Mtui Devolent R.Pdf

EVALUATING LANDSCAPE AND WILDLIFE CHANGES OVER TIME IN

TANZANIA’S PROTECTED AREAS

A DISSERTATION SUBMITTED TO THE GRADUATE DIVISION OF THE

UNIVERSITY OF HAWAI‘I AT MĀNOA IN PARTIAL FULFILLMENT OF THE

REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

NATURAL RESOURCE AND ENVIRONMENTAL MANAGEMENT

DECEMBER 2014

Devolent Tomas Mtui

Dissertation committee:

Christopher A. Lepczyk, Chairperson

Qi Chen

Linda Cox

Tomoaki Miura

Norman Owen-Smith

Andrew Taylor

Keywords: Wildlife, Protected areas, National Park

Dedication:

To my beloved mother Maria Aminiel Mrai for showing me the light of the world. It is

sad that you didn’t live long enough to witness my education and life achievements.

To my loving and caring father, Tomas Kirimia Mtui, for encouraging me to pursue

graduate studies, and supporting me throughout this dissertation journey.

My step-mother Subira Njaala, and my siblings Norah, Hazel, Hellen, Onasia, Engerasia,

Nancy, Kirimia and Anderson, for your love and prayers.

Luc Leblanc, my husband and best friend for your love and caring.

ACKNOWLEDGEMENTS

I am indebted to all the good people who provided me with genuine support throughout the time of writing this dissertation.

In addition to members of my dissertation committee, I am grateful to the following people at the University of Hawai‘i at Mānoa, who were not members of the dissertation committee, but gave their priceless time to help me: Dr. Travis Idol

(Department of Natural Resource and Environmental Management) who kindly provided access to the FLAASH software, used for atmospheric correction of the satellite images used in this research; Dr. Orou Gaoe (Department of Botany), Dr. Russell Yost

(Department of Tropical Plant and Soil Sciences), and Dr. Ronald Heck (Department of

Educational Administration) for support with statistical analysis, and advice on various analytical approaches; Dr. Luc Leblanc (Department of Plant and Environmental

Protection Sciences) for huge assistance in the field and grammatical edits of my drafts of this dissertation.

I am thankful to the following institutions who without them I wouldn’t have been able to achieve this work: USGS Earth Resources Observation and Science Center for providing Landsat satellite images used for analysis of land cover; Tanzania National

Parks Headquarters in Arusha Tanzania, for granting a research permit to obtain ground

Truthing data for the analysis of the satellite images at Tarangire, Ruaha and Katavi

National Parks; the management teams and ecology department of the three national parks, and particularly James Wakibara and Abel Mtui (Tarangire), Houston Njamasi and

Davis Mushi (Katavi), and Stephano Qolli and Paul Banga (Ruaha), who kindly provided logistic information and accommodation. Joseph Mhina (Game Ranger, KNP) and

iii

Samuel Mareju (Game Ranger, Ruaha) provided safety keeping from wild animals during the course of field work; Richard Ndaskoi (Tanzania Wildlife Research Institute

(TAWIRI) and Yusuff Mkwanga (Morogoro) who safely drove us around during the field work.

I am grateful to the management teams of Tanzania Wildlife Research Institute -

Tanzania Wildlife Conservation Monitoring (TAWIRI-CIMU) for providing the wildlife count survey data used in this study. Particularly, I thank TAWIRI-CIMU’s Director

General, Dr. Simon Mduma, the Head of CIMU Mr. H. Maliti, and CIMU’s database manager, Mr. Machoke Mwita, for their support during the process of securing the data. I am grateful to Richard Ogoshi and Richard Kablan (Department of Tropical Plant and

Soil Sciences) for providing me with a part time job, under the Jatropha project.

I would like to thanks my great friends and family friends for their support, encouragement and cheering me up all the time I was down: Kiran Sagoo, Hang Nguyen,

Louise Lo, Ilse Layau, Bonnie Tolson, Gaoussou Diarra, Angela Nyaki, the Harris family particularly Betty-Jo, Ernest and Tanya Harris, Nancy Wond and Charles Huxel, Rudolph

Putoa, Aubert Ruzigandekwe and Faïna Iligoga, Roger Vargas, Fr. Leonard Ssempija,

Mark and Joan Helbling, and the family of Lutheran Church of Honolulu.

My colleagues at the University of Dar es Salaam: Kim Howell, Wirlik Ngalason, and Henry Ndangalasi for their support with advice, and field equipment, and identification of types of vegetation cover respectively.

I am indebted to my father, step mother, siblings (to whom this dissertation is dedicated) for the love and prayers all these years of being away from the family. My new family: Monique Desroches, Laure Leblanc, Lorne Ross, Pierre Leblanc, Francois

Leblanc, Marie DeBrienne, Christina Lavergne-Leblanc, Jean-Philippe, Violette and

William Labelle for their love and kindness. Luc Leblanc, thank you for who you are to me!

This study would not have been achieved without financial support from the

International Ford Foundation Fellowship Program Tanzania (IFP-Tanzania), in collaboration with the East-West Centre (EWC) at Honolulu, Hawai‘i in United States of

America.

ABSTRACT Declines in wildlife and their habitats associated with land cover changes are documented worldwide. Wildlife protected areas are established as a strategy to maintain and protect viable wildlife populations. International and national policies and regulations are set-forth in countries across the globe to emphasize wildlife species protection.

Tanzania has allocated 26.5% (250,425 km2) of its total land area for wildlife protection, and its government established in 1998 a National Wildlife Policy to ensure the maintenance of viable protected areas and survival of important species, habitat and ecosystems. After more than a decade of its implementation, the level and rate of anthropogenic activities experienced in and around protected areas, and the consequent declines of wildlife species, was expected to be reduced to a minimum. Yet, recent studies show that degradation and isolation of wildlife habitats and declines in species populations continue to occur in Tanzanian protected areas.

I evaluated changes in landscape and wildlife in three protected areas in Tanzania from the 1980s to the 2010s. Specifically, we investigated changes in land cover and species abundance over time, inside and outside the protected areas, and determined the effects of changes in the types of land cover on wildlife abundance. First, I used

Maximum Likelihood classification procedure to derive land cover classes from Landsat

TM and ETM+ satellite images of the 1980s, 1990s and 2010s, and to detect changes using post-classification comparison technique and landscape metrics approach. Second,

I analysed animal density data for six species or groups of large herbivores, from 1991 to

2012. Thirdly, I evaluated the effect of land cover change on three species of large herbivores.

The results show evidence of loss and degradation of types of land covers utilized by large herbivores. Habitats for large herbivores species have shrunken, inside and outside protected areas, resulting in declines of large herbivore populations. Wildlife protected areas in Tanzania appear not effective in protecting and maintaining wildlife populations and their habitats. As a result, revision of the existing wildlife policy is needed in order to come up with techniques that could assist protected areas achieve their intended goal.

vii

LIST OF TABLES

Table 2.1. Definition of land cover classes as identified at the Tarangire (TNP) and

Katavi (KNP) National Parks.

Table 2.2. Error matrix of the classification map derived from the 2009 and 2011 Landsat

TM images of TNP and KNP and adjacent areas.

Table 2.3. Determining land cover trends over the past 20 years in KNP and TNP.

Table 2.4. Overall estimates for open shrubland and barren land which was observed to

have significant change in amount of coverage over time.

Table 2.5. Land covers change (km2) inside TNP from 1988 to 1999 and from 1999 to

2009.

Table 2.6. Land covers change (km2) outside TNP from 1988 to 1999 and from 1999 to

2009.

Table 2.7. Land covers change (km2) inside the KNP from 1984 to 1999 from 1999 to

2011.

Table 2.8. Land cover change (km2) outside the KNP from 1984 to 1999 from 1999 to

2011.

Table 3.1: Large herbivore species recorded in Ruaha-Rungwa, Katavi-Rukwa and

Tarangire protected areas.

Table 3.2: General Linear Model (GLM) analysis of results on temporal changes on

average densities of large herbivores in Katavi, Ruaha and Tarangire protected

areas.

Table 3.3: Estimates of changes in mean density (Indiv/km2) for species observed to have

changed significantly over time in Table 2 above.

viii

Table 3.4: General Linear Model (GLM) analysis results on temporal changes on number

of cells occupied by large herbivores in Katavi, Ruaha and Tarangire protected

areas.

Table 3.5: Estimates of changes in mean number of cells occupied by large herbivores

observed to have changed significantly over time in Table 4 above.

Table 4.1: Large herbivore species recorded by TAWIRI in KNP and TNP.

LIST OF FIGURES

Figure 2.1. The map of Tanzania showing location of the study sites KNP and TNP.

Figure 2.2. Total annual rainfall and average annual recorded in KNP and TNP.

Figure 2.3. Land cover maps of TNP and KNP and their adjacent areas.

Figure 3.1. The three wildlife protected areas covered in this study: Katavi-Rukwa (A)

Ruaha- Rungwa (B) and Tarangire (C) protected areas.

Figure 3.2. Annual rainfall for Katavi (A) Ruaha (B) and Tarangire (C) wildlife

ecosystems.

Figure 3.3: Scatter plots showing Log10 (n+1) transformed density of large herbivores

over time in Katavi, Ruaha and Tarangire protected areas.

Figure 3.4: Scatter plots showing number of grid cells occupied by large herbivores over

time, in Katavi, Ruaha and Tarangire protected areas.

Figure 3.5: Bar chart showing average densities of elephant from 1991 to 2012 in Katavi,

Ruaha and Tarangire protected areas.

Figure 4.3. Land cover change detection in Tarangire National park using class area

metric.

Figure 4.4. Land cover change detection in Katavi National park using class area metric.

Figure 4.5. Land cover change detection in Tarangire National park using number of

patches.

Figure 4.6. Land cover change detection in Tarangire National park using Euclidean

nearest neighbor mean distance metric,

Figure 4.7. Land cover change detection in Katavi National park using Euclidean nearest

neighbor mean distance metric.

Figure 4.8. Effect of land covers change in terms of class area (km2) on population

densities of large herbivores in KNP and TNP.

Figure 4.9. Effect of land covers change in terms of number of patches on population

densities of species of large herbivores in TNP.

Figure 4.10. Effect of land covers change in terms of mean distance between nearest

neighbor patches on population densities of large herbivores’ species or groups, in

KNP and TNP.

Figure 4.11. Bar chart summarizes average densities of elephant from 1991 to 2012

recorded inside the Katavi and Tarangire National Parks.

Figure 4.12. Influence of rainfall on number of patches of open shrubland, savannah and

mean distance between patches of open shrubland in TNP.

Figure 4.13. Influence of rainfall on densities of zebra, giraffe and elephant in TNP.

LIST OF ABBREVIATIONS

ENN_MN Euclidean nearest neighbor mean distance

GR Game Reserve

KNP Katavi National Park

NP National Park

RNP Ruaha National Park

TANAPA Tanzania National Parks

TAWIRI Tanzania Wildlife Research Institute

TNP Katavi National Park

CIMU Conservation Information and Monitoring Unit

GCA Game Controlled Area n.d. No date

FR Forest Reserve

B. land Bare or less vegetated land

C. shrubs closed shrubland

O. shrubs Open shrubland

W. savannah Woody savannah

xii

TABLE OF CONTENTS

ACKNOWLEDGMENTS………………..……………..…………..………...iii

ABSTRACT………………….…………………….………...……….…..…...vi

LIST OF TABLES………………………….…………….….……….…...…viii

LIST OF FIGURES………………………..….…………...…………..……….x

LIST OF ABBREVIATIONS……..…………..…………………….…….….xii

Chapter 1. Introduction ...... 4

1.0 Problem, goal and objectives ...... 4

2.0 References ...... 11

Chapter 2. Assessing Multi-decadal Land Cover-Land Use Change in Wildlife Protected Areas in Tanzania Using Landsat Imagery ...... 14

Abstract ...... 14

1.0 Introduction ...... 15

2.0 Material and Methods...... 20

3.0 Results ...... 30

4.0 Discussion ...... 32

5.0 Conclusion ...... 37

6.0 Acknowledgments ...... 38

Appendix A ...... 62

Appendix B ...... 63

Appendix C ...... 65

Appendix D ...... 67

Appendix E ...... 68

Appendix G ...... 69

Appendix H ...... 71

8.0 References ...... 73

Chapter 3. Assessment of Wildlife Populations Trends in the Tanzania’s Protected Areas from 1991 to 2012 ...... 81

Abstract ...... 81

1.0 Introduction ...... 82

2.0 Materials and Methods ...... 86

3.0 Results ...... 95

4.0 Discussion ...... 96

5.0 Conclusion ...... 101

6.0 Acknowledgments ...... 102

Appendix I: ...... 114

Appendix J...... 115

Appendix K: ...... 121

8.0 References ...... 122

Chapter 4. Effect of Land Cover–Land use Change on Abundance of Large Herbivores in the National Parks in Tanzania ...... 130

Abstract ...... 130

1.0 Introduction ...... 131

2.0 Materials and Methods ...... 137

3.0 Results ...... 146

4.0 Discussion ...... 150

5.0 Conclusion ...... 155

Appendix L ...... 171

Appendix M...... 176

7.0 References ...... 183

Chapter 5. Conclusions and Recommendations ...... 194

1.0 Introduction ...... 194

2.0 Limitations ...... 194

3.0 Summary ...... 196

4.0 Synthesis...... 200

5.0 Recommendations ...... 201

6.0 References ...... 204

Chapter 1. Introduction

1.0 Problem, goal and objectives

Protected areas are established as a strategy to maintain and protect biodiversity and ecosystems (IUCN 1994, Margules & Pressey 2000). Six categories of protected areas are recognized by the World Conservation Union (IUCN) and the categories depend upon the management goal. These categories include those managed for science or wilderness (category I), ecosystem protection and recreation (II), conservation of specific, unique natural features (III), preservation of specific species and to maintaining habitats (IV), landscape/seascape protection and recreation (V), and sustainable use of natural ecosystems (VI) (IUCN 1994). Categories I and II have strict restrictions on resource use, while categories III to VI have some flexibility in the restrictions. This dissertation focuses on protected areas in categories II and IV, because my interest was in wildlife protected areas (i.e. national parks and game reserves). Permanent human settlement, cultivation, and livestock grazing are prohibited in game reserves, although hunting, fishing and logging are allowed with a special license. The national parks are given the highest level of protection, meaning that no permanent settlement, hunting, livestock grazing, or logging is allowed.

Many protected areas, particularly those in developing countries, occur in locations with high levels of poverty (Fisher & Christopher 2007), where communities depend on wildlife resources to sustain their livelihoods (Mfunda & Røskaft 2010).

Maintaining and protecting wildlife species when economic development activities are the highest priority remains a challenge. Wildlife species continue to be threatened, inside

and outside protected areas, by habitat destruction caused by land clearing (Nagendra

2008, Ottichilo et al. 2001, Clerici et al. 2007), road construction (Newmark 2008), and illegal hunting (Brashares et al. 2004). Such activities block animal movement, preventing wildlife from accessing dispersal, calving and breeding sites, potentially resulting in population declines. Protected areas remain a valuable approach implemented across the globe to minimize the decline of wildlife inside the areas (Bruner et al. 2001,

Sánchez-Azofeifa et al. 2003, Gaveau et al. 2009, Nagendra 2008). As human population growth fuels the demand to sustain livelihoods, fears mount that protected areas might fail to achieve their long-term conservation goals (Wittemyer et al. 2008, Mora et al.

2011).

International and national strategies and regulations are set forth in countries across the globe to ensure protection of viable biological diversity. Effective execution of such strategies, whether in developed or developing countries, requires adequate funding.

Three major factors that hinder effective wildlife protection in developing countries are a growing human population (Wittmeyer et al. 2008, Mora et al. 2011), extreme poverty conditions in communities bordering protected areas (Fisher and Christopher 2007), and limited resources for conservation activities and management of the protected areas

(Balmford et al. 2003, Bruner et al. 2004). These factors challenge the reallocation of wildlife revenues obtained from tourism and licensed hunting from socio-economic and developmental activities to conservation and management of wildlife.

Like other developing countries, Tanzania is experiencing rapid population growth, which increased from an estimate of about 30,000,000 in 2002, to 45,000,000 in

2012 (TNBS 2013). Despite the growing population and its associated socio-economic

demands, Tanzania has demonstrated a commitment to protecting its wildlife resource by allocating 26.5% (250,425 km2) of its total land area to some type of protected area for wildlife, including national parks (NPs) (5.5%), the Ngorongoro Conservation area (1%), game reserves (10.4%), and game controlled areas (9.6%). Tanzania National Parks

(TANAPA) regulates NPs, while the Tanzania Wildlife Division Authority regulates GR and GCA. Moreover, in 1998, the Tanzanian government established a National Wildlife

Policy to emphasize maintenance of viable protected areas for important habitats and viable populations of important species, and ensure the survival of species classified as endangered, endemic, or rare. The policy was amended in 2007 to also include wetland resources of national and international importance for biodiversity and water catchments, and it states that “Wildlife and wetlands are natural resources of great biological, economical, environmental cleaning, climate ameliorating, water and soil conservation, and nutritional values that must be conserved. It can be used indefinitely if properly managed” (MNRT 2007).

The policy covers four major areas: wildlife protection, wildlife utilization, management and development of protected areas, and international cooperation (MNRT

2007). The strategy, aimed at achieving the objective of wildlife protection, is to maintain the existing wildlife protected areas and create new ones where necessary. For example, a new category of protected area in Tanzania known as Wildlife Management Areas

(WMAs) was formed to promote community conservation outside the parks and economic development through tourism enterprises and tourist hunting. Formulation of

WMAs was expected to reduce species loss by involving the local community in the protection of wildlife outside parks boundaries while also demonstrating the benefits of

conservation. With such increased emphasis in protecting wildlife species, after ten years of implementation of the wildlife policy, the level and rate of anthropogenic activities experienced in and around protected areas, and the usual subsequent declines of wildlife species were expected to be reduced to a minimum. However, recent scientific studies conducted in Tanzania national parks and game reserves show that habitat degradation and isolation caused by cultivation, livestock rearing, and illegal hunting (for food and income), continue to occur inside and outside the protected areas, and wildlife populations continue to decline (Stoner et al. 2007, Caro 2008, Msoffe et al. 2011).

The goal of this dissertation was to evaluate changes on landscape and wildlife in protected areas (National Parks and Game Reserves) in Tanzania, from the early 1980s to the early 2010s. The assessment focused on two ecological indicators of success for protected areas, developed by the Millennium Ecosystem Assessment (MEA) in 2005, which are quantity of land cover utilized by wildlife inside and in areas adjoining the parks, and abundance of wildlife species. To address the overarching goal major objectives were to: 1) quantify the extent to which land cover change has occurred between the 1980s and the 2010s, inside and outside the protected areas, 2) quantify the extent to which wildlife abundance has changed during that period, inside and outside the protected areas, and the direction and rate of change, and 3) determine the influence of changes on land cover spatial component on wildlife species abundance inside and outside the protected areas.

The three objectives are addressed in three chapters as standalone manuscripts prepared for submission to three different scientific journals: Journal of Land-use Science

(Chapter 2), African Journal of Ecology (Chapter 3) and Landscape Ecology (Chapter 4).

Therefore each chapter is written following the style and format of its respective journal.

Below is a brief summary of each chapter.

The objective of chapter 2 was to quantify the extent to which changes have taken place in the different types of land covers from the late 1980s to the early 2010s, potentially detrimental to wildlife conservation, inside and outside Tarangire and Katavi

National Parks in Tanzania. Based upon this objective two main questions were addressed. First, have the areas of different types of land cover inside and outside the national parks changed over the years? Second, were the changes in each type of land cover consistent inside and outside the parks? In order to answer these questions, Landsat

TM and ETM+ satellite images from 1980s, 1990s, and 2010s were analysed using

Maximum Likelihood classification procedures to derive types of land cover for the three time periods. Then, a post-classification comparison technique was used to identify and determine the extent and direction of changes in the identified types of land cover. In order to understand the dynamics of land cover change, land cover transition matrix tables were computed. Temporal changes in the amount of change for each land cover were tested using a general linear model (GLM). The satellite data used in this objective were provided by U.S. Geological Survey (USGS).

The objective of chapter 3 was to quantify population trends of large herbivores in three wildlife protected areas in Tanzania, namely Katavi-Rukwa, Ruaha-Rungwa, and

Tarangire, in order to determine the effectiveness of protected areas after the establishment of the wildlife policy in 1998. In order to achieve this objective, four questions were addressed: 1) Has the population density of large herbivores and area they occupied, in Katavi-Rukwa, Ruaha-Rungwa and Tarangire protected areas in Tanzania,

changed between 1991 and 2012; 2) If change has occurred, was the degree of change, among species and the areas they occupy, consistent across protected areas; 3) Were the patterns of change similar inside and outside of the protected areas; and, 4) Have the protected areas been effective in achieving the goal of protecting species diversity? If the densities of large herbivores, and/or the areas occupied by large herbivores, have remained unchanged or increased over time, then the protected areas would be considered as effective, otherwise not. To answer these questions animal density data computed from aerial census data that were collected by Tanzania Wildlife Research Institute from 1991 to 2012 were analyzed.

The objective of chapter 4 was to determine the effect of changes on the types of land cover components, in Katavi and Tarangire National parks in Tanzania, on the populations of large herbivores over the past two decades. In order to achieve this objective, three main questions were addressed. First, have landscape metrics undergone significant change between 1980s and 2010s inside and outside the national parks.

Second, if changes have occurred, are the patterns similar inside and outside the parks.

Finally, are the populations of large herbivores influenced by change in landscape metrics. To answer the first two questions I quantified four metrics, class area (total area), number of land cover patches, mean patch size, and mean distance between patches) from land cover raster maps (30 m resolution) of the 1980s, 1990s and 2010s for each park, inside and outside, within 5 km buffer zone from the park boundary. General linear models (GLM) were used to test whether each of the four land cover components had changed over time. Using GLM, the land cover metrics found to be significant were

evaluated against densities of three species of large herbivores (elephant, giraffe and zebra) inside the parks and within their 5 km buffer zone.

2.0 References

Balmford, A., Gaston, K. J., Blyth, S., James, A., & Kapos, V. (2003). Global variation in

terrestrial conservation costs, conservation benefits, and unmet conservation

needs. Proceedings of the National Academy of Sciences of the United States of

America, 100, 1046–1050.

Brashares, J. S., Arcese, P., Sam, M. K., Coppolillo, P. B., & Balmford, A. (2004).

Bushmeat Hunting, Wildlife Declines, and Fish Supply in West Africa. Science,

306, 1180–1183.

Bruner, A. G., Gullison, R. E., Rice, R. E., & da Fonseca, G. (2001). Effectiveness of

parks in protecting tropical biodiversity. Science, 291, 125–128.

Bruner, A. G., Gullison, R. E., & Balmford, A. (2004). Financial Costs and Shortfalls of

Managing and Expanding Protected-Area Systems in Developing Countries.

BioScience, 54, 1119–1126.

Caro, T. (2008). Decline of large mammals in the Katavi-Rukwa ecosystem of western

Tanzania. African Zoology, 43, 99–116.

Clerici, N., Bodini, A., Eva, H., Grégoire, J.-M., Dulieu, D., & Paolini, C. (2007).

Increased isolation of two Biosphere Reserves and surrounding protected areas

(WAP ecological complex, West Africa). Journal for Nature Conservation, 15,

26–40.

Fisher, B., & Christopher, T. (2007). Poverty and biodiversity: Measuring the overlap of

human poverty and the biodiversity hotspots. Ecological Economics, 62, 93–101. 11

Gaveau, D. L. a., Epting, J., Lyne, O., Linkie, M., Kumara, I., Kanninen, M., & Leader-

Williams, N. (2009). Evaluating whether protected areas reduce tropical

deforestation in Sumatra. Journal of Biogeography, 36, 2165–2175.

IUCN. (1994). Guidelines for Protected Area Management Categories: CNPPA with the

assistance of WCMC.IUCN (pp. 1–257). Gland, Switzerland and Cambridge, UK.

Margules, C. R., & Pressey, R. L. (2000). Systematic conservation planning. Nature, 405,

243–253.

Mfunda, I. M., & Røskaft, E. (2010). Bushmeat hunting in Serengeti, Tanzania : An

important economic activity to local people. International Journal of Biodiversity

and Conservation, 2, 263–272.

MNRT (Ministry of Natural Resource and Tourism). (2007). The Wildlife Policy of

Tanzania, Dar es Salaam Tanzania, Dar es Salaam, Tanzania: Government

Printer.

Mora, C., Aburto-Oropeza, O., Ayala Bocos, A., Ayotte, P. M., Banks, S., Bauman, A.

G., … Zapata, F. A. (2011). Global human footprint on the linkage between

biodiversity and ecosystem functioning in reef fishes. PLoS Biology, 9, e1000606.

Msoffe, F. U., Kifugo, S. C., Said, M. Y., Neselle, M. O., Van Gardingen, P., Reid, R. S.,

… de Leeuw, J. (2011). Drivers and impacts of land-use change in the Maasai

Steppe of northern Tanzania: an ecological, social and political analysis. Journal

of Land Use Science, 6, 261–281.

Nagendra, H. (2008). Do Parks Work? Impact of Protected Areas on Land Cover

Clearing. Ambio, 37, 330–337.

Newmark, W. D. (2008). Isolation of African protected areas. Frontiers in Ecology and

the Environment, 6, 321–328.

Ottichilo, W. K., Leeuw, J. De, & Prins, H. H. T. (2001). Population trends of resident

wildebeest [Connochaetes taurinus hecki (Neumann)] and factors influencing

them in the Masai Mara ecosystem, Kenya. Biological Conservation, 97, 271–

282.

Sánchez-Azofeifa, A. G., Daily, G. C., Pfaff, A. S. P., & Busch, C. (2003). Integrity and

isolation of Costa Rica’s national parks and biological reserves: examining the

dynamics of land-cover change. Biological Conservation, 109, 123–135.

Stoner, C., Caro, T., Mduma, S., Mlingwa, C., Sabuni, G., Borner, M., & Schelten, C.

(2007). Changes in large herbivore populations across large areas of Tanzania.

African Journal of Ecology, 45, 202–215.

TNBS (Tanzania National Bureau Standard). (2013). Population Distribution by age.

Internal Report. Dar es Salaam.

Wittemyer, G., Elsen, P., Bean, W. T., Burton, a C. O., & Brashares, J. S. (2008).

Accelerated human population growth at protected area edges. Science, 321, 123–

126.

Chapter 2. Assessing Multi-decadal Land Cover-Land Use Change in Wildlife Protected Areas in Tanzania Using Landsat Imagery

Abstract

This research sought to understand the extent to which changes have taken place in the different types of land covers, potentially detrimental to wildlife conservation, in

Tarangire and Katavi National Parks, in Tanzania. Two main questions were asked: have the areas of different types of land cover in the national parks changed over the past 27 years; and were the changes consistent between inside and outside the parks? Maximum

Likelihood classification procedures were used to derive eight land cover classes from

Landsat TM and ETM+ satellite images: woody savannah, savannah, grassland, open and closed shrublands, swamps and water, and bare lands. Post-classification comparison technique was used to identify and determine the extent and direction of changes for all land cover classes. The results show only two land cover classes, barren land in

Tarangire, and open shrubland in Katavi, exhibited significant increases and decreases, respectively, over the past 27 years. The open shrubland were replaced by the savannah and woody savannah, while the bare land increased due to degradation of savannah woodlands possibly due to human encroachment by cultivation. These changes should be monitored to prevent detrimental effects on wildlife populations.

Keywords: Conservation; Katavi; Maximum Likelihood; Tarangire

1.0 Introduction

Changes in quality and quantity of land cover-land use have implications for the continual existence of wildlife species (Mas 2005; Greenwald, Purrenhage, and Savage

2009; Haberl et al. 2009). In Africa, land cover-land use change is driven by people

(Frost 1996; Kikula 1986; Lawton 1978), high densities of large herbivores when their movement is restricted (Cumming et al. 1997; Owen-Smith 1988; Holdo 2006), fires, or through a combination of herbivores and fire (Holdo, Holt, and Fryxell 2009; Langevelde et al. 2003; Skarpe 1991). People directly affect land cover by clearing land for cultivation and harvesting trees or grasses, and may initiate fires (Frost 1996). In non-fire prone ecosystems, fire can still damage vegetation cover and prevent or slow down its recovery (Frost 1996; Kikula 1986). Cultivation may reduce dispersal areas for wildlife and block migration corridors, forcing animals to concentrate in one location, hence degrading their habitat by overgrazing or browsing (Kahurananga and Silkiluwasha

1997). One strategy used worldwide to prevent degradation of wildlife habitat and ensure the long-term survival of wildlife species is creating protected areas, such as national parks. But the effectiveness of protected areas is frequently questioned in developing countries because they often adjoin poor communities that rely on the wildlife and their habitat to sustain their livelihoods (Fisher and Christopher 2007; Mfunda and Røskaft

2010). Nevertheless, protected areas are effective at decreasing land clearing, logging, and grazing inside the protected areas, compared to areas outside their boundaries

(Bruner et al. 2001; Sánchez-Azofeifa et al. 2003; Gaveau et al. 2009).

Converting woodlands to grassland (Owen Smith 1988; Brenneman, Bagine,

Brown, Ndetei, and Louis 2009) or grassland to woodlands (Packer et al. 2005; Owen- 15

Smith 1988) may reduce pastures available for browsing and grazing, decrease shade, increase vulnerability of prey to predators, reduce dispersal areas for migratory species

(Owen-Smith 1988; Western, Russell, and Cuthill 2009), and block wildlife migratory routes (Mwalyosi 1992; Mundia and Murayama 2009), which have all resulted in declining wildlife populations. In East Africa, wildebeest (Connochaetes taurinus),

Grant’s gazelle (Gazella granti), Thomson’s gazelle (Gazella thomsonii), and zebra

(Equus burchellii) populations declined by 60% between 1975 and 2007, due to conversion of grasslands to modern agriculture and tourism (Mundia and Murayama

2009). Similarly, the giraffe (Giraffa camelopardalis rothschildi) population in Nakuru

National Park has declined over the past 10 years, partly due to a reduction of forage resulting from severe drought (Brenneman et al. 2009). On the other hand, in the

Serengeti National Park, the population of lions (Panthera leo) increased in the 1980s and 1990s, due to increases in woodland vegetation and tall grasses that enhanced prey catchability (Packer et al. 2005).

Land cover and land use change have affected wildlife species in and around protected areas in Tanzania, due to the establishment and expansion of villages and changes in agricultural policies that were established in 1974 and 1983 to improve social welfare (Prins 1987). These policies increased land degradation via increased settlements, livestock herds, farming, and mining (Prins 1987; Mwalyosi 1992; Msoffe et al. 2011).

Between 1957 and 1987, 77% (from 630 km2 to 144 km2) of woody vegetation in the

Masai Steppe was converted into grassland and cultivated farms (Mwalyosi 1992). For instance, bean farms and settlements increased from 3.4 km2 to 15 km2 from 1984 to

2000, replacing woodland and grassland within a distance of less than 40 km from the

Tarangire National Park (TNP) boundary. As a result, four of the nine wildlife corridors on the western and southern sides of TNP were blocked, causing the wildebeest population to decrease from 40,000 to 5,000 (Msoffe et al. 2011). Furthermore, human migrants from northern Tanzania settled the open area south of the Katavi National Park

(KNP) in the late 1990, with their cattle grazing inside the protected area (Tim Caro, personal communication, May 2012). In the northwest of KNP the flow of Katuma River has been reduced due to upstream rice cultivation, reducing the water available to maintain wildlife species in the wetlands of Lakes Chada and Katavi and the Katisunga flood plain, which harbor the highest animal density in the park during dry seasons (Caro

1999; Manase, Gara, and Wolanski 2010). These activities all have notable effects on the land cover types utilized by wildlife species, and ultimately on these species populations.

However, no recent evaluation of land cover change in the TNP and KNP ecosystems has been conducted.

In 1996 the Tanzanian government ratified the Convention on Biological

Diversity that established national laws and policies in order to protect biodiversity.

Specifically, the 1998 National Wildlife Policy was established to emphasize maintenance of viable conservation areas for important habitats and viable population of all endangered, endemic and rare species, ensuring the survival. The policy was amended in 2007 to include wetland resources that are nationally and internationally important habitats for biodversity and water catchment (MNRT 2007). Various conservation programs involving local communities in protecting wildlife, such as community conservation schemes and wildlife management programs were introduced to support these legislative actions. With the implementation of the National wildlife policy,

protected areas were expected to increase protection efforts to substantially reduce human activities inside and around parks, allowing restoration of the main types vegetation cover in national parks (woody savannah, savannah, open and closed shrublands, and swamps) that were previously converted or transformed to other types, as previously reported

(Msoffe et al. 2011; Mwalyosi 1992; Manase et al. 2010). In this study we use Landsat

Thematic Mapper (TM) and Enhanced Thematic Mapper plus (ETM+) data imagery to identify and analyse land cover/land use change in wildlife protected areas in Tanzania.

The remote sensing technique is widely used, especially in the monitoring of land cover changes (Jensen 2007). The generated imagery data is an important means of detecting and analysing spatial and temporal dynamics of land cover (Jensen 2005, 2007), as it can provide information over large areas including those that are inaccessible due to dangerous wild animals or rough terrain. A number of studies have used this approach to assess land cover -land use change in protected areas in Tanzania, but only a few have focused on the Tarangire and Katavi ecosystems, which are the main focus of our study.

For example the most recent study by Msoffe et al. (2011) used Landsat (TM and ETM+) images of 1984 and 2000 to assess land use changes in Tarangire wildlife ecosystem; and

Pelkey et al. (2000, 2003) used Advanced Very High Resolution Radiometer (AVHRR) to assess habitat changes in terms of greenness, in protected areas across Tanzania, comparing 1982 and 1994 images. In order to understand the current condition of land cover-land use in wildlife protected areas and the likely impact it may have on sustaining wildlife species, analysis of current satellite images is necessary. The goal of this research was to understand the extent to which changes have taken place in the different types of land covers, potentially detrimental to wildlife conservation, inside and outside

the protected areas in Tanzania. Based upon this goal we sought to address two main questions. First, have the areas of different types of land cover inside and outside the national parks changed significantly over the past 27 years? Second, were the changes in each type of land cover consistent between the inside and outside the parks?

Based upon the aforementioned questions we predicted that, because of increased protection efforts through implementation of the wildlife policy 1) the area occupied by closed and open shrubland inside and outside the parks would decrease, as savannah or woody savannah covers that were forced to remain as shrubs due to over browsing or frequent fires, would be restored, and that the rate of decrease of shrublands would be higher inside than outside the park; 2) that the areas of swamps inside and outside the park would decrease because of drought conditions following reduction of river flow, and the rate of decrease would be lower inside than outside the park; 4) that grassland inside and outside the parks would decrease through natural recovery of areas once covered with closed woody vegetation, and that the decrease should be more pronounced inside than outside the park; and 5) that the barren land inside and outside the parks would decrease, and at a higher rate inside than outside the parks. The proximity to human settlements is the primary reason for the expected changes in land cover inside and outside the parks, with areas in close proximity to people (outside protected areas) expected to have less improvement than more distant areas (inside protected areas)

(Metzger et al. 2010).

2.0 Material and Methods

2.1 Study sites

To achieve the goal of this study, land cover was investigated at the TNP and

KNP in Tanzania (Figure 1). The TNP covers 2,600 km2 (between 3o40’ to 5o35’ latitude, and 35o45’ to 37o longitude) at an elevation of ranging from1,200 m to 1,600 m above sea level (Figure 1). The park was established as a Game Reserve in 1957 and declared a

National Park in 1970. The TNP is bordered by Lake Natron and the Mto-wa-Mbu Game

Controlled Area (GCA) to the north, the Lolkisale Game Controlled Area and the

Simanjiro Plains to the east, the Mkungunero Game Reserve (GR) to the south, and Lake

Burunge, the Burunge Game Controlled Area, the Kwakuchinja Open Area, and Lake

Manyara National Park to the west (Figure 1). The GCA is one category of protection in

Tanzania where human activities such as settlements and hunting with permits by hunting companies and local residents, can occur. Average annual rainfall is about 655 mm

(Figure 2) with short rains between October and December and heavy rains between

February and April or May. Hot season (December to February) temperatures range from

17 oC to 29 oC with cold season (June and July) temperatures ranging from 14 oC to 25 oC. The TNP is situated in the semi-arid wooded steppe. The major types of vegetation in this park are riparian woodland, wetlands and seasonal flood plain, Acacia-Commiphora woodland, riverine grassland, Combretum-Dalbergia woodland, Acacia drepanolobium woodland, and grasslands with scattered baobab trees (TANAPA n.d.).

Pastoralism was the major land use in and around the TNP over the past two centuries (Prins 1987). However, in the past two decades agricultural activities have

increased on the northeast and eastern side of outside the park, blocking wildlife migration routes and dispersal areas (Mwalyosi 1992; Msofe et al. 2011).

The KNP covers 4,238 km2 (between at 6◦63’ to 7◦30’ latitude, and 30◦75’ to

31◦74’Longitude) at an elevation of ranging from 800 m to 1,600 m above sea level

(Figure 1). The park was established in 1974 with 1,816 km2, but was enlarged in 1998 to the current size to reduce pressure from settlements and grazing by cattle. The KNP is bordered by Msanginia Forest Reserve (FR) and Mlele GCA to the north, Lwafi GR and

Nkamba FR to the west, Usevya Open Area to the south, and Rukwa GR to the south and south east (Figure 1). The inhabitants include the Pimbwe people, who are native to the area and the Sukuma people who emigrated from the northern regions. Both tribes practice horticultural activities and pastoralism (Caro 1999). The integrity of the park is threatened by increased damming of River Katuma for rice cultivation, farming inside and adjacent to the park, and wildfires caused by poachers, farmers and people travelling by the main road which cuts across the park (Manase et al. 2010).

Average annual rainfall is about 955 mm (Figure 2), which falls between

November to April or May. The vegetation consists of grassland interspersed with miombo woodlands and mixed woodlands. Miombo forms a single story, with a light, closed canopy of deciduous woodland usually greater than 15 m tall dominated by trees of the genera Brachystegia, Julbernadia, and Isoberlinia (Kikula 1987; Frost 1996).

Underneath the trees are layers of scattered shrubs, grasses, and forbs that grow to a height of 0.3 to 100 cm with 50 - 75% ground cover (Lawton 1978). The genera

Markhamia, Grewia, Terminalia, Combretum, Syzygium, and Acacia also occur in the miombo woodland of the KNP (Banda, Schwartz, and Caro 2006). During the dry season,

large mammals feed on swampy vegetation occurring in the large seasonal lakes of

Katavi and Chada, and the Katisunga flood plain (Caro 1999), which is maintained by

Katuma River (Figure 2.1).

Of the various types of protected areas, national parks (NP) are given the highest level of protection, meaning that no permanent settlement, hunting, livestock grazing, or logging is allowed. Like the NP, permanent human settlement, cultivation and livestock grazing are prohibited in the grazing areas (GR), although hunting, fishing and logging are allowed with a special license. However, in the GCAs human settlement is allowed, along with grazing, cultivation, and licensed hunting. In the forest reserves (FR), settlements, cattle grazing, and hunting by tourists is prohibited, although selective logging with a license is allowed. NPs are regulated by Tanzania National Park

(TANAPA), GR and GCA are regulated by the Tanzania Wildlife Division authority, and the FR by the Tanzania Forest and Bee Keeping Department. Finally the open areas are under the local governments and allow public use, including settlements, farming, and livestock grazing.

2.2 Study approach

To address the main research questions, land cover was evaluated inside the TNP and KNP and within a 5 km buffer zone around them (Figure 2.1). Landsat images were selected to assess land cover change in the two landscapes. Specifically, the 1988, 1999, and 2009 images (path 168, row 63) were selected for TNP and the 1984, 1999, and 2011 images (path 171, row 65) for KNP. Images were selected based upon similar times of the year to minimize seasonal (vegetation phenology cycle) and sun angle effects, which could affect multi-temporal comparisons (Singh 1989; Jensen 2005). Specifically, we

used Thematic Mapper (TM) and Enhanced Thematic Mapper plus (ETM+) images during a period of short rains on 01/10/1988, 01/11/1999, and 04/11/2009 for TNP and during the dry season on 29/06/1984, 9/03/1999, and 07/26/2011 for KNP. All images were obtained from the U.S. Geological Survey (USGS) (www.earthexplorer.usgs.gov and www.glovis.usgs.gov).

Variation in rainfall may have implications for the change detection analysis

(Jensen 2007) and caution must be used in interpreting the results. In TNP, the annual total rainfall varied among time periods when the data imagery were acquired (Figure

2.2). In KNP, there was no variation in rainfall (Figure 2.2) between 1999 and 2011, and no rainfall data were available for 1984.

2.3 Image pre-processing

The six images were geometrically corrected by the USGS distributor. The digital numbers (DN) were calibrated to radiance. Then, atmospheric effects were removed using a MODTRAN 5+-based atmospheric correction algorithm known as Fast-Line-of

Sight Atmospheric Analysis of Spectral Hypercube (FLAASH) (Exelis Visual

Information Solution, Boulder, Colorado) to derive reflectance. The tropical atmospheric model was used for the images with visibility set to 100 km.

2.4 Training data for the 2009 and 2011 Landsat images

A total of ten land cover classes (barren land, cropland, closed shrubland, open shrubland, grassland, savannah, swamp, built-up/natural vegetation mosaic, water bodies, woody savannah) were identified during the preliminary ground survey in December

2011, based on the land cover definitions used by the International Geosphere-Biosphere

Program (IGBP) (Friedl 2002). Each land cover class was photographed and a field guide

(Table 1.1, Appendix A) was developed.

Training sites included a total of 59 polygons (431 pixels) for TNP and 54 polygons (351 pixels) for KNP, all collected in the field in June and July 2012. The sampling sites were located approximately 150 m from roadside, and waypoints recorded using Garmin GPSmap 62s, which has an accuracy of ± 3.658 m. The vegetation type at each site was entered into the GPS, and a series of photographs taken to aid in data interpretation.

Although ground truth data should be from the same time frame as the remotely sensed image to prevent bias, this is often not possible. One alternative is to use images for the same time of the year with higher spatial resolution (Jensen 2005). Images such as

SPOT and Geoeye which have spatial resolutions of 20 m × 20 m and 1.36 m × 1.36 m, respectively are freely available on Google Earth. Training sites for classifying the 2009

(TNP) and 2011 (KNP) Landsat TM images were obtained using higher resolution images of SPOT and Geoeye from Google Earth. The field data obtained in 2012 aided in the interpretation of the SPOT and Geoeye images, and served as a guide to locate the training sites in Google Earth.

The 2012 locations of each of the 59 TNP field polygons were visually checked with the SPOT or Geoeye images (depending on availability) for the 2009 image. We suspected that only 31 (84 pixels) of the 59 polygons (431 pixels) drawn in 2012 had the same land cover type as in 2009. These included polygons that represented woody savannah, closed shrubland, open shrubland, and grassland. Some parts of savannah and swamp were obscured by shadow and clouds or had different land cover on it. Therefore,

25 additional sample polygons (285 pixels) for savannah, barren land, water, and swamp were defined on SPOT and Geoeye images for 2009, increasing the total training sites to

56 polygons (369 pixels). Through visual inspection 9 polygons (114 pixels) for cloud and shadows were delineated on the 2009 Landsat image, increasing the number of training sites to 67 polygons (493 pixels), representing 10 land cover classes (Appendix

B). No cropland sample location was found on SPOT or Geoeye photos for the 2009 TNP image.

Following the same procedures used for TNP, 38 (234 pixels) out of the 54 polygons (351 pixels) obtained for the KNP in 2012 had the same land cover types as in

2011. The 38 polygons were located in barren land, closed shrubland, cropland, open shrubland, grassland, savannah, swamp, water bodies, and woody savannah. The locations observed as swamp and water in 2012 were barren and vegetated in 2011, respectively, and therefore, 9 sample polygons (36 pixels) were defined in 2011 SPOT and Geoeye images to represent swamps (5 polygon, 23 pixels) and water (4 polygons, 13 pixels). The total number of training sites was 270 pixels (47 polygons) (Appendix C).

The 493 pixels for TNP and 270 pixels for KNP were used as training sites for classification of the 2009 and 2011 Landsat images, respectively, using Maximum

Likelihood Classification (MLC). The minimum number of training pixels required to perform MLC is N+1 per class, where N represents the number of wavebands. Six bands were used (bands 1 to 5, and 7), yielding a minimum number of pixels per class as 7.

Only land cover classes with >10 pixels were used as training data, which included barren land, grassland, open shrubland, closed shrubland, savannah, woody savannah, water bodies, swamp, cloud, and shadow. Cropland and built-up natural vegetation

mosaic in TNP had <10 pixels and thus were omitted. Omitting these two land cover classes was considered negligible as the built-up natural vegetation mosaic was likely to be classified as woody savannah, because most buildings were surrounded by big trees while cropland as barren, savannah, or grassland.

An exploratory analysis determined statistical separation between classes using the Jefferies-Matushita (J-M) distance in order to identify and correct overlapping spectral classes. Values ranged from 1.80 to 2.00 for spectral class pairs, indicating that the pairs were separable enough for classifying the 2009 Landsat image. As for the 2011

Landsat image, the J-M distance values ranged from 0.984 to 1.999. Woody savannah vs. natural vegetation mosaic, and savannah vs. cropland had J-M value of 0.984 and 1.287, respectively. A J-M value of below 1.0 may indicates the class spectral pair is inseparable, while a value of below 1.9 indicates that the class pairs have lower separability (Swain and Davis 1978). Therefore, these two pairs were merged, and the separability for the remaining spectral pairs ranged from 1.5 to 2.0, that contained 39 polygons (201 pixels) considered separable for classifying the 2011 Landsat image.

2.5 Assessment of Classification Accuracy of the 2009 and 2011 Landsat TM images

The 2009 Landsat TM image was classified using the MLC Classifier with the

483 training pixels. A random sample of 1,500 points was generated inside the TNP and within the 5 km buffer zone and overlaid on the SPOT and/or GeoEye images in Google

Earth. Only 1.5% (22 points) fell on SPOT/Geoeye images of 2009. Some of the remaining 98.5% fell on SPOT/Geoeye images for other years than 2009 and therefore were left out. A total of 22 polygons comprising 1,116 pixels was obtained, which covered seven land cover classes (bare land, savannah, grassland, open and closed

shrublands, woody savannah, and swamp). Through visual inspection of the Landsat image, 11 additional polygons comprising of 217 pixels were directly delineated to represent water, cloud, and shadow. A total of 1,333 testing pixels were used to assess the accuracy of the MLC-classified map of the 2009 TM image using a confusion matrix

(Richards and Jia 2006).

The 2011 Landsat TM image was classified using the MLC Classifier with the

270 training pixels. The same procedures used for accuracy assessment of the 2009 TM

Landsat image were followed and a total of 32 random polygons comprising 1,211 pixels were obtained that represented eight land cover classes (bare land, closed shrubland, grassland, open shrubland, savannah, swamp, water, and woody savannah). Therefore, a total of 1,211 testing pixels were used to assess the accuracy of the MLC-classified map of the 2011 TM image using a confusion matrix.

2.6 Classification of Landsat TM image of 1984, 1988 and ETM+ of 1999

The classified maps for 2009 (TNP) and 2011 (KNP) were used as a reference for classification of the 1999, 1988, and 1984 Landsat images because training data for these historical images could not be obtained. Land cover classes that existed on the 2009 and

2011 Landsat images were assumed to exist in the 1999 and 1988 (TNP), and 1999 and

1984 (KNP) Landsat images, respectively, based upon similar historical classification approaches (Munyati 2000; Msoffe et al. 2011).

All of the 2009 training pixels, excluding clouds and shadows, were overlaid on the 1999 ETM+ and the 1988 TM image in TNP. The locations for each of the 2009 training pixels were visually examined for land cover changes using a combination of colour composites of bands 4, 3, 2 (Near Infrared, Red and Green respectively) and 7, 4,

2 (Shortwave Infrared, Near Infrared and Green). These colour composites, enabled identifying areas with different vegetation types in terms of colour and texture. The 4, 3,

2 combination makes vegetation appear as red tones, with brighter reds indicating dense or more growing vegetation, bare soils appear as white to brown depending on moisture and organic matter, and clear waters appear as dark blue while shallow waters as lighter blue. Where changes occurred, the pixels were deleted. For example, in 2009, the number of training pixels for water was 63. On the 1988 and 1999 images, six pixels from two polygons were vegetated and therefore were deleted from both images leaving 57 pixels

(4 polygons) for water. A total of 347 pixels (53 polygons) and 370 pixels (54 polygons) were obtained for 1988 and 1999, respectively, and these were used to classify the corresponding images using the MLC. The classification of 1984 and 1999 Landsat images for KNP followed the same procedures as were used in classifying the 1988 and

1999 in TNP. A total of 137 and 189 training pixels were obtained for 1999 and 1984, respectively, and were used to classify the corresponding images using the MLC.

2.7 Land cover change detection

Various techniques are available for land cover change detection (Singh 1989; Lu et al. 2003; Coppin et al. 2004) but we used the post classification comparison technique.

This technique compares independently produced classified images and provides detailed information of land cover change, including the amount of change, and location and the nature of change (Singh 1989; Lu et al. 2003; Coppin et al. 2004). Although the use of post classification technique provides a complete matrix of the nature of changes, the accuracy of results may be affected by possible misclassification and mis-registration errors (Verbyla and Boles 2000; Coppin et al. 2004; Lu et al. 2003). False classification

change increases with number of classes and landscape heterogeneity (Verbyla and Boles

2000).

On the thematic maps derived from the Landsat TM and ETM+ for 1984, 1988,

1999, 2009, and 2011, the amount of land cover and percentages of area for each class, inside and outside the respective parks (TNP and KNP), were calculated and the amount of change measured. Each land cover class was coded in ArcGIS for each year and then eight transition matrices were established using the spatial analyst-Tabulation Area Tool, inside and outside the two parks. Specifically four matrices were calculated for TNP

(1988 vs. 1999 inside, 1988 vs. 1999 outside, 1999 vs. 2009 inside, and 1999 vs. 2009 outside) and four matrices for KNP (1984 vs. 1999 inside, 1984 vs. 1999 outside, 1999 vs. 2011 inside, and 1999 vs. 2011 outside).

2.8 Statistical analysis

The analyses were performed using SAS software version 9.2. Prior to the analyses, data were examined using scatter plots and Ryan-Joiner test to test for linearity and normality of residuals.

A general linear model (GLM) was used to test if the amount of 8 land cover classes (barren land or less vegetated land, closed shrubland, open shrubland, grassland, savannah, swamps, woody savannah and water) had changed over the time, inside and outside the two national parks. For each park, the effect of three models which included year, location, and their interaction were tested on the amount of land cover (km2), first as individual class cover and then all classes combined. Because our main research question was to test if the amounts of land cover types were increasing or decreasing over time, the year variable was treated as a quantitative rather than categorical variable, and 29

location treated as categorical. Where the results were significant (P ≤ 0.05) for the year variable or the interaction of year and location, estimates of the slopes were obtained using ESTIMATE statement in PROC GLM (Appendix H) to determine the general trend of change. Results are presented as mean ± error, unless otherwise noted.

Additionally, the same GLM method was used to test for temporal change of total annual rainfall from 1988 to 2009 (TNP) and 1997 to 2012 (KNP). Also, the amount of types of land cover in TNP were tested against total rainfall to see if any observed changes in the first part of the analysis could be influenced by rainfall. This test was only done for TNP because of availability of rainfall data comparable to the time period when the Landsat images were acquired.

3.0 Results

3.1 Thematic maps derived from Landsat images and accuracy assessment

Thematic maps with 10 land cover classes including cloud and shadow (Figure

2.3) were derived from 2009 and 2011 Landsat TM for TNP and KNP, respectively, with an overall classification accuracy of 86.8% and kappa coefficient of 0.85 (TNP) and

84.6% and 0.80 (KNP) (Table 2.2). Except for water in the KNP, 62.5% to 100% of sample pixels representing all land cover classes in the TNP and KNP maps were correctly identified. As for water in KNP, only 33% of sample pixels were correctly identified. The chances that map users find all land cover classes (including water) on the ground range from 69% to 99% (Table 2.2a and b).

3.2 Change detection results

Out of eight classes of land cover analysed in this research, only open shrubland in KNP and bare land (or less vegetated land) in TNP exhibited evidence of change over time (Tables 2.3 and 2.4). The open shrubland in the KNP decreased significantly over the past 27 years (Table 2.3) at an overall rate of 7.2 ± 0.9 km2 per year (Table 2.4).

However, the level of decrease was inconsistent across locations, as the interaction of year and location was significant (Table 2.3). Inside the KNP the open shrubland decreased significantly over time, at a higher rate of 12.2 ± 1.2 km2 per year (Table 2.4) than outside the park where shrubland decreased at a rate of 2.3±1.2 km2 (Table 2.4), but the rate of decrease outside the KNP was not significant (Table 2.4). The open shrubland inside the KNP decreased by 62%, 81% and 93% from 1984 to 1999, 1999 to 2011, and from 1984 to 2011 respectively; while on the outside remained unchanged from 1984 to

1999, but from 1999 to 2011 and from 1984 to 2011 it decreased by 56% and 65.5% respectively (Table 2.8abc and 2.9abc).

The barren land in the TNP increased significantly over the past 20 years (from

1988 to 2009) at a rate of 2.2 ± 0.2 km2 per year (Table 2.4), as the main effect of year was significant (Table 2.3). The amount of increase was similar inside and outside of the park (Table 2.4). The bare land increased inside the park by about 370%, 100% and

860% from 1988 to 1999, 1999 to 2009 and 1988 to 2009 respectively; and outside the park by about 700%, 60% and 1180% from the first, second and the third periods, respectively (Tables 2.6abc and 2.7abc). The remaining land cover classes exhibited no change across the three time periods, either inside or outside the two parks (Table 2.3).

3.3 Effect of rainfall on land cover change

There was no significant change in annual rainfall both KNP and TNP for over 15 and 17 years respectively (Figure 2.2 C and D). Also rainfall did not have any significant effect on the changes in the amount of types of land cover in TNP (Table 2.5).

4.0 Discussion

The findings of this research showed only two out of eight land cover classes, open shrubland in KNP and barren land (or less vegetated land) in TNP, exhibited significant changes across the three time periods. As expected, the open shrubland was replaced by savannah and woody savannah both inside and outside the KNP (Table 3.0a).

Total coverage by woody savannah, savannah, grassland, swamps, closed shrubland, and water did not change significantly over time. The lack of significant changes in savannah and woody savannah coverage, in both parks, was unexpected because from the transition matrices, the dynamics of these classes show that more than 50% of their baseline coverage was converted into other land cover types (Table 2.6bc – 2.9bc, Figure 2.3).

Contrary to our expectation, bare land in TNP increased over time, and at the same rate inside and outside the park, mainly due to the degradation of grassland, open shrubland, savannah, and woody savannah (Table 2.5abc, 2.6abc and 3.0ab). Previous studies reported conversion of woody vegetation and grassland into cultivated farms and settlements in the TNP ecosystem (Mwalyosi 1992; Msoffe et al. 2011). In our research we were not able to identify and assess cultivated or cropland covers (see section 2.4), but using high resolution photos of Geoeye and SPOT of 2012 on Google Earth we observed farm plots inside and outside TNP (Appendix D) which confirmed that the

observed increase in bare land inside and outside park boundaries was likely caused by conversion of woodlands to farm lands.

In addition, elephant density recorded in the Tarangire ecosystem (including the

330 km2 Lake Manyara National Park) was about 0.7 (indiv/km2) during the 2006 dry season (TAWIRI 2007), which is above the maximum density of 0.5 (indiv/km2) recommended by Cumming et al. (1997). When large herbivores as elephant exceed the carrying capacity of their preserve, and their movements restricted to the parks, are likely to degrade the vegetation covers into bare land (Cumming et al. 1997), through overgrazing and/or over browsing and trampling (Barnes 1983), and accelerating erosion

(Owen-Smith 1988). In Kruger National Park (South Africa), where wildebeest movement restriction by fencing resulted in overgrazing (Beale et al. 2013). It is possible that, the observed increased amount of bare land was contributed by large densities of elephants in TNP.

Large extent of woody vegetation was reported to have declined between 1971 and 1996 in TNP due to severe drought that occurred in 1993, and possibly also an earlier drought, from 1991 to 1992 (Vijver et al. 1999). During that period, i.e. 1993, the total annual rainfall was well below the 655 mm annual average (Figure 2.2). The amount of annual rainfall received in TNP between 2007 and 2009 was even lower than recorded in

1993 (Figure 2.2 A). However, these changes in the amount of total rainfall from 1988 to

2009 are not significant (Figure 2.2 D), and there was no significant relationship between the increase of bare or less vegetated land in TNP with amount of rainfall (Table 2.5).

Therefore, significant increases of bare or less vegetated land inside and outside the TNP

(Table 2.6abc – 2.7abc) may have not been caused by drought. Overall, cultivation by

humans and the large density of elephants were likely the main factors responsible for conversion of vegetation cover into bare land.

As expected, the open shrubland declined inside and outside the KNP, and at a higher rate inside the park. The decline of open shrubland from 1984 to 2011 was due to recovery and increase of savannah and woody savannah (Table 3.0a). Most of the area previously covered with open shrubland was replaced by woodlands inside and outside the park (Table 3.0a) were replaced by savannah and woody savannah. Noteworthy, however, the savannah class was merged with croplands during the analysis of the satellite images, as the two were spectrally inseparable (see section 2.6). Therefore, reports of savannah increasing or decreasing at KNP may in part be confounded with increase or decrease in cropland areas.

Structural and compositional dynamics of miombo woodlands as such of KNP is determined by fire (Lawton 1978; Kikula 1986) and herbivory, mediated by precipitation,

(Holdo et al. 2009; Guy 1989; Chidumayo 2004; Mapaure and Moe 2009; Mapaure and

Campbell 2002; Langevelde et al. 2003). Rainfall promotes tree growth cover, and in the absence of fire and sufficient browsing, the tree cover increases in biomass and height, with the more extensive canopy shading and reducing grass production, hence lowering fuel to fires (Holdo et al. 2009). On the other hand, rainfall promotes grass cover, and without sufficient grazing, fuel load builds up, therefore fuelling fires, damaging and killing trees and shrubs and the suppressing recruitment of saplings to the canopy trees

(Holdo et al. 2009). Conversely, longer fires, with less fuel load, allow woodland regeneration, while more frequent fire events, together with elephant impact, convert woodlands to grassland and hinder regeneration (Chidumayo 2004, Cumming et al. 1997, 34

Guy 1989, Gambiza, Bond, Frost, and Higgins 2000). Controlling fuel load and over exploitation by large herbivores helps the woodlands recover (Asner and Levick 2012).

In the Sengwa Wildlife Research Area (Zimbabwe), 48% of large trees in miombo woodland declined because of elephant impacts and fire, and following a decrease in elephant populations and fires, natural recruitment helped the woodlands recover

(Mapaure and Moe 2009). In KNP there was no significant changes in rainfall, from 1997 to 2012 (Figure 2.2 C). Hence, the observed restoration of savannah woodlands at KNP may be due increased protection effort and expansion of the national park in 1998 and possibly decrease of frequent wildfires.

Contrasting with our results, Banda et al. (2006), who assessed woody vegetation structure and composition in the Katavi ecosystem in 2000 and 2003, reported declining densities of woody trees, at a higher rate inside than outside the KNP, due to over browsing by large herbivores. The increase in canopy trees reported in the present study likely resulted from their recovery after 13 years of park protection in the extension area

(Figures 2.1 and 2.3). Banda et al. (2006) had likely not detected this recovery because their assessment was conducted only 5 years of after the park extension.

The reported increase of savannah, on the other hand, cannot be ascertained to be the result of increased protection effort or by human destruction, because savannah was merged with cropland in the imagery interpretation (see section 2.6). Observations made on Google Earth images as of 2013 confirm the existence of cultivated land inside the

KNP (Appendix E).

Our results show opposite trends in woody savannah between the two parks: decreasing in TNP and increasing in KNP. These findings are similar to those of Pelkey 35

et al. (2003) who reported differing amount of vegetation greenness (NDVI index) during the dry season across Tanzania between 1982 and 1994, which decreased in northern

Tanzania, where TNP is located, but increased in the other parts of the country, including the KNP ecosystem, where there the savannah woodlands were increasing. Huete (2002) correlated NDVI index with various canopy covers and found that the greener and denser the cover type was, the higher was the NDVI values were. Taking into account these two studies, the changes in amount of greenness, declining inside and outside TNP and increasing in KNP area, imply that vegetation condition is worsening in TNP and improving in KNP.

Wildlife species utilize different types of vegetation covers at different seasons.

For example, some species of herbivores such as elephants, prefer open woodland, shrubland and grassland during the wet season (Loarie et al. 2009) and closed woodlands during the dry season (Loarie et al. 2009; Owen-Smith 1988; Owen-Smith et al. 2006).

Migratory species as zebra and wildebeest migrate seasonally between inside and outside the parks for forage, breeding or calving (Western et al. 2009, Mwalyosi 1992; Mundia and Murayama 2009). The population of wildebeest in TNP declined from 40,000 to

5,000 between 1984 and 2000, due to degradation and conversion of woodland and grassland into farmland and settlement (Msoffe et al. 2011). In Kenya, wildebeest

(Connochaetes taurinus), Grant gazelle (Gazella granti), Thomson gazelle (Gazella thomsonii) and zebra (Equus burchellii) populations declined by 60% between 1975 and

2007 due to conversion of grasslands to modern agriculture and urbanization (Mundia and Murayama 2009). The observed increased bare land in TNP, a result of conversion of

grassland and woodlands, degrade their seasonal habitats and threatening their continual survival.

Skarpe et al. (2004) reported increase of buffalo, impalla, and greater kudu with increase of shrubland. The decline of open shrublands in KNP therefore may be of concern as may result in food shortage for small and medium sized browsing animals.

5.0 Conclusion

In this research we expected an increase of woodlands (savannah and woody savannah), via succession of shrublands, grasslands, swamps and bare lands, following increased efforts to protect wildlife species through implementation of the wildlife policy from 1999 onwards. The results indicate that the effect of protection on the dynamics of land cover classes differed between the two national parks, and inside and outside the parks. In TNP, we found significant increases of bare land inside and outside the park due to degradation of savannah woodlands, but neither of these changes was related to variations in the amount of rainfall. These results were contrary to our expectations, indicating that TNP may not be as effective as hoped after implementation of the wildlife policy. In KNP the woodlands, which were the dominant cover of the original ecosystem, have been restored, mostly inside the park as expected. However, this finding was possibly due to the extension of the protected area. Considering we could not separate savannah from farmed lands due to low resolution of the Landsat imagery data, and given the fact that farming activities were clear through Google Earth inside and outside the parks, we failed to conclude the findings of KNP as a success of the 1998 policy in protecting wildlife habitats. Management action is required to stop human encroachment

activities inside and outside the parks to ensure the maintenance of a healthy wildlife habitat for species survival.

6.0 Acknowledgments

The lead author wishes to thank Tanzania National Parks Headquarters in Arusha

Tanzania, for granting a research permit. TNP and KNP park management teams and ecology department, particularly James Wakibara and Abel Mtui (TNP), and Houston

Njamasi and Davis Mushi (KNP), kindly provided logistic information and accommodation. Joseph Mhina (Game Ranger, KNP) for safety keeping from wild animals. Luc Leblanc, Department of Plant and Environmental Protection Sciences,

University of Hawai‘i at Mānoa, provided huge assistance in the field. Norman Owen-

Smith, emeritus professor at the School of Animal, Plant and Environmental Sciences,

University of the Witwatersrand, Johannesburg, South Africa, gave constructive suggestions in designing the study. Travis Idol of the Department of Natural Resource and Environmental Management, University of Hawai‘i at Mānoa, kindly provided access to the FLAASH software, used for atmospheric correction of the satellite images used in this research. Landsat satellite images are a courtesy from the USGS Earth

Resources Observation and Science Centre. This study would not have been achieved without financial support from the Ford Foundation International Fellowships Program.

Table 2.1 Definition of land cover classes as identified at the Tarangire (TNP) and Katavi (KNP) National Parks (adapted and modified from Friedl et al. 2002) Land cover type Definition Bare land or less Lands with exposed soils, sand or rocks, with no more than 10% vegetation cover at any time of the year. vegetated land Built up/natural Lands with houses at least 40%, and 60% composed of natural vegetation (trees, shrubs grass, forbs), bare vegetation mosaic land (rock, soils or sand) (modified definition). Closed shrubland Land with woody vegetation (evergreen or deciduous) < 2 meters tall and with shrub canopy cover >60%. Lands covered with temporary crops followed by harvest and bare-soil period e.g. single and multiple Cropland cropping systems. Perennial woody crops are classified as the appropriate shrubland cover type. Grassland Lands with variety of grasses and forbs. Woody vegetation cover (trees and shrubs) is less than 10%. Lands with woody vegetation (evergreen or deciduous) < 2 meters tall and with shrub canopy cover Open shrubland between 10-60%. Lands with grasses, forbs, and other understory systems and with tree canopy cover between 10 and 30% Savannah The tree canopy cover height exceed 2 meters. Lands with mixture of water and grasses and forbs, or woody vegetation. The water often moving Swamp perceptibly, supporting low vegetation (e.g. sedges), reeds and woody vegetation. Water bodies Lakes, streams, rivers and water holes (boreholes). Lands with grasses, forbs and other understory systems, with canopy cover between 30 and 60%. The Woody savannah trees height exceeds 2m and can be deciduous or evergreen.

Table 2.2 (a) Error matrix of the classification map derived from the 2009 Landsat TM image of the TNP and adjacent areas.

Reference Data

% %

land

Row Row

Open Open

Total

Water Water

Cloud

bodies

Barren Barren

Closed

Woody Woody

Swamp

Shadow

% User User %

savannah

accuracy

Savannah

shrubland shrubland

Producer Producer Grassland Thematic classes Accuracy Barren land 83 0 0 0 0 0 0 1 0 0 84 84.69 98.81 Cloud 0 70 0 0 0 0 0 0 0 0 70 100 100 Grassland 11 0 141 1 0 0 1 5 1 2 162 71.94 87.04 Savannah 4 0 0 66 0 0 0 0 0 0 70 92.96 94.29 Shadow 0 0 0 0 90 0 0 0 0 0 90 100 100 Swamp 0 0 0 0 0 55 0 0 4 3 62 100 88.71 Water bodies 0 0 0 0 0 0 56 0 0 0 56 98.25 100 Open shrubland 0 0 12 4 0 0 0 185 68 0 269 95.36 68.77 Closed shrubland 0 0 2 0 0 0 0 1 221 7 231 73.67 95.67 Woody savannah 0 0 41 0 0 0 0 2 6 190 239 94.06 79.5 Column Total 98 70 196 71 90 55 57 194 300 202 1333

Overall accuracy = 86.7967%; Kappa Coefficient (K-hat) = 0.8479.

Table 2.2 (b) Error matrix of the classification map derived from the 2011 Landsat TM image of the KNP and adjacent areas.

Reference Data

% %

land

vannah

Row Row

Open Open

Total

Water

Barren Barren

Closed

Woody Woody

Swamp

% User User %

savannah

accuracy accuracy

shrubland shrubland Producer Producer Thematic classes Grassland Barren land 89 0 0 0 0 4 0 0 93 93.68 95.7 Closed shrubland 0 104 0 1 0 14 0 1 120 87.39 86.67 Grassland 0 0 126 0 30 0 0 21 177 77.3 71.19 Open shrubland 0 2 0 5 0 0 0 0 7 62.5 71.43 Savannah 0 5 14 0 150 2 0 33 204 79.37 73.53 Swamp 0 7 0 0 9 154 8 2 180 87.5 85.56 Water 0 0 0 0 0 1 4 0 5 33.33 80 Woody savannah 6 1 23 2 0 1 0 392 425 87.31 92.24 Column Total 95 119 163 8 189 176 12 449 1211

Overall accuracy = 84.6%, and Kappa coefficient = 0.804.

Table 2.3 Determining land cover trends over the past 20 years in Katavi and Tarangire National parks [Sample size for individual and all land cover classes combined are 6 and 36, respectively. Df Error = 2 for individual classes and 32 for all classes combined.

Park Land cover type

Variables Bare or less Closed Grassland Open Savannah Swamps Woody Water vegetated land shrubland shrubland savannah

KNP Year ns ns F 1,2 = 69.56, ns ns ns ns ns p = 0.014 Location ns ns F 1,2 = 33.22, ns ns ns ns ns p = 0.029 Location*Year ns ns F 1,2 = 32.89, ns ns ns ns ns p = 0.029 TNP Year F 1,2 = 189.34, ns ns ns ns ns n/a p = 0.005 Location ns ns ns ns ns n/a ns Location*Year ns ns ns ns ns n/a ns

Note: ns means not significant at α ≤ 0.05; n/a indicate where analysis was not performed because of sufficient data.

Table 2.4 Overall estimates for open shrubland and barren land which was observed to have significant change in amount of coverage over time.

Land cover class Park Parameter Estimate ± Error Open shrubland KNP Trend ‘Year’ -7.214 ± 0.865 Trend ‘Inside’ -12.175 ± 1.223 Trend ‘Outside’ -2.254 ± 1.223 Bare / less vegetated land TNP Trend ‘Year’ 2.239 ± 0.163

Note: Bolded values were significant at 95% confidence intervals.

Table 2.5 Relationship between type of Log10[land cover] vs. log10[rainfall] in Tarangire National Parks [For all parameters Df= 1, DF error = 2].

Variable Land cover type Bare Closed Open Open Woody land shrubs Grassland shrubs shrubs Savannah Swamps savannah ns ns ns ns ns ns ns ns Rain ns ns ns ns ns ns ns ns Location ns ns ns ns ns ns ns ns Rain*Location

Note: ns means not significant at α ≤ 0.05.

Table 2.6 (a) Land covers change (km2) inside TNP from 1988 to 1999. The rows and the columns present land cover classes for 1988 and 1999 respectively. Change (km2) on land cover classes between the years is shown at the last two rows of the table. The diagonal values (bolded) show amount of cover (km2) that remained unchanged over 10 years period while the off diagonal values show amount that was changed to another class.

Total (km2) per Change (1999-

1999 year 1988)

)

(%)

1988 1999

Open Open

(km

Closed

Woody Woody

Swamp

savannah

Bare land Bare

Savannah

shrubland shrubland 1988 Grassland Bare land 0.8 0.1 2.8 0.3 2.0 0.0 0.2 6.1 28.6 22.5 368.9 Closed shr. 2.2 158.2 52.7 138.4 40.6 94.8 143.4 630.4 489.1 -141.3 -22.4 Grassland 15.3 106.1 211.5 149.0 128.9 105.1 160.0 875.8 446.9 -428.8 -49.0 Open shr. 0.3 72.4 16.8 94.7 36.0 25.9 48.0 294.1 496.6 202.5 68.9 Savannah 2.4 4.2 18.0 9.3 24.0 1.2 5.7 64.8 308.5 243.7 376.1 Swamp 0.9 50.8 26.6 34.3 11.1 75.1 46.6 245.3 367.1 121.7 49.6 Woody sav. 6.8 97.3 118.5 70.6 66.0 64.9 69.3 493.4 473.1 -20.3 -4.1 Total 2609.9 2609.9 0.0

Table 2.6 (b) Land covers change (km2) inside TNP from 1999 to 2009. The rows and the columns present land cover classes for 1999 and 2009, respectively. Change (km2) on land cover classes between the years is shown at the last two rows of the table. The diagonal values (bolded) show amount of cover (km2) that remained unchanged over 10 years period while the off diagonal values show amount that was changed to another class.

Total (km2) per Change (2009

2009 year – 1999)

)

pen pen

(%)

1999 2009

(km

Water

Cloud

Closed

Woody Woody

Swamp

Shadow

savannah

Bare land Bare

Savannah

shrubland shrubland 1999 Grassland Bare land 2.0 0.3 20.1 0.5 1.3 2.0 0.7 0.5 0.5 0.8 28.6 58.5 29.9 104.8 Closed shr. 8.7 155.4 136.9 75.6 1.9 60.3 28.7 0.1 7.8 13.8 489.1 456.7 -32.4 -6.6 Grassland 10.5 23.7 225.8 23.8 13.8 89.3 31.7 1.9 7.5 18.9 446.9 910.8 463.8 103.8

Open shr. 16.6 128.6 120.8 90.0 6.6 53.8 64.1 0.2 4.3 11.6 496.6 301.0 -195.5 -39.4

Savannah 9.3 45.5 124.9 48.0 20.7 14.9 33.3 0.1 4.0 7.8 308.5 46.0 -262.5 -85.1 Swamp 3.6 48.0 125.7 31.0 0.3 123.9 19.6 0.2 6.0 9.0 367.1 502.2 135.1 36.8

Woody sav. 7.9 55.3 156.7 32.1 1.4 158.1 44.3 0.3 5.8 11.3 473.1 222.3 -250.8 -53.0 Water ------3.2 3.2 Shadow ------0.0 35.9 35.9 Cloud ------0.0 73.2 73.2 Total 2609.9 2609.9 0.0

Table 2.6 (c) Land covers change (km2) inside TNP from 1988 to 2009. The rows and the columns present land cover classes for 1988 and 2009 respectively. Change (km2) on land cover classes between the years is shown at the last two rows of the table. The diagonal values (bolded) show amount of cover (km2) that remained stable over 21 years period while the off diagonal values show amount that was changed to another class.

Total (km2) per Change (2009

2009 year – 1999)

)

(%)

1988 2009

Open Open

(km

Water

Cloud

Closed

Woody Woody

Swamp

Shadow

savannah

Bare land Bare

Savannah

shrubland shrubland 1988 Grassland Bare land 0.6 0.0 3.1 0.1 1.6 0.2 0.2 0.0 0.0 0.2 6.1 58.5 52.4 860.3 Closed shr. 12.0 161.7 159.3 70.1 1.5 162.9 36.4 0.4 9.4 16.6 630.4 456.7 -173.6 -27.5 Grassland 21.3 107.6 364.8 89.0 22.1 160.2 72.1 1.2 12.0 25.3 875.8 910.8 35.0 4.0

Open shr. 8.8 72.9 73.8 65.4 2.8 23.6 38.0 0.0 2.3 6.5 294.1 301.0 7.0 2.4

Savannah 4.4 3.4 25.5 7.6 12.9 1.7 8.1 0.0 0.3 1.0 64.8 46.0 -18.8 -28.9 Swamp 4.7 42.7 72.8 24.3 0.2 72.1 16.1 0.2 4.6 7.7 245.3 502.2 256.9 104.7

Woody sav. 6.6 68.5 211.6 44.5 4.8 81.7 51.5 1.2 7.2 15.8 493.4 222.3 -271.1 -54.9 Water 0.0 3.2 3.2 Shadow 0.0 35.9 35.9 Cloud 0.0 73.2 73.2 2609.9 2609.9 0.0

Table 2.7 (a) Land covers change (km2) outside TNP from 1988 to 1999. The rows and the columns present land cover classes for 1988 and 1999 respectively. Change (km2) on land cover classes between the years is shown at the last two rows of the table. The diagonal values (bolded) show amount of cover (km2) that remained stable over 10 years period while the off diagonal values show amount that was changed to another class.

Total (km2) per

1999 year Change (1999 - 1988)

)

(%)

1988 1999

Open Open

(km

Water

Closed

Woody Woody

Swamp

savannah

Bare land Bare

Savannah

shrubland shrubland 1988 Grassland Bare land 0.8 0.0 1.3 0.3 0.8 0.0 0.2 0.0 3.5 28.6 25.0 710.8 Closed shr. 0.4 94.7 21.4 137.7 9.1 15.5 67.4 0.0 346.2 182.2 -163.9 -47.4 Grassland 17.4 23.8 118.5 119.9 64.6 13.8 91.2 4.5 453.7 206.8 -246.9 -54.4 Open shr. 0.1 22.6 5.8 76.6 4.2 4.0 20.9 0.0 134.2 398.1 263.9 196.7 Savannah 5.1 1.2 13.7 6.6 22.2 0.3 3.7 0.0 52.6 112.4 59.8 113.6 Swamp 0.5 24.8 9.1 27.6 2.2 11.7 33.3 0.2 109.3 50.5 -58.8 -53.8 Woody sav. 4.3 15.1 36.9 29.5 9.3 5.1 29.4 9.9 139.6 246.0 106.3 76.2 Water 0.0 0.0 0.0 0.0 0.0 0.0 0.0 22.1 22.1 36.8 14.6 66.1 Total 1261.3 1261.3 0.0

Table 2.7 (b) Land covers change (km2) outside TNP from 1999 to 2009. The rows and the columns present land cover classes for 1999 and 2009, respectively. Change (km2) on land cover classes between the years is shown at the last two rows of the table. The diagonal values (bolded) show amount of cover (km2) that remained stable over 10 years period while the off diagonal values show amount that was changed to another class.

Total (km2) per Change (2009 –

2009 year 1999)

(%)

1999 2009

Open Open

Wat

(km2)

Cloud

Closed

Woody Woody

Swamp

Shadow

savannah

Bare land Bare

Savannah

shrubland shrubland 1999 Grassland Bare land 3.5 0.6 14.9 0.3 4.9 1.2 2.3 0.1 0.1 0.7 28.6 45.2 16.7 58.4 Closed shr. 3.3 106.3 16.2 10.6 1.4 30.3 8.4 0.4 2.8 2.5 182.2 410.7 228.4 125.3 Grassland 8.8 20.3 93.2 8.2 15.9 35.5 13.6 1.5 2.0 7.7 206.8 282.9 76.1 36.8 Open shr. 17.1 195.3 53.9 43.6 7.6 30.2 40.6 0.2 3.0 6.7 398.1 83.8 -314.3 -79.0 Savannah 4.3 9.0 49.3 4.7 24.0 7.3 9.6 0.0 1.0 3.2 112.4 56.3 -56.1 -49.9 Swamp 0.9 11.3 9.5 3.3 0.2 19.0 3.0 1.0 1.0 1.4 50.5 200.9 150.4 297.7 Woody sav. 7.3 67.9 45.8 13.1 2.5 76.9 22.0 0.6 3.8 6.1 246.0 99.5 -146.5 -59.5 Water 0.0 0.0 0.1 0.0 0.0 0.4 0.0 36.2 0.0 0.0 36.8 40.1 3.3 9.1 Shadow ------0.0 13.7 13.7 Cloud ------0.0 28.3 28.3 Total 1261.3 1261.3 0.0

Table 2.7 (c) Land covers change (km2) outside TNP from 1988 to 2009. The rows and the columns present land cover classes for 1988 and 2009, respectively. Change (km2) on land cover classes between the years is shown at the last two rows of the table. The diagonal values (bolded) show amount of cover (km2) that remained stable over 21 years period while the off diagonal values show amount that was changed to another class.

Total (km2) per Change (2009 –

2009 year 1988)

)

pen pen

(%)

19 2009

(km

Water

Cloud

Closed

Woody Woody

Swamp

Shadow

savannah

Bare land Bare

Savannah

shrubland shrubland 1988 Grassland Bare land 0.6 0.0 1.6 0.0 1.1 0.0 0.2 0.0 0.0 0.0 3.5 45.2 41.7 1184.3 Closed shr. 8.8 181.0 28.9 28.6 2.6 65.5 20.6 0.2 4.2 5.8 346.2 410.7 64.5 18.6 Grassland 18.1 80.5 162.3 30.6 31.6 69.5 37.2 6.6 5.7 11.7 453.7 282.9 -170.9 -37.7 Open shr. 6.3 72.2 14.9 14.0 1.0 8.1 14.8 0.0 0.9 2.0 134.2 83.8 -50.4 -37.6 Savannah 3.6 1.5 17.5 1.7 16.0 1.2 9.1 0.0 0.3 1.6 52.6 56.3 3.7 7.0 Swamp 3.9 43.8 9.9 3.2 0.9 34.8 7.6 0.4 1.6 3.2 109.3 200.9 91.6 83.8 Woody sav. 3.9 31.6 47.7 5.7 3.2 21.8 10.1 10.7 1.1 4.0 139.6 40.1 -99.5 -71.3 Water 0.0 0.0 0.0 0.0 0.0 0.0 0.0 22.1 0.0 0.0 22.1 99.5 77.4 349.6 Shadow 13.7 13.7 Cloud 28.3 28.3 Total 1261.3 1261.3 0.0

Table 2.8 (a) Land covers change (km2) inside the KNP from 1984 to 1999. The rows and the columns present land cover classes for 1984 and 1999 respectively. The total change between land cover classes between the years is presented on the last two columns. The diagonal values (bolded) show amount of cover that remained stable over 10 years period, while the off diagonal values show amount of cover that was converted to another class. Cloud and shadow are not included in the matrix.

Total (km2) per Change (1999

1999 year – 1984)

)

(%) (%)

land land

Bare

1984 1999

Open Open

(km

Water

Cloud

Closed

Woody Woody

Swamp

Shadow

savannah

Savannah

shrubland shrubland 1984 Grassland Bare land 27.2 9.6 26.9 7.2 30.5 2.0 20.0 1.7 0.0 0.0 125.1 286.4 161.4 129.0 Closed shr. 20.4 88.4 227.1 17.0 42.6 5.2 112.2 7.4 0.0 0.0 520.2 582.3 62.1 11.9 Grassland 90.7 217.5 609.4 36.6 240.8 13.5 196.0 25.5 0.0 0.0 1430.0 1587.0 157.0 11.0 Open shr. 24.4 23.5 116.3 14.8 110.6 2.2 54.2 4.5 0.0 0.0 350.5 132.5 -218.0 -62.2 Savannah 49.1 46.1 236.1 20.3 184.7 4.3 88.2 9.8 0.0 0.0 638.6 770.3 131.7 20.6 Swamp 11.8 25.6 57.2 8.2 33.8 30.3 34.0 11.1 0.0 0.0 212.2 66.9 -145.3 -68.5 Woody sav. 57.4 163.4 258.0 24.7 107.4 7.6 208.9 23.5 0.0 0.0 851.0 726.8 -124.2 -14.6 Water 0.2 0.2 1.8 0.1 0.3 0.1 0.4 0.4 0.0 0.0 3.5 86.1 82.6 2379.7 Shadow 2.2 4.2 27.1 1.6 9.2 1.0 6.9 1.2 0.0 0.0 53.5 0.0 -53.5 Cloud 3.0 3.7 27.2 1.9 10.4 0.6 6.0 1.1 0.0 0.0 53.8 0.0 -53.8 Total 4238.2 4238.2 0.0

Table 2.8 (b) Land covers change (km2) inside the KNP from 1999 to 2011. The rows and the columns present land cover classes for 1999 and 2011, respectively. The total change between land cover classes between the years is presented on the last two columns. The diagonal values (bolded) show amount of cover that remained stable over 10 years period, while the off diagonal values show amount of cover that was converted to another class. Cloud and shadow are not included in the matrix.

2 Total (km ) per Change (2011 –

Year 2011 year 1999)

)

(%)

2011

1999

Open Open

(km

Water

Closed

Woody Woody

Swamp

savannah

Bare land Bare

Savannah

shrubland shrubland 1999 Grassland Bare land 33.9 11.5 37.5 2.3 135.5 5.5 60.2 0.0 286.4 173.4 -113.0 -39.5 Closed shr. 14.6 60.2 31.4 6.4 122.6 20.2 326.9 0.0 582.3 222.5 -359.7 -61.8 Grassland 8.4 73.9 143.2 2.7 574.5 43.4 740.8 0.2 1587.0 418.6 -1168.4 -73.6 Open shr. 14.4 12.3 15.3 3.8 56.9 5.5 24.2 0.1 132.5 24.5 -108.0 -81.5 Savannah 39.3 27.9 115.4 3.3 438.8 13.8 131.6 0.2 770.3 1563.7 793.4 103.0 Swamp 27.8 0.8 2.8 0.0 17.1 6.7 11.6 0.1 66.9 133.3 66.4 99.4 Woody sav. 26.3 33.1 69.2 5.3 189.2 21.9 381.7 0.2 726.8 1700.3 973.5 133.9 Water 8.7 3.0 3.7 0.7 29.1 16.2 23.4 1.1 86.1 1.9 -84.2 -97.8 Total 4238.2 4238.2 0.0

Table 2.8 (c) Land covers change (km2) inside the KNP from 1984 to 2011. The rows and the columns present land cover classes for 1984 and 2011 respectively. The total change between land cover classes between the years is presented on the last two columns. The diagonal values (bolded) show amount of cover that remained stable over 27 years period, while the off diagonal values show amount of cover that was converted to another class. Cloud and shadow are not included in the matrix.

Total (km2) per Change (2011 –

2011 year 1984)

)

(%)

land

Bare

2011

1984

Open Open

(km

Water

Cloud

Closed

Woody Woody

Swamp

Shadow

savannah

Savannah

shrubland shrubland 1984 Grassland Bare land 16.9 7.4 15.9 2.0 40.7 1.7 40.4 0.0 0.0 0.0 125.1 173.4 48.4 38.7 Closed shr. 18.5 38.9 45.9 4.4 141.5 17.9 252.9 0.1 0.0 0.0 520.2 222.5 -297.7 -57.2 - Grassland 22.9 73.3 117.6 2.5 629.5 41.9 541.7 0.5 0.0 0.0 1430.0 418.6 1011.4 -70.7 Open shr. 16.8 18.6 47.2 3.9 153.7 6.2 104.1 0.0 0.0 0.0 350.5 24.5 -326.0 -93.0 Savannah 23.5 29.1 94.1 2.1 281.8 12.5 195.3 0.1 0.0 0.0 638.6 1563.7 925.1 144.9 Swamp 31.8 9.0 10.8 0.7 79.9 21.3 58.5 0.3 0.0 0.0 212.2 133.3 -78.9 -37.2 Woody sav. 39.2 42.3 76.4 8.0 182.2 26.1 476.0 0.7 0.0 0.0 851.0 1700.3 849.4 99.8 Water 0.1 0.1 0.3 0.0 1.4 0.7 0.7 0.1 0.0 0.0 3.5 1.9 -1.6 -45.6 Shadow 1.9 2.0 4.8 0.4 27.1 2.4 14.9 0.0 0.0 0.0 53.5 0.0 -53.5 -100.0 Cloud 1.7 1.9 5.4 0.5 26.0 2.4 15.9 0.0 0.0 0.0 53.8 0.0 -53.8 -100.0 Total 4238.2 4238.2 0.0

Table 2.9 (a) Land cover change (km2) outside the KNP from 1984 to 1999. The rows and the columns present land cover classes for 1984 and 1999 respectively. The total change between land cover classes between the years is presented on the last two columns. The diagonal values (bolded) show amount of cover that remained stable over 10 years period, while the off diagonal values show amount of cover that was converted to another class. Cloud and shadow are not included in the matrix.

Total (km2) per Change (1999

1999 year - 1984)

)

nah

(%)

land

Bare

1984 1999

Open Open

(km

Water

Cloud Closed

1984 Woody

Swamp

Shadow

savan

Savannah

shrubland shrubland Grassland Bare land 6.1 6.2 11.2 5.8 9.1 0.5 11.1 4.0 0.0 0.0 54.1 92.1 38.0 70.3 Closed shr. 9.7 47.0 75.2 17.7 20.9 1.5 65.1 12.1 0.0 0.0 249.3 344.6 95.3 38.2 Grassland 27.6 135.3 214.7 9.9 55.7 0.5 137.1 12.8 0.0 0.0 593.6 628.6 35.0 5.9 Open shr. 5.7 8.5 33.0 5.5 17.8 0.2 19.6 4.4 0.0 0.0 94.8 77.3 -17.4 -18.4 Savannah 9.5 14.5 67.9 4.6 27.5 0.1 30.9 5.8 0.0 0.0 161.0 203.1 42.1 26.2 Swamp 5.7 13.7 45.7 8.3 22.2 1.9 20.4 5.7 0.0 0.0 123.5 7.9 -115.6 -93.6 Woody sav. 24.3 109.9 122.6 23.0 35.9 3.1 199.7 17.2 0.0 0.0 535.7 496.0 -39.6 -7.4 Water 0.1 0.6 3.9 0.3 0.3 0.0 0.6 0.1 0.0 0.0 5.9 64.5 58.6 993.1 Shadow 0.6 3.4 15.4 0.6 3.2 0.0 3.5 1.6 0.0 0.0 28.3 0.0 -28.3 Cloud 2.8 5.4 39.0 1.6 10.5 0.0 8.0 0.8 0.0 0.0 68.1 0.0 -68.1 Total 1914.3 1914.3 0.0

Table 2.9 (b) Land cover change (km2) outside the KNP from 1999 to 2011. The rows and the columns present land cover classes for 1999 and 2011, respectively. The total change between land cover classes between the years is presented on the last two columns. The diagonal values (bolded) show amount of cover that remained stable over 10 years period, while the off diagonal values show amount of cover that was converted to another class. Cloud and shadow are not included in the matrix.

Total (km2) per Change (1999 -

2011 year 1984)

)

(%)

2011

1999

Open Open

(km

Water

Closed Woody Woody

1999 Swamp

savannah

Bare land Bare

Savannah

shrubland shrubland Grassland Bare land 13.49 4.73 9.78 1.53 35.82 2.37 24.35 0.02 92.1 193.3 101.2 109.9 Closed shr. 19.89 23.17 11.73 8.72 51.00 9.10 220.96 0.05 344.6 108.1 -236.5 -68.6 Grassland 23.91 39.68 55.52 8.48 241.45 26.08 232.86 0.65 628.6 125.7 -502.9 -80.0 Open shr. 33.92 5.54 3.18 2.34 12.58 4.74 15.02 0.01 77.3 32.7 -44.7 -57.8 Savannah 37.42 7.41 20.04 2.99 98.84 3.54 32.85 0.07 203.1 525.6 322.5 158.7 Swamp 3.89 0.27 0.05 0.05 0.22 0.65 2.78 0.00 7.9 65.2 57.3 723.9 Woody sav. 30.55 23.68 24.38 6.67 78.18 16.49 315.96 0.12 496.0 862.7 366.7 73.9 Water 30.18 3.66 1.02 1.87 7.52 2.27 17.97 0.02 64.5 0.95 -63.6 -98.5 Total 1914.3 1914.3 0.0

Table 2.9 (c) Land cover change (km2) outside the KNP from 1984 to 2011. The rows and the columns present land cover classes for 1984 and 2011 respectively. The total change between land cover classes between the years is presented on the last two columns. The diagonal values (bolded) show amount of cover that remained stable over 27 years period, while the off diagonal values show amount of cover that was converted to another class. Cloud and shadow are not included in the matrix.

Total (km2) per Change (2011 -

2011 year 1984)

)

(%)

land

Bare

2011

1984

Open Open

(km

Water

Cloud

Closed

Woody Woody Swamp

1984 Shadow

savannah

Savannah

shrubland shrubland Grassland Bare land 17.1 4.7 3.7 1.5 10.7 1.2 15.2 0.0 0.0 0.0 54.1 193.3 139.2 257.3 Closed shr. 40.8 19.4 16.6 8.0 54.6 10.6 99.2 0.0 0.0 0.0 249.3 108.1 -141.1 -56.6 Grassland 26.2 30.3 30.9 6.5 183.9 19.5 295.8 0.5 0.0 0.0 593.6 125.7 -467.9 -78.8 Open shr. 14.8 7.0 11.1 2.5 30.4 1.8 27.2 0.0 0.0 0.0 94.8 32.7 -62.1 -65.5 Savannah 14.0 10.3 19.9 2.0 58.2 2.8 53.8 0.1 0.0 0.0 161.0 525.6 364.6 226.4 Swamp 21.5 4.5 4.5 1.3 49.5 8.0 34.4 0.1 0.0 0.0 123.5 65.2 -58.3 -47.2 Woody sav. 56.3 28.9 28.5 9.7 86.3 15.1 310.7 0.2 0.0 0.0 535.7 862.7 327.1 61.1 Water 0.0 0.2 0.7 0.0 2.6 0.8 1.5 0.0 0.0 0.0 5.9 0.9 -5.0 -84.0 Shadow 1.8 1.2 2.1 0.5 11.0 2.1 9.6 0.0 0.0 0.0 28.3 0.0 -28.3 -100.0 Cloud 0.7 1.8 7.7 0.6 38.5 3.3 15.4 0.0 0.0 0.0 68.1 0.0 -68.1 -100.0 Total 1914.3 1914.3 0.0

Table 3.0 (a) Change from open shrubland into other types of land cover inside and outside Katavi National Park.

From To Inside Outside 1984-2011 (km2) %cover 1984-2011 (km2) %cover Open shrubland Bare / less vegetated land 16.8 4.8 14.8 16.0 Closed shrubs 18.6 5.4 7.0 7.6 Grassland 47.2 13.6 11.1 12.0 Savannah 153.7 44.3 30.4 32.9 Swamps 6.2 1.8 1.8 2.0 Woody savannah 104.1 30.0 27.2 29.4 Bare / less vegetated land 16.8 4.8 14.8 16.0

Table 3.0 (b) Change in the types of land cover that contributed to increase in the area of bare or less vegetated land inside and outside Tarangire National Park.

From To Inside Outside 1988-2009 (km2) %cover 1988-2009 (km2) %cover Closed shrubs Bare or less 12.0 20.6 8.8 19.4 Grassland vegetated land 21.3 36.4 18.1 40.0 Open shrubs 8.8 15.1 6.3 13.9 Savannah 4.4 7.5 3.6 8.0 Swamps 4.7 8.0 3.9 8.6 Woody savannah 6.6 11.3 3.9 8.6

Figure 2.1 Tarangire National Park (TNP, right) and Katavi National Park (KNP, left) in Tanzania. The Katavi map show the main

Katuma river, which flow from north west towards south east, the major swamps that harbor high density of large mammals, particularly during dry seasons, and the adjacent areas. The Tarangire map also shows the major swamps, the Tarangire River, and adjacent areas bordering the park. 59

F1,15 =3.83, p = 0.123

F1,12 =0.06, p = 0.818

Figure 2.2 Total annual rainfall (dotted line) and average annual rainfall (straight line) recorded in Tarangire (A) and Katavi (B),

National Parks head offices and the scatter plots of Katavi (C) and Tarangire (D) showing change over time.

Figure 2.3 Land cover maps of TNP (below) and KNP (upper) and their buffer areas derived from the Landsat imagery of 1988, 1999 and 2009 for TNP, and 1984, 1999 and 2011 for KNP.

Appendices

Appendix A: Land cover photos produced as field guide after the ground preliminary survey in 2011.

Appendix B (i): Checking the 2012 field data on 2009 SPOT/Geoeye photos in order to obtain training data for classification of the 2009 TM Landsat image for Tarangire National park.

Remained data after checking on 2009 2012 field data SPOT/Geoeye photos Land cover class Polygon Pixels Polygon Pixels Barren land 7 37 0 0 Built-up/ Natural vegetation mosaic 6 57 0 0 Closed shrubland 5 20 5 20 Cropland 3 23 0 0 Grassland 11 33 8 13 Open shrubland 7 18 6 19 Savannah 7 22 5 8 Swamp 4 5 3 4 Water 2 175 0 0 Woody savannah 7 41 4 20 Total 59 431 31 84

Appendix B (ii): Training and reference data used for classification of the 2009 TM Landsat image and accuracy assessment.

Training data Reference data Land cover class Polygon Pixels Polygon Pixels Barren land 4 45 4 98 Closed shrubland 7 45 3 300 Cloud 5 51 5 70 Grassland 8 30 2 196 Open shrubland 8 30 3 194 Savannah 9 79 2 71 Shadow 4 63 5 90 Swamp 8 51 2 55 Water 6 63 1 57 Woody savannah 8 36 6 202 Total 67 493 33 1333

Appendix C (i): Checking the 2012 field data on 2011 SPOT /Geoeye photos in order to obtain training data for classification of the 2011 TM Landsat image for Katavi National park.

Remained data after checking on 2011 2012 field data SPOT/Geoeye photos Land cover class Polygon Pixels Polygon Pixels Bare land 6 20 5 24 Closed shrubland 5 16 5 16 Cropland 3 31 3 31 Grass 6 37 5 51 Natural vegetation mosaic 5 38 4 13 Open shrubland 4 17 5 28 Savannah 6 31 5 23 Swamp 2 32 5 38 Water 2 11 4 13 Woody savannah 15 118 6 33 Total 54 351 47 270

Appendix C (ii): Training and testing dataset used to classify the 2011 Landsat image.

Training data Reference data Land cover class Polygon Pixels Polygon Pixels Bare land 5 24 4 95 Closed shrubland 5 16 2 119 Grassland 5 51 8 163 Open shrubland 4 13 1 8 Savannah 5 28 4 189 Swamp 5 23 4 176 Water 4 13 4 12 Woody savannah 6 33 5 449 Total 39 201 32 1211

Appendix D: Determining land cover trends over the past 20 years in Katavi and Tarangire National parks [Sample size for individual and all land cover classes combined are 6 and 36, respectively. Df Error = 2 for individual classes and 32 for all classes combined.

Katavi Tarangire Land cover Type III Mean Type III Mean class Source DF SS Square F Value Pr > F SS Square F Value Pr > F Closed Year 1 42054.14 42054.14 1.550 0.340 3428.95 3428.95 0.250 0.668 shrubland Location 1 6057.39 6057.39 0.220 0.684 13905.14 13905.14 1.000 0.422 Year*Location 1 5840.69 5840.69 0.210 0.689 13643.42 13643.42 0.980 0.426 Open Year 1 38096.79 38096.79 69.560 0.014 195.90 195.90 0.000 0.951 shrubland Location 1 18196.42 18196.42 33.220 0.029 695.74 695.74 0.020 0.909 Year*Location 1 18013.13 18013.13 32.890 0.029 740.40 740.40 0.020 0.906 Year 1 495190.87 495190.87 2.470 0.257 6009.34 6009.34 0.080 0.804 Grassland Location 1 68302.17 68302.17 0.340 0.619 9231.31 9231.31 0.120 0.759 Year*Location 1 65863.65 65863.65 0.330 0.624 9670.88 9670.88 0.130 0.754 Year 1 392022.15 392022.15 6.870 0.120 6.82 6.82 0.000 0.988 Savannah Location 1 71783.03 71783.03 1.260 0.379 72.18 72.18 0.000 0.960 Year*Location 1 74326.30 74326.30 1.300 0.372 66.41 66.41 0.000 0.962 Year 1 5710.36 5710.36 1.000 0.423 29728.30 29728.30 7.720 0.109 Swamp Location 1 131.85 131.85 0.020 0.893 6864.37 6864.37 1.780 0.314 Year*Location 1 120.91 120.91 0.020 0.898 7085.09 7085.09 1.840 0.308 Woody Year 1 312645.33 312645.33 2.330 0.267 23014.76 23014.76 2.230 0.274 savannah Location 1 59884.03 59884.03 0.450 0.573 13657.61 13657.61 1.320 0.369 Year*Location 1 61424.03 61424.03 0.460 0.569 13371.24 13371.24 1.290 0.373 Year 1 9498.243 9498.243 1.460 0.351 2212.851 2212.851 189.340 0.005 Location 1 1573.933 1573.933 0.240 0.672 26.977 26.977 2.310 0.268 Barren land Year*Location 1 1530.342 1530.342 0.230 0.676 27.268 27.268 2.330 0.266 Year 1 4.292 4.292 0.000 0.976 - - - - Location 1 6.008 6.008 0.000 0.971 - - - - Water Year*Location 1 6.234 6.234 0.000 0.970 - - - -

Appendix E: Relationship between type of Log10[land cover] vs. log10[rainfall] in Tarangire National Parks [For all parameters Df = 1, DF error = 2].

Closed Woody Bare land shrubs Grassland Open shrubs Savannah Swamps savannah F P F P F P F P F P F P F P Rain 0.22 0.687 2.16 0.279 1.17 0.393 10.99 0.080 6.04 0.133 7.07 0.117 0.07 0.821 Location 0.01 0.948 2.99 0.226 0.38 0.599 3.61 0.198 1.28 0.375 1.69 0.324 1.96 0.297 Rain*Location 0.00 0.955 3.53 0.201 0.25 0.669 3.05 0.223 1.38 0.361 2.36 0.265 2.67 0.244

A B

C D

Appendix G (i): Cultivated land inside and outside the Tarangire National Park (TNP). The green line shows park boundary, white lines show the farmed plots lettered A to C inside, and D outside the park. See geographical locations of each plot on Appendix G (ii).

Appendix G (ii): Satellite imagery date, geographical coordinates (taken approximately at the centre of the plot), elevation (m), and eye altitude (km) for each farmed plots in TNP.

TNP Image plot Google Earth Imagery Date Latitude Longitude Elevation (m) Eye altitude (km) A 1/8/2012 -4.075711° 35.833514° 1243 2.24 B 1/8/2012 -4.105499° 35.873422° 1164 2.36 C 10/24/2011 -4.520173° 36.040004° 1190 2.84 D 8/13/2013 -4.245198° 36.363452° 1343 2.38

Appendix H (i): Cultivated land inside and outside the Katavi National Park (KNP). The green line shows park boundary and white lines show the farmed plots inside the park. See geographical locations of each plot on Appendix H (ii).

Appendix H (ii): Satellite imagery date, geographical coordinates (taken approximately at the centre of the plot), elevation (m), and eye altitude (km) for the KNP farmed plots.

Google Earth Imagery Date Latitude Longitude Elevation (m) Eye altitude (km) 7/5/2013 -6.642068° 31.177376° 1067 1.46

8.0 References

Asner, G.P., and Levick, S.R. (2012), “Landscape-scale effects of herbivores on treefall

in African savannas,” Ecology Letters, 15, 1211–1217.

Banda, T., Schwartz, M.W., and Caro, T. (2006), “Woody vegetation structure and

composition along a protection gradient in a miombo ecosystem of western

Tanzania,” Forest Ecology and Management, 230, 179–185.

Barnes, R.F.W. (1983), “The Elephant Problem in Ruaha National Park, Tanzania,”

Biological Conservation, 26, 127–148.

Beale, C.M., Rensberg, S. van, Bond, W.J., Coughenour, M., Fynn, R., Gaylard, A.,

Grant, R., Harris, B., Jones, T., Mduma, S., Owen-Smith, N., and Sinclair, A.R.E.

(2013), “Ten lessons for the conservation of African savannah ecosystems,”

Biological Conservation, 167, 224–232.

Brenneman, R.A., Bagine, R.K., Brown, D.M., Ndetei, R., and Edward, E.L. Jr. (2009),

“Implications of closed ecosystem conservation management : the decline of

Rothschild’s giraffe (Giraffa camelopardalis rothschildi) in Lake Nakuru

National Park, Kenya,” African Journal of Ecology, 47, 711–719.

Bruner, A.G., Gullison, R.E., Rice, R.E., and Fonseca, G.A. (2001), “Effectiveness of

parks in protecting tropical biodiversity,” Science, 291, 125–128.

Caro, T.M. (1999), “Densities of Mammals in Partially Protected Areas: The Katavi

Ecosystem of Western Tanzania,” Journal of Applied Ecology, 36, 205–217.

Chidumayo, E.N. (2004), “Development of Brachystegia-Julbernardia woodland after

clear-felling in central Zambia: Evidence for high resilience,” Applied Vegetation

Science, 7, 237–242.

Coppin, P., Jonckheere, I., Nackaerts, K., and Muys, B. (2004), “Review Article Digital

change detection methods in ecosystem monitoring : a review,” International

Journal of Remote Sensing, 25, 1565–1596.

Cumming, D.H.M., Fenton, M.B., Rautenbach, I.L., Taylor, R.D., Cumming, G.S.,

Cumming, M.S., Dunlop, J.M., Ford, G.S., Hovorka, M.D., Johnston, D.S.,

Kalcounis, M.C., Mahlanga, Z., and Portforts, C.V. (1997), “Elephants,

woodlands and Biodiversity in Southern Africa,” South African Journal of

Science, 93, 231–236.

Fisher, B., and Christopher, T. (2007), “Poverty and biodiversity: Measuring the overlap

of human poverty and the biodiversity hotspots,” Ecological Economics, 62, 93–

101.

Friedl, M. (2002), “Global land cover mapping from MODIS: algorithms and early

results,” Remote Sensing of Environment, 83, 287–302.

Frost, P. (1996). The ecology of miombo woodlands: In: Campbell, B. (Ed), The Miombo

on Transition: Woodlands and Welfare in Africa. Center for International Forestry

Research, Bogor, Indonesia.

Gambiza, J., Bond, W., Frost, P.G.H., and Higgins, S. (2000), “Land use options in dry

tropical woodland ecosystems in Zimbabwe, a simulation model of miombo 74

woodland dynamics under different management regimes,” Ecological

Economics, 33, 353–368.

Gaveau, D.L.A., Epting, J., Lyne, O., Linkie, M., Kumara, I., Kanninen, M., and Leader-

Williams, N. (2009), “Evaluating whether protected areas reduce tropical

deforestation in Sumatra,” Journal of Biogeography, 36, 2165–2175.

Greenwald, K.R., Purrenhage, J.L., and Savage, W.K. (2009), “Land cover predicts

isolation in Ambystoma salamanders across region and species,” Biological

Conservation, 142, 2493–2500.

Guy, P.R. (1989), “The influence of elephants and fire on a woodland in Zimbabwe,”

Journal of Tropical Ecology, 5, 215–226.

Haberl, H., Gaube, V., Díaz-Delgado, R., Krauze, K., Neuner, A., Peterseil, J., Plutzar,

C., Singh, S., and Vadineanu, A. (2009), “Towards an integrated model of

socioeconomic biodiversity drivers, pressures and impacts. A feasibility study

based on three European long-term socio-ecological research platforms,”

Ecological Economics, 68, 1797–1812.

Holdo, R.M. (2006), “Elephant herbivory, frost damage and topkill in Kalahari sand

woodland savanna trees,” Journal of Vegetation Science, 17, 509-518.

Holdo, R.M., Holt, R.D., and Fryxell, J.M. (2009), “Grazers, browsers, and fire influence

the extent and spatial pattern of tree cover in the Serengeti,” Ecological

Applications, 19, 95–109.

Huete, A. (2002), “Overview of the radiometric and biophysical performance of the

MODIS vegetation indices,” Remote Sensing of Environment, 83, 195–213.

Jensen, J. (2005), Introductory Digital Image Processing: A Remote Sensing Perspective

(3rd ed.), Upper Saddle River, NJ: Pearson Prentice Hall.

Jensen, J. (2007), Remote Sensing of the Environment: An Earth Resource Perspective,

Upper Saddle River, NJ: Pearson Prentice Hall.

Kahurananga, J., and Silkiluwasha, F. (1997), “The migration of zebra and wildebeest

between Tarangire National Park and Simanjiro Plains, northern Tanzania, in

1972 and recent trends,” African Journal of Ecology, 35, 179–185.

Kikula, I.S. (1986), “The Influence of Fire on the Composition of Miombo Woodland of

SW Tanzania,” Oikos, 46, 317–324.

Langevelde, F. Vijver, C.A.D.M. van de, Kumar, L., Koppel, J. van de, Ridder, N. de,

Andel, J. van., Skidmore, A.K., Hearne, J.W., Stroosnijder, L., Bond, W.J., Prins,

H.H.T., and Rietkerk, M. (2003), “Effects of Fire and Herbivory on the Stability

of Savanna Ecosystems,” Ecology, 84, 337–350.

Lawton, R. (1978), “A study of the Dynamic Ecology of Zambian Vegetation,” Journal

of Ecology, 66, 175-198.

Loarie, S.R., Aarde, R.J.Van, and Pimm, S.L. (2009), “Elephant seasonal vegetation

preferences across dry and wet savannas,” Biological Conservation, 142, 3099–

3107.

Lu, D., Mausel, P., Brondizio, E., and Moran, E. (2003), “Change detection techniques,”

International Journal of Remote Sensing, 25, 2365–2401.

Manase, E., Gara, J.I., and Wolanski, E. (2010), “A review of the water crisis in

Tanzania’s protected areas, with emphasis on the Katuma River-Lake Rukwa

ecosystem,” Ecohydrology and Hydrobiology, 10, 153–166.

Mapaure, I., and Moe, S.R. (2009), “Changes in the structure and composition of miombo

woodlands mediated by elephants (Loxodonta africana) and fire over a 26-year

period in north-western Zimbabwe,” African Journal of Ecology, 47, 175–183.

Mapaure, I.N., and Campbell, B.M. (2002), “Changes in miombo woodland cover in and

around Sengwa Wildlife Research Area, Zimbabwe, in relation to elephants and

fire,” African Journal of Ecology, 40, 212–220.

Mas, J.F. (2005), “Assessing protected area effectiveness using surrounding (buffer)

areas environmentally similar to the target area,” Environmental Monitoring and

Assessment, 105, 69–80.

Metzger, K.L., Sinclair, A.R.E., Hilborn, R., Hopcraft, J.G.C., and Mduma, S.A.R.

(2010), “Evaluating the protection of wildlife in parks: the case of African buffalo

in Serengeti,” Biodiversity and Conservation, 19, 3431–3444.

Mfunda, I.M., and Røskaft, E. (2010), “Bushmeat hunting in Serengeti, Tanzania : An

important economic activity to local people,” International Journal of

Biodiversity and Conservation, 2, 263–272.

MNRT (Ministry of Natural Resource and Tourism). (2007), The Wildlife Policy of

Tanzania, Dar es Salaam Tanzania, Dar es Salaam, Tanzania: Government

Printer.

Msoffe, F.U., Kifugo, S.C., Said, M.Y., Neselle, M.O., Gardingen, P. van, Reid, R.S.,

Ogutu, J.O., Herero, M., and Leeuw, J.de. (2011), “Drivers and impacts of land-

use change in the Maasai Steppe of northern Tanzania: an ecological, social and

political analysis,” Journal of Land Use Science, 6, 261–281.

Mundia, C., and Murayama, Y. (2009), “Analysis of land use/cover changes and animal

population dynamics in a wildlife sanctuary in East Africa,” Remote Sensing, 1,

952–970.

Munyati, C. (2000), “Wetland change detection on the Kafue Flats, Zambia, by

classification of a multitemporal remote sensing image dataset,” International

Journal of Remote Sensing, 21, 1787–1806.

Mwalyosi, R.B.B. (1992), “Land-use Changes and Degradation in South-West Masaai

Land, Tanzania,” Environmental Conservation, 19, 145–152.

Owen-Smith, N. (1988), Megaherbivores: The Influence of Very Large Body Size

Ecology, Cambridge, UK: Cambridge University Press.

Owen-Smith, N., Kerley, G.I.H. Kerley, Page, B., Slotow, R., and Aarde, R.J.van. (2006),

“A scientific perspective on the management of elephants in the Kruger national

park and elsewhere,” South African Journal of Science, 102, 389–395.

Packer, C., Hilborn, R., Mosser, A., Kissui, B., Borner, M., Hopcraft, G., Wilmshurst, J.,

Mduma, S., and Sinclair, A. (2005), “Ecological change, group territoriality, and

population dynamics in Serengeti lions,” Science, 307, 390–393.

Pelkey, N.W., Stoner, C.J., and Caro, T.M. (2000), “Vegetation in Tanzania : assessing

long term trends and effects of protection using satellite imagery,” Biological

Conservation, 94, 297–309.

Pelkey, N.W., Stoner, C.J., and Caro, T.M. (2003), “Assessing habitat protection regimes

in Tanzania using AVHRR NDVI composites : comparisons at different spatial

and temporal scales,” International Journal of Remote Sensing, 24, 2533–2558.

Prins, H.H.T. (1987), “Nature conservation as an integral part of optimal land use in East

Africa: The case of the Masai ecosystem of Northern Tanzania,” Biological

Conservation, 40, 141–161.

Richards, J., and Jia, X. (2006), Remote Sensing Digital Image Analysis: An Introduction,

Berlin: Springer.

Sánchez-Azofeifa, A.G., Daily, G.C., Pfaff, A.S.P., and Busch, C. (2003), “Integrity and

isolation of Costa Rica’s national parks and biological reserves: examining the

dynamics of land-cover change,” Biological Conservation, 109, 123–135.

Singh, A. (1989), “Review article: Digital change detection techniques using remotely

sensed data,” International Journal of Remote Sensing, 10, 989–1003.

Skarpe, C. (1991), “Impact of Grazing in Savanna Ecosystems,” Ambio, 20, 351–356.

Skarpe, C., Aarrestad, P.A., Andreassen, H.P., Dhillion, S.S., Dimakatso, T., du Toit,

J.T., Halley, D.J., Hytteborn, H., Makhabu, S., Mari, M., Marokane, W.,

Masunga, G., Ditshoswane, M., Moe, S.R., Mojaphoko, R., Mosugelo, D.,

Motsumi, S., Neo-Mahupeleng, G., Ramotadima, M., Rutina, L., Sechele, L.,

Sejoe, T.B., Stokke, S., Swenson, J.E., Taolo, C., Vandewalle, M., and Wegge, P.

(2004), “The return of the giants: ecological effects of an increasing elephant

population,” Ambio, 33, 276–82.

Swain, P.H., and Davis, S.M. (1978), Remote Sensing: The Quantitative Approach, West

Lafayette, IN: Purdue University.

TANAPA (Tanzania National Parks). (no date), Tarangire General Management Plan,

Arusha, Tanzania: Tarangire National Park.

TAWIRI (Tanzania Wildlife Research Institute). (2007), Total count of in the Tarangire

ecosystem (including Lake Manyara National Park), dry season, 2006, Arusha,

Tanzania: internal report.

Verbyla, D.L., and Boles, S.H. (2000), “Bias in land cover change estimates due to

misregistration,” International Journal of Remote Sensing, 21, 3553–3560.

Vijver, C.A.D.M. van de, Foley, C.A., and Olff, H. (1999), “Changes in the woody

component of an East African savanna during 25 years,” Journal of Tropical

Ecology, 15, 545–564.

Western, D., Russell, S., and Cuthill, I. (2009), “The status of wildlife in protected areas

compared to non-protected areas of Kenya,” PLos One, 4, e6140. 80

Chapter 3. Assessment of Wildlife Populations Trends in the Tanzania’s Protected Areas from 1991 to 2012

Abstract

Wildlife populations have been declining across eastern and western Africa. A

Tanzanian national wildlife policy was instituted in 1998 that emphasized the protection and maintenance of viable wildlife populations. Nevertheless wildlife populations have continued to decline inside and outside protected areas. Using large herbivores as representative of other species, four main questions were addressed: 1) Has the population density of large herbivores and area they occupied, in Katavi-Rukwa, Ruaha-

Rungwa and Tarangire protected areas in Tanzania, changed between 1991 and 2012; 2)

If change has occurred, was the degree of change, among species and area occupied, consistent across protected areas; 3) Were the patterns of change similar inside and outside of the protected areas; and, 4) Have the protected areas been effective in achieving the goal of protecting population of various species? Wildlife aerial census data from 1991 to 2012 was analyzed to understand the temporal change in populations of six large herbivore’s species or groups and the areas they occupied in the three protected areas. Between 1991 and 2012 three of the six species or groups exhibited decreasing population trends inside and outside the three protected areas. The size of areas occupied by two of the three species also decreased inside and outside the three protected areas.

The population trends for the remaining three species, two remained static in the three protected areas, and one group decreased in one protected area while increasing in the other two protected areas. The observed shrinkage in size of habitat size and subsequent declining trends for 50% of the six large herbivores examined, suggests that protected

areas appear not to be effective in protecting the populations of various species. In order to avoid continual loss of habitats and species population, a revision of the existing policy, to come up with techniques that would help protected areas achieve their intended goal is recommended.

1.0 Introduction

Protected areas are established to ensure the long-term viability of wildlife species, habitat, and ecosystems. However, the long-term effectiveness of protected areas, particularly in developing regions (Africa, Asia, Latin America), is threatened by habitat destruction through land clearing (Nagendra, 2008; Ottichilo, Leeuw & Prins,

2001; Clerici et al., 2007), fencing (Berry, 1997; Beale et al., 2013), road construction

(Newmark, 2008) and illegal hunting (Brashares et al., 2004). Such activities block animal movements, preventing them from accessing dispersal, calving and breeding areas, potentially resulting in their decline. Nevertheless, protected areas are implemented worldwide to reduce the rate of species decline inside, relative to the areas outside the boundary of protection (Bruner et al. 2001, Sánchez-Azofeifa et al., 2003; Gaveau et al.,

2009; Nagendra, 2008). As human population growth fuels the increases in demand for settlement, farmland and income, fears mount that protected areas might fail to achieve long-term conservation goals (Wittemyer et al., 2008; Mora et al., 2011).

Like other countries in the globe, Tanzania has demonstrated a commitment to protecting wildlife resource by allocating 23.8% (224,958 km2) of its total land area to some type of protected area for wildlife, including national parks (4.4%), Ngorongoro

Conservation area (0.9%), game reserves (13%), and game controlled areas (5.5%).

National parks (NP) are given the highest level of protection, meaning that no permanent settlement, hunting, livestock grazing, or logging is allowed. Like the NP, permanent human settlement, cultivation, and livestock grazing are prohibited in the game reserves

(GR), although hunting, fishing and logging are allowed with a special license. Game controlled areas (GCAs) allow human settlement, grazing, cultivation, and licensed hunting. The NPs are regulated by Tanzania National Parks (TANAPA) while the GR and GCA are regulated by the Tanzania Wildlife Division Authority. The Ngorongoro

Conservation Area allows human settlement by local Maasai people and their subsistence agriculture, and no hunting is permitted.

Overall, wildlife populations in Tanzania’s protected areas have been declining

(Stoner et al., 2007), similar to other protected areas in eastern and western Africa

(Craigie et al., 2010; Western et al., 2009). The primary reason for wildlife population declines in Tanzania are anthropogenic activities, such as illegal hunting (Caro, 1999,

2008; Andimile & Caro, 2012) and farming (Msoffe et al., 2011) in and around protected areas. For instance, wildebeest (Connochaetes taurinus) declined from 40,000 to 5,000 individuals between 1984 and 2000 due to the conversion of 15 km2 of woodland and grassland into crop farms and settlements that blocked four of the nine wildlife migration corridors on the western and southern sides of Tarangire National Park (Msoffe et al.,

2011). In addition, many herbivores, including waterbuck (Kobus ellipsiprymnus), topi

(Damaliscus lunatus), Grant’s gazelle (Gazella granti), Thomson’s gazelle (Gazella thomsonii), reedbuck (Redunca sp.), hartebeest (Alcelaphus buselaphus), roan antelope

(Hippotragus equines), sable antelope (Hippotragus niger), warthog (Phacochoerus africanus) and zebra (Equus burchellii) have declined in four of the eight wildlife zones

surveyed across Tanzania, between 1989 and 2001 (Stoner et al., 2007, Caro, 2011), due to migratory routes being blocked (Msoffe et al., 2011) and/or illegal hunting (Caro,

2008; Andimile & Caro 2012). In contrast, the populations of the mega herbivores, such as giraffe (Giraffa camelopardalis) and elephant (Loxodonta africana), have increased or remained unchanged (Stoner et al., 2007).

Elephants in Tanzania are protected under CITES appendix I, and have increased from 55,000 (0.23 indiv/km2) in 1989 to 136,753 (0.6 indiv/km2) in 2006 (CITES, 2010).

Recent land use land cover change analysis in Tarangire and Katavi National Parks in

Tanzania (Mtui et al. under preparation) indicates that cultivation continues to encroach on wildlife habitats. When densities of elephants exceed 0.5 individuals/km2 and are restricted in their movements, they may contribute to the local conversion of savanna woodlands into shrubland or grassland (Cumming et al., 1997). Changes in vegetation structure and composition may result in a decline or extirpation of species, as happened to gerenuk and giraffe in Amboseli National Park in Kenya (Parker, 1983, cited in

Kerley, 2008), as elephants created food shortages by overexploiting available resource and damaging trees by uprooting or breaking tree stems (Barnes, 1983). Removal of large trees by elephants may reduce availability of nutritional grass for grazers because grass growing under the canopy of trees is more nutritious than in open grassland (Owen-

Smith, 1988; Ludwig et al., 2008). In contrast, the opening up of shrub thickets by elephants has led to increased abundance of grazing animals (oryx, warthog and zebra) in

Tsavo National Park and browsing herbivores (eland and kudu) in Addo National Park

(South Africa) (Parker, 1983, cited in Kerley 2008). The high densities of large herbivores are claimed to impact on other species through competition for food, such as

between buffalo and elephant in Lake Manyara National Park (Tanzania) (De Boer &

Prins, 1990). However, competition for food is not seen consistently in national parks.

For example, in Chobe National Park in Botswana, which hosts the largest population of elephants in Africa (Owen-Smith, 2006), no evidence of food competition between elephant and buffalo has been found (Skarpe et al., 2004).

In 1998, the Tanzanian government established a National Wildlife Policy to emphasize maintenance of viable protected areas of important habitats and viable populations of important species, and ensure the survival of species classified as endangered, endemic, or rare. The policy was amended in 2007 to include wetland resources of national and international importance for biodiversity and water catchments, and it states that “Wildlife and wetlands are natural resources of great biological, economical, environmental cleaning, climate ameliorating, water and soil conservation, and nutritional values that must be conserved. It can be used indefinitely if properly managed” (MNRT, 2007). The policy was implemented under four major themes including wildlife protection, wildlife utilization, management and development of protected areas, and international cooperation (see MNRT, 2007 for details). The strategy put forward was to maintain the existing wildlife protected areas and create new ones where necessary. For example, the policy established a Wildlife Management Area as a new category of protected area to involve community in protecting wildlife outside parks boundaries.

With such increased efforts on wildlife protection through the implementation of the National Wildlife Policy, protected areas were expected to improve their performance on protection on wildlife populations by alleviating and preventing threats to their

existence inside protected areas, as well as dispersal areas and migratory routes and therefore increase their numbers inside and outside the protected areas. However, anthropogenic threats such as land clearing for cultivation, reduction of river flow, illegal hunting and livestock grazing, continue to occur within wildlife protected areas (Mtui et al., under preparation; Andimile & Caro, 2012; Manase, Gara, & Wolanski, 2010).

Given the importance of protected areas such as National parks and game reserves for wildlife conservation, evaluating the current population trends of wildlife species in them is critical in order to make informed management decisions.

Four main questions were addressed in relation to performance of protected areas as a result of the wildlife policy: 1) Has the population density of large herbivores and area occupied by large herbivores, across wildlife protected areas in Tanzania changed between 1990s and 2010s; 2) If change has occurred, was the degree of change among species and areas occupied by species consistent across different protected areas in the country; 3) Were the patterns of change similar inside and outside of the protected areas; and, 4) are protected areas effective in protecting species diversity? If the trends densities of large herbivores, and/or area occupied by large herbivores, are similar to the levels of

1998, or show increasing patterns over time across protected areas, then the protected areas would be considered as effective, otherwise not.

2.0 Materials and Methods

2.1 Site descriptions

To achieve the goal of this study, densities of seven species or groups (see below) of large herbivores were investigated at the Tarangire, Ruaha-Rungwa, and Katavi-

Rukwa wildlife regions (Figure 3.1). Because of the varying sizes of the two parks, density estimates were used as a comparative measure of population abundance. The study is restricted to these three regions because of problems with availability of data, limited time and funding to conduct the research.

Located in northern Tanzania the Tarangire ecosystem covers 12,749 km2 (Figure

3.1) and is comprised of the Tarangire National Park (TNP), the Lake Manyara National

Park, the Lolksale Game Controlled Area (GCA), the Mkunganero Game Reserve (GR), the Simanjiro plains, and the surrounding areas. The largest component of the ecosystem is the TNP itself, which covers 2,600 km2 at an elevation of ranging from 1,200 to 1,600 m. The park was established as a Game Reserve in 1957 and declared a National Park in

1970. The Simanjiro Plains include the grasslands of the Maasai Steppe and serve as dispersal areas and areas for lactating and calving for wildebeests and zebras during the wet season. The perennial Tarangire River serves as a dry season refuge for large mammals. The park receives short rains, usually between October and December, which may start as early as August, and heavy rains between February and April or May.

Average annual rainfall, from 1979 to 2009, was approximately 656 mm. The major types of vegetation in the Tarangire ecosystem are riparian woodland, wetlands and seasonal flood plain, Acacia-Commiphora woodland, riverine grassland, Combretum-

Dalbergia woodland, Acacia drepanolobium woodland, and grasslands with scattered baobab trees (TANAPA n.d., Mwalyosi, 1992; Table 1). Pastoralism has been the major land use in and around the Tarangire ecosystem over the past two centuries (Prins, 1987).

However, in the past two decades, agricultural activities have been increasing on the

northeast and eastern side of ecosystem, blocking wildlife migration routes and dispersal areas (Mwalyosi, 1992; Msofe et al., 2011).

Located in southwestern Tanzania, the Katavi-Rukwa ecosystem covers 13,378 km2 (Figure 1) and is comprised of the Katavi National Park (KNP), Rukwa Game

Reserve (GR), Mlele GCA, Msanginia Forest Reserve (FR), Lwafi GR, Nkamba FR, and

Usevya Open Area. The largest component of the ecosystem is the KNP itself, which covers 4,238 km2 at an elevation ranging from 800 m to 1600 m. The park was established in 1974 with 1,816 km2 (Figure 1) and enlarged in 1998 to the current size.

The park receives single annual rains from November to April/May. Average annual rainfall from 1997 to 2012 was approximately 955 mm. The large seasonal Katavi and

Chada lakes and the Katisunga flood plain (Caro, 1999) which are maintained by Katuma

River (Figure 3.1), support large mammals during the dry season. The major types of vegetation in Katavi ecosystem are grassland interspersed with miombo woodlands and mixed woodlands. Miombo forms a single story, with a light, closed canopy of deciduous woodland, usually greater than 15 m tall, dominated by trees of the genera Brachystegia,

Julbernadia, and Isoberlinia (Kikula, 1987; Frost, 1996). The genera Markhamia,

Grewia, Terminalia, Combretum, Syzygium, and Acacia also occur in the miombo woodland of the KNP (Banda et al., 2006). Pastoralism and horticulture are the major land use around the Katavi ecosystem (Caro, 1999). The integrity of the park is threatened by increased damming of River Katuma for rice cultivation, farming inside and adjacent to the park, and wildfires caused by poachers, farmers and people travelling on the main road, which cuts across the park (Manase et al., 2010).

Located in the south-central Tanzania, the Ruaha-Rungwa ecosystem covers

43,641 km2 (Figure 3.1) and is comprised of the Ruaha National Park (RNP), Rungwa

GR, Muhesi GR and Kizigo GR. The largest component of the ecosystem is RNP itself, which covers 20,226 km2 at an elevation ranging from 750 m to 1830 m. The park was established in 1964 with 10,300 km2 (Figure 1), and enlarged in 2008 to the current size.

The park receives single annual rains from December to April. Average annual rainfall as recorded at Msembe Headquarters, from 1996 to 2011, was approximately 470 mm. The

Great Ruaha River used to be perennial until 1993, when the flow became scarce due to excessive use of water upstream. Since then, the river has been drying up in mid dry season, leaving the ecosystem to be sustained by springs and small water sources

(TANAPA, n.d.). The major types of vegetation in Ruaha ecosystem are Brachystegia

(miombo) woodlands, Acacia woodland, Commiphora-Combretum woodlands and bushed grassland and evergreen forest, dominated by Drypetes gerrardii (Bjørnstad,

1976). The integrity of the park is threatened by damming of river for farming irrigation, wild fires, overgrazing by livestock, blockage of dispersal areas by settlements and agriculture, poaching, and uncontrolled settlements along the river basin (Figure 1)

(TANAPA, n.d.).

2.2 Animal density data

Animal density data from 1991 to 2012 were obtained from the Tanzania

Wildlife Research Institute (TAWIRI). TAWIRI conducted aerial surveys using

Systematic Reconnaissance Flights (SRFs), following the guidelines designed by Norton-

Griffiths (1978). The surveys were typically conducted during the dry season (September to October) every 1 to 3 years, and sometimes during the wet season (March to May).

The surveys were carried out using two aircrafts, each with two parallel rods mounted on the wing struts. The aircrafts were flown along parallel transects oriented in the east-west or north-south direction, 5 km apart from each other and about 106 m above ground.

Each transect was divided into sub-units defined by 30 seconds flying time, which translates to 1.8 km on the ground. Animals visible between the two parallel rods were counted within each sub-unit by two experienced observers positioned in the rear seats of the aircraft, while the pilot recorded the geographical position (GPS) codes at the beginning and end of each transect and informed observers as the flights entered a new subunit (TAWIRI, 2010a). Species density (individuals/km2) in each subunit was derived by the survey teams based on the number of animals observed and the strip width spanned at a range of 150 and 279 (depending on the flight used and surveyed area) of each transect using using SISTA software, developed at TAWIRI’s Conservation and

Information Monitoring Unit (CIMU), specifically for SRFs surveys (TAWIRI, 2010b).

The survey team also grouped the subunits into a grid system of cells covering the surveyed area.

For decades aerial surveys have been used to estimate the abundance of wildlife species. The estimates have been criticized as inaccurate (Caughley, 1974) due to biases in sightability (Steinhorst & Samuel, 1989) and visibility (Samuel, 1981). However, aerial surveys are the only practicable method of estimating the number of large animals over a large land area and provide estimates with a level of approximation judged acceptable (Caughley, 1977). According to Rabe et al., (2002) the aerial survey is the most effective method estimating large animal populations when sources of biases are substantially minimized. The sighting bias relates to low probability of sighting single

animals and small groups of animals, and this can be minimized by keeping operational factor such as searching rate and height above ground level within reasonable limits. The method recommends counting large and conspicuous animals, at a height of between 100 m to 200 m above ground, and at a speed varying between 180 and 250 km/h. Visibility bias relates to animals not being visible to the observer because they are concealed by other animals or obstructions such as tree canopies or because their color makes them blend imperceptibly with the background. Visibility bias can be reduced by using a double counting technique (Jachman, 2002). These visibility and sighting biases were accounted for by TAWIRI as they used two aircrafts, each with two observers, flown about 106 m above ground (see the first paragraph on this section) at a speed of between

212 km/h. While observers did change between surveys, an overlap in observer personnel occurred across surveys which provided important continuity in counting techniques

(TAWIRI, 2010a, 2010b).

2.3 Approach to statistical analysis

Only data from surveys carried out during the dry seasons and that covered similar extensive land areas in each ecosystem was considered. Wildlife density data were selected to include only those which were within the extent of the ecosystem and a

10 km buffer zone (Figure 2) [Note: In Tanzania, human activities are observed up to 2 km from PA boundaries, despite the fact that 7 km is considered as an official buffer zone

(Caro et al. 2011). However, a 10 km buffer was used here because it is considered the standard buffer size and has been widely used in other studies (Bruner et al., 2001;

Gaveau et al., 2009). Furthermore, only grid cells from transects that cut across the entire park and its buffer zones were selected for analysis (Figure 2). Each borderline grid cell

was treated as inside or outside a protected area boundary only when 75% of the cell was on the respective side, otherwise it was omitted from the analysis. Wildlife density data for 22 species (Table 3.1) were available for total of 774, 1261, and 3383 grid cells inside and outside the Tarangire, Katavi-Rukwa, and Ruaha-Rungwa ecosystems, respectively

(Appendix I).

Only species recorded in at least four grid cells per region were considered for statistical analysis, and those recorded in less than 4 grid cells were grouped together, based on genus and female body weight. As a result elephant, giraffe, buffalo and zebra were analyzed individually, and the remaining species were grouped as small and medium antelopes. If two or more species of the same grouping were encountered in the same grid cell the sum of their densities was computed. The small antelopes included only bovines with female mean body weight ranging between 40 and 100 kg (lesser kudu, bushbuck, impala, topi, warthog, puku, and Grant gazelle; Table 3.1). Although warthog is not a bovine it was included with small antelopes based on its body size. Medium antelopes included those bovines with 100-500 kg female mean body weight range

(waterbuck, hartebeest, wildebeest, roan antelope, sable antelope, greater kudu, eland and

Oryx). To avoid bias small antelopes with female mean body weight less than 40 kg, hippopotamus and bush pigs were excluded from the analysis because they are difficult to detect using aerial surveys.

Average annual densities (indiv/km2) were computed for the six species or groups by taking the sum of densities in grid cells where species were recorded and dividing it by: (i) the total number of grid cells i.e. including cells where the species or group was not recorded (density equals zero) across the park area (inside or buffer zone); and (ii)

total number of grid cells where the species or groups were present. The first procedure used to obtain the average densities was considered given an assumption that each of the grid cell in the park or buffer zone had equal chance of being utilized by large herbivores, and therefore would provide general information on the status of large herbivores across the entire protected areas. The second procedure assumed that species or groups of large herbivores would most likely occur in particular areas (grid cells) that had more suitable forage for each species or group (Macandza et al. 2004, Ryan et al. 2006). This second approach was expected to provide information on the status of large herbivores in areas where they occur. Both procedures (i) and (ii) provide complementary information. In addition, the number of grid cells occupied by species or groups of large herbivores over time was analyzed, in order to determine their trends over the years. The decrease or increase in number of cells occupied by large herbivores would indicate the shrinkage or expansion of wildlife distribution across the parks or buffer zones, information which would guide on judging the effectiveness of the protected areas.

2.4 Statistical analysis

Where necessary, data were transformed prior to analysis to meet the assumptions of linearity. Specifically, the density data were transformed using Log10 (n+1) transformations. The final estimates were back transformed to the original values for reporting purposes, except for the figures. Data were also inspected for non-linearity by considering quadratic terms, which were ultimately not found in the data.

A general linear model (GLM) was used to test if: (i) the average densities of the six large herbivores species or groups, and (ii) number of grid cells occupied by large

herbivores species or groups, had changed over the past two decades, inside and outside

(buffer zones) the three protected areas. In considering how the (i) average densities of large herbivores, and (ii) number of occupied grid cells, might change over time, the analyses included three independent variables, year, protected areas (Katavi, Ruaha and

Tarangire) and location (inside and within 10 km buffer the region), and their two and three way interactions.

The year variable was treated as a quantitative rather than categorical variable and location and protected area as categorical variables. The results using Type III tests are reported although the conclusions using Type II tests were very similar (see Kutner,

Nachtsheim et al. 2005 for details). Wherever the statistical results were significant (p ≤

0.05) for year, or for year, or interaction of year and protected area, the overall slopes were estimated using ESTIMATE statement in PROC GLM (appendix H) to determine the general trend of change. Results are presented either as Log10 [(Mean +1) ± (SE + 1)]

(for average density analyses), or as [mean ± SE] (for grid cells analyses) unless otherwise noted. Analyses were performed using SAS software version 9.2.

Additionally, the same GLM method was used to test for temporal change of total annual rainfall i) from 1988 to 2009 (Tarangire) and 1997 to 2012 (Katavi) and from

1996 to 2011 (Ruaha). The rainfall data were recorded in the national park head offices and therefore although may not reflect rainfall conditions in their surrounding game reserves, we still used them as an approximation.

3.0 Results

The results show no significant changes in amount of annual rainfall received in the three parks over the past two decades (Figure 3.2-ii). Analysis of changes in average densities of large herbivores species or groups, considering the entire areas of the parks and their buffer zones, show that, except for buffalo, elephant and small antelopes, the population densities of all other large herbivore species and groups changed negatively between 1991 and 2012 (Table 3.2). The changes in population density for giraffe, medium antelopes, and zebra did not differ significantly across protected area, both inside and outside of the protected area (Table 3.2). The average densities of these species or groups on Log10 (density+1) scale have been declining at a rate ranging between 0.016 ±

0.005 and 0.004 ± 0.002 indiv/km2 (Table 3.3, Figure 3.3). These rates of change were similar across protected areas, both inside and outside the protected areas (Table 3.3). In contrast to giraffe, medium antelopes and zebra, the population density change for the small antelopes differed significantly across protected areas, but not between inside and outside the protected areas (Table 3.2). In the Tarangire protected area, the average density of small antelopes on Log10[density+1) scale were declining at a rate of 0.02 ±

0.01 indiv/km2, whereas in the Katavi and Ruaha protected areas these antelopes remained unchanged (Table 3.3, Figure 3.3). Finally, buffalo and elephant exhibited no significant changes over time (Table 3.2, Figure 3.3).

Analysis of changes in average densities considering only areas (grid cells) occupied by large herbivores species or groups show no significant changes in the densities of either of the six large herbivores species or groups over the years, across the

three protected areas, both inside and on their buffer zones (Appendix G2). On the other hand, except for elephant, small and medium antelopes, the analysis of changes in the size of area occupied by large herbivores between 1991 and 2012 exhibited significant changes over time, in areas occupied by buffalo, giraffe and zebra (Table 3.4). Levels of change, however, differed significantly across the three regions for buffalo, but not for giraffe (Table 3.4). The area occupied by buffalo decreased over the years at an overall rate of [mean ± SE] 0.374 ± 0.1 (occupied grid cells per year), but the decreases were most evident in Katavi and Ruaha regions, at rates of 0.7 ± 0.2 and 0.6 ± 0.2, respectively, while at Tarangire the number of grid cells occupied buffalo increased at a non-significant rate of 0.2 ± 0.2 (Table 3.5, Figure 3.4). Similarly, the area occupied by giraffe decreased over the years, at an overall rate of 0.5 ± 0.2 (occupied grid cells per year). The decrease was most evident in Katavi region (0.833 ±0.3), both inside and outside, while the decrease in Ruaha and Tarangire regions was not so evident (Table 3.5,

Figure 3.4). Level of changes in area occupied by zebra also differed across the regions, both inside and outside (Table 3.4). The number of grid cells occupied by zebra decreased over the years, at an overall rate of 0.7 ± 0.2 (Table 3.5 and Figure 3.4).

However, the rates of decreases in Ruaha (1.1 ± 0.3) and Tarangire (0.8 ± 0.3) were about

4 and 3 times higher than in Katavi region (0.3 ± 0.2) and respectively.

4.0 Discussion

The results presented here indicate that in the 20 year period from 1991 to 2012 habitat occupied by buffalo, giraffe and zebra declined but remained stable for other species or groups of large herbivores (Table 3.4). Three of the six species or groups of

large herbivores decreased in population across the entire areas of the three protected areas of Tanzania. Giraffes, medium antelopes, and zebras declined at similar rates both inside and outside across the entire Katavi, Ruaha, and Tarangire protected areas. The population of small antelopes decreased inside and outside the entire area of Tarangire, and increased inside and outside the Katavi and Ruaha (Table 3.3, Figure 3.3). Elephants and buffalo populations remained static in the entire areas of the three protected areas

(Table 3.2). On the other hand, the results indicate that the population of the large herbivores examined, remained unchanged in areas of the protected areas where they occurred over the past 20 years. These suggest that, the declining populations of the species of groups of large herbivores are probably due to shrinkage of their area of occupancy. The contribution of rainfall to the declining trends of these species or groups is not significant changes across the three regions from 1988 to 2009 (Tarangire) and

1997 to 2012 (Katavi) and from 1996 to 2011 (Ruaha) (Figure 3.2-ii).

The accuracy of the results, may be limited by the small number of protected areas examined and the lack of inclusion of potentially confounding variables (see below), which have likely also had influence on the observed trends. Three protected areas out of the fourteen wildlife ecosystems were examined, all harboring large herbivores in Tanzania and may have failed to detect the real changes in densities of large herbivores. Densities of species that did not meet the criteria for statistical analysis were grouped as small antelopes or medium antelopes. Such aggregation might have affected the results because species tends respond differently to environmental changes (Marion

Valeix et al., 2011). Other variables which might have influenced the trends of large herbivores over time, such availability of water (Gereta et al. 2004, Epaphras et al. 2007), 97

predation by carnivores (Ogutu and Owen-Smith, 2005), and illegal hunting (Caro 1999) were not included in the analysis. Nevertheless, some of this information are available for reference, such as for water in Tarangire (Gereta et al. 2004) and Ruaha (Epaphras et al.

2007), and illegal hunting in Katavi (Caro 1999, 2008 and 2011). Despite the limitations, the data presented here are valid and useful to consider in decision making for the three study sites, even if they cannot be generalized across the country.

The results show that giraffe, medium antelopes and zebra populations have decreased in the Katavi, Ruaha, and Tarangire wildlife protected areas; and the decreases in giraffe and zebra parallel the decline of the area they occupy in the three regions

(Tables 3.3 and 3.6, Figures 3.4 and 3.5). The trends for giraffe differ from previous reports on assessment of changes in population density of large herbivores across protected areas in Tanzania (including Ruaha, Katavi and Tarangire). Giraffe populations were reported as stable across the country from the late 1988 to 2001 (Stoner et al., 2007) and by 2009 species density was declining marginally (p < 0.1) inside the Katavi

National Park (Caro, 2011). Our study confirmed the decreases in medium antelopes and zebra populations, which were also observed in more half of eight protected areas in the country (Stoner et al., 2007), and in Katavi region (Caro, 2008, 2011). Contrary to previous reports that the population densities of elephant across protected areas in the country (CITES, 2010), and in Tarangire National Park (Foley & Faust, 2010) were increasing, the results showed stable populations for this species across the three protected areas (Table 3.3). Lack of significant trends for elephant found here may be due to the small number of protected areas analyzed and the lack of data from Selous Game

Reserve, which may harbor the largest concentration (54%) of the Tanzania’s estimated 98

136,753 elephants in 2006 (CITES, 2010). Elephant densities have actually remained at higher densities, about 4 to10 times more, than the recommended carrying capacity of 0.5 indiv/km2 (Cumming et al. 1997), since the 1990s (after the ivory ban in 1989) (CITES,

2010) (Figure 3.6).

The data analyzed in this study were collected during the dry seasons when wildlife species restrict their movements to areas with availability of surface water

(Owen-Smith, 1988). Wildlife populations in Ruaha and Tarangire are known for their seasonal movements over long distances searching for water (Epaphras et al., 2007;

Kahurananga &Silkiluwasha, 1997, Gereta et al., 2004, Voeten et al., 2009). In Katavi, on the other hand, these migrations do not occur (Caro, 1999), though Kiffner et al.,

(2013) observed that elephant and buffalo were more frequently in the core area rather than on its periphery during the dry season, presumably driven by water availability inside the park. Elephants and buffalo are known to cause overgrazing and damage of trees and shrubs around waterholes at a radius of 5 km in Ruaha ecosystem (Epaphras et al., 2007). Such levels of habitat utilization and damage might have created food shortages and/or competition among species that led to decreases in giraffe, medium antelopes and zebra populations across the three regions. The elimination of buffalo from northern and western Serengeti National park through poaching, allowed an increase in abundance of topi, impala and oribi (Arsenault and Owen-Smith, 2002). In Kidepo

Valley National Park in Uganda, Field & Ross (1976) observed overlaping of food resources between elephant and giraffe. More than two thirds of plant species found in elephant diet were species that were important for giraffe during dry season. Such extensive overlap of food requirement makes it possible that the declines if density of 99

giraffe was due to food shortage and/or competition by the elephants. Zebra are considered to be mobile animals and can move long distances to access forage (Smuts

1975). Zebra can co-exist with the elephant without any competition (Valeix et al., 2011).

The declines in these species are therefore probably due to habitat loss possibly due to buffalo overgrazing. Stoner et al. (2007) concluded that specilized feeders browsers or grazers did not fare as well across protected areas in Tanzania as mixed feeders.

Dispersal is important for wildlife species because it prevents overabundance and depletion of their seasonal resources (Owen-Smith, 1988). In Tarangire ecosystem, buffalo, elephant, zebra and wildebeest (medium antelope) migrate every year from the park to the Simanjiro plains (Figure 3.1), at the start of wet season (November /

December), and return in the dry season (June/July) (Kahurananga & Silkiluwasha, 1997,

Gereta et al., 2004, Voeten et al., 2009). Increased farming activities outside the protected areas now restricts their migration to the wet season forage, causing them to over utilize their dry season pastures (Kahurananga & Silkiluwasha, 1997; Voeten et al.,

2009; Voeten & Prins, 1999; Gereta et al., 2004). Therefore, declining densities of giraffe, zebra and medium antelopes may be due to their confinement inside the parks which prevented their access to seasonal dispersal areas.

Interviews (Caro, 2008) and surveys (Andimile & Caro, 2012) suggested that illegal hunting was the major cause of wildlife species decline in protected areas.

Andimile & Caro (2012) reported that smaller antelopes (bushbuck) and medium antelopes (greater kudu, roan and sable antelope) were impacted more by poaching than giraffe, buffalo or zebra. Poaching may also have contributed to the decline in giraffe,

100

zebra and medium antelopes as observed in Katavi, Ruaha and Tarangire, and small antelopes in Tarangire.

5.0 Conclusion

Our results show that the populations of giraffe, zebra and medium antelopes have been declining over time while the populations of buffalo and elephant remained stable.

The declining population of these species or groups, combined with the compression in their distributional range, raises a concern for their long-term existence, especially giraffe since they have low fertility rates (Owen-Smith, 1988). Giraffe reach puberty stage at the mean age of 4.5 years and can live up to 28 years, with a mean birth interval of about 1.7 years (Owen-Smith, 1988). The declining trend of giraffe and shrinkage of its habitats therefore needs attention.

One of the objectives of the Tanzania’s National Wildlife Policy of 1998 was to encourage communities living adjacent protected areas to establish WMAs on communal land, from which they would directly benefits from wildlife resource through tourism enterprises and tourist hunting, while preventing degradation and loss of wildlife habitats

(MNRT 2007). While economic benefits from WMAs in Tanzania have increased (Sulle et al. 2011, USAID 2013), wildlife populations continue to decline inside and outside protected areas. Similar approaches to WMAs were adopted in Namibia, Zimbabwe and

Zambia and have been successful in achieving both the economic development and conservation objectives (Shaw & Platts 2004, Boudreaux, 2010). In Namibia for example, the wildlife populations are increasing inside and outside protected areas

(Boudreaux, 2010). The shrinkage in the area occupied by buffalo, giraffe and zebra, and

101

the declining densities of the latter two species together with medium antelopes across the three regions suggests that protected areas are still not effective in protecting and maintaining species populations. In order to avoid continual loss of habitats and species population, a revision of the existing policy is recommended to come up with techniques that would help protected areas achieve their intended goal as observed.

6.0 Acknowledgments

We are grateful to the management teams of Tanzania Wildlife Research Institute

- Tanzania Wildlife Conservation Monitoring (TAWIRI-CIMU) for providing the wildlife count survey data used in this study. Particularly, we thank TAWIRI-CIMU’s

Director General, Dr. Simon Mduma, the Head of CIMU Mr. H. Maliti, and CIMU’s database manager, Mr. Machoke Mwita, for their support during the process of securing the data. This study would not have been achieved without financial support from the

International Ford Foundation Fellowship Program in collaboration with the East-West

Centre at Honolulu, Hawai‘i in United States of America.

102

Table 3.1 Large herbivore species recorded in Ruaha-Rungwa, Katavi-Rukwa and Tarangire protected areas.

Scientific name Common name Ruaha Katavi Tarangire Aepyceros melampus Impala    Alcelaphus buselaphus Hartebeest    Connochaetes taurinus Wildebeest (blue)  Damaliscus lunatus Topi    Equus burchellii Zebra    Gazella granti Grant gazelle  Gazella thomsonii Thomson gazelle  Giraffa camelopardalis Giraffe    Hippopotamus amphibius Hippopotamus   Hippotragus equinus Roan antelope   Hippotragus niger Sable antelope   Kobus ellipsiprymus Waterbuck    Kobus vardoni Puku  Loxodonta africana Elephant    Madoqua spp. Dikdik    Oreotragus oreotragus Klipspringer   Oryx gazella Oryx  Ourebia ourebi Oribi   Phacochoerus africanus Warthog    Potamochoerus larvatus Bushpig  Redunca spp. Reedbucks    Sylvicapra grimmia Duiker    Syncerus caffer Buffalo    Taurotragus oryx Eland    Tragelaphus imberbis Lesser Kudu   Tragelaphus scriptus Bushbuck   Total 22 21 19

103

Table 3.2 General Linear Model (GLM) analysis of results on temporal changes on average densities of large herbivores inside Katavi, Ruaha and Tarangire protected areas and in their 10 km buffer zones. [Sample size for individual species = 36; PA = Protected Area].

Source

Species Year Location Year * PA Year*PA PA Year* Location *location PA*Location

Buffalo ns ns ns ns ns ns ns Elephant ns ns ns ns ns ns ns

Giraffe F1,24 = 6.16, ns ns ns ns ns ns p = 0.021 Medium ns ns ns ns ns ns antelopes F1,24 = 8.35, p = 0.008 Small ns ns ns F1,24 =4.23, F1,24 = 4.21, ns ns antelopes p = 0.027 p = 0.027

Zebra F1,24 = 5.66, ns ns ns ns ns ns p = 0.026

Note: ns means not significant at α ≤ 0.05.

104

Table 3.3: Estimates of mean density (Indiv/km2) for species observed to have changed significantly over time in Table 3.2. For small antelope the mean density is shown for each region separately because of the interaction between year and region, as shown in Table 3.2 above.

Species Parameter Log10[Mean ± Error] Giraffe Year -0.004± 0.002 Zebra Year -0.016± 0.007 Medium antelopes Year -0.015± 0.005 Small antelopes Year -0.004±0.003 Year (Katavi) 0.000±0.005 Year (Ruaha) 0.004± 0.006 Year (Tarangire) -0.016± 0.005

Note: bolded values were significant at 95% confidence limits.

105

Table 3.4: General Linear Model (GLM) analysis results on temporal changes on number of cells (5 x 5) km2 occupied by large herbivores inside Katavi, Ruaha and Tarangire protected areas and their 10 km buffer zones. [Sample size for individual species = 36, PA = Protected Area]. Source Species Year Location Year * PA Year*PA PA *location Year* Location PA*Location

Buffalo F1,24 = 15.03, F1,24 = 9.02, F1,24 = 8.73, F1,24 = 9.29, F1,24 = 9.18, ns ns p = 0.001 p = 0.006 p = 0.007 p = 0.001 p = 0.001 Elephant ns ns ns ns ns ns ns

Giraffe F1,24 = 7.11, ns ns ns ns ns ns p = 0.014 Medium ns ns ns ns ns ns ns antelopes Small ns ns ns ns ns ns ns antelopes

Zebra F1,24 = 19.33, F1,24 = 13.47, F1,24 = 13.13, ns ns ns ns p = 0.000 p = 0.001 p = 0.001

Note: ns means not significant at α ≤ 0.05.

106

Table 3.5 Proportion estimates for number of grid cells (5 x 5) km2, occupied by large herbivores observed to have changed significantly over time in Table 3.4 above. Species Parameter Mean± Error Buffalo Year -0.371 ±0.096 Year (Katavi) -0.673 ±0.151 Year (Ruaha) -0.649 ±0.178 Year (Tarangire) 0.208 ±0.168 Giraffe Year -0.508 ±0.191 Year (Katavi) -0.833 ±0.300 Year (Ruaha) -0.402 ±0.355 Year (Tarangire) -0.290 ±0.333 Zebra Year -0.711 ±0.162 Year (Katavi) -0.311 ±0.254 Year (Ruaha) -1.063 ±0.301 Year (Tarangire) -0.758 ±0.283

Note: bolded values were significant at 95% confidence limits.

107

Figure 3.1 The three wildlife protected areas covered in this study: Katavi-Rukwa (A) [Geographical location: between 6o to 7o Latitude, and 31o to 33o Longitude]; Ruaha-Rungwa (B) [5o to 8o’ Latitude, and 34o to 36o Longitude] and Tarangire (C) [3o to 5o’ Latitude, and 35o to 37o Longitude]. The red line shows the boundary of the buffer zone (10 km) used in this study.

108

Figure 3.2 (i) Annual rainfall for Katavi (A) Ruaha (B) and Tarangire (C) wildlife ecosystems [Source: Tanzania National Parks Head offices: Tarangire, Katavi, and Ruaha.

109

(A) (B) (C)

F1,15 =3.83, p = 0.123

F1,12 =0.06, p = 0.818 F1,12 =0.25, p = 0.623

Figure 3.2 (ii) Scatter plots showing temporal changes in annual rainfall for Katavi (A) Ruaha (B) and Tarangire (C) wildlife ecosystems [data source: Tanzania National Parks Head offices: Tarangire, Katavi and Ruaha].

110

(a) (b) 1.6 1.6 (c) Bufallo Elephant 0.30 Giraffe 1.4 1.4 0.25

1.2 )

) 1.2

)

1.0

1.0 y

y y

t 0.20

s s

0.8 n

n 0.8

D D

( 0.15

(

0.6 0.6

L L 0.4 0.4 o

L 0.10 0.2 0.2 0.0 0.0 0.05 1989 1992 1995 1998 2001 2004 2007 2010 2013 1989 1992 1995 1998 2001 2004 2007 2010 2013 1989 1992 1995 1998 2001 2004 2007 2010 2013 Year Year Year (d) (e) (f) 1.4 1.8 0.6 Medium antelopes Zebra Small antelopes 1.6 1.2 0.5

) 1.4

)

1 )

1.0 1

1.2 0.4

t t

s 0.8 i

i 1.0

s n

e 0.3

e e

D 0.8

( 0.6

(

g g

1 0.6 0.2

o o

g 0.4

L L

L 0.4 0.1 0.2 0.2 0.0 0.0 0.0 1989 1992 1995 1998 2001 2004 2007 2010 2013 1989 1992 1995 1998 2001 2004 2007 2010 2013 1989 1992 1995 1998 2001 2004 2007 2010 2013 Year Year Year (a) Bufallo: F = 0.81, p = 0.378 1,24 0.30 Scatterplot of Log Density vs Year (b) Elephant F = 0.01, p = 0.919 Site 1,24 0.6 (c) Giraffe: F1,24 = 6.16, p = 0.02, slope = -0.004 0.002 KTV_10km_Bf Region (d) Med. antelopes: F1,24 = 8.35, p = 0.01,0.25 slope = -0.015 0.005 ) 0.5 KTV_inside KTV (e) Zebra: F = 5.66, p1 = 0.03, slope = -0.02 0.01

1,24 RU_10km_Bf RU +

(f)Small antelopes :F2,24 = 4.21, p = 0.027 Slopes: y 0.4 RU_inside TA t 0.20 i Katavi: = 0.000 0.005

s t

i TA_10km_Bf s

n Ruaha: = 0.004 0.006 Site n e 0.3

Tarangire: = -0.016 e 0.005 TA_inside D

D KTV_10km_Bf

(

0.15 g

0 KTV_inside

L 0.2 1

g RU_10km_Bf Figure 3.3 Scatter plots showing average densityo 0.10 on Log10 (density +1) scale for large herbivores’ species or group over time, across L 0.1 RU_inside Katavi (KTV), Ruaha (RU) and Tarangire (TA) protected areas, and within their 10km buffer zones (Bf)TA_10km_Bf.The bolded values were significant at 95% confidence limits. 0.05 0.0 TA_inside 1990 1995 2000 2005 2010 2015 9 2 5 8 1 4 7 0 3 Year 8 9 9 9 0 0 0 1 1 9 9 9 9 0 0 0 0 0 1 1 1 1 2 2 21112 2 Year

Bufallo Giraffe Zebra 1.75 2.2 2.0 1.50 1.8 1.8

1.25

) )

) 1.6

e e

e 1.00 1.4

1.4

r r

0.75 G

G G

1.2 (

(

g 1.0

o L

0.50 L 1.0 L

0.25 0.8 0.6

0.00 0.6 0.4 0.2 1989 1992 1995 1998 2001 2004 2007 2010 2013 1989 1992 1995 1998 2001 2004 2007 2010 2013 1989 1992 1995 1998 2001 2004 2007 2010 20131.75 Year Year Year Location_KTV

KTV_10km_Bf Rates of change 1.50between locations in a region KTV_inside Katavi (KTV) Ruaha (RU) Tarangire (TA) RU_10km_Bf 1.25 Species Inside 10km_Bf Inside 10km_Bf Inside 10km_Bf RU_inside

Bufallo -1.01 ± 0.28 -0.34 ± 0.28 -1.091.00 ± 0.20 -0.21 ± 0.20 0.13 ± 0.16 0.28 ± 0.16 TA_10km_Bf d

Giraffe -1.13 ± 0.29 -0.53 ± 0.29 -0.96i ± 0.82 0.16 ± 0.82 -0.41 ± 0.31 -0.20 ± 0.31

r TA_inside

Zebra -0.38 ± 0.19 -0.25 ± 0.19 -2.12G ± 0.72 -0.01 ± 0.72 -1.40 ± 0.28 -0.12 ± 0.28

0.75

g o Figure 3.4 Scatter plots showing grid cells L on log scale occupied by large herbivores over time, inside Katavi, Ruaha and Tarangire 0.50 protected areas and within their 10 km buffer zones. The rates of changes between presented here were obtained from non-transformed grid cells data because transformed and non-transformed statistical results were similar. 0.25

0.00

1990 1995 2000112 2005 2010 2015 Year

7 9 9

) )

) (c)

2 2 6 (a) 2 8 (b) 8 7 7 5 6 6 4 5 5 3 4 4 3 3 2 2 2

Average density (indiv/km density Average Average density density (indiv/km Average Average density (indiv/km density Average 1 1 0 0 0 1991 1998 2000 2002 2006 2009 2012 1993 1999 2002 2009 2011 1994 1999 2004 2007 2011 2012 Year Year Year

Katavi Ruaha Tarangire

Figure 3.4 Bar chart showing average densities of elephant from 1991 to 2012 in Katavi, Ruaha, and Tarangire.

113

Appendices

Appendix I: Total number of grid cells (5 km x 5 km) assessed and analyzed during the study inside the protected areas and within 10 km buffer.

Ecosystem Year Inside 10 km buffer Total Tarangire 1994 79 49 128 1999 48 23 71 2004 71 55 126 2007 75 63 138 2011 84 75 159 2012 78 74 152 Total 435 339 774 1991 156 55 211 Katavi-Rukwa 1998 128 55 183 2000 168 72 240 2002 132 42 174 2006 83 29 112 2009 126 57 183 2012 114 44 158 Total 907 354 1261 1993 590 100 690 1999 534 91 625 Ruaha-Rungwa 2002 582 79 661 2009 574 85 659 2011 624 124 748 Total 2904 479 3383

114

Appendix J1: General Linear Model (GLM) analysis of results on temporal change on average densities of large herbivores inside Katavi, Ruaha and Tarangire protected areas and in their 10 km buffer zones. [Sample size for individual species = 36 and DF error = 24]. Significant p-values are in bold font. Abbreviations: SS = Sum of Squares and MS = Mean square, PA = Protected Area.

Type III Model ANOVA Species Source DF Type III SS MS F Value Pr > F Buffalo Year 1 0.11193 0.11193 0.81 0.378 Location 1 0.11196 0.11196 0.81 0.378 Year*Location 1 0.10988 0.10988 0.79 0.383 PA 2 0.21625 0.10813 0.78 0.471 Year*PA 2 0.21553 0.10777 0.78 0.472 PA*Location 2 0.41603 0.20802 1.5 0.244 Year*PA*Location 2 0.41631 0.20816 1.5 0.244 Elephant Year 1 0.00036 0.00036 0.01 0.919 Location 1 0.02399 0.02399 0.7 0.410 Year*Location 1 0.02476 0.02476 0.72 0.403 PA 2 0.06846 0.03423 1.0 0.382 Year*PA 2 0.06820 0.03410 1.0 0.383 PA*Location 2 0.01077 0.00538 0.16 0.855 Year*PA*Location 2 0.01075 0.00538 0.16 0.855 Giraffe Year 1 0.02096 0.02096 6.16 0.021 Location 1 0.00018 0.00018 0.05 0.822 Year*Location 1 0.00018 0.00018 0.05 0.822 PA 2 0.01084 0.00542 1.59 0.224 Year*PA 2 0.01081 0.00540 1.59 0.225 PA*Location 2 0.00111 0.00056 0.16 0.850 Year*PA*Location 2 0.00109 0.00055 0.16 0.853

115

Appendix J1 continues. Type III Model ANOVA Species Source DF Type III SS MS F Value Pr > F Medium antelopes Year 1 0.33469 0.33469 8.35 0.008 Location 1 0.00887 0.00887 0.22 0.642 Year*Location 1 0.00867 0.00867 0.22 0.646 PA 2 0.23968 0.11984 2.99 0.069 Year*PA 2 0.23658 0.11829 2.95 0.071 PA*Location 2 0.06852 0.03426 0.85 0.438 Year*PA*Location 2 0.06779 0.03390 0.85 0.442 Small antelopes Year 1 0.02565 0.02565 1.9 0.181 Location 1 0.00008 0.00008 0.01 0.941 Year*Location 1 0.00006 0.00006 0 0.946 PA 2 0.11440 0.05720 4.23 0.027 Year*PA 2 0.11385 0.05692 4.21 0.027 PA*Location 2 0.00237 0.00118 0.09 0.917 Year*PA*Location 2 0.00243 0.00122 0.09 0.914 Zebra Year 1 0.36559 0.36559 5.66 0.026 Location 1 0.00988 0.00988 0.15 0.699 Year*Location 1 0.00950 0.00950 0.15 0.705 PA 2 0.32996 0.16498 2.55 0.099 Year*PA 2 0.32647 0.16324 2.53 0.101 PA*Location 2 0.07374 0.03687 0.57 0.573 Year*PA*Location 2 0.07357 0.03678 0.57 0.573

116

Appendix J2: General Linear Model (GLM) analysis results on temporal change on average densities of large herbivores’ species or groups, based on in cells (5 x 5) km2 that were occupied, inside Katavi, Ruaha and Tarangire protected areas and their 10 km buffer zones. [Sample size for individual species = 36 and DF error = 24]. Abbreviations: SS = Sum of Squares and MS = Mean square, PA = Protected Area.

Type III Model ANOVA Species Source DF Type III SS MS F Value Pr > F Buffalo Year 1 0.05697 0.05697 0.24 0.631 Location 1 0.55843 0.55843 2.33 0.141 Year*Location 1 0.55363 0.55363 2.31 0.143 PA 2 0.76442 0.38221 1.60 0.225 Year*PA 2 0.76075 0.38038 1.59 0.227 PA*Location 2 1.50340 0.75170 3.14 0.063 Year*PA*Location 2 1.50219 0.75110 3.14 0.063 Elephant Year 1 0.08570 0.08570 1.41 0.246 Location 1 0.04955 0.04955 0.82 0.375 Year*Location 1 0.04995 0.04995 0.82 0.373 PA 2 0.15814 0.07907 1.30 0.290 Year*PA 2 0.15718 0.07859 1.30 0.292 PA*Location 2 0.00949 0.00474 0.08 0.925 Year*PA*Location 2 0.00939 0.00469 0.08 0.926 Giraffe Year 1 0.00442 0.00442 0.38 0.542 Location 1 0.00061 0.00061 0.05 0.820 Year*Location 1 0.00060 0.00060 0.05 0.822 PA 2 0.00953 0.00477 0.41 0.666 Year*PA 2 0.00957 0.00478 0.42 0.665 PA*Location 2 0.00004 0.00002 0.00 0.998 Year*PA*Location 2 0.00005 0.00002 0.00 0.998

117

Appendix J2 continues. Type III Model ANOVA Species Source DF Type III SS MS F Value Pr > F Medium antelopes Year 1 0.11117 0.11117 1.29 0.267 Location 1 0.00371 0.00371 0.04 0.837 Year*Location 1 0.00371 0.00371 0.04 0.837 PA 2 0.14495 0.07247 0.84 0.443 Year*PA 2 0.14120 0.07060 0.82 0.452 PA*Location 2 0.06757 0.03379 0.39 0.679 Year*PA*Location 2 0.06818 0.03409 0.40 0.677 Small antelopes Year 1 0.02663 0.02663 0.48 0.497 Location 1 0.01841 0.01841 0.33 0.572 Year*Location 1 0.01803 0.01803 0.32 0.575 PA 2 0.18784 0.09392 1.68 0.208 Year*PA 2 0.18644 0.09322 1.67 0.210 PA*Location 2 0.03291 0.01646 0.29 0.748 Year*PA*Location 2 0.03264 0.01632 0.29 0.750 Zebra Year 1 0.24871 0.24871 2.10 0.160 Location 1 0.03759 0.03759 0.32 0.579 Year*Location 1 0.03823 0.03823 0.32 0.575 PA 2 0.01853 0.00927 0.08 0.925 Year*PA 2 0.01919 0.00960 0.08 0.923 PA*Location 2 0.14687 0.07343 0.62 0.547 Year*PA*Location 2 0.14745 0.07373 0.62 0.545

118

Appendix J3: General Linear Model (GLM) analysis results on temporal change on number of grid cells (5 x 5) km2 that were occupied by large herbivores’ species or groups inside Katavi, Ruaha and Tarangire protected areas and their 10 km buffer zones. [Sample size for individual species = 36 and DF error = 24]. Abbreviations: SS = Sum of Squares and MS = Mean square, PA = Protected Area.

Type III Model ANOVA Species Source DF Type III SS MS F Value Pr > F Buffalo Year 1 207.332 207.332 15.03 0.001 Location 1 124.432 124.432 9.02 0.006 Year*Location 1 120.388 120.388 8.73 0.007 PA 2 256.202 128.101 9.29 0.001 Year*PA 2 253.139 126.569 9.18 0.001 PA*Location 2 35.350 17.675 1.28 0.296 Year*PA*Location 2 34.208 17.104 1.24 0.307 Elephant Year 1 752.216 752.216 2.31 0.142 Location 1 545.833 545.833 1.68 0.208 Year*Location 1 568.264 568.264 1.74 0.199 PA 2 58.363 29.182 0.09 0.915 Year*PA 2 63.024 31.512 0.1 0.908 PA*Location 2 99.549 49.774 0.15 0.859 Year*PA*Location 2 107.510 53.755 0.16 0.849 Giraffe Year 1 388.180 388.180 7.11 0.014 Location 1 168.999 168.999 3.1 0.091 Year*Location 1 160.162 160.162 2.94 0.100 PA 2 88.973 44.487 0.82 0.454 Year*PA 2 90.975 45.488 0.83 0.447 PA*Location 2 50.376 25.188 0.46 0.636 Year*PA*Location 2 45.305 22.652 0.42 0.665

119

Appendix J3 continues. Type III Model ANOVA Species Source DF Type III SS MS F Value Pr > F Medium antelopes Year 1 224.719 224.719 3.16 0.088 Location 1 225.894 225.894 3.18 0.087 Year*Location 1 218.250 218.250 3.07 0.092 PA 2 378.119 189.060 2.66 0.090 Year*PA 2 384.069 192.034 2.7 0.087 PA*Location 2 10.851 5.425 0.08 0.927 Year*PA*Location 2 12.404 6.202 0.09 0.917 Small antelopes Year 1 162.020 162.020 1.62 0.215 Location 1 119.788 119.788 1.2 0.284 Year*Location 1 112.911 112.911 1.13 0.298 PA 2 222.960 111.480 1.12 0.344 Year*PA 2 229.836 114.918 1.15 0.333 PA*Location 2 1.477 0.739 0.01 0.993 Year*PA*Location 2 1.400 0.700 0.01 0.993 Zebra Year 1 758.751 758.751 19.33 0.000 Location 1 528.707 528.707 13.47 0.001 Year*Location 1 515.320 515.320 13.13 0.001 PA 2 154.296 77.148 1.97 0.162 Year*PA 2 149.116 74.558 1.9 0.172 PA*Location 2 262.593 131.297 3.34 0.052 Year*PA*Location 2 256.786 128.393 3.27 0.055

120

Appendix K: SAS codes used during the course of analyses performed in this dissertation.

Assessments of change in the types of land cover over time.

Proc GLM data = Land cover; Class Location; model (Land cover type e.g. grassland) = time|location/clparm solution; estimate 'trend' time 1/E; estimate 'trend inside' time 1 time*location 0 1/E; estimate 'trend outside' time 1 time*location 1 0 /E; run;

Assessment of change in average density of large herbivores over time.

Proc glm data = Widlife; class region location; model Log(Density+1) = Year|location|region/ ss2 ss3 clparm solution; estimate 'trend' year 1/E; estimate 'trend Katavi' year 1 year*region 1 0 0 /E; estimate 'trend Ruaha' year 1 year*region 0 1 0 /E; estimate 'trend TNP' year 1 year*region 0 0 1 /E; run;

Effect of land cover-land use on densities of large herbivores.

Step 1: Landscape metric

Proc GLM data = Landscape metric; Class Location; model (# grassland patches) = time|location/clparm solution; estimate 'trend' time 1/E; estimate 'trend inside' time 1 time*location 0 1/E; estimate 'trend outside' time 1 time*location 1 0 /E; run;

Step 2: Average density versus landscape metric Proc GLM data = A; class location; model Zebra density = # grassland patches|location/clparm solution; estimate 'trend' OPN 1/E; estimate 'trend Inside' OPN 1 OPN*location 0 1/E; estimate 'trend Buffer' OPN 1 OPN*location 1 0/E; run;

121

8.0 References

Andimile, M. & Caro, T. (2012) Illegal hunting in the Katavi-Rukwa ecosystem. Afr. J.

Ecol., 51, 172–175.

Arsenault, R and Owen-Smith, N. (2002) Facilitation versus competition in grazing

herbivore assemblages. Oikos, 97, 313–318.

Banda, T., Schwartz, M. W. & Caro, T. (2006) Woody vegetation structure and

composition along a protection gradient in a miombo ecosystem of western

Tanzania. For. Ecol. Manage., 230, 179–185.

Barnes, R. F.W. (1983) The Elephant Problem in Ruaha National Park, Tanzania. Biol.

Conserv., 26, 127–148.

Beale, C. M., Rensberg, S. Van, Bond, W. J., Coughenour, M., Fynn, R., Gaylard, A. &

Sinclair, A. R. E. (2013) Ten lessons for the conservation of African savannah

ecosystems. Biol. Conserv., 167, 224–232.

Berry, H. H. (1997) Aspects of wildebeest Connochaetes taurinus ecology in the Etosha

National Park - a synthesis for future management. Madoqua., 20, 137–148.

Bjørnstad, A. (1976) The Vegetation of Ruaha National Park, Tanzania: Annotated

Check-list of the Plant Species. Serengeti Research Institute. Oslo: Serengeti

Research Institute, S.R.I. Publication No. 215.

Boudreaux, K. C. (2010) Community Conservation in Namibia: Devolution as a Tool for

the Legal Empowerment of the Poor. Working Paper No. 10 - 64. Mercatus

Center, George Mason University.

122

Brashares, J. S., Arcese, P., Sam, M. K., Coppolillo, P. B. & Balmford, A. (2004)

Bushmeat hunting, wildlife declines, and fish supply in West Africa. Science, 306,

1180–1183.

Bruner, A. G., Gullison, R. E., Rice, R. E. & Da Fonseca, G. A. (2001) Effectiveness of

parks in protecting tropical biodiversity. Science, 291, 125–128.

Caro, T. (2008) Decline of large mammals in the Katavi-Rukwa ecosystem of western

Tanzania. Afr. Zool., 43, 99–116.

Caro, T. (2011) On the merits and feasibility of wildlife monitoring for conservation: a

case study from Katavi National Park, Tanzania. Afr. J. Ecol., 49, 320–331.

Caro, T. M. (1999) Densities of mammals in partially protected areas: The Katavi

ecosystem of Western Tanzania. J. Appl. Ecol., 36, 205 – 217.

Caughley, G. (1974) Bias in Aerial Survey. Wildl. Manage., 38, 921 - 933.

Caughley, G. (1977) Sampling in Aerial Survey. Wildl. Manage., 41, 605 - 615.

CITES (Convention on International Trade in Endangered Species) (2010) Consideration

of Proposals for amendment of appendices I and II. In Consideration of Proposal s

for Amendment of Appendices I and II. Fifteen meeting of Conference of Parties

(CoP15). Doha (Qatar) 13 - 25 March 2010.

Clerici, N., Bodini, A., Eva, H., Grégoire, J. M., Dulieu, D. & Paolini, C. (2007)

Increased isolation of two Biosphere Reserves and surrounding protected areas

(WAP ecological complex, West Africa). J. Nat. Conserv., 15, 26–40.

Craigie, I. D., Baillie, J. E. M., Balmford, A., Carbone, C., Collen, B., Green, R. E., &

Hutton, J. M. (2010) Large mammal population declines in Africa’s protected

areas. Biol. Conserv., 143, 2221-2228. Elsevier Ltd.

123

Cumming, D. H. M., Fenton, M. B., Rautenbach, I. L., Taylor, R. D., Cumming, G. S.,

Cumming, M. S., Dunlop, J. M., Ford, G. S., Hovorka, M. D., Johnston, D. S.,

Kalcounis, M. C., Mahlanga, Z. & Portforts, C. V. (1997) Elephants, woodlands

and Biodiversity in Southern Africa. S. Afr. J. Sci., 93, 231 – 236.

De Boer W. F. & Prins, H. H. T. (1990) Large herbivores that strive mightily but eat and

drink as friends. Oecol. 82, 264 – 274.

Epaphras, A. M., Gereta, E., Lejora, I. A., Ole Meing’ataki, G. E., Ng’umbi, G.,

Kiwango, Y., … Mtahiko, M. G. G. (2007) Wildlife water utilization and

importance of artificial waterholes during dry season at Ruaha National Park,

Tanzania. Wetl. Ecol. Manage., 16, 183–188.

Field, C., & Ross, I. (1976). The savannah ecology of Kidepo valley National Park. Afr.

J. Ecol. 14, 1–15

Foley, C., & Faust, L. (2010) Rapid population growth in an elephant Loxodonta africana

population recovering from poaching in Tarangire National Park, Tanzania. Oryx,

44, 205–212.

Frost, P. (1996) The Ecology of Miombo Woodlands. In B. Campbell, The Miombo on

Transition: Woodlands and Welfare in Africa. Bogor, Indonesia: Center for

International Forestry Research (CIFR).

Gaveau, D. L. a., Epting, J., Lyne, O., Linkie, M., Kumara, I., Kanninen, M. & Leader-

Williams, N. (2009) Evaluating whether protected areas reduce tropical

deforestation in Sumatra. J. Biogeogr., 36, 2165–2175.

124

Gereta, E., Ole Meing’ataki, G. E., Mduma, S. & Wolanski, E. (2004) The role of

wetlands in wildlife migration in the Tarangire ecosystem, Tanzania. Wetl. Ecol.

Manage., 12, 285–299.

Jachman, H. (2002) Comparison of aerial counts with ground counts for large herbivores.

J. Appl. Ecol., 39, 841–852.

Kahurananga, J. & Silkiluwasha, F. (1997) The migration of zebra and wildebeest

between Tarangire National Park and Simanjiro Plains, northern Tanzania, in

1972 and recent trends. Afr. J. Ecol., 35, 179–185.

Kerley, G. I. H., Landman, M., Kruger M. L. & Owen-Smith, N. (2008) Effects of

elephants on ecosystem biodiversity. In Scholles, R.J. and Mennell, K. G. (Eds.),

Elephant management. A scientific assessment for South Africa. Johannesburg,

South Africa. Wits University Press.

Kiffner, C., Stoner, C., and Caro, T. (2013) Edge effects and large mammals. In Animal

Ludwig, F., De Kroon, H. & Prins, H. H. T. (2008) Impacts of savanna trees on

forage quality for a large African herbivore. Oecologia, 155, 487–96.

Kikula, I. S. (1986) The Influence of Fire on the Composition of Miombo Woodland of

SW Tanzania. Oikos, 46, 317–324.

Kutner, M., Nachtsheim, C., Neter, J., & Li, W. (2005) Applied linear statistical models.

In M. Kutner, C. Nachtsheim, J. Neter, & W. & Li. New York: McGraw-Hill.

Ludwig, F., De Kroon, H., & Prins, H. H. T. (2008) Impacts of savanna trees on forage

quality for a large African herbivore. Oecologia, 155, 487–96.

125

Macandza, V. A., Owen-smith, N., & Cross, P. C. (2004) Forage selection by African

buffalo in the late dry season in two landscapes. South Afr. J. Wildl. Res., 34,

113–121.

Manase, E., Gara, J. I. & Wolanski, E. (2010) A review of the water crisis in Tanzania’s

protected areas, with emphasis on the Katuma River-Lake Rukwa ecosystem.

Ecohydrol. Hydrobiol., 10, 153–166.

MNRT (Ministry of Natural Resource and Tourism), United Republic of Tanzania.

(2007) The wildlife policy of Tanzania Revised March, 2007. The Government

Printer, Dar es Salaam, Tanzania.

Mora, C., Aburto-Oropeza, O., Ayala Bocos, A., Ayotte, P. M., Banks, S., Bauman, A.

G., … Zapata, F. A. (2011) Global human footprint on the linkage between

biodiversity and ecosystem functioning in reef fishes. PLoS Biol., 9, e1000606.

Msoffe, F. U., Kifugo, S. C., Said, M. Y., Neselle, M. O., Van Gardingen, P., Reid, R. S.

… & de Leeuw, J. (2011) Drivers and impacts of land-use change in the Maasai

Steppe of northern Tanzania: an ecological, social and political analysis. J. Land

Use Sci., 6, 261–281.

Mwalyosi, R. (1992) Land-use Changes and Degradation in South-West Masaai Land,

Tanzania. Environ. Conserv., 19, 145–152.

Nagendra, H. (2008) Do parks work? Impact of protected areas on land cover clearing.

Ambio, 37, 330–337.

Newmark, W. (2008) Isolation of African protected areas. For. Ecol. Environ., 6, 321–

328.

126

Norton-Griffiths, M. (1978) Counting Animals. (2nd Edition) Handbook no. 1.

Techniques in African Wildlife Ecology. African Wildlife Foundation, Nairobi.

Ottichilo, W. K., Leeuw, J. De. & Prins, H. H. T. (2001) Population trends of resident

wildebeest [Connochaetes taurinus hecki (Neumann)] and factors influencing

them in the Masai Mara ecosystem, Kenya. Biol. Conserv., 97, 271–282.

Owen-Smith, N. (1988) Megaherbivores: The Influence of very large body size ecology.

Cambridge University Press, Cambridge, U.K.

Owen-Smith, N., Kerley, G. I. H., Page, B., Slotow, R., & van Aarde R. J. (2006) A

scientific perspective on the management of elephants in the Kruger national park

and elsewhere. S.Afr. J. Sci., 102, 389–395.

Prins, H. H. T. (1987), Nature conservation as an integral part of optimal land use in East

Africa: The case of the Masai ecosystem of Northern Tanzania. Biol. Conserv. 40,

141–161.

Rabe, M. J. Rosenstock, S. S., & deVos, J. C. (2002) Review of Big-Game Survey

Methods Used by Wildlife Agencies of the Western United States. Wild. Soc.

Bul., 30, 46–52.

Ryan, S. J., Knechtel, C. U., & Getz, W. M. (2006) Research Article Range African. J.

Wild. Manage., 70, 764–776.

Samuel M. (1981) Correction of Visibility bias in Aerial Surveys where Animals Occurs

in Groups. J. Wild. Manage., 45, 993–997.

127

Sánchez-Azofeifa, A. G., Daily, G. C., Pfaff, A. S. P. & Busch, C. (2003) Integrity and

isolation of Costa Rica’s national parks and biological reserves: examining the

dynamics of land-cover change. Biol. Conserv., 109, 123–135.

Shaw, J. & Platts, L. (2004) Building the CAMPFIRE Paradigm. Property and

Environment Research Center Report (PERC). Montana, USA.

Skarpe, C., Aarrestad, P. A., Andreassen, H. P., Dhillion, S. S., Dimakatso, T., du Toit, J.

T., & Wegge. (2004) The return of the giants: ecological effects of an increasing

elephant population. Ambio, 33, 276–82.

Smuts, G. L. (1975) Home range sizes for Burchell’s Zebra Equus burchellii antiquorum

from the Kruger National Park. Koedoe, 18, 139–146.

Steinhorst, R. K. & Samuel M. D. (1989) Sightability adjustment methods for aerial

surveys of wildlife populations. Biom., 45, 415–425.

Stoner, C., Caro, T., Mduma, S., Mlingwa, C., Sabuni, G., Borner, M., & Schelten, C.

(2007) Changes in large herbivore populations across large areas of Tanzania. Afr.

J. Ecol., 45, 202–215.

Sulle, E., Lekaita, E., & Nelson, F. (2011) From Promise to Performance ? Wildlife

Management Areas in Northern Tanzania. Tanzania Natural Resource Forum.

TANAPA (Tanzania National Park). (nodate) Tarangire General Management Plan.

Simanjiro: Tarangire National Park. Unpublished report.

TAWIRI (Tanzania Wildlife Research Institute). (2010a) Aerial Census in the Katavi-

Rukwa ecosystem in dry Season 2009. Internal Report. Njiro, Arusha.

128

TAWIRI (Tanzania Wildlife Research Institute). (2010b) Aerial Census in the Ruaha-

Rungwa ecosystem in dry Season 2009. Internal Report. Njiro, Arusha.

USAID. (2013). Tanzania Wildlife Management Areas Evaluation. United States Agency

for International Development. Report.

Valeix, M., Fritz, H., Sabatier, R., Murindagomo, F., Cumming, D., & Duncan, P. (2011)

Elephant-induced structural changes in the vegetation and habitat selection by

large herbivores in an African savanna. Biol. Conserv., 144, 902–912.

Voeten, M. M. & Prins, H. H. T. (1999) Resource partitioning between sympatric wild

and domestic herbivores in the Tarangire region of Tanzania. Oecol., 120, 287–

294

Voeten, M. M., Vijver, C. A. D. M. Van De, Olff, H. & Langevelde, F. Van. (2009)

Possible causes of decreasing migratory ungulate populations in an East African

savannah after restrictions in their seasonal movements. Afr. J. Ecol., 48, 169–

179.

Western, D., Russell, S., & Cuthill, I. (2009) The status of wildlife in protected areas

compared to non-protected areas of Kenya. PLos One, 4, 6140.

Wittemyer G., Elsen P., Bean W. T., Coleman A., Burton O., & Brashers J. S. (2008)

Accelerated Human Population Growth at Protected Area Edges. Science, 321,

123–126.

129

Chapter 4. Effect of Land Cover–Land use Change on Abundance of Large

Herbivores in the National Parks in Tanzania

Abstract

Large herbivores prefer certain types of land cover and/or land cover components during different seasons. Stability of their populations is therefore determined largely by availability of heterogeneous habitats they utilize. The objective of the chapter was to understand the effect of changes in land cover spatial components at Katavi and

Tarangire National Parks in Tanzania, on the populations of elephant, giraffe and zebra from 1980s to 2010s. To achieve this objective, I addressed the following questions: i) have landscape metrics undergone significant change over time, inside and outside the national parks, ii) if changes have occurred, are the patterns of change similar across locations, and iii) are the populations of large herbivores influenced by change on land cover components? Seven types of land cover utilized by large herbivores were identified and four landscape metrics (class area, number of patches, mean patch size, mean distance between land cover patches) were computed for each. The metrics for each type were analyzed to determine if significant changes occurred over time using general linear models (GLM). Open shrubland, closed shrubland, woody savannah, and grassland, experienced significant change over the years. The populations of elephant, giraffe and zebra were positively influenced by a change in land cover class and/or number of patches. Closed shrubland and woody savannah were lost between 1980s and 2010s, and included, of which one that appeared to influence the densities of zebra and elephant. The results suggest that land cover homogenization, reduction, or loss of can have a profound

130

effect on species population and that land cover heterogeneity is needed in order to stop further populations decreases.

1.0 Introduction

Stability of large herbivore populations is determined by the quantity and quality of heterogeneous habitats available for their utilization (Owen-Smith 2004). Large herbivores acquire their food by browsing and/or grazing, depending on species. Some of the mixed feeders, particularly elephants, prefer open woodland, shrubland and grassland during the wet season (Loarie et al. 2009, Ott 2007, de Beer & van Aarde 2008), and prefer closed woodlands during the dry season (Field & Ross 1976, Owen-Smith 1988,

Owen-Smith et al. 2006, Loarie et al. 2009). Primary browsers (e.g., giraffe) depend on open woodlands throughout the year, and frequently select sites with high abundance of leafy vegetation and high visibility (Field & Ross 1976, Pellew 1984, Riginos & Grace

2008, Valeix et al. 2011). Grazers (e.g. zebra) feed on short grass often associated with sites of high quality forage, and therefore depend on woodlands and grasslands, as year- round forage resource (Pienaar 1963, Voeten & Prins 1999). Heterogeneity of such habitats is however being reduced through destruction and isolation (Nagendra et al.

2007, Mwalyosi 1992, Newmark 2008, Msoffe et al. 2011), raising a concern on their ability to support wildlife populations on the long-term. One strategy used worldwide to prevent degradation of wildlife habitats and ensure the long-term viability of wildlife population is the creation of protected areas, such as national parks. But the effectiveness of protected areas, particularly in developing countries, is frequently questioned because they often border poor communities that rely on wildlife and their habitat to sustain their

131

livelihoods (Fisher & Christopher 2007, Mfunda & Røskaft 2010). Nevertheless, protected areas are effective at reducing habitat loss (Bruner et al. 2001, Sánchez-

Azofeifa et al. 2003, Gaveau et al. 2009).

Habitat loss refers to removal of spatial components of habitat from a landscape, which may result into reduced size and number of suitable habitat patches, and increased isolation of the remaining patches (Fahrig 1997, 2003). Decreasing the number of habitat patches reduces the amount and variety of resource available for species utilization, while reduced size can affect species population viability, because a patch may be too small to sustain a local population, hence causing species to decline (Bender et al. 1998, Fahrig

2003). Increased distance between nearest neighbor patches of similar habitat reduces the amount of habitat available for wildlife utilization or make habitats on a landscape inaccessible to migrant animals (Bender et al. 2003). These can affect specialist feeders and species of small home range more negatively than those with large home range or mixed feeders (Andren 1994, Bender et al. 1998).

Habitat loss occurs by conversion or transformation of land cover from one form to another through land clearing (Wilcove et al. 1986, Hobbs et al. 2008), or by compression of large densities of wildlife species in one location by fencing of protected areas, resulting into habitat degradation through overutilization (Beale et al. 2013). In the

Masai-Mara ecosystem (Kenya) populations of resident wildbeest (Connochaetes taurinus) declined from 119,000 in 1977 to about 22,000 in 1997 due to conversion of grasslands, which served as their wet season forage, calving and breeding ranges, into agricultural farms (Ottichilo et al. 2001). Likewise, Grant’s gazelle (Gazella granti),

132

Thomson’s gazelle (Gazella thomsonii) and zebra (Equus burchellii) populations have also declined by 60% between 1975 and 2007, due to loss of grasslands, converted into modern agriculture and urbanization (Mundia & Murayama, 2009).

When densities of large herbivores such as elephant (Cumming et al. 1997), deer

(Côté et al. 2004) and wildebeest (Beale et al. 2013) exceed the carrying capacity of their preserve, and their movement is restricted, conversion or transformation of types of land cover from one form to another occurs due to overgrazing or over browsing, uprooting trees, and breaking stems (Barnes 1983). Such changes result into reduction or even loss of habitat, as evidenced in Tsavo in Kenya (Leuthold 1977), Ruaha in Tanzania (Barnes

1983), Sengwa in Zimbabwe (Mapaure and Moe 2009), Chobe and Linyanti in Botswana

(Kalwij et al. 2010, Teren & Owen-smith 2010), and Kruger in South Africa (Beale et al.

2013, Asner & Levick 2012). In Amboseli National Park in Kenya, structural change of woody vegetation by elephants has resulted in extirpation of gerenuks and giraffes

(Parker 1983, cited in Kerley et al. 2008). In Kruger National Park, fencing to prevent spread of disease from wildlife to cattle has blocked widebeest migration and resulted in land cover degradation due to overgrazing as the wildebeest densities increased inside the park (Beale et al. 2013).

Wildlife protected areas are being degraded and isolated in Tanzania, mostly by cultivation, harvesting of trees, and expansion of settlements (Prins 1987, Mwalyosi

1992, Msoffe et al. 2011). Between 1984 and 2000, 15 km2 of woodland and grassland, within a distance of less than 40 km from the Tarangire National Park (TNP) boundary, were converted into crop farms and settlements, blocking four of the nine wildlife

133

corridors on the western and southern sides of TNP, and resulting in the decline of wildebeest population from 40,000 to 5,000 (Msoffe et al. 2010). In the late 1990s, due to human immigration from northern Tanzania, large communities settled in open areas south of the Katavi National Park (KNP), within the park buffer zone, degrading wildlife habitat through cattle grazing (Tim Caro, pers. comm.) and irrigation farming that required diversion of the Katuma River (Figure 1), hence reducing water flow into Lakes

Chada and Katavi and the Katisunga flood plain, which harbor the highest animal density in the park during dry seasons (Caro 1999, Manase et al. 2010).

Between the late 1980s and early 2000s populations of large herbivores, including waterbuck, topi, gazelle, reedbuck, hartebeest, roan antelope, sable antelope, warthog and zebra, were declining in four of the eight Tanzanian wildlife ecosystems (including TNP and KNP) surveyed, whereas populations of elephants and giraffes remained stable or increased (Stoner et al. 2007, Caro 2008, 2011). The decline of large herbivores was attributed to illegal hunting and the large quotas allocated to hunting companies (Caro

2008, Andimile et al. 2012). However, the loss of wildlife habitats may also have contributed to the declines. For instance, from 1988 to 2009 a significant increase in bareland occurred due to the degradation of woody vegetation and grassland inside and outside TNP (Mtui et al. under review). These findings concur with Pelkey et al. (2003), who reported the loss of woodlands in protected areas, hence decreased amount of vegetation greenness during the dry season in northern Tanzania, where TNP is located, between 1982 and 1994. However, no recent studies have assessed habitat loss in the national parks in Tanzania and its potential impacts on large herbivores.

134

The main objective of the study was to understand the effect of changes in the types of land-cover in Tanzanian national parks on the populations of large herbivores over the past two decades. To achieve this objective I addressed the following questions: i) Have the landscape metrics undergone significant change between 1980s and 2010s inside and outside the national parks, ii) if changes have occurred, are the patterns similar inside and outside the parks, and iii) are the populations of large herbivores influenced by change on land cover components?.

To answer these questions, class area (total area), number of land cover patches, mean patch size, and mean distance between patches of land cover were evaluated. These four metrics can provide complete information about landscape changes over time

(Turnner et al. 2001, McGarical & Mark 1995, Fahriq 2003). Similar metrics have been used in other studies to assess the extent of land cover change and their effect on animals

(Grossmann & Mladenoff, 2007, Bender et al. 1998, Kie et al. 2002, Grainger et al.

2005). The class area metric provides information about how much of a particular land cover is available, number of patches shows how many patches of that particular land cover are available for species utilization, mean patch size metric provides information about size distribution of all patches of that a particular land cover, and the nearest neighbor distance shows how isolated are patches suitable for species utilization

(McGarigal & Mark 1995, Gustafson 1998). If amount of class area was significantly lower, fewer land cover patches, reduced and smaller-sized patches, and longer distance between suitable patches, at present than in the previous years (2010s vs. 1980s or

1990s), then habitat loss has occurred.

135

Three species of large herbivores (elephants, giraffe and zebra) were evaluated in relation to changes in landscape metrics. Elephant was chosen because it is a mixed feeder, and its population has remained stable over the past 20 years (Chapter 3), whereas giraffe and zebra are specialist feeders (Riginos & Grace 2008, Valeix et al. 2011, Voeten

& Prins 1999) and their populations have declined over time (Chapter 3). Hence evaluation of the populations of the three species vs. changes in land cover components would be interesting to see if land cover changes may have contributed to their population stability of instability (Chapter 3).

Both Inside and outside KNP and TNP, the densities of the three species were expected to respond differently to changes in different types of land cover given their differences in terms of habitat utilization. First, because elephants prefer closed woodlands during the dry season, their density is expected to increase with increase in class area and number of patches of woody savannah, closed shrubland and swamps, whereas they are expected to decrease with increasing distance between patches of the woody savannah, closed shrubland and swamps, and decrease with increases in class areas and number of patches for grassland, open shrubland and savannah, and bare or less vegetated land.

Because of their selection for habitats with fine leaves and their preference for open areas as avoidance to predation, the density of giraffes is expected to increase with an increase in class area and number of patches of savannah and open shrubland, decrease with an increase in distance between patches of the two classes, decrease with the increase in class area and number of patches of grassland, woody savanna and closed

136

shrubland, in class area and number of patches of grassland, woody savannah and closed shrubland.

Because zebras forage on short fine grasses, their density is expected to increase with increases in class area and number of patches of woody savannah, savannah, open shrubland and swamps, decrease with increases in distance between patches of these four classes, and decrease with decreases in class area and numbers of patches of grassland.

All three species were expected to decrease with increases in class area and number of patches of bare or less vegetated land, increase with increases in distance between patches of bare or less vegetated land.

2.0 Materials and Methods

2.1 Study sites descriptions

To achieve the goal of this chapter, changes in densities of elephant, giraffe and zebra were investigated in relation to change in landscape metrics inside and outside the

Tarangire (TNP) and Katavi (KNP) national parks (Figure 4.1). Because of the varying sizes of the two parks, density estimates were used as a comparative measure of population abundance. The study was restricted to these two parks because of availability and suitability of data and limited funding to conduct the research.

The TNP covers 2,600 km2 (between 3o40’ to 5o35’ latitude, and 35o45’ to 37o longitude) with elevation ranging from 1,200 m to 1,600 m above sea level (Figure 1).

The park was established as a Game Reserve in 1957 and declared a National Park in

1970. The TNP is bordered by Lake Natron and the Mto-wa-Mbu Game Controlled Area

(GCA) to the north, the Lolkisale GCA and the Simanjiro Plains to the east, the

137

Mkungunero Game Reserve (GR) to the south, and Lake Burunge, the Burunge GCA, the

Kwakuchinja Open Area, and Lake Manyara National Park to the west (Figure 4.1). The

GCA is one category of protection in Tanzania where human activities, such as settlements and hunting with permits can occur. Average annual rainfall is about 655 mm

(Figure 4.2) with short rains between October and December and heavy rains between

February and April or May. Hot season (December to February) temperatures range from

TNP, pastoralism has been primary form of land use over the past two centuries (Prins

1987). However, in the past two decades agricultural activities have increased on the north east and eastern regions adjacent to the park, blocking wildlife migration routes and dispersal areas (Mwalyosi 1992, Msofe et al. 2011).

The KNP covers 4,238 km2 (between at 6◦63’ to 7◦30’ latitude, and 30◦75’ to

31◦74’ longitude) with elevation ranging from 800 m to 1,600 m above sea level (Figure

4.1). The park was established in 1974 with 1,816 km2 of land, and enlarged in 1998 to reduce pressure from settlements and cattle grazing. The KNP is bordered by Msanginia

Forest Reserve (FR) and Mlele GCA to the north, Lwafi GR and Nkamba FR to the west,

Usevya Open Area to the south, and Rukwa GR to the south and south east (Figure 4.1).

The inhabitants include the Pimbwe people, who are native to the area and the Sukuma people who emigrated from the northern regions. Both tribes practice horticultural 138

activities and pastoralism (Caro 1999a). The integrity of the park is threatened by increased damming of River Katuma for rice cultivation, farming inside and adjacent to the park, and wildfires caused by poachers, farmers and people traveling the main road across the park (Manase et al. 2010).

Average annual rainfall is about 955 mm (Figure 4.2), which falls from November to April or May. The vegetation consists of grassland interspersed with miombo woodlands and mixed woodlands. Miombo forms a single story, with a light, closed canopy of deciduous woodland usually greater than 15 m tall and dominated by trees of the genera Brachystegia, Julbernadia, and Isoberlinia (Kikula 1987, Frost 1996).

Underneath the trees are layers of scattered shrubs, grasses, and forbs that grow to a height of 0.3 to 100 cm with 50 - 75% ground cover (Lawton 1978). The genera

Markhamia, Grewia, Terminalia, Combretum, Syzygium, and Acacia also occur in the miombo woodland of the KNP (Banda et al. 2006). During the dry season, large mammals feed on swampy vegetation occurring in the large seasonal Katavi and Chada lakes, and the Katisunga flood plain (Caro 1999), which is maintained by Katuma River

(Figure 4.1).

National parks (NP) are given the highest level of protection, meaning that no permanent settlement, hunting, livestock grazing, or logging is allowed. Like the NP, permanent human settlement, cultivation and livestock grazing are prohibited in the GR, although hunting, fishing and logging are allowed with a special license. However, in the

GCAs human settlement is allowed, along with grazing, cultivation, and licensed hunting.

In the forest reserves (FR), settlements, cattle grazing, and hunting by tourists is prohibited, although selective logging with a license is allowed. NPs are regulated by

139

Tanzania National Park (TANAPA), GR and GCA are regulated by the Tanzania

Wildlife Division authority, and the FR by the Tanzania Forest and Bee Keeping

Department. Finally the open areas are managed by local governments and allow public use, including settlements, farming, and livestock grazing.

2.2 Study approach

The analysis in this study was conducted in three parts: (i) derivation of land cover raster maps (30 m × 30 m resolution), and computation of the four landscape metrics, (ii) analysis of animal density data and their relation to land cover components, and (iii) analysis of land cover components and species densities vs. rainfall. Since rainfall is a fundamental driver in semi-arid ecosystems, the third part of analysis was required in order to disentangle the effect of change on land cover components and on population of species, from that of rainfall.

2.2.1 Derivation of land covers maps

We selected a set of six Landsat images of 1988, 1999, and 2009 images (path

168, row 63) for TNP, and the 1984, 1999, and 2011 images (path 171, row 65) for KNP, at based upon similar times of the year to minimize seasonal and sun angle effects, which could affect multi-temporal comparisons (Singh 1989, Jensen 2005). Specifically,

Thematic Mapper (TM) and Enhanced Thematic Mapper plus (ETM+) images acquired during a period of short rains on 01/10/1988, 01/11/1999, and 04/11/2009 for TNP and dry season on 29/06/1984, 9/03/1999, and 07/26/2011 for KNP were chosen. All images were obtained from the U.S. Geological Survey (USGS) (www.earthexplorer.usgs.gov and www.glovis.usgs.gov). Rainfall variation may have implications for the change detection analysis (Jensen 2007) and caution must be used in interpreting the results. In

140

TNP, the annual total rainfall varied during the timeframe during which the data imagery were acquired (Figure 4.2). In KNP, no variation in rainfall existed (Figure 2) between

1999 and 2011, and no rainfall data were available for 1984.

The six Landsat images were classified using Maximum Likelihood Classification

(MLC) to generate maps with eight classes of land cover, including barren land, closed shrubland, grasslands, open shrubland, savannah, swamps, woody savanna and water, for the years 1988, 1999 and 2009 for TNP, and 1984, 1999 and 2011 for KNP. A detailed description of the ground truthing methodology, processing the Landsat images, classification and assessment of accuracy of classification of the images are provided in

Mtui et al. (under review) and chapter 2 in Mtui (2014).

Four metrics, including class area (km2), number of patches, mean patch size

(km2) and mean distance between nearest neighbor patches (m), were computed on raster maps (30 m resolution) for barren land, grasslands, open and closed shrublands, savanna and woody savanna) for each date with a four-cell neighborhood definition using

FRAGSTATS software (version 4) (McGarigal et al. 2012).

2.3 Assessment of large herbivores population densities in relation to land covers change

2.3.1 Animal density data

Animal density data from 1991 to 2012 were obtained from the Tanzania

Wildlife Research Institute (TAWIRI). TAWIRI conducted aerial surveys using

Systematic Reconnaissance Flights (SRFs), following the guidelines designed by Norton-

Griffiths (1978). The surveys were typically conducted during the dry season (September to October) every 1 to 3 years, and sometimes during the wet season (March to May).

The surveys were carried out using two aircrafts, each with two parallel rods mounted on

141

the wing struts. The aircrafts were flown along parallel transects oriented in the east-west or north-south direction, 5 km apart from each other and about 106 m above ground.

Information Monitoring Unit (CIMU), specifically for SRFs surveys (TAWIRI, 2010b).

The survey team also grouped the subunits into a grid system of cells covering the surveyed area.

For decades aerial surveys have been used to estimate the abundance of wildlife species. The estimates have been criticized as inaccurate (Caughley 1974) due to biases in sightability (Steinhorst & Michael, 1989) and visibility (Samuel, 1981). However, aerial surveys are the only practicable method of estimating the number of large animals over a large land area and provide estimates with a level of approximation judged acceptable (Caughley, 1977). The level of acceptability depends on the objective of the study because for short term assessments, impresion can be too large (Leatherwood et al.

1978). According to Rabe et al. (2002) the aerial survey is the most effective method to estimate large animal populations when sources of biases are substantially minimized.

142

The sighting bias relates to low probability of sighting single and small groups of animals, and can be minimized by keeping operational factor such as searching rate and height above ground level within reasonable limits. Counting large and conspicuous animals at a height of between 100 m to 200 m above ground, and at a speed varying between 180 and 250 km/h is the recommended method. Visibility bias relates to animals not being visible to the observer because they are concealed by other animals, obstructions such as tree canopies or because their color makes them blend with the background. Visibility bias can be reduced by using a double counting technique

(Jachman, 2002). These biases were accounted for by TAWIRI as they used two aircrafts, each with two observers, flown about 106 m above ground at a speed of between 212 km/h. While observer identity did change between surveys, an overlap in observer personnel across surveys did occur which provided important continuity in counting techniques (TAWIRI, 2010a, 2010b).

Because the purpose of this was to determine the effect of land cover change, between the years of 1984, 1999 and 2011 for KNP, and 1988, 1999 and 2009 for TNP

(section 2.2), on densities of wildlife species, density survey data from similar periods would have been most suitable. However, the unavailability of density data for some of these time periods meant that in some cases wildlife density data from up to 7 years before the acquisition of the land cover data had to be used, assuming that no significant land cover changes had occurred subsequent to the wildlife count. Animal density data covering surveys carried out in 1991, 2000 and 2012 for KNP, and 1993, 1999 and 2011 for TNP were selected from among the data set obtained from Tanzania Wildlife

Research Institute (TAWIRI).

143

Only animal density data from surveys carried out during the dry seasons and that covered similar extensive land areas in each ecosystem were considered. Wildlife count data were selected to include only those which were within the park and a 5 km buffer zone. In Tanzania, human activities are observed up to 2 km from PA boundaries, despite the fact that 7 km is considered as an official buffer zone, as human population around park boundaries has been drastically increasing since 1999 (Caro et al. 2011). A five km buffer was selected to enable comparison with land cover data that were sampled within the same extent of geographic coverage (see section 1). Furthermore, only grid cells from transects that cut across the entire park and its buffer zones were selected for analysis.

Each borderline grid cell was treated as inside or outside a protected area boundary only when 75% of the cell was on the respective side, otherwise it was omitted from the analysis. Wildlife density data for 22 species (Table 4.1) were available for total of 211 and 91 grid cells inside and outside TNP, and 438 and 130 grid cells inside and outside

KNP. However, to achieve this objective, only densities data for elephant, giraffe and zebra were considered.

2.4 Assessment of large herbivores population densities in relation changes in rainfall

Other studies have shown that wildlife densities can be influenced by the amount of rainfall (Ogutu & Owen-Smith 2003,Valeix et al. 2008). Therefore, the impact of rainfall on land cover spatial components, and on the densities of wildlife species was also assessed. Rainfall data that were available and acquired during time periods as close as possible to land cover data and wildlife count years were selected. For TNP, rainfall data were available for the exact same years as land cover data (i.e. 1988, 1999 and

2009), and also for the exact same years as density data (1993, 1999, 2011). But for KNP

144

rainfall data covered 1996, 1999 and 2011, while land cover data were of 1984, 1999 and

2011 while density data of 1991, 1999 and 2011. Because of the variation among the three data sets, KNP data were not included in the analysis.

2.5 Statistical analysis

Analyses were performed using SAS software version 9.2. Prior to the analyses, data for land cover and animal density were examined using scatter plots and Ryan-Joiner test to test for normality of residuals. Data were Log10 transformed to ensure the assumptions of normality of residuals and linearity were met.

Using general linear models (GLM), the two-step analysis was comducted. Each of the four landscape metrics was evaluated to determine if change had occurred over time, for each land cover type inside the two National Parks and within their 5 km buffer.

Each park location was treated as independent and contained one observation from each of the three different periods i.e. 1980s, 1990s and 2010s. The analyses included one response variable (i.e. a landscape metric), and three independent variables (i.e. time, location (inside/buffer zone), and their interaction). Second, the temporal effect of changes in the landscape metrics for each land cover type on the densities of the three species of large herbivores was examined. Only three landscape metrics, class area, number of patches, and mean nearest neighbor distance, were because they demonstrated evidence of changes over time (analysis in step 1). The mean patch size was omitted because it remained unchanged over time (in step 1), and therefore was not considered further. The analyses were performed separately for each species, and each analysis

145

included one response variable as average density and three independent variables which are landscape metric, location and their 2-ways interactions.

Lastly, the temporal effect of changes in amount of rainfall was tested on the: i) the landscape metrics that exhibited changes over time for each land cover type and ii) densities of the three large herbivore species inside and outside the two national parks.

The analyses were performed separately for each metric and for each species. Each analysis included one response variable as landscape metric or average density of species, and three independent variables which are rainfall, location (inside/outside) and their 2- way interactions [Note: rainfall values were the same inside and outside].

Wherever the statistical results were significant (P ≤ 0.05) in step one or step two, estimates of the slopes were obtained using ESTIMATE statement in PROC GLM

(Appendix H) to determine the general trend of change. Results are presented as mean ± error, unless otherwise noted.

3.0 Results

3.1 Land cover change in terms of landscape metrics

Three of the four landscape metrics changed over time in one or both Katavi

(KNP) and Tarangire (TNP) National Parks (Figure 4.3 – 4.10, (i-viii)). Changes were observed in at least five types of land cover, as indicated by one or more of the metrics class area, number of patches and mean nearest neighbor distance metrics (Figure 4.3-

4.7).

146

Class area changed significantly over time for open shrubland in KNP and bare or less vegetated land in TNP (Figure 4.3-i and 4.4-iv). In TNP, bare or less vegetated land increased both inside and outside the park at a rate of [Mean ± S.E] 2.2 ± 0.2 km2/year. In

KNP, open shrubland decreased at a rate of 7.2 ± 0.9 km2/year (Table 3b). However, the rate of decline was higher inside (12.2 ± 1.2 km2/year) (Figure 4.4), than outside (2.3 ±

1.2 km2/year) the park.

The number of patches for closed shrubland and grassland changed significantly over time in TNP (Figure 4.5-ii-iii). The patches of closed shrubland and grassland decreased at rates of 2196.7 ± 200.1 and 1701 ± 21.5 patches per year, respectively

(Table 4.2a). However, the levels of decrease for both classes were not similar across locations (Table 4.2a). Inside the park, the closed shrubland patches decreased by 3161.3

± 283.0, and outside by 1232 ± 283.0 (Figure 4.5-ii). Similarly, the number of grassland patches decreased at a higher rate inside (2491 ± 30.5) than outside (910.63 ± 30.5) the park (Figure 4.5-iii).

The mean distances (m) between nearest neighbor patches were significant only for woody savannah in TNP (Figure 4.6-viii), closed and open shrublands, and grassland in KNP (Figure 4.7-ii to iv). In TNP, the mean distance between woody savannah nearest neighbor patches increased at a rate of 0.586 ± 0.1 (m/year), but the increase was three times higher outside (0.879 ± 0.083) than inside (0.292 ± 0.08) the park (Figure 4.6-viii).

In KNP, the mean distances between the nearest neighbor patches for the three land cover classes increased, both inside and outside the park, at similar rates of 0.403 ± 0.1

147

(m/year) (closed shrubland), 0.302 ± 0.03 (m/year) (grassland) and 1.693 ± 0.336

(m/year) (open shrubland) (Figure 4.7-ii to iv).

3.2 Effect of land cover change on the densities of large herbivores

In TNP, elephant density was significantly related to area covered by open shrubland (Figure 4.8-i), number of savannah patches (Figure 4.9-iv), and to the mean distance between nearest neighbor patches of swamps (Figure 4.10-iv). The density increased both inside and outside the park at a rate of about 0.01 ± 0.001 indiv/km2 with increase of open shrubland area, and (1.2 ± 2.3) x 10-5 indiv/km2 with the increase of number of savannah patches, although the rate of increase was six times higher outside than inside the park (Figure 4.9-iv). On the other hand, density decreased inside while increasing outside the park at rates of 0.23 ± 0.03 indiv/km2 and 0.30 ± 0.02 indiv/km2, respectively, with the increase in distance between swamps patches (Figure 4.10-iv).

In TNP, the density of giraffe was significantly related to area covered with grassland (Figure 4.8-ii), number of swampy patches (Figure 4.9-v), and distance between nearest neighbor patches of bare or less vegetated land (Figure 4.10-i) and to mean distance between patches of savannah (Figure 4.10-ii). Density decreased inside but increased outside the park at rates of 0.001 ± 0.000 and 0.003 ± 0.000 indiv/km2 respectively, as the area of grassland increased (Figure 4.8-ii). Also density increased both inside and outside the park at a rate of about (3.7 ± 0.8) × 10-5 indiv/km2 with the increase of number of swampy patches (Figure 4.9-v). Density increased inside and outside the park at a rate of 0.06 ± 0.01 indiv/km2 with increase in distance between patches of bare or less vegetated land (Figure 4.10-i). On the other hand, density

148

decreased inside the park and increased outside at rates of 0.02 ± 0.04 and 0.04 ± 0.01 indiv/km2, respectively, as the distance between nearest neighbor patches of savannah increased (Figure 4.10-ii).

In TNP, the density of zebra was significantly related to the number of patches for open shrubland, savannah and woody savannah (Figure 4.9-i to iii), and distance between nearest neighbor patches for open shrubland (Figure 4.10-v). Density increased both inside and outside the park at rates of about (6.70 ± 1.03) x 10-4, (6.10 ± 1.2) x 10-4, and

(4.2 ± 0.8) x 10-4 indiv/km2 with increasing number of patches of open shrubland, savannah, and woody savannah, respectively (Figure 4.9-i to iii). Density decreased both inside and outside the park at the rates of -13.62 ± 1.15 and -0.56 ± 0.21 indiv/km2 respectively, with the increase of distance between patches of open shrubland (Figure

4.10-v). In KNP, the density decreased both inside and outside the park at a rate of 0.01 ±

0.003 indiv/km2 (Figure 4.8-iii).

3.3 Effect of rainfall on types of land cover and on the densities of large herbivores

Number of patches of savannah and open shrubland, and mean distance between patches of open shrubland were significantly influenced by rainfall (Figure 4.12). The number of patches for open shrubland and savannah increased as rainfall increased, while the distance between patches of open shrubland decreased as rainfall increased inside and outside the park (Figure 4.12). Among the three species (i.e. elephant, giraffe and zebra), only the density of zebra was influenced by rainfall (Figure 4.13). The density increased inside and outside the TNP as rainfall increased (Figure 4.13).

149

4.0 Discussion

4.1 Limitations of the study

Before discussing the results the shortcomings of data and analyses will be discussed. The small number of protected areas and limited number temporal data points for land cover, animal density and rainfall analyses, may have limited the ability to detect real changes. The time difference in terms of acquisition or collection of the land cover and animal density may have had effect on the precision of the results. Potentially confounding variables such as availability of water (Gereta et al. 2004, Epaphras et al.

2007), predation by carnivores (Owen-Smith et al. 2005, Hopcraft et al. 2005, Grange &

Duncan, 2006) poaching, (Caro 1999), and wildfires (Holdo et al. 2009) were not included, which also are likely to influence changes on the types of land cover, as well as densities of elephant, giraffe and zebra. Previous research does indicate that water availability (Gereta et al. 2004) and Ruaha (Epaphras et al. 2007), and illegal hunting in

Katavi (Caro 1999, 2008, 2011) do influence species density.

4.2 Changes in land cover over time

The results show evidence of loss in five out of the eight types land cover utilized by wildlife in KNP and TNP, between the 1980s and the 2010s (Figures 4.3 – 4.7). In

KNP, the area covered by open shrubland decreased, and the distance between patches of closed shrubland, open shrubland and grassland increased. In TNP, the area covered by bare or less vegetated land increased, the number of closed shrubland and grassland patches decreased, and the distance between patches of woody savannah increased. The number of patches and distance between patches are a function of total landscape (Fahrig

150

1997, 2003), meaning that the observed decreases in the amount of woodlands and grassland, and increased area of bare or less vegetated in the parks indicates a loss of habitats has occurred.

As expected, the loss of woody savannah, closed shrubland and grassland, and the increase in bare or less vegetated land occurred during the dry season, when no vegetation regrows and is still consumed by large herbivores. Since rainfall remained unchanged over the years, its contribution to changes on land cover is not significant. A possible primary cause for these land cover changes is land clearing for cultivation inside and outside the parks (Mtui et al. under review, Mwalyosi 1992, and Msoffe et al. 2011), which might have restricted large herbivores’ access to their dispersal areas (Voeten et al.

2009). The density of elephant in TNP is about 4 to 10 times the recommended carrying capacity of 0.5 indiv/km2 (Cumming et al. 1997) (Figure 4.11). If their movement is restricted then land cover degradation is unavoidable. Similar impacts of elephant on habitats were reported in Ruaha National Park in Tanzania (Barnes 1983).

Vijver et al. (1999) concluded that the decline in much of the woody vegetation cover, between 1971 and 1996 in TNP, was caused by severe drought conditions that occurred in 1993, and possibly also an earlier drought, from 1991 to 1992. This is contrary to the present findings where closed woodlands (woody savannah and closed shrubland) that declined in Tarangire had no significant relationship with rainfall, and the open woodlands (savannah and open shrubland) that were influenced by rainfall did not show evidence of change.

151

4.3 Influence of rainfall and land cover change on densities of large herbivores

The results show that of the three species (elephant, giraffe and zebra) of large herbivores only the density of zebra was influenced by rainfall in TNP, and the densities of all species in TNP, and only zebra in KNP were influenced by changes in the types of land cover.

The density of elephants in TNP was not influenced by amount of rainfall. This lack of relationship is contradictory to other studies that conclude rainfall influenced elephant abundance. In Hwange National Park in Zimbabwe (Valeix et al., 2008) elephant density increased as dry season rainfall increased.

Contrary to expectations the density of elephants increased inside and outside

TNP, with an increase in open shrubland area, and number of patches of savannah

(Figures 8-i and 9iv), and decreased inside the park but increased outside as the distance between swampy patches increased. This latter increase outside the park (Figure 10-iv) was consistent with expectations. Osborn (2004) observed that regardless of season, elephants tend to select the maximum amount of highly nutritious food available. Loarie et al. (2009) concluded that throughout the year, elephants selected sites with highest amount of greenness and tracked the greener vegetation, despite being constrained by water availability. In this study, the closed woodlands, which would have been greener and preferred by elephant during dry season, were substantially lost. The second most green alternative vegetation cover was open woodlands (savannah and open shrubland) which in this study were positively influenced by amount of rainfall. Therefore these results support the findings of Loarie et al. (2009) and Osborn (2004).

152

In Sengwa Wildlife Research Area, Osborn (2004) noted that the nutritional quality of wild grasses during dry season was lower compared to the quality of crop species on communal lands, and that the movement of elephant out of protected areas into crop fields corresponded to the quality of crop species. This suggests that their differing movements inside and outside the park might be related to the quality of grazing pasture or greener vegetation, as suggested by Loarie et al. (2009), rather than distance.

Highly mobile animals with a large home range are affected less by distance between patches (Andren 1994, Bender et al. 1998), and elephants in Tarangire are reported to have a home range varying between 477 to 5060 km2 (Galanti et al. 2006). Therefore the distance between patches is not likely to be an important factor.

The density of giraffe was not influenced by amount of rainfall. This result is consistent to Ogutu & Owen-Smith (2003, 2005) who in Southern Africa found that giraffes in high abundance remained so through dry seasons. As expected the density of giraffes inside and outside TNP increased with an increase in number of swampy patches

(figure 4.9-v) and distance between patches of bare or less vegetated land (figure 4.10-i).

Also the density of giraffe inside the park decreased with the increase in area of grassland

(Figure 4.8-ii) and distance between patches of savannah as expected (Figure 4.10-ii).

However under these same conditions, density increased outside the park (Figures 4.8-ii and 10ii), which is inconsistent with expectation. These patterns may be due to a shortage of food and competition inside the park (Field & Ross 1976), and/or by predation

(Riginos & Grace 2008). In Kidepo Valley National Park in Uganda, Field & Ross (1976) observed an overlaping of food resources between elephant and giraffe, with over two thirds of plant species in elephant diet also important for giraffe during dry season. With 153

such an extensive overlap in food resources, and the loss of closed woodlands observed in the parks (objective 1), giraffes may have faced a food shortage, and/or competition with elephants in Tarangire. Giraffe tends to forage in more open habitat to avoid predation, especially when accompanied by juveniles, spending more time at the sites with the fewest trees for visibility (Riginos & Grace 2008, Valeix et al. 2011). In this study, giraffe may have fed in swamps to avoid predation or competition from elephants, in addition to seeking green fine vegetation.

The density of zebras in TNP was positively influenced by amount of rainfall.

This positive relationship is contradictory to other studies. In Southern Africa, Ogutu and

Owen-Smith (2003, 2005) and Grange and Duncan (2006) reported that amount of rainfall did not prove to have an effect on the abundance of zebra. The densities of zebra in KNP and TNP responded differently to changes of the types of land cover. Inconsistent with expectations, zebra in KNP decreased, inside and outside the park, with increase of swampy area (Figure 4.8-iii), while in TNP no significant response to the any of the swamp spatial components was found. On the other hand, as expected, zebra density increased, inside and outside the TNP, as patches of open shrubland, savannah and woody savannah increased (Figure 4.9-i to iii), and decreased as the distance between patches of open shrubland increase (Figure 4.10-v). This result is interesting because the open shrubland and savannah were positively related to the amount of rainfall suggesting that zebra were selecting these patches to have fine green grasses to eat (Pienaar 1963,

Voeten & Prins 1999).

154

Zebra densities increased on types of land cover that were also utilized by elephants, and no indication of zebras avoiding elephants due to competition for food, as was suspected between giraffe and elephant. Similar observations were reported by

Valeix et al. (2011). Zebras are mobile animals and can access forage as long as it is available (Smuts 1975). These findings therefore suggest that, as long as heterogeneity of land cover and their patches is maintained, the zebra abundance of zebra will likely be sustained (Hobbs & Gordon 2010).

Zebras were positively influenced by rainfall, and the most of land covers utilized by zebras were positively influenced by rainfall. The amount of rainfalls in TNP has remained the same for about two decades while the density of zebra has substantially declined (Mtui 2014). Such declines may be due to food shortages given the reported loss of grassland and the closed woodlands.

5.0 Conclusion

These results show that during dry season, the densities of elephant, zebra and giraffe depend on areas and/or number of patches of open shrubland, savannah, woody savannah, and swamps. The number of land cover patches is more important in maintaining the population of zebra than a total land cover area. As for elephant and giraffe, both number of patches and total area of the type of land cover seem to be important to maintain their populations. The significant relationships between land cover components and densities of large herbivores suggest that homogenization of these types of land cover as a result of any activity that may cause its reduction or loss, such as

155

human encroachment, cultivation or mismanagement of fires, can have profound consequences for species population viability (Owen-Smith 2004).

The observed losses in the amounts of grassland, closed shrubland, and woody savannah in TNP, and of grassland and open and closed shrubland in KNP, are of great conservation concern because heterogeneity of habitats needed for species utilization is reduced. The observed shrinkage of wildlife habitats and subsequent decline of species populations reported in chapter 2 of Mtui (2014) was possibly due to loss of land cover.

Considering that rainfall in both national parks did not change considerably for about two decades, the losses of land cover components utilized by large herbivores have occurred, and have had a negative effect on species or groups of large herbivores. Active management of protected areas is required to restore and maintain availability of heterogeneous forage in order to stabilize species populations.

156

Table 4.1 Large herbivore species recorded by TAWIRI in KNP (22 species) and TNP (19 species). The tick marks indicate the presence of species in the park.

Scientific name Common name Feeding type KNP TNP   Aepyceros melampus Impala Mixed   Alcelaphus buselaphus Hartebeest Grazer Connochaetes taurinus Wildebeest, blue Grazer    Damaliscus lunatus Topi Grazer   Equus burchellii Zebra Grazer Gazella granti Grant gazelle Mixed  Gazella thomsonii Thomson gazelle Mixed  Giraffa camelopardalis Giraffe Browser    Hippopotamus amphibius Hippopotamus Grazer  Hippotragus equinus Roan antelope Grazer  Hippotragus niger Sable antelope Grazer

Kobus ellipsiprymnus Waterbuck Grazer   Kobus vardoni Puku Grazer    Loxodonta africana Elephant Mixed   Madoqua spp. Dikdik Browser  Oreotragus Klipspringer Browser Oryx gazella Oryx Mixed   Ourebia ourebi Oribi Mixed   Phacochoerus africanus Warthog Grazer   Redunca sp Reedbuck Mixed

Sylvicapra grimmia Duiker, red Mixed     Syncerus caffer Buffalo Browser   Taurotragus oryx Eland Mixed Tragelaphus imberbis Lesser Kudu Browser   Tragelaphus scriptus Bushbuck Browser   Tragelaphus strepsiceros Grater Kudu Browser Total 22 19

157

Figure 4.1 Tarangire National Park (TNP, right) and Katavi National Park (KNP, left) in Tanzania. The Katavi map show the main

158

F1,15 =3.83, p = 0.123 F1,12 =0.06, p = 0.818

Figure 4.2 Total annual rainfall (dotted line) and average annual rainfall (straight line) recorded in Tarangire (A) and Katavi (B),

National Parks head offices and the scatter plots of Katavi (C) and Tarangire (D) showing change over time. 159

(iii) (i) (ii)

(iv) (v) (vi)

(vii) (i) F 1, 2 190.13, p < 0.05 Slopes 2.8 Location Overall = 2.240 0.2

2.6 Buffer inside = 2.488 0.2 Inside outside = 1.991 0.2 2.4 (ii) – (viii) = p > 0.05 (See appendix L2 for details) 2.2

N V

S 2.0 D

Figure 4.3 LandW cover change detection in Tarangire National park using class area metric [Abbreviations: B.land = bare land, C. 1.8 shrubs = closed shrubland, O. shrubs = open shrubland and W. savannah = woody savannah]. The slopes are obtained from untransformed1.6 data because the GLM results of transformed and non-transformed were similar. The bolded values were significant at 95% confidence intervals. 1.4

1.2 1990 1995 2000 2005 2010 Time 160

(i) (ii) (iii)

(vi) (v) (iv)

Time 2.8 (viii) (vii) Location 2.6 Buffer Inside 2.4 (iv) F 1, 2 69.39, p < 0.05 2.2 Overall = -7.214 0.9

N V

S 2.0 inside = -12.175 1.2 D W outside = -2.254 1.2 Time1.8 Time (i – iii, v – viii) = p > 0.05 (See appendix L3 for details) 1.6 Figure 4.4 Land cover change detection1.4 in Katavi National park using class area metric [Abbreviations: B.land = bare land, C. shrubs = closed shrubland, O. shrubs = open shrubland1.2 and W. savannah = woody savannah]. The slopes are obtained from untransformed data because the GLM results of transformed1990 and non1995-transformed2000 were2005 similar2010. The bolded values were significant at 95% Time confidence intervals. 161

(i) (ii) (iii)

(iv) (v) (vi)

Time (vii) (ii) F 1, 2 120.54, p < 0.05 Overall = -2196.687 200.1 (viii) inside = -3161.275 283.0 outside = -1232.100 283.0

(iii) F 1, 2 6237.97, p =0.000 Overall = -1701.225 21.5 inside = -2491.817 30.5 2.8 outside = -910.633 30.5 2.8 Location LocationTime 2.6 Time Buffer (i v – viii) = p > 0.05 2.6 Buffer Inside (See appendix L2 for details) 2.4 Inside 2.4 Figure 4.5 Land cover change detection2.2 in Tarangire National park using number of patches [Abbreviations: B.land = bare land, C.

2.2 N V

shrubs = closed N shrubland, O. shrubs = open shrubland and W. savannah = woody savannah]. The slopes are obtained from

S 2.0

V D

S 2.0 W untransformed dataD because the GLM results of transformed and non-transformed were similar. The bolded values were significant at

W 1.8 95% confidence inte1.8 rvals. 1.6 1.6 1.4 1.4 1.2 162 1.2 1990 1995 2000 2005 2010 1990 1995 2000 2005 Time2010 Time

(ii) (iii) (i)

(iv) (v) (vi)

(vii) Time Time

2.8 Location (viii) F 1, 2 98.9, p < 0.05 2.6 Buffer Overall = 0.586 0.1 Inside inside = 0.292 0.08 2.4 outside = 0.879 0.1 (i – vii) = p > 0.05 2.2

N (See appendix L2 for details) V

S 2.0 D

W Time 1.8

Figure 4.6 Land cover1.6 change detection in Tarangire National park using Euclidean nearest neighbor mean distance (ENN_MN) metric [Abbreviations: B.land = bare land, C. shrubs = closed shrubland, O. shrubs = open shrubland and W. savannah = woody savannah]. The slopes1.4 are obtained from untransformed data because the GLM results of transformed and non-transformed were similar. The bolded1.2 values were significant at 95% confidence intervals. 1990 1995 2000 2005 2010 Time 163

(i) (ii) (iii)

(iv) (v) (vi)

(vii) (viii) (ii) F 1,2 635.86 , p < 0.05 Overall = 0.403 0.1 inside = 0.395 0.1 outside = 0.410 0.1 (iii) F 1, 2 111.74, p < 0.05 Overall = 0.302 0.03 inside = 0.288 0.04 outside = 0.316 0.04 (iv) F 1,2 25.37, p < 0.05 2.8 Overall = 1.693 0.34 Location Time inside = 2.043 0.5 2.8 Time 2.6 Location Buffer outside = 1.343 0.5 Inside 2.6 Buffer (i v – viii) = p > 0.05. (See appendix L3) 2.4 Inside 2.4 2.2

FigureN 4.7 Land cover change detection in Katavi National park using Euclidean nearest neighbor mean distance (ENN_MN) metric

2.2 V

S 2.0 N

[Abbreviations:D B.land = bare land, C. shrubs = closed shrubland, O. shrubs = open shrubland and W. savannah = woody savannah].

V W S 2.0

D The1.8 slopes are obtained from untransformed data because the GLM results of transformed and non-transformed were similar. The W 1.8 bolded1.6 values were significant at 95% confidence intervals. 1.6 1.4 1.4 1.2 164 1.2 1990 1995 2000 2005 2010 Time 1990 1995 2000 2005 2010 Time

(i) (ii) (iii)

F1,2 =18.92, p = 0.049 F1,2 =43.94, p = 0.022 Slopes Slopes Overall = 0.001 0.000 F1,2 =18.89, p = 0.049 Overall = 0.007 0.001 inside = -0.001 0.000 Slopes inside = 0.006 0.002 outside = 0.003 0.000 Overall = -0.010 0.003 outside = 0.009 0.001 inside = -0.013 0.003 outside = -0.008 0.004

0.8 Location Buffer Inside 0.7 2

] Figure 4.8 Effect of land covers change in terms of class area (km ) on population densities of large herbivores’ species in Tarangire (i

t i

s 0.6 – ii) and Katavi (iii) National Parks. The figures displayed are only for relationships that were significant. For details see appendix B n

e 1-6. The slopes are obtained from untransformed data because the GLM results of transformed and non-transformed were very similar.

n 0.5 The bolded values were significant at 95% confidence intervals.

p e

l 0.4

[

0 1

g 0.3

o L 0.2 165

0.1

6 7 8 9 0 1 2 3 4 5

......

1 1 1 1 2 2 2 2 2 2 Log 10 [Savannah CA (kmsq)]

(i) (ii) (iii)

F1,2 =42.45, p = 0.023 F1,2 =27.86, p = 0.034 F1,2 =25.73, p = 0.037 Slopes Slopes Slopes Overall = (6.7 1.0) x 10-4 Overall = (6.1 1.2) x 10-4 Overall =( 4.2 0.8) x 10-4 inside = (9.6 0.95) x 10-4 inside = (5.7 0.5) x 10-4 inside = (6.5 1.1) x 10-4 outside = (3.8 1.8) x 10-4 outside = (6.5 2.0) x 10-4 outside = (1.9 1.3) x 10-4

(iv) (v)

0.8 Location Buffer Inside 0.7

]

t i

s 0.6 F =19.86,n p = 0.047 (v) F1,2 =23.57, p = 0.040 1,2 e

Slopes Slopes D -4 -5 Overall = (1.2t 2.3) x 10 Overall = (3.7 0.8) x 10 inside = (1.9n 0.5 1.0) x 10-5

a inside = (1.8 0.7) x 10-5

outside = h (2.3 0.5) x 10-4 outside = (5.5 1.0) x 10-5

p e

l 0.4

[

g 0.3 o Figure 4.9 Effect of land covers changeL in terms of number of patches on population densities of species of large herbivores in Tarangire National Parks. The table displays0.2 statistical results only for relationships that were significant. The figures displayed are only for relationships that were significant. For details see appendix B 1-6. The slopes are obtained from untransformed data because

the GLM results of transformed and 0.1non-transformed were very similar. The bolded values were significant at 95% confidence

6 7 8 9 0 1 2 3 4 5

......

1 1 1 1 2 2 2 2 2 intervals. 2 Log 10 [Savannah CA (kmsq)] 166

(i) (ii)

F1,2 =14.71, p = 0.050 Overall = 0.010 0.003 inside = -0.020 0.004 outside = 0.038 0.005

F1,2 =20.61, p = 0.045 Slopes Overall = 0.06 0.01 inside = 0.04 0.02 outside = 0.07 0.02

(iv) (v) F =21.08, p = 0.044 0.8 1,2 Location Overall = -7.09 0.56 inside = -13.62 1.15 Buffer outside = -0.56 0.21 Inside

0.7 ]

F1,2 =50.98, p = 0.019 y

Overall = 0.04 0.02 t i

inside = -0.23 0.03 s 0.6 n

outside = 0.3 0.02 e

n 0.5

a h

p e

l 0.4

[

0 1

g 0.3 o

L 0.2

Figure 4.10 Effect of land covers change in terms of mean0.1 distance between nearest neighbor patches on population densities of large

6 7 8 9 0 1 2 3 4 5

......

1 1 1 1 2 2 2 2 2 herbivores’ species or groups, in Katavi and Tarangire National Parks. The table displays statistical2 results only for relationships that were significant. For details see appendix B 1- 6. The slopes are obtaineLog 10 d[Savannah from untransformed CA (kmsq)] data because the GLM results of transformed and non-transformed were very similar. The bolded values were significant at 95% confidence intervals.

167

Figure 4.11 Bar chart summarizes average densities of elephant from 1991 to 2012 recorded inside the Katavi (a) and Tarangire (b) National Parks. The dashed line shows the recommended density of 0.5 (indiv/km2) by Cumming et al. (1997) to avoid habitat degradation.

168

(i) (ii) (iii)

F1,2 =52.48, p = 0.019 F1,2 =24.06, p = 0.039 F1,2 =18.96, p = 0.049 Slopes Slopes Slopes Overall = 68.177 10.577 Overall = 90.536 21.339 Overall = -0.020 0.005 inside = 92.180 14.930 inside = 150.359 30.178 inside = -0.007 0.008 outside = 44.173 14.930 outside = 30.712 30.178 outside = -0.033 0.008

0.8 Location Buffer Inside

0.7

]

t i

s 0.6 Figure 4.12 Influence of rainfall on number of patches of open shrubland (i) savanna (ii) and mean distance between patches of open n

e shrubland (iii) in Tarangire National park. The slopes are obtained from untransformed data because the GLM results of transformed

n 0.5 and non-transformed were very similar. The bolded values were significant at 95% confidence intervals.

p e

l 0.4

E [

169

0 1 g 0.3

o L 0.2

0.1

6 7 8 9 0 1 2 3 4 5

......

1 1 1 1 2 2 2 2 2 2 Log 10 [Savannah CA (kmsq)]

(i) (ii) (iii)

F1,2 =1.82, p = 0.310

F1,2 =28.28, p = 0.034 F1,2 =0.92, p = 0.439 Slopes Overall = 0.048 0.020 inside = 0.076 0.028 outside =0.019 0.028

Figure 4.13 Influence of rainfall on densities of zebra (i), giraffe (ii) and elephant (iii) in Tarangire National Park. The slopes are obtained from untransformed data because the GLM results of transformed and non-transformed were very similar. The bolded values were significant at 95% confidence intervals.

170

Appendices

Appendix L 2(a): Changes in class area (km2), number of patches, mean patch size (MPS) (km2) and mean nearest neighbor distance (m) for various types of land covers over 21 years (1988 to 2009) in Tarangire National Park [Sample size for individual class are 6. Df Error = 2]. All significant changes at P ≤ 0.05 are highlighted in bold.

Land cover Class Area (km2) No. of Patches MPS (km2) MNN (m) type Source DF F-value P-value F-value P-value F-value P -value F -value P -value Time 1 190.130 0.005 15.620 0.059 0.030 0.872 2.650 0.245 Barren land Location 1 2.310 0.268 0.470 0.565 0.010 0.922 0.000 0.990 Time*Location 1 2.340 0.266 0.470 0.563 0.010 0.921 0.000 0.997 Closed Time 1 0.250 0.668 120.540 0.008 4.580 0.166 3.110 0.220 shrubland Location 1 1.000 0.422 23.690 0.040 0.700 0.491 0.080 0.809 Time*Location 1 0.980 0.426 23.240 0.040 0.710 0.489 0.070 0.810 Time 1 0.080 0.804 6237.970 0.000 1.210 0.385 0.650 0.504 Grassland Location 1 0.120 0.759 1386.520 0.001 0.370 0.604 0.200 0.695 Time*Location 1 0.130 0.754 1347.180 0.001 0.370 0.604 0.210 0.693 Open Time 1 0.000 0.951 0.560 0.531 0.180 0.710 1.100 0.405 shrubland Location 1 0.020 0.909 0.000 0.977 0.000 0.954 0.760 0.475 Time*Location 1 0.020 0.906 0.000 0.971 0.000 0.955 0.760 0.475 Time 1 0.000 0.988 0.050 0.845 9.920 0.088 0.280 0.649 Savannah Location 1 0.000 0.960 0.030 0.869 1.580 0.336 0.520 0.545 Time*Location 1 0.000 0.962 0.030 0.871 1.560 0.338 0.520 0.545 Time 1 7.710 0.109 1.700 0.322 6.680 0.123 0.050 0.848 Swamp Location 1 1.780 0.314 0.090 0.797 0.030 0.876 0.020 0.907 Time*Location 1 1.840 0.308 0.080 0.807 0.030 0.874 0.020 0.905 Time 1 1.770 0.315 0.710 0.488 0.010 0.927 98.900 0.010 Woody Location 1 5.860 0.137 0.000 0.951 0.010 0.925 24.880 0.038 savannah Time*Location 1 5.740 0.139 0.000 0.956 0.010 0.928 24.850 0.038

171

Appendix L 2(b): Estimates of changes (Mean ± error) for landscape metrics (class area, number of patches, mean patch size and standard deviation, and nearest neighbor distance) over 21 years (1988 to 2009) in Tarangire National Park, the bolded values identify trends that were significant at 95% confidence intervals.

Mean Nearest Class Area Neighbor (MNN) Land cover type Slope (km2) Number of Patches Mean patch size (m2) (m) Overall 2.240 ± 0.2 725.498 ± 183.6 16.25 ± 89.3 -0.399 ± 0.2 Barren land Inside 2.488 ± 0.2 851.551 ± 259.6 6.29 ± 126.3 -0.401 ± 0.4 Outside 1.991 ± 0.2 599.446 ± 259.6 26.21 ± 126.3 -0.398 ± 0.4 Overall -2.786 ± 5.6 -2196.687 ± 200.1 889.19 ± 415.6 0.438 ± 0.3 Closed shrubland Inside -8.343 ± 7.9 -3161.275 ± 283.0 540.16 ± 587.8 0.506 ± 0.4 Outside 2.772 ± 7.9 -1232.100 ± 283.0 1238.23 ± 587.8 0.370 ± 0.4 Overall -3.691 ± 13.0 -1701.225 ± 21.5 306.08 ± 277.7 0.167 ± 0.2 Grassland Inside 0.992 ± 18.4 -2491.817 ± 30.5 475.26 ± 392.8 0.073 ± 0.3 Outside -8.374 ± 18.4 -910.633 ± 30.5 136.90 ± 392.8 0.261 ± 0.3 Overall -0.665 ± 9.7 -669.284 ± 891.1 82.61 ± 192.8 0.259 ± 0.3 Open shrubland Inside 0.631 ± 13.7 -632.384 ± 1260.1 94.92 ± 272.6 0.043 ± 0.4 Outside -1.961 ± 13.7 -706.184 ± 1260.1 70.31 ± 272.6 0.474 ± 0.4 Overall -0.124 ± 7.1 -334.708 ± 1509.3 158.68 ± 50.4 -0.361 ± 0.7 Savannah Inside -0.510 ± 10.1 -611.639 ± 2134.5 221.55 ± 71.2 0.131 ± 1.0 Outside 0.263 ± 10.1 -57.778 ± 2134.5 95.81 ± 71.2 -0.854 ± 1.0 Overall 8.205 ± 3.0 -682.486 ± 523.4 458.57 ± 177.4 -0.060 ± 0.3 Swamp Inside 12.212 ± 4.2 -828.464 ± 740.2 490.34 ± 251.0 -0.098 ± 0.4 Outside 4.199 ± 4.2 -536.509 ± 740.2 426.79 ± 251.0 -0.022 ± 0.4 Overall -4.540 ± 3.4 -1208.425 ± 1432.4 -4443.09 ± 42665.3 0.586 ± 0.1 Inside -12.726 ± 4.8 -1297.346 ± 2025.7 -68.74 ± 60337.9 0.292 ± 0.08 Woody savannah Outside 3.645 ± 4.8 -1119.505 ± 2025.7 -8817.45 ± 60337.9 0.879 ± 0.1

172

Appendix L 3(a): Changes in class area (km2), number of patches, mean patch size (MPS) (km2) and mean nearest neighbor distance (m) for various types of land covers over 27 years (1984 to 2011) in Katavi National Park [Sample size for individual class are 6. Df Error = 2]. All significant changes highlighted in bold.

Land cover Class Area (km2) No. of Patches MPS (km2) MNN (m) type Source DF F-value P-value F-value P-value F-value P -value F -value P -value Time 1 1.460 0.351 1.560 0.338 5.480 0.144 4.500 0.168 Barren land Location 1 0.24 0.672 0.240 0.673 1.020 0.418 0.040 0.862 Time*Location 1 0.23 0.676 0.230 0.677 1.030 0.417 0.040 0.861 Closed Time 1 1.55 0.340 5.730 0.139 0.050 0.848 35.860 0.027 shrubland Location 1 0.22 0.683 1.340 0.367 0.010 0.917 0.010 0.918 Time*Location 1 0.21 0.689 1.320 0.369 0.010 0.916 0.010 0.926 Time 1 2.47 0.257 5.610 0.141 0.150 0.732 111.740 0.009 Grassland Location 1 0.34 0.619 0.310 0.632 0.000 0.989 0.230 0.677 Time*Location 1 0.33 0.624 0.300 0.640 0.000 0.990 0.250 0.669 Open Time 1 69.39 0.014 13.690 0.066 0.130 0.751 25.370 0.037 shrubland Location 1 33.14 0.029 5.020 0.154 0.350 0.616 1.080 0.408 Time*Location 1 32.81 0.029 4.980 0.155 0.350 0.615 1.080 0.407 Time 1 6.87 0.120 0.000 0.966 4.370 0.172 0.230 0.679 Savannah Location 1 1.26 0.379 0.020 0.896 0.010 0.925 0.010 0.947 Time*Location 1 1.30 0.372 0.020 0.901 0.010 0.922 0.010 0.950 Time 1 1.00 0.423 0.060 0.830 0.030 0.876 0.010 0.931 Swamp Location 1 0.02 0.893 0.000 0.986 0.010 0.948 0.010 0.943 Time*Location 1 0.02 0.898 0.000 0.984 0.010 0.945 0.010 0.941 Time 1 0.00 0.975 0.000 0.989 0.280 0.650 1.850 0.307 Location 1 0.00 0.971 0.000 0.976 0.380 0.601 0.020 0.910 Water Time*Location 1 0.00 0.970 0.000 0.976 0.380 0.602 0.020 0.911 Time 1 2.33 0.266 3.240 0.214 9.200 0.094 0.050 0.849 Woody Location 1 0.45 0.573 0.180 0.716 0.040 0.854 0.200 0.699 savannah Time*Location 1 0.46 0.569 0.160 0.731 0.050 0.850 0.190 0.702

173

Appendix L 3(b): Estimates (Mean ± error) of changes in class area, number of patches, mean patch size and nearest neighbor distance) over 27 years (1984 to 2011) in Katavi National Park, the bolded values were significant at 95% confidence intervals.

Mean Nearest Land cover type Slope Class Area (km2) Number of Patches Mean patch size (m2) Neighbor (m) Overall 3.603 ± 3.0 -1004.373 ± 804.7 403.470 ± 172.4 0.666 ± 0.3 Barren land Inside 2.159 ± 4.2 -1392.290 ± 1139.0 228.290 ± 243.8 0.604 ± 0.4 Outside 5.047 ± 4.2 -616.456 ± 1138.0 578.650 ± 243.8 0.728 ± 0.4 Overall -7.578 ± 6.1 -4280.035 ± 1787.7 41.120 ± 189.4 0.403 ± 0.1 Closed shrubland Inside -10.403 ± 8.6 -6336.739 ± 2528.2 63.660 ± 267.9 0.395 ± 0.1 Outside -4.753 ± 8.6 -2223.331 ± 2528.2 18.580 ± 267.9 0.410 ± 0.1 Overall -26.010 ± 16.6 -2408.489 ± 1016.4 -140.160 ± 356.2 0.302 ± 0.03 Grassland Inside -35.495 ± 23.4 -2963.009 ± 1437.4 -145.360 ± 503.8 0.288 ± 0.04 Outside -16.524 ± 23.4 -1853.970 ± 1437.4 -134.970 ± 503.8 0.316 ± 0.04 Overall -7.214 ± 0.9 -4526.687 ± 1223.4 12.300 ± 33.8 1.693 ± 0.34 Open shrubland Inside -12.175 ± 1.2 -7256.671 ± 1730.2 -7.650 ± 47.8 2.043 ± 0.5 Outside -2.254 ± 1.2 -1796.702 ± 1730.2 32.240 ± 47.8 1.343 ± 0.5 Overall 23.141 ± 8.8 -143.057 ± 2995.1 215.850 ± 103.3 -0.083 ± 0.2 Savannah Inside 33.217 ± 12.5 -565.112 ± 4235.7 227.320 ± 146.1 -0.071 ± 0.2 Outside 13.065 ± 12.5 278.998 ± 4235.7 204.370 ± 146.1 -0.096 ± 0.2 Overall -2.792 ± 2.8 -324.571 ± 1331.8 -37.430 ± 211.9 0.133 ± 1.4 Swamp Inside -3.198 ± 3.9 -293.737 ± 1883.4 -21.040 ± 299.6 0.020 ± 1.9 Outside -2.386 ± 3.9 -355.405 ± 1883.4 -53.830 ± 299.6 0.247 ± 1.9 Overall 0.077 ± 2.2 -4.643 ± 284.3 -20.800 ± 39.4 4.713 ± 3.5 Inside 0.170 ± 3.1 5.073 ± 402.041 3.390 ± 55.7 4.276 ± 4.9 Water Outside -0.016 ± 3.1 -14.359 ± 402.0 -44.990 ± 55.7 5.150 ± 4.9 Overall 20.670 ± 13.5 -1621.037 ± 900.9 266.120 ± 87.8 0.030 ± 0.1 Inside 29.830 ± 19.2 -1977.204 ± 1274.1 247.270 ± 124.0 -0.031 ± 0.2 Woody savannah Outside 11.510 ± 19.2 -1264.870 ± 1274.1 284.970 ± 124.0 0.091 ±0.2

174

Appendix L 4: Relationship between landscape metrics vs. rainfall in Tarangire National Parks [For all parameters Df= 1, DF error = 2]. All significant changes at P ≤ 0.05 are highlighted in bold.

Closed Woody Bare land shrubs Grassland Open shrubs Savannah Swamps savannah Class Area (km2) F P F P F P F P F P F P F P Rain 0.22 0.687 2.16 0.279 1.17 0.393 10.99 0.080 6.04 0.133 7.07 0.117 0.07 0.821 Location 0.01 0.948 2.99 0.226 0.38 0.599 3.61 0.198 1.28 0.375 1.69 0.324 1.96 0.297 Rain*Location 0.00 0.955 3.53 0.201 0.25 0.669 3.05 0.223 1.38 0.361 2.36 0.265 2.67 0.244 No. patches Rain 0.88 0.446 1.60 0.333 1.36 0.364 24.06 0.039 52.48 0.019 0.43 0.578 13.85 0.065 Location 0.02 0.896 0.03 0.875 0.01 0.927 1.02 0.419 7.80 0.108 0.59 0.524 6.62 0.124 Rain*Location 0.02 0.904 0.07 0.815 0.00 0.967 0.44 0.577 8.84 0.097 0.87 0.450 5.83 0.137 Mean Euclidean Distance (m) Rain 0.660 0.503 0.120 0.760 0.470 0.564 40.960 0.024 0.960 0.431 0.320 0.627 0.700 0.490 Location 0.880 0.446 0.740 0.480 0.090 0.794 19.070 0.049 1.360 0.363 2.800 0.236 0.050 0.848 Rain*Location 1.080 0.408 0.760 0.475 0.060 0.826 18.960 0.049 1.370 0.362 3.060 0.222 0.050 0.842

175

Appendix M 1: Relationship between land covers (in terms of class area) and population densities of elephant, giraffe, and zebra in Katavi and Tarangire National Parks. All significant changes at P ≤ 0.05 are highlighted in bold.

Katavi National Park Tarangire National Park Elephant Giraffe Zebra Elephant Giraffe Zebra Land cover type F P F P F P F P F P F P Barren land 1 0.64 0.508 0.00 0.981 0.86 0.452 0.08 0.803 0.94 0.435 0.28 0.650 Location 1 0.08 0.805 11.27 0.078 1.25 0.380 0.18 0.712 0.29 0.644 0.34 0.621 Barren land *Location 1 0.06 0.828 11.18 0.079 1.59 0.334 0.00 0.953 0.21 0.693 0.18 0.714 Closed shrubs 1 0.01 0.944 2.68 0.243 0.45 0.572 14.09 0.064 0.14 0.744 0.06 0.836 Location 1 0.11 0.775 0.00 0.957 0.57 0.530 0.78 0.471 0.00 0.969 0.45 0.572 Closed shrubs *Location 1 0.06 0.826 0.01 0.932 0.43 0.580 1.75 0.317 0.01 0.947 0.49 0.558 Grass 1 0.10 0.785 4.17 0.178 0.19 0.707 1.32 0.370 0.00 0.984 2.05 0.289 Location 1 0.03 0.881 0.00 0.983 0.37 0.606 0.11 0.774 18.92 0.049 1.05 0.414 Grass*Location 1 0.00 0.964 0.05 0.838 0.23 0.678 0.27 0.656 19.61 0.047 0.79 0.468 Open shrubs 1 0.40 0.592 3.25 0.213 0.08 0.801 43.94 0.022 1.01 0.420 3.52 0.202 Location 1 0.03 0.881 1.49 0.346 0.82 0.460 1.93 0.299 2.28 0.270 1.54 0.340 Open shrubs*Location 1 0.12 0.765 1.97 0.296 0.45 0.573 1.43 0.354 2.23 0.274 1.50 0.345 Savanna 1 0.35 0.616 3.12 0.219 0.00 0.964 15.65 0.058 0.45 0.572 3.28 0.212 Location 1 0.00 0.955 0.13 0.750 0.17 0.721 13.14 0.068 5.76 0.139 0.00 0.971 Savanna*Location 1 0.00 0.961 0.25 0.669 0.30 0.641 10.49 0.084 5.26 0.149 0.00 0.989 Swamps 1 2.08 0.286 0.53 0.541 18.89 0.049 2.02 0.291 0.04 0.855 1.41 0.357 Location 1 0.00 0.964 0.59 0.524 8.06 0.105 3.28 0.212 0.26 0.662 0.40 0.591 Swamps*Location 1 0.17 0.720 0.71 0.489 3.28 0.212 5.98 0.134 0.30 0.638 0.25 0.667 Water 1 0.23 0.678 2.50 0.255 4.92 0.157 Location 1 0.82 0.461 0.65 0.505 3.94 0.186 Water*Location 1 0.13 0.755 0.13 0.757 0.11 0.770 Woody savanna 1 0.03 0.882 4.44 0.170 0.56 0.534 0.08 0.801 0.42 0.583 2.85 0.233 Location 1 0.00 0.989 0.46 0.567 0.31 0.635 0.01 0.935 1.90 0.302 5.74 0.139 Woody savanna*Location 1 0.00 0.965 0.51 0.550 0.42 0.582 0.10 0.779 1.87 0.305 5.83 0.137

176

Appendix M 2: Relationship between land covers (in terms of number of patches) and population densities of elephant, giraffe and zebra in Katavi National Park and Tarangire National Parks. The bolded values were significant at α ≤ 0.05.

Katavi National Park Tarangire National Park Land cover type (number of Elephant Giraffe Zebra Elephant Giraffe Zebra patches) F P F P F P F P F P F P Barren land 1 0.13 0.752 2.58 0.249 0.51 0.548 0.12 0.764 1.28 0.375 0.66 0.503 Location 1 1.26 0.379 0.00 0.966 0.57 0.529 0.58 0.524 0.29 0.647 1.07 0.409 Barren land *Location 1 0.60 0.519 0.61 0.517 0.18 0.711 0.13 0.749 0.23 0.677 0.20 0.698 Closed shrubs 1 1.94 0.299 1.99 0.294 0.69 0.494 0.01 0.924 2.40 0.262 0.08 0.806 Location 1 0.26 0.659 1.24 0.381 2.00 0.293 0.57 0.529 0.25 0.664 0.00 0.975 Closed shrubs *Location 1 0.27 0.657 2.90 0.231 0.03 0.885 0.00 0.980 1.61 0.333 0.03 0.882 Grass 1 1.60 0.333 1.84 0.308 1.06 0.411 0.00 0.963 1.57 0.337 0.07 0.818 Location 1 0.74 0.481 1.67 0.326 3.37 0.208 0.49 0.557 0.33 0.623 0.00 0.978 Grass*Location 1 0.01 0.935 3.81 0.190 0.37 0.603 0.05 0.846 1.17 0.392 0.00 0.959 Open shrubs 1 1.00 0.422 2.92 0.230 0.18 0.715 2.24 0.273 0.56 0.531 42.45 0.023 Location 1 0.64 0.507 0.84 0.457 2.08 0.286 0.99 0.425 0.43 0.578 21.76 0.043 Open shrubs*Location 1 0.14 0.743 3.33 0.210 0.02 0.900 0.82 0.460 0.03 0.875 7.58 0.111 Savanna 1 2.26 0.272 0.32 0.631 7.95 0.106 27.82 0.034 0.37 0.606 27.86 0.034 Location 1 0.11 0.774 0.06 0.834 2.75 0.239 35.76 0.027 1.97 0.295 0.03 0.880 Savanna*Location 1 1.52 0.343 0.00 0.956 0.98 0.426 19.86 0.047 1.20 0.388 0.12 0.760 Swamps 1 3.24 0.214 0.04 0.859 8.93 0.096 0.27 0.653 23.57 0.040 1.25 0.379 Location 1 0.00 0.990 0.22 0.687 6.36 0.128 0.04 0.858 0.00 0.981 3.99 0.184 Swamps*Location 1 1.65 0.328 0.61 0.515 0.04 0.866 0.66 0.502 5.93 0.135 2.50 0.255 Water 1 2.05 0.289 0.48 0.560 7.89 0.107 Location 1 2.52 0.253 0.81 0.464 5.17 0.151 Water * Location 1 1.08 0.408 0.26 0.659 0.00 0.973 Woody savanna 1 2.08 0.286 1.36 0.363 1.83 0.309 4.60 0.165 0.25 0.666 25.73 0.037 Location 1 0.22 0.685 2.38 0.263 2.63 0.246 2.29 0.269 0.97 0.428 9.90 0.088 Woody savanna*Location 1 0.10 0.784 3.51 0.202 0.31 0.635 1.58 0.336 0.47 0.562 7.58 0.110

177

Appendix M 3: Relationship between land covers (in terms of mean nearest neighbor) and population densities of elephant, giraffe and zebra in Katavi National Park and Tarangire National Parks. The bolded values were significant at α ≤ 0.05.

Land cover type Katavi National Park Tarangire National Park Elephant Giraffe Zebra Elephant Giraffe Zebra (mean distance between patches) DF F P F P F P F P F P F P Barren land 1 0.26 0.663 7.79 0.108 0.15 0.732 0.18 0.712 20.61 0.045 3.53 0.201 Location 1 0.05 0.843 0.44 0.575 0.08 0.799 0.32 0.630 0.02 0.913 5.39 0.146 Barren land *Location 1 0.04 0.860 0.40 0.593 0.11 0.775 0.35 0.615 0.00 0.961 5.35 0.147 Closed shrubs 1 2.11 0.283 0.99 0.425 0.50 0.552 0.57 0.528 4.41 0.171 0.92 0.439 Location 1 0.40 0.591 1.89 0.303 0.52 0.547 0.46 0.567 0.05 0.850 2.05 0.288 Closed shrubs *Location 1 0.38 0.600 1.87 0.305 0.56 0.533 0.41 0.586 0.04 0.862 2.02 0.291 Grass 1 1.10 0.405 2.14 0.281 0.09 0.796 1.00 0.423 0.01 0.942 0.91 0.442 Location 1 0.15 0.738 0.98 0.426 0.32 0.630 0.08 0.801 6.88 0.120 0.39 0.597 Grass*Location 1 0.13 0.750 0.95 0.432 0.34 0.617 0.07 0.821 6.81 0.121 0.40 0.590 Open shrubs 1 0.98 0.427 3.27 0.212 0.00 0.964 0.77 0.472 2.21 0.275 21.08 0.044 Location 1 0.32 0.629 1.66 0.327 0.18 0.710 0.01 0.940 2.26 0.271 15.10 0.060 Open shrubs*Location 1 0.26 0.659 1.57 0.337 0.26 0.663 0.00 0.950 2.27 0.271 15.03 0.061 Savanna 1 0.15 0.738 2.86 0.233 3.04 0.223 0.25 0.669 0.39 0.598 0.95 0.432 Location 1 0.14 0.740 0.01 0.942 0.57 0.530 0.00 0.951 14.45 0.063 0.98 0.426 Savanna*Location 1 0.14 0.748 0.01 0.943 0.52 0.547 0.01 0.919 14.71 0.050 0.95 0.434 Swamps 1 0.89 0.445 1.07 0.409 6.55 0.125 9.60 0.090 1.98 0.294 0.17 0.721 Location 1 0.48 0.561 0.94 0.434 0.08 0.802 53.09 0.018 0.11 0.768 1.45 0.351 Swamps*Location 1 0.40 0.594 0.90 0.442 0.18 0.711 50.98 0.019 0.12 0.758 1.44 0.353 Water 1 0.00 0.978 5.11 0.152 0.92 0.439 Location 1 0.00 0.971 0.00 0.969 0.04 0.857 Water*Location 1 0.01 0.914 0.00 0.994 0.15 0.740 Woody savanna 1 6.19 0.131 0.28 0.648 2.91 0.230 0.00 0.986 1.89 0.303 3.11 0.220 Location 1 5.31 0.148 1.38 0.361 0.08 0.808 0.01 0.950 0.47 0.564 2.42 0.260 Woody savanna*Location 1 5.27 0.149 1.36 0.364 0.07 0.817 0.01 0.936 0.48 0.562 2.39 0.262

178

Appendix M 4: Estimates of change on mean densities of large herbivores in KNP and TNP, as influenced by change in land cover in terms of class area (km2) over time. The table displays the proportions of change only for relationships between densities vs. class areas that appeared significant in appendix B1.

Park Large herbivore’s Class area of: Overall change Change inside the Change on buffer species/group (Mean ± Error) park (Mean ± zone (Mean ± Error) Error)

KNP Zebra Swamps -0.010 ± 0.003 -0.013 ± 0.003 -0.008 ± 0.004 TNP Elephant Open shrubland 0.007 ± 0.001 0.006 ± 0.002 0.009 ± 0.001 Giraffe Grassland 0.001 ± 0.000 -0.001 ± 0.000 0.003 ± 0.000

Note: The bolded values were significant at 95% confidence intervals.

179

Appendix M 5: Estimates of change on densities of large herbivores in TNP, as influenced by change in land cover in terms of number of land cover patches over time. The table displays the proportions of change only for relationships between densities vs. land cover patches that appeared significant in Appendix J2.

Park Large herbivore’s Number of patches Overall change Change inside the park Change on buffer species/group for: (Mean ± Error) (Mean ± Error) zone (Mean ± Error)

TNP Elephant Savannah 0.000122 ± 0.000023 0.000019 ± 0.000010 0.000225 ± 0.00005 Giraffe Swamp 0.000037 ± 0.000008 0.000018 ± 0.000007 0.000055 ± 0.00001 Zebra Open shrubland 0.00067 ± 0.000103 0.000955± 0.0000946 0.000388± 0.000183

Savannah 0.000606± 0.000115 0.000567 ± 0.000048 0.00065± 0.0002 Woody savannah 0.00042 ± 0.000083 0.00065 ± 0.00011 0.00019 ± 0.00013

Note: The bolded values were significant at 95% confidence intervals.

180

Appendix M 6: Estimates of change on densities large herbivores in TNP, as influenced by change in land cover in terms of mean distance (m) between nearest neighbor land cover patches, over time. The table displays the proportions of change only for relationships between densities vs. inter-patch mean distance that appeared significant in appendix B3.

Park Large herbivore’s Nearest distance between Overall change Change inside the Change on buffer species/group neighbor patches (m) of: (Mean ± Error) park (Mean ± zone (Mean ± Error) Error)

TNP Giraffe Bare or less vegetated land 0.0559 ± 0.0146 0.0396 ± 0.0202 0.0723 ± 0.0213 Giraffe Savannah 0.009 ± 0.003 -0.020 ± 0.004 0.038 ± 0.005 Elephant Swamps 0.037 ± 0.016 -0.229 ± 0.027 0.303 ± 0.017 Zebra Open shrubland -7.091 ± 0.585 -13.622 ± 1.149 -0.559 ± 0.214

Note: The bolded values were significant at 95% confidence intervals.

181

Appendix M 7: Relationship between rainfall and population densities of elephant, giraffe and zebra Tarangire National Parks [Df error = 2]. All significant changes at P ≤ 0.05 are highlighted in bold.

Tarangire National Park Elephant Giraffe Zebra Land cover type (number of patches) DF F P F P F P Rain 1 1.820 0.310 0.920 0.439 28.280 0.034 Location 1 1.550 0.339 0.810 0.462 4.350 0.172 Rain*Location 1 1.260 0.379 0.770 0.473 4.800 0.160

182

7.0 References

Andren, H. (1994). Effects of habitat fragmentation on birds and mammals in Landscapes with

different proportions of suitable habitat : A review. Oikos, 71, 355–366.

Asner, G. P., & Levick, S. R. (2012). Landscape-scale effects of herbivores on tree fall in

African savannas. Ecology Letters, 15, 1211–1217.

Banda, T., Schwartz, M. W., & Caro, T. (2006). Woody vegetation structure and composition

along a protection gradient in a miombo ecosystem of western Tanzania. Forest Ecology

and Management, 230, 179–185.

Barnes, R. F. W. (1983). The Elephant Problem in Ruaha National Park, Tanzania. Biological

Conservation, 26, 127–148.

Beale, C. M., Rensberg, S. Van, Bond, W. J., Coughenour, M., Fynn, R., Gaylard, A., …

Sinclair, A. R. E. (2013). Ten lessons for the conservation of African savannah

ecosystems. Biological Conservation, 167, 224–232.

Bender D. J., Thomas A. Contres, and L. F. (1998). Habitat Loss and Population decline: A

meta-analysis of the patch size effect. Ecology, 79, 517–533.

Bender, D. J., Tischendorf, L., & Fahrig, L. (2003). Using patch isolation metrics to predict

animal movement in binary landscapes. Landscape Ecology, 18, 17–40.

Bruner, A.G., Gullison, R.E., Rice, R.E., and Fonseca, G.A. (2001). Effectiveness of parks in

protecting tropical biodiversity. Science, 291, 125–128.

183

Caro, T. M. (1999). Densities of Mammals in Partially Protected Areas: The Katavi Ecosystem

of Western Tanzania. Journal of Applied Ecology, 36, 205–217.

Caro, T. (2008). Decline of large mammals in the Katavi-Rukwa ecosystem of western Tanzania.

African Journal of Zoology, 43, 99–116.

Caro, T. (2011). On the merits and feasibility of wildlife monitoring for conservation: a case

study from Katavi National Park, Tanzania. African Journal of Ecology, 49, 320–331.

Caughley, G. (1974). Bias in Aerial Survey. Wildlife Management, 38, 921–933.

Caughley, G. (1977). Sampