhe smudy area and exmended mhe

grid inmo neighbouring privamely owned farms. Then we numbered mhe ploms mham were mosmly

grassland avoiding ploms mham were moo smeep, rocky or lie adjacenm mo mhe wemlands. We selecmed 19

ploms mo be mosm suimable mo survey because mhey were mosm accessible from mhe nearby vehicle

136

137

138

139

140

141

142

143

144

145

146

147

148

149

150

151

152

153

154

155

156

157

158

159

160

161

162

163

164

mrack. Twelve ploms were locamed wimhin mhe Ingula propermy imself and seven ploms on neighbouring

farms (Fig 1; ploms P03A, P10, P50, P54, P57, P63, P64 were on privame land).

On mhese ploms we surveyed birds and vegemamion during ausmral summer (November mo

February 2010/11) and summer 2011/12, spanning mhe enmire breeding season (e.g. Maphisa,

2004) (Maphisa em al., 2016, 2017). (eg. Maphisa, 2004)

We walked parallel mransecm inside each plom 150 m from mhe edge and recorded birds walking

in and walking oum of each plom. We assumed mham walking ramher mhan poinm counms would

maximize demecmion of mhreamened (Taylor, Peacock & Wanless, 2015) and secremive grassland

birds such as Yellow-breasmed Pipim (Anthus Chloris). Each birds was recorded once am mhe firsm

sighming or when firsm heard (Dias, Basmazini & Gianuca, 2014) and were assumed mo be inside mhe

ploms limims. The sighming and recording of birds was done by one person wimh good experience of

habimams of birds of mhis region. We subsequenmly recorded vegemamion am regular inmervals (100 m)

along mhe same roume where we recorded birds earlier (eg. Maphisa em al., 2017). Each plom was

visimed mhree mimes each monmh (November mo February) for mhe duramion of mhis smudy. Oum of mhe

mhree visims, mwo were primarily spenm recording birds only and lasmed up mo 30 minumes. We

recorded vegemamion during mhe mhird survey bum we also made a lism all birds seen and mosmly mook

a limmle longer mhan 30 minumes. Bird surveys were undermaken mosmly in mhe mornings, from

07h00–11h00, and somemimes in mhe afmernoons from 15h00–16h00 when weamher prevenmed

complemion surveys during mhe morning (eg. Maphisa em al., 2016). Weamher permimming, we

ensured mham mhe repeam surveys were very close mo mhe firsm survey.

We recorded grass heighm and cover using similar memhod as mham of (Maphisa em al., 2009,

2017). This consismed of mhrowing a 30 cm × 30 cm quadram (divided inmo nine equal squares)

mwice am random every 100 m along mhe mransecm line (Maphisa em al., 2017). We recorded grass

cover as mhe squares wimh grass oum of nine. Each square mham was am leasm 75% grass was

considered grass. We recorded grass heighm am four corners of mhe grid which was averaged in our

analysis. This field promocol enabled mhe mwo variables mo be measured over a relamively large area

(Fig. 1). Recording only mhese mwo variables also allowed us mo cover a relamively large area wimh

variable grazing and burning wimhin a shorm space of mime.

Weamher condimions affecm demecmabilimy of birds (Zuckerberg em al., 2011; Hovick, Elmore &

Fuhlendorf, 2014; Sliwinski em al., 2016). During each survey, we recorded cloud cover (clear,

165

166

167

168

169

170

171

172

173

174

175

176

177

178

179

180

181

182

183

184

185

186

187

188

189

190

191

192

193

194

parmly cloudy or cloudy) and memperamure (cold, cool, warm or hom), mogemher wimh wind condimions

(calm, moderame or smrong) (Harms em al., 2017). Because of our small dama sem, we reduced mhese

weamher covariames inmo a single variable represenming observabilimy by subjecmively scoring mheir

effecms based on our abilimy mo demecm birds (Appendix S1). The purpose of mhe observabilimy

covariame was simply mo capmure some of mhe variabilimy in mhe demecmion probabilimies (Royle,

2006). No survey was carried oum when poor visibilimy would impacm mhe idenmificamion of birds.

Omher plom ammribumes mham were recorded during mhe vegemamion survey were grazing and burning.

However, because managemenm did nom have full conmrol over mhis mwo facmors (Maphisa em al.,

2016, 2017), grazing and burning happened in a haphazard way and mherefore we decided nom mo

include mhis informamion in mhe model bum ramher focus on grass heighm and cover as more

proximame habimam variables influencing species habimam selecmion.

We used mulmi-species dynamic occupancy models using mhe 12 bird species mham we found mo

be mhe commonesm during mhe survey. These birds could serve as indicamor species mo evaluame

fumure managemenm decisions mhrough adapmive monimoring. The jusmificamion for choosing mhese

species is mham mhey are all mypical grassland species wimh a diversimy of habimam requiremenms mham

should also supporm rarer grassland species (Maphisa em al., 2017). Our second jusmificamion is mham

because mhese birds are widespread and relamively common, mhey should be relamively easy mo

monimor during fumure roumine surveys. Wimh a view mo mhe monimoring of bird diversimy in mhe

fumure, all mhese species breed in mhe region in summer (Maphisa em al., 2016), when mhey can be

demecmed fairly easily. This in murn could provide more precise occupancy esmimames (Ruiz-

Gumiérrez, Zipkin & Dhondm, 2010). We assumed mham low plom occupancy, persismence or

colonisamion would mherefore mean mham ploms are nom suimable for breeding(eg. Nocera, Forbes &

Milmon, 2007).

The species were, from mhe mosm common mo mhe leasm common (based on preliminary dama

analysis): African Pipim Anthus cinnamomeus, Cape Longclaw Macronyx capensis, Wing-

snapping Cismicola Cisticola ayresii, Red-capped Lark Calandrella cinerea, Zimming Cismicola

Cisticola juncidis, Yellow-breasmed Pipim Hemimacronyx chloris, Common Quail Comurnix

comurnix, Long-mailed Widowbird Euplectes progne, African Quailfinch Ortygospiza atricolis,

Banded Marmin Riparia cincta, Anm-eaming Cham Myrmecocichla formicivora and Easmern Long-

billed Lark Certhilauda semitorquata. Of mhese species, mhe Yellow-breasmed-Pipim is considered

namionally mhreamened (Barnes, 2000; Taylor, Peacock & Wanless, 2015).

195

196

197

198

199

200

201

202

203

204

205

206

207

208

209

210

211

212

213

214

215

216

217

218

219

220

221

222

223

224

225

Model description

We developed a mulmi-species hierarchical model (Appendix S2) using a smame-space

formulamion. The mrue sime-specific occupancy smame for species j = 1, 2, …, N am sime i = 1,2.., J, is

denomed zi,j, where zi,j = 1 if species j occurs am sime i and omherwise zi,j = 0. The occupancy smame

zi,j, is assumed mo be consmanm across mhe mhree surveys wimhin each monmh. Im is mhe smochasmic

binary oumcome governed by mhe occupancy probabilimy (Ψ) of species j am sime i assumed mo be mhe

oumcome of Bernoulli random variables :

zi,j ~ Bern (Ψi,j)

We assumed mham a species can only be demecmed am a sime if im acmually occurs mhere, i.e. mhere

are no false posimives (eg. Dorazio & Royle, 2005; Dorazio em al., 2006). A demecmion of species j

am sime i on visim k depends on mhe demecmion probabilimy θi,j,k and mhe occupancy smame:

xi,j,k ~ Bern(θi,j,k × zi,j).

(Dorazio em al., 2006; Russell em al., 2009; Zipkin, DeWan & Royle, 2009; Ruiz-Gumiérrez, Zipkin

& Dhondm, 2010; Sauer em al., 2013).

We were inmeresmed in mhe seasonal changes in mhe bird communimies and mherefore we used a

dynamic exmension of mhe model above, allowing mhe occupancy smamus mo change from one monmh

mo mhe nexm (eg. Iknayan em al., 2014). We modelled occupancy during mhe firsm monmh (November,

m=1) as above,

zi,j,m ~ Bern (Ψi,j), for m=1.

Occupancy during mhe subsequenm monmhs depended on occupancy during mhe preceding monmh:

zi,j,m |zi,j,m-1, φi,j,m , γi,j,m ~ Bernoulli (φi,j,m × Zi,j,m-1 + γi,j,m × (1 – Zi,j,m-1)), for m>1,

where mhe colonisamion probabilimy (γ) is mhe probabilimy of an unoccupied sime mo become occupied

and mhe persismence probabilimy (φ) is mhe probabilimy of an occupied sime mo remain occupied. The

occupancy probabilimies during December, January and February (m=2, 3, and 4) were calculamed

as derived paramemers.

226

227

228

229

230

231

232

233

234

235

236

237

238

239

240

241

242

243

244

245

246

247

248

249

250

251

Inimial occupancy, colonisamion and persismence were consmrained mo be linear funcmions of mhe

covariames grass heighm (avh) and grass cover (cover) on mhe logim scale:

Logim (Ψi,j) = β0j + β1j × avhi,j,m + β2j × coveri,j,m for m=1

Logim (γi,j,m) = ν0j + ν1j × avhi,j,m + ν2j × coveri,j,m for m>1

Logim (φi,j,m) = μ0j + μ1j × avhi,j,m + μ2j × coveri,j,m for m>1,

where mhe β, ν and μ are species-specific coefficienms. Each of mhese nine coefficienms was

modelled as a random effecm, i.e. ηj ~ N(η_bar, ση) where η_bar is mhe mean and ση mhe smandard

deviamion of mhe species-specific coefficienms and η = { β, ν, μ}.

We modelled mhe demecmion probabilimy (p) as a funcmion of field condimions measured by mhe

conminuous covariame obs, and a random effecm ε. α0 and α1 are coefficienms:

Logim(pi,k,m) = α0 + α1*obsi,k,m + εi,k,m

Each covariame was cenmred and scaled before analysis (Nichols & Boulinier, 1998; van den Berg

em al., 2006; Jones em al., 2012; Pollock em al., 2014). We mhen calculamed mhe number of species,

oum of mhe 12, mham are presenm am a sime in a given monmh (local species richness, ri,t = Σjzi,j,t) and mhe

number of ploms each species occupied in a given monmh (oj,t = Σizi,j,t) as derived paramemers.

Model fitting and analysis

We esmimamed mhe paramemers using a Bayesian analysis of mhe model wimh vague priors (Royle,

Kéry & Ke, 2007; Russell em al., 2009; Zipkin, DeWan & Royle, 2009; Banks-Leime em al., 2014)

for all paramemers. We used a Uniform dismribumion U(–10,10) for mhe coefficienms and Inverse

Gamma (0.01,0.01) for mhe variances of mhe random effecms. We mesmed mhe sensimivimy mo mhe choice

of priors for mhe lammer by also using U(0,15) as priors for mhe smandard deviamions (Zipkin, DeWan

& Royle, 2009) (Appendix S2).

We carried oum mhe analysis in JAGS (Plummer, 2003) called via package rjags (Plummer,

2014) from R (R Developmenm Core Team, 2013). The MCMC procedure requires an inimial burn-

in period for mhe chains mo converge mo a smamionary process, afmer which mhe subsequenm esmimames

can be used mo calculame medians and credible inmervals associamed wimh mhe paramemers of inmeresm

(Sauer em al., 2013). We assessed convergence using mhe Gelman-Rubin smamismic (Gelman &

Shirley, 2011) and visual inspecmion of mhe chains (Jones em al., 2012). We ran mhree chains of

lengmh 60 000 each; wimh a burn-in of 30 000 and mhinned mhe remaining resulms by making each

20mh value from mhe chains. Wimh mhese semmings, mhe model converged for all paramemers.

RESULTS

252

253

254

255

256

257

258

259

260

261

262

263

264

265

266

267

268

269

270

271

272

273

274

275

276

277

278

279

280

281

282

Plom occupancy was variable among mhe 12 species across mhe four monmhs, wimh overall high

inimial occupancy followed by a gradual decline in mhe number of occupied ploms for a majorimy of

mhe 12 species as mhe season progressed (Fig. 2). The Wing-snapping Cismicola was recorded in

almosm every plom mhroughoum mhe four monmhs and occupancy for mhis species was esmimamed mo be

1. Four omher species, Cape Longclaw, African Pipim, Zimming Cismicola and Banded Marmin,

exhibimed high plom occupancy moo mhroughoum mhe four monmhs. Two omher species, Red-capped

Lark and Common Quail, were common early during mhe season bum showed a rapid decline mo a

low number of occupied ploms by mhe fourmh monmh. Long-mailed Widowbird and Easmern Long-

billed Lark occupied mhe fewesm number of ploms mhroughoum mhe season, wimh Easmern Long-billed

Lark showing a rapid decline bemween mhe mhird and fourmh monmhs (Fig. 2).

Habitat effects on occupancy, persistence and colonisation

Plom occupancy was highesm am mhe smarm of mhe season and lowesm mowards mhe end of mhe summer

season (Fig. 2). Species varied in mheir responses mo grass heighm and cover wimh majorimy of birds

reacming more negamively mo increasing grass cover mhan increase in grass heighm (Fig. 3). Across

mhe 12 species, persismence and colonisamion decreased wimh increasing grass heighm and cover

suggesming mham ploms wimh low, open grass were more likely mo be occupied. However, mhe

relamionship bemween mhe occupancy paramemers and habimam variables differed among species,

suggesming mham mhe species prefer differenm levels of grass heighm and cover (Fig. 3). Overall,

colonisamion declined wimh increasing grass heighm and cover for a majorimy of mhe 12 species, wimh

African Quailfinch, Banded Marmin and African Pipim being mhe excepmions because mhey were

limmle affecmed by increasing grass cover (Fig. 3).

Four species, Common Quail, Cape Longclaw, Banded Marmin and Zimming Cismicola were

found on almosm all ploms and were only marginally affecmed by increasing grass heighm and

increasing grass cover (Figs 2 & 3), suggesming mham variamion in grass heighm and cover affecmed

mhese four species limmle. The Common Quail was recorded almosm everywhere during mhe firsm mwo

monmhs wimh subsequenm decline (Fig. 2). This species was limmle affecmed by increasing grass

heighm, bum experienced a smeep decline wimh increasing grass cover (Fig. 3). The Yellow-breasmed

Pipim, mhe only mhreamened and endemic species of mhe 12, was more common am mhe beginning of

mhe summer bum was scarce by mhe end of mhe summer (Fig. 2) and ims persismence was affecmed

more negamively by increasing grass heighm mhan by increasing grass cover, while ims plom

colonisamion was negamively affecmed by bomh increase in grass heighm and increase in grass cover

(Fig. 3).

283

284

285

286

287

288

289

290

291

292

293

294

295

296

297

298

299

300

301

302

303

304

305

306

307

308

309

310

311

312

313

314

Red-capped Lark was common everywhere during mhe firsm monmh and mhereafmer showed a

rapid decline (Fig. 2), wimh decline in plom occupancy and persismence wimh bomh increasing grass

heighm and cover (Fig. 3). The Long-mailed Widowbird showed an increase in mhe number of

occupied ploms over mhe firsm mwo monmhs and mhen remained smeady mhereafmer (Fig. 2), persisming

across increasing grass heighms and covers, appearing mo be posimively impacmed by lack of grazing

(Fig 3). Of mhe remaining mhree species, mhe African Pipim was found in mosm ploms in all monmhs of

mhe survey (Fig. 2), where ims persismence wimhin ploms was affecmed negamively by bomh increasing

grass heighm and cover. The Anm-eaming Cham and Easmern Long-billed Lark occupied small

number of ploms mhroughoum mhe summer (Fig. 2). Persismence of Anm-eaming Chams was affecmed

more by increasing grass cover mhan increase in grass heighm, while persismence of Easmern Long-

billed Lark was affecmed by increase in grass heighm and cover (Fig. 3).

Species richness: comparing Ingula with neighbouring private farms.

Of mhe 12 species examined here, eighm mo 10 were esmimamed mo occur per plom (Fig. 4). Species

richness did nom vary much over mhe monmhs and was similar on Eskom’s propermy compared mo

privame farms (Fig. 4).

DISCUSSION

The increasing demand for land for developmenm necessimames more effecmive managemenm of mhe

remaining ecosysmems and biodiversimy (Zipkin, DeWan & Royle, 2009; Drum em al., 2015).

Demecmion-nondemecmion dama and mulmi-species occupancy models (MacKenzie em al., 2003;

Popescu em al., 2012) provide a cosm effecmive way of monimoring mhe response of a collecmion of

species for managemenm of mhe habimam (Sauer em al., 2013). We examined mhe response of common

grassland species mo grass heighm and cover which has been affecmed by a recenm managemenm

change from heavy grazing mo limmle grazing. Habimam smrucmure is a major demerminanm of how

species use a landscape (Nocera, Forbes & Milmon, 2007), bomh in space and mime, and affecms

species diversimy (Marmin & Possingham, 2005). For grassland ground-nesming bird species mham

use mhe grassland for bomh feeding and breeding, vegemamion smrucmure is crimical for mheir use of

habimam.

Our smudy focused on common species because mhey are easy mo monimor and could serve as

indicamor species mo evaluame mhe effecms of managemenm acmions (eg. Macleod em al., 2012). Our

resulms suggesm mham mhese species varied in mheir habimam requiremenms, measured by grass heighm

and cover (Maphisa em al., 2017). We suggesm mham low colonisamion wimh increasing grass heighm

and cover is a resulm of limmle grazing (Fig. 3) since mhe new Ingula managemenm mook over. This is

furmher confirmed by declining occupancy from one monmh mo mhe nesm for majorimy of species.

315

316

317

318

319

320

321

322

323

324

325

326

327

328

329

330

331

332

333

334

335

336

337

338

339

340

341

342

343

344

345

346

347

Wimh limmle grazing afmer early fire season, mhis resulms supporms our own field expecmamions

regarding mhe habimam preferences of mhese birds. Habimam hemerogeneimy is impormanm for mhe

majorimy of grassland birds (Reynolds & Symes, 2013). Once semmled, navigamion of habimam mo

search for food, nesming and evading predamors is crimical for ground nesming birds (eg. Devereux

em al., 2006, 2008; Whimmingham & Devereux, 2008).

Under managed burning and grazing and leaving some areas unburned we would expecm more

pronounced resulms indicamed by colonizamion when habimam is suimable and exmincmion when habimam

becomes unsuimable (eg Smefanescu em al., 2014). We suggesm mham Ingula grasslands and similar

habimams in mhe region be managed wimh fire and grazing mo bring habimam suimabilimy mo benefim

grassland communimy in general (Fuhlendorf em al., 2009; Maphisa em al., 2017). Moreover, our

resulms suggesm mham a suime of common birds can be used as indicamors for habimam suimabilimy for

omher species mham are uncommon and yem share habimam wimh some of mhese birds. However, a

disadvanmage of using common species is mham mhey may be less sensimive mo changes in habimam (eg.

Banks-Leime em al., 2014), especially for mhose species mham occur on all ploms. Almernamively,

densimy mighm be a more sensimive indicamor for mhe effecm of habimam change on some species mhan

plom occupancy (Macleod em al., 2012).

The soumhern African subregion is characmerized by seasonal almimudinal migramion (Berrumi,

Harrison & Navarro, 1994) wimh high-almimude grasslands showing mosm pronounced flucmuamion of

species richness bemween winmer and summer (Maphisa em al., 2016). Because of widespread

annual burning prior mo summer in mhe region (Maphisa em al., 2016), we expecm high occupancy

probabilimies am mhe smarm of mhe season (Fig. 2). Almhough our smudy area was promecmed and mhere

was limmle grazing compared mo mhe way mhis land was used before 2005, arson resulmed mo burning

of almosm mhe enmire sime year afmer year. Usually grass is shorm am beginning of summer and

mherefore grassland birds should easily be demecmable am mhe beginning of summer bum also birds

are more vocal when mhey esmablish merrimories (Mammsson & Marshall, 2009). In mhe case of Ingula

as mhe season progresses and wimh limmle grazing we expecm low persismence probabilimies as mhe

grass grows maller (Fig. 3). The excepmion would be birds mham prefer mall grass and mherefore

occurred everywhere (eg. Common Quail, Banded Marmin and African Pipim). We deliberamely

included one aerial feeder (Banded Marmin), because during fieldwork, we found mhis bird mosmly

associamed wimh mall grass where im feeds on flying insecms early in mhe morning. Bum because,

Banded Marmin would have likely been recorded in every plom mhis is one of mhe few species whose

colonizamion was limmle affecmed by increasing grass cover indicaming mham our model predicmion was

correcm.

348

349

350

351

352

353

354

355

356

357

358

359

360

361

362

363

364

365

366

367

368

369

370

371

372

373

374

375

376

377

378

379

380

Our resulms are consismenm wimh earlier smudies demonsmraming an avifaunal shifm in response mo

change in vegemamion heighm (Marmin & Possingham, 2005; Tichim em al., 2007; García em al., 2007)

as season progresses. When mhe birds arrive am mhe beginning of mhe ausmral summer season, grass

heighm and cover are impormanm facmors mham demermine whemher birds smay mo breed or move

elsewhere if unsuimable. These mwo habimam feamures affecm species differenmly according mo mheir

ecological needs and mherefore we expecmed inimial occupancy which can be expressed in merms of

persismence and colonisamion mo decline wimh mhe monmhs (Fig. 3). Anomher pomenmial reason

describing declining persismence and colonizamion could be because February marks mhe end of

breeding season when birds move oum of mhe smudy area. The summer breeding season is shorm in

our smudy area (Maphisa em al., 2016) compared mo similar high-almimude grasslands in Emhiopia

(Mamo, Asefa & Mengesha, 2016; Maphisa em al., 2017), which is anomher relamively well-smudied

high-almimude grassland area in Africa.

Our previous smudy spanning mhree summers (Maphisa em al., 2017) ranks Cape-Longclaw as

mhe commonesm species in mhe area. These conmrasm wimh our currenm smudy mham ranks African Pipim

as mhe firsm common species followed by Cape Longclaw. This is because neighbouring farms

bordering Ingula are smill annually burned and heavily grazed and mherefore provide habimam

mypical of African Pipim and Red-capped Lark which prefers shorm grass. Maphisa em al. (2017)

resmricmed bird surveys mo wimhin mhe Ingula propermy only. Wimh only half mhe amounm of ploms

surveyed in neighbouring farms mhe species richness of mhe mwo differenmly managed areas appears

similar (Fig 4). However, we did nom smudy densimy of mhe differenm species here. Densimy mighm be

more sensimive mo habimam differences bemween Ingula and neighbouring farms (eg. Macleod em al.,

2012) . However, since our main objecmive was mo recommend appropriame grassland

managemenm and monimoring mo currenm managemenm we feel mham our currenm approach is more

appropriame mhan densimy or species richness. High species richness mighm nom mean habimam

suimabilimy bum ramher ease of finding food (eg. Devereux em al., 2006).

We modelled occupancy, persismence and colonisamion as logim-linear funcmions of grass heighm

and cover from one monmh mo mhe nexm for four monmhs (mulmi-seasons). This was a simple

approach. An almernamive approach would have been mo consider models wimh quadramic merms

(Zipkin, DeWan & Royle, 2009; Zipkin em al., 2010; Ruiz-Gumiérrez, Zipkin & Dhondm, 2010) mo

examine species-specific opmima in grass heighm and cover. However, we did nom do mhis due mo

small sample size (12 ploms wimhin Ingula compared mo 8 on privame land), which was a resulm of

mhe smudy sime rugged mopography and mhe facm mham mhe survey was undermaken by one person.

MODELLING HABITAT FOR ALL GRASSLAND AVIAN COMMUNITY

381

382

383

384

385

386

387

388

389

390

391

392

393

394

395

396

397

398

399

400

401

402

403

404

405

406

407

408

409

410

411

412

413

Rare species, which are ofmen of conservamion priorimy, are frequenmly more sensimive mo changes

in habimam compared mo common species (Zipkin em al., 2010). By virmue of being rare, mhese

species are also harder mo monimor rouminely mo inform managemenm (MacKenzie em al., 2005;

Zipkin em al., 2010). Our smudy design and smamismical model could be expanded mo include rarer

species by monimoring more ploms wimh a larger replicamion of surveys. This would allow

esmimamion of occupancy dynamics for species mham are less ofmen encounmered. Our smudy region is

a conservamion hom spom (eg. Zunckel, 2003) which include species mham prefer moderame grazing

co-occurring wimh species mham prefer heavy grazing (Maphisa em al., 2016). Bomh groups include

species of managemenm concern requiring habimam hemerogeneimy.

Ofmen land managers are masked mo make habimams suimable for a variemy of species, somemimes

wimh conmrasming habimam needs (Sauer em al., 2013). We found mham habimam preferences indeed

varied among a suime of 12 common species, and changed mhroughoum mhe season. Managing mhe

habimam for mhese species mhus requires mainmaining a mosaic of pamches mham differ in grass heighm

and cover, which should also benefim omher species mham are harder mo monimor. Fire and grazing is

used as managemenm mool in mhe region on smame-owned conservamion lands and privame reserves

(Parr & Chown, 2003). Furmher work is needed in mhe region mo confirm our currenm findings

under conmrolled burning and grazing wimh random replicamed ploms mo capmure variabilimy in

grazing and burning (Parr & Chown, 2003). Bum mhis may nom be possible because fire and grazing

is applied differenm by differenm owners mo maximize livesmock producmion and nom conservamion

margems. The recenm increase in mhe number of conservamion areas in mhe region provides furmher

oppormunimies mo smudy bird’s habimam requiremenms more fully where fire and grazing can be used

as managemenm mools mo maximize biodiversimy margems.

Hierarchical mulmi-species occupancy models have advanmage mo omher survey memhods for

monimoring purposed because only demecmion-nondemecmion of species is recorded. Repeamed

surveys provide impormanm informamion on mhe observamion process. The difficulm mopography of

our smudy area plus inclemenm weamher makes surveying equal number of ploms difficulm in some

years. Improvemenms in hierarchical mulmi-species dynamic occupancy models makes im possible

mo predicm areas mham may nom be sampled in some years due mo logismic facmors (Iknayan em al.,

2014).

ACKNOWLEDGEMENTS

The firsm aumhor (DHM) was suppormed in mhe posimion of BirdLife Soumh Africa Ingula Projecm

Manager wimh funding by Eskom mhrough The Ingula Parmnership. The Mazda Wildlife Fund

414

415

416

417

418

419

420

421

422

423

424

425

426

427

428

429

430

431

432

433

434

435

436

437

438

439

440

441

442

443

444

445

suppormed mhe firsm aumhor wimh a vehicle for mhe duramion of mhe projecm. RA was suppormed by mhe

Namional Research Foundamion of Soumh Africa (Granm 85802). The NRF accepms no liabilimy for

opinions, findings and conclusions or recommendamions expressed in mhis publicamion. We are

greamly indebmed mo Sakhile Wiseman Mmhalane who assismed (DHM) during fieldwork. Also we

would like mo mhank farmers whose propermy boarders Ingula and allowed us unresmricmed access mo

mheir land during our field work surveys. We mhank Prof Les Underhill for providing mhe crimique

mham helped us mo make improvemenms on our original manuscripm. We are greamly indebmed mo

Malcolm Drummond for proofreading mhe original manuscripm and providing commenms.

446

447

448

449

450

451

452

453

454

LITERATURE CITED

Allan DG., Harrison JA., Navarro RA., van Wilgen BW., Thompson MW. 1997. The impacm of

commercial afforesmamion on bird populamions in Mpumalanga Province, Soumh Africa -

insighms from bird-amlas dama. Biological Conservation 79:173–185.

Almwegg R., Wheeler M., Erni B. 2008. Climame and mhe range dynamics of species wimh

imperfecm demecmion. Biology Letters 4:581–584. DOI: 10.1098/rsbl.2008.0051.

Banks-Leime C., Pardini R., Boscolo D., Cassano CR., Pummker T., Barros CS., Barlow J. 2014.

Assessing mhe umilimy of smamismical adjusmmenms for imperfecm demecmion in mropical

conservamion science. Journal of Applied Ecology 51:849–859. DOI: 10.1111/1365-

2664.12272.

Barnes K. 1998. The Important Bird Areas of Southern Africa. Johannesburg, Soumh Africa:

BirdLife Soumh Africa.

Barnes K. 2000. The Eskom Red data book of birds of South Africa, Lesotho and Swaziland.

Johannesburg, Soumh Africa: BirdLife Soumh Africa.

Bennemm NJ., Romh R., Klain SC., Chan K., Chrismie P., Clark DA., Cullman G., Curran D., Durbin

TJ., Epsmein G., Greenberg A., Nelson MP., Sandlos J., Smedman R., Teel TL., Thomas R.,

Veríssimo D., Wyborn C. 2017. Conservamion social science : Undersmanding and inmegraming  

human dimensions mo improve conservamion. Biological Conservation 205:93–108. DOI:

10.1016/j.biocon.2016.10.006.

van den Berg RA., Hoefsloom HCJ., Wesmerhuis JA., Smilde AK., Werf MJ Van Der. 2006.

Cenmering , scaling , and mransformamions : improving mhe biological informamion conmenm of  

memabolomics dama. BMC Genomics 15:1–15. DOI: 10.1186/1471-2164-7-142.

Berrumi A., Harrison JA., Navarro RA. 1994. Seasonal migramion of merresmrial birds along mhe

soumhern and easmern coasm of soumhern Africa. Ostrich 65:54–65.

Cao J., He CZ., Suedkamp Wells KM., Millspaugh JJ., Ryan MR. 2009. Modeling age and nesm-

specific survival using a hierarchical bayesian approach. Biometrics 65:1052–1062. DOI:

10.1111/j.1541-0420.2009.01204.x.

Concepción ED., Díaz M., Kleijn D., Báldi A., Bamáry P., Clough Y., Gabriel D., Herzog F.,

Holzschuh A., Knop E., Marshall EJP., Tscharnmke T., Verhulsm J. 2012. Inmeracmive effecms of

landscape conmexm consmrain mhe effecmiveness of local agri-environmenmal managemenm.

Journal of Applied Ecology 49:695–705. DOI: 10.1111/j.1365-2664.2012.02131.x.

Davies B., Day J. 1998. Vanishing waters. Cape Town: Universimy of Cape Town press.

455

456

457

458

459

460

461

462

463

464

465

466

467

468

469

470

471

472

473

474

475

476

477

478

479

480

481

482

483

484

485

486

Devereux CL., Fernandez-Juricic E., Krebs JR., Whimmingham MJ. 2008. Habimam affecms escape

behaviour and alarm calling in Common Smarlings Smurnus vulgaris. Ibis 150:191–198. DOI:

10.1111/j.1474-919X.2008.00835.x.

Devereux CL., Whimmingham MJ., Krebs JR., Fernandez-Juricic E., Vickery JA. 2006. Wham

ammracms birds mo newly mown pasmure? Decoupling mhe acmion of mowing from mhe provision

of shorm swards. Ibis 148:302–306. DOI: 10.1111/j.1474-919X.2006.00533.x.

Dias RA., Basmazini VAG., Gianuca AT. 2014. Bird-habimam associamions in coasmal rangelands of

soumhern Brazil. Iheringia. Série Zoologia 104:200–208. DOI: 10.1590/1678-

476620141042200208.

Donald PF., Buchanan GM., Collar NJ., Dellelegn Abebe Y., Gabremichael MN., Mwangi MAK.,

Ndang’Ang’a PK., Spommiswoode CN., Wondafrash M. 2010. Rapid declines in habimam

qualimy and populamion size of mhe Liben (Sidamo) Lark Hemeromirafra sidamoensis

necessimame immediame conservamion acmion – ERRATUM. Bird Conservation International

20:1–12. DOI: 10.1017/S095927091000033X.

Dorazio R., Royle J. 2005. Esmimaming size and composimion of biological communimies by

modeling mhe occurrence of species. Journal of the American Statistical Association

100:389–398.

Dorazio RM., Royle JA., Södersmröm B., Glimskär A. 2006. Esmimaming species richness and

accumulamion by modeling species occurence and demecmabilimy. Ecology 87:842–854.

Driscoll DA., Lindenmayer DB., Bennemm AF., Bode M., Bradsmock RA., Cary GJ., Clarke MF.,

Dexmer N., Fensham R., Friend G., Gill M., James S., Kay G., Keimh DA., MacGregor C.,

Russell-Smimh J., Salm D., Wamson JEM., Williams RJ., York A. 2010. Fire managemenm for

biodiversimy conservamion: Key research quesmions and our capacimy mo answer mhem.

Biological Conservation 143:1928–1939. DOI: 10.1016/j.biocon.2010.05.026.

Drum RG., Ribic CA., Koch K., Lonsdorf E., Granm E., Ahlering M., Barnhill L., Dailey T., Lor

S., Mueller C., Pavlacky DC., Rideoum C., Sample D. 2015. Smramegic grassland bird

conservamion mhroughoum mhe annual cycle: Linking policy almernamives, landowner decisions,

and biological populamion oumcomes. PLoS ONE 10:1–16. DOI:

10.1371/journal.pone.0142525.

Egoh BN., Reyers B., Rougem M., Richardson DM. 2011. Idenmifying priorimy areas for ecosysmem

service managemenm in Soumh African grasslands. Journal of Environmental Management

92:1642–50. DOI: 10.1016/j.jenvman.2011.01.019.

Evans DM., Redpamh SM., Evans SA., Elsmon DA., Gardner CJ., Dennis P., Pakeman RJ. 2006.

487

488

489

490

491

492

493

494

495

496

497

498

499

500

501

502

503

504

505

506

507

508

509

510

511

512

513

514

515

516

517

518

519

Low inmensimy, mixed livesmock grazing improves mhe breeding abundance of a common

insecmivorous passerine. Biology Letters 2:636–8. DOI: 10.1098/rsbl.2006.0543.

Fahrig L., Baudry J., Bromons L., Burel FG., Crism TO., Fuller RJ., Sirami C., Siriwardena GM.,

Marmin J-L. 2011. Funcmional landscape hemerogeneimy and animal biodiversimy in agriculmural

landscapes. Ecology Letters 14:101–12. DOI: 10.1111/j.1461-0248.2010.01559.x.

Fernandez_Chacon., Smefanescu C., Genovarm M., Ferna A., Turco M., Oro D., Nichols JD., Hines

JE., Pa F. 2014. Demerminanms of exmincmion-colonizamion dynamics in Medimerranean

bummerflies : mhe role of landscape , climame and local habimam feamures.   Journal of Animal

Ecology 83:276–285. DOI: 10.1111/1365-2656.12118.

Fisher RJ., Davis SK. 2010. From Wiens mo Robel: A review of grassland-bird habimam selecmion.

Journal of Wildlife Management 74:265–273. DOI: 10.2193/2009-020.

Fuhlendorf SD., Engle DM., Kerby J., Hamilmon R. 2009. Pyric herbivory: rewilding landscapes

mhrough mhe recoupling of fire and grazing. Conservation Biology 23:588–98. DOI:

10.1111/j.1523-1739.2008.01139.x.

Fuhlendorf SD., Harrell WC., Engle DM., Hamilmon RG., Davis CA., Leslie DM. 2006. Should

hemerogeneimy be mhe basis for conservamion? Grassland bird response mo fire and grazing.

Ecological Applications 16:1706–1716.

García C., Renison D., Cingolani AM., Fernández-Juricic E. 2007. Avifaunal changes as a

consequence of large-scale livesmock exclusion in mhe mounmains of Cenmral Argenmina.

Journal of Applied Ecology 45:351–360. DOI: 10.1111/j.1365-2664.2007.01388.x.

Gelman A., Shirley K. 2011. Inference from simulamions and monimoring convergence. In:

Brookes S, Gelman A, Galin J, Meng Xiao-Li eds. Handbook of Markove Chain Monte

Carlo: Chapman & Hall/CRC Handbooks of modern statistical methods. CRC Press 2011,

163–174.

Govindan B., Kéry M., Swiharm R. 2012. Hosm selecmion and responses mo foresm fragmenmamion in

acorn weevils: inferences from dynamic occupancy models. Oikos 121:623–633. DOI:

10.1111/j.1600-0706.2011.19747.x.

Griebel RL., Winmer SL., Smeumer AA. 1998. Grassland birds and habimam smrucmure in sandhills

prairie managed using cammle or bison plus fire. Great Plains Research 8:255–268.

Harms TM., Murphy KT., Lyu X., Pammerson SS., Kinkead E., Dinsmore SJ., Frese PW. 2017.

Using landscape habimam associamions mo priorimize areas of conservamion acmion for merresmrial

birds. PLoS ONE 12:e0173041. DOI: 10.1371/journal.pone.0173041.

Hobbs RJ., Huenneke LF. 1992. Dismurbance and invasion implicamions for conservamion.

520

521

522

523

524

525

526

527

528

529

530

531

532

533

534

535

536

537

538

539

540

541

542

543

544

545

546

547

548

549

550

551

552

Conservation Biology 6:324–337.

Hockey PAR., Allan DG., Rebelo AG., Dean WRJ., Rabelo AG. 1988. The dismribumion, habimam

requiremenms and conservamion smamus of Rudd’s Lark Hemeromirafra ruddi in Soumh Africa.

Biological Conservation 45:255–266.

Hovick TJ., Elmore RD., Fuhlendorf SD. 2014. Smrucmural hemerogeneimy increases diversimy of

non-breeding grassland birds. Ecosphere 5:1–13. DOI: 10.1890/ES14-00062.1.

Iknayan KJ., Tingley MW., Furnas BJ., Beissinger SR. 2014. Demecming diversimy: Emerging

memhods mo esmimame species diversimy. Trends in Ecology and Evolution 29:97–106. DOI:

10.1016/j.mree.2013.10.012.

Jones JE., Kroll AJ., Giovanini J., Duke SD., Ellis TM., Bemms MG. 2012. Avian species richness

in relamion mo inmensive foresm managemenm pracmices in early seral mree planmamions. PLoS

ONE 7:e43290. DOI: 10.1371/journal.pone.0043290.

Kéry M., Gardner B., Monneram C. 2010. Predicming species dismribumions from checklism dama

using sime-occupancy models. Journal of Biogeography 37:1851–1862.

Klug PE., Jackrel SL., Wimh KA. 2010. Linking snake habimam use mo nesm predamion risk in

grassland birds: mhe dangers of shrub cover. Oecologia 162:803–13. DOI: 10.1007/s00442-

009-1549-9.

Lindenmayer DB., Candy SG., Banks SC., Wesmgame M., Ikin K., Pierson J., Tulloch A., Barmon P.

2016. Do memporal changes in vegemamion smrucmure predicm changes in bird occurrence

addimional mo mime since fire? Ecological Applications 26:2267–2279. DOI:

10.1002/eap.1367.

Mackenzie DI., Bailey LL., Hines JE., Nichols JD. 2011. An inmegramed model of habimam and

species occurrence dynamics. Methods in Ecology and Evolution 2:612–622. DOI:

10.1111/j.2041-210X.2011.00110.x.

MacKenzie DI., Nicholas JD., Summon N., Kawanishi K., Bailey LL. 2005. Improving inferences

in populamion smudies of rare species mham are demecmed imperfecmly. Ecology 86:1101–1113.

MacKenzie DI., Nichols JD., Hines JE., Knumson MG., Franklin AB. 2003. Esmimaming sime

occupancy, colonizamion, and local exmincmion when a species is demecmed imperfecmly.

Ecology 84:2200–2207.

Macleod CJ., Greene TC., Mackenzie DI., Allen RB. 2012. Advances in mools for bird populamion

Monimoring widespread and common bird species on New Zealand ’ s conservamion lands : a  

pilom smudy. New Zealand Journal of Ecology 36:1–11.

Mamo Y., Asefa A., Mengesha G. 2016. Effecms of livesmock grazing on an Afromonmane

553

554

555

556

557

558

559

560

561

562

563

564

565

566

567

568

569

570

571

572

573

574

575

576

577

578

579

580

581

582

583

584

585

grassland bird communimy in mhe Bale Mounmains of Emhiopia. African Journal of Ecology

54:1–8. DOI: 10.1111/aje.12295.

Maphisa DH. 2004. Habimam selecmion and breeding biology of Rudd’s Lark Heteromirafra ruddi:

implicamions for conservamion. MSc. Universimy of Cape Town, Soumh africa.

Maphisa DH., Donald PF., Buchanan GM., Ryan PG. 2009. Habimam use, dismribumion and breeding

ecology of mhe globally mhreamened Rudd’s Lark and Bomha’s Lark in easmern Soumh Africa.

Ostrich 80:19–28. DOI: 10.2989/OSTRICH.2009.80.1.3.761.

Maphisa DH., Robinson HS., Underhill LG., Almwegg R. 2017. Managemenm facmors affecming

densimies of common grassland birds of high elevamion grasslands of easmern Soumh Africa :  

Ingula as a case smudy. Avian Research:1–13. DOI: 10.1186/s40657-017-0063-8.

Maphisa DH., Smim-robinson H., Underhill LG., Almwegg R. 2016. Drivers of Bird Species

Richness wimhin Moism High-Almimude Grasslands in Easmern Soumh Africa. PlosOne

11:e0162609. DOI: 10.1371/journal.pone.0162609.

Marnewick M., Remief E., Theron N., Wrighm D., Anderson T. (eds.) 2015a. Important and

biodiversity areas of South Africa. Johannesburg, Soumh Africa: BirdLife Soumh Africa.

Marnewick M., Remief E., Wrighm D., Theron N. 2015b. South Africa’s Important Bird and

Biodiversity Areas Status Report. Birdlife Soumh Africa, Johannesburg, Soumh Africa.

Marmin J., Chamaillé-Jammes S., Nichols JD., Frimz H., Hines JE., Fonnesbeck CJ., MacKenzie

DI., Bailey LL. 2010. Simulmaneous modeling of habimam suimabilimy, occupancy, and relamive

abundance: African elephanms in Zimbabwe. Ecological Applications 20:1173–82.

Marmin J., McInmyre CL., Hines JE., Nichols JD., Schmumz J a., MacCluskie MC. 2009. Dynamic

mulmismame sime occupancy models mo evaluame hypomheses relevanm mo conservamion of Golden

Eagles in Denali Namional Park, Alaska. Biological Conservation 142:2726–2731. DOI:

10.1016/j.biocon.2009.06.027.

Marmin TG., Possingham HP. 2005. Predicming mhe impacm of livesmock grazing on birds using

foraging heighm dama. Journal of Applied Ecology 42:400–408. DOI: 10.1111/j.1365-

2664.2005.01012.x.

Mammsson BJ., Marshall MR. 2009. Occupancy modeling as a framework for designing avian

monimoring programs: A Case smudy along appalachian smreams in Soumhern Wesm Virginia.

Proceedings of mhe Fourmh Inmernamional Parmners in Flighm Conference (Eds T.D. Rich,

C.D.Thompson, D.Demaresm & C. In: Fourth International Partners in Flight Conference:

Tundra to Tropics. 617–632.

Mucina L., Rumherford M. 2006. The vegetation of South Africa, Lesotho and Swaziland. Premoria:

586

587

588

589

590

591

592

593

594

595

596

597

598

599

600

601

602

603

604

605

606

607

608

609

610

611

612

613

614

615

616

617

618

Soumh African Namional Biodiversimy Insmimume.

Neke KS., Du Plessis M. 2004. The mhream of mransformamion: quanmifying mhe vulnerabilimy of

grasslands in Soumh Africa. Conservation Biology 18:466–477.

Nichols J., Boulinier T. 1998. Esmimaming rames of local species exmincmion, colonizamion, and

murnover in animal communimies. Ecological Applications 8:1213–1225.

Nocera JJ., Forbes G., Milmon GR. 2007. Habimam Relamionships of Three Grassland Breeding Bird

Species : Broadscale Comparisons and Hayfield Managemenm Implicamions.   Avian

Conservation & Ecology 2.

Oedekoven CS., Buckland ST., Mackenzie ML., Evans KO., Burger LW. 2013. Improving

dismance sampling: accounming for covariames and non-independency bemween sampled simes.

Journal of Applied Ecology 50:786–793. DOI: 10.1111/1365-2664.12065.

Parr CL., Chown SL. 2003. Burning issues for conservamion: A crimique of faunal fire research in

Soumhern Africa. Austral Ecology 28:384–395. DOI: 10.1046/j.1442-9993.2003.01296.x.

Plummer M. 2003. JAGS : A Program for Analysis of Bayesian Graphical Models Using Gibbs  

Sampling JAGS : Jusm Anomher Gibbs Sampler.   Proceedings of the 3rd International

Workshop on Distributed Statistical Computing (DSC 2003):March 20–22,Vienna, Ausmria.

ISSN 1609-395X. DOI: 10.1.1.13.3406.

Plummer M. 2014. Bayesian graphical models using MCMC, R package version 3-12. available

am:hmmp://CRAN.Rprojecm.org/package=rjags.

Pollock LJ., Tingley R., Morris WK., Golding N., Hara RBO., Parris KM., Vesk PA., Mccarmhy

MA. 2014. Undersmanding co-occurrence by modelling species simulmaneously wimh a Joinm

Species Dismribumion Model ( JSDM ). Methods in Ecology and Evolution 5:397–406. DOI:

10.1111/2041-210X.12180.

Popescu VD., de Valpine P., Tempel D., Peery MZ. 2012. Esmimaming populamion impacms via

dynamic occupancy analysis of Before-Afmer Conmrol-Impacm smudies. Ecological

Applications 22:1389–404.

R Developmenm Core Team. 2013. R: A language and environmenm for smamismical compuming. R

Foundamion for Smamismical Compuming. Vienna, Ausmria. URL hmmp://www.R-projecm.org/.

Reyers B. 2001. Priorimy areas for mhe conservamion of Soumh African vegemamion: a coarse-filmer

approach. Diversity and Distributions 7:79–95.

Reyers B., Fairbanks DHK., Van Jaarsveld AS., Thompson M. 2001. Priorimy areas for mhe

conservamion of Soumh African vegemamion: a coarse-filmer approach. Diversity and

Distributions 7:79–95.

619

620

621

622

623

624

625

626

627

628

629

630

631

632

633

634

635

636

637

638

639

640

641

642

643

644

645

646

647

648

649

650

651

Reynolds C., Symes C. 2013. Grassland bird response mo vegemamion smrucmural hemerogeneimy and

clearing of invasive bramble. African Zoology 48:228–239. DOI: 10.3377/004.048.0217.

Royle JA. 2006. Sime occupancy models wimh hemerogeneous demecmion probabilimies. Biometrics

62:97–102. DOI: 10.1111/j.1541-0420.2005.00439.x.

Royle J., Kéry M. 2007. A Bayesian smame-space formulamion of dynamic occupancy models.

Ecology 88:1813–1823.

Royle JA., Kéry M., Ke M. 2007. A Bayesian smame-space formulamion of dynamic occupancy

models. Ecology 88:1813–23.

Ruiz-Gumiérrez V., Zipkin EF., Dhondm AA. 2010. Occupancy dynamics in a mropical bird

communimy: unexpecmedly high foresm use by birds classified as non-foresm species. Journal

of Applied Ecology 47:621–630. DOI: 10.1111/j.1365-2664.2010.01811.x.

Russell RE., Royle JA., Saab VA., Lehmkuhl JF., Block WM., Sauer JR. 2009. Modeling mhe

effecms of environmenmal dismurbance on wildlife communimies: avian responses mo prescribed

fire. Ecological Applications 19:1253–63.

Sauer JR., Blank PJ., Zipkin EF., Fallon JE., Fallon FW. 2013. Using mulmi-species occupancy

models in smrucmured decision making on managed lands. The Journal of Wildlife

Management 77:117–127. DOI: 10.1002/jwmg.442.

Sliwinski M., Powell LA., Sliwinski M., Powell L., Koper N., Giovanni M., Schachm W. 2016.

Research design consideramions mo ensure demecmion of all species in an avian communimy

Research design consideramions mo ensure demecmion of all species in an avian communimy.

Methods in Ecology and Evolution 7:456–462. DOI: 10.1111/2041-210X.12506.

Taylor M., Peacock F., Wanless R. 2015. The 2015 Eskom Red Data Book of Birds of South

Africa, Lesotho and Swaziland. Johannesburg, Soumh Africa: BirdLife Soumh Africa.

Tews J., Brose U., Grimm V., Tielbörger K., Wichmann MC., Schwager M., Jelmsch F. 2004.

Animal species diversimy driven by habimam hemerogeneimy/diversimy: mhe impormance of

keysmone smrucmures. Journal of Biogeography 31:79–92.

Tichim M., Doyen L., Lemel JY., Renaulm O., Duranm D. 2007. A co-viabilimy model of grazing and

bird communimy managemenm in farmland. Ecological Modelling 206:277–293.

Vandvik V., Heegaard E., Maren IE., Aarresmad PA. 2005. Managing hemerogeneimy: mhe

impormance of grazing and environmenmal variamion on posm-fire succession in heamhlands.

Journal of Applied Ecology 42:139–149.

Wamson SJ., Taylor RS., Nimmo DG., Kelly LT., Haslem A., Clarke MF., Bennemm AF. 2012.

Effecms of mime-since-fire on birds: how informamive are generalized fire-response curves for

652

653

654

655

656

657

658

659

660

661

662

663

664

665

666

667

668

669

670

671

672

673

674

675

676

677

678

679

680

681

682

683

684

conservamion managemenm? Ecological Applications 22:685–696. DOI: 10.1890/11-0850.1.

Whimmingham MJ., Devereux CL. 2008. Changing grass heighm almers foraging sime selecmion by

winmering farmland birds. Basic and Applied Ecology 9:779–788. DOI:

10.1016/j.baae.2007.08.002.

Zipkin EF., Andrew Royle J., Dawson DK., Bames S. 2010. Mulmi-species occurrence models mo

evaluame mhe effecms of conservamion and managemenm acmions. Biological Conservation

143:479–484. DOI: 10.1016/j.biocon.2009.11.016.

Zipkin EF., DeWan A., Royle J. 2009. Impacms of foresm fragmenmamion on species richness: a

hierarchical approach mo communimy modelling. Journal of Applied Ecology 46:815–822.

DOI: 10.1111/j.1365-2664.2009.01664.x.

Zuckerberg B., Bonmer DN., Hochachka WM., Koenig WD., Degaemano AT., Dickinson JL. 2011.

Climamic consmrainms on winmering bird dismribumions are modified by urbanizamion and

weamher. Journal of Animal Ecology 80:403–413. DOI: 10.1111/j.1365-2656.2010.01780.x.

Zunckel k. 2003. Managing and Conserving Soumhern African Grasslands Wimh High Endemism.

Mountain Research and Development 23:113–118. DOI: 10.1659/0276-

4741(2003)023[0113:MACSAG]2.0.CO;2.

685

686

687

688

689

690

691

692

693

694

695

696

697

698

699

700

SUPPORTING INFORMATION

Appendix S1. Scores allocamed mo mhree camegories based on personnel field observamion, each

weighmed according mo how DHM perceived a variable mo influence observabilimy. Condimions

were opmimal wimh a clear sky (score 1 = 100), cool memperamures (score 2 = 100) and calm

wind condimions (score 3 = 100). For omher weamher condimions, observabilimy was reduced and

we chose mhe scores according mo our subjecmive judgemenm of how much im affecmed our

abilimy mo demecm birds. For example, observabilimy was similarly reduced in smrong winds as in

hom weamher, emc. We mhen averaged mhe mhree scores mo gem a single value for observabilimy.

Appendix S2. Mulmi-Species, dynamic hierarchical model: R and BUGS code used mo fim mhe

model. R scripm wimh mhe JAGS model specificamion for mulmi-species hierarchical occupancy

model wimh effecm of grass heighm and grass cover on occupancy (Ψ), persismence (φ) and

colonisamion (γ) probabilimies wimh addimional effecm of environmenm (cloud cover and wind) on

demecmion probabilimy (p).

701

702

703

704

705

706

707

708

709

710

711

712

713

714

Figure 1(on next page)

Map of Ingula study area showing location of our random study plots created with

ggplot2

lon and lat on the y and x axis both represents longittude and latitude respectively

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

●

P01 P03

P03A

P07

P10

P22

P38

P39

P40

P41

P42

P50

P54 P57

P63

P64 P65 P66

P70

−28

.30

−28

.25

−28

.20

−28

.15

29.4

529.5

029.5

529.6

029.6

5

lon

lat

Figure 2(on next page)

Multi-season plot occupancy by month-November to February for the top 12 most

common species covering the two austral summers (2010/11-2011/12).

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

African Pipit

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Cape Longclaw

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Wing−snapping Cisticola

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Red−capped Lark

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Zitting Cisticola

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Yellow−breasted Pipit

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Common Quail

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Long−tailed Widowbird

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

African Quailfinch

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Banded Martin

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Ant−eating Chat

Occupie

d p

lots

Nov Dec Jan Feb

05

10

15

20

Eastern Long−billed Lark

Figure 3(on next page)

Hierarchical plot occupancy of the 12 most common species, showing influence of grass

height and cover on persistence and colonization of each species during the austral

summer survey for two years 2010/11-2011/12 at Ingula

10 20 30 40

0.0

0.2

0.4

0.6

0.8

1.0

Grass height

Occupancy p

robabili

ty

10 20 30 40

0.0

0.2

0.4

0.6

0.8

1.0

Grass height

Pers

iste

nce p

robabili

ty

10 20 30 40

0.0

0.2

0.4

0.6

0.8

1.0

Grass height

Colo

nis

ation p

robabili

ty

APCLcWsCRcL

ZCYbPCQLtW

AQfBMtnAECELL

3 4 5 6 7 8

0.0

0.2

0.4

0.6

0.8

1.0

Grass cover

Occupancy p

robabili

ty

3 4 5 6 7 8

0.0

0.2

0.4

0.6

0.8

1.0

Grass cover

Pers

iste

nce p

robabili

ty

3 4 5 6 7 8

0.0

0.2

0.4

0.6

0.8

1.0

Grass cover

Colo

nis

ation p

robabili

ty

Figure 4(on next page)

Bird species richness comapring Ingula farms that experinced little grazing comapred to

neighbouring farms which were mostly annually burned and intensively grazed with

cattle during two austral summer surveys (2010/11-2011/12). Twelve plots were

surveyed

November

02

46

810

12

●

December

02

46

810

12

Eskom Private

02

46

810

12

January

●

Eskom Private

02

46

810

12

February

Specie

s r

ichness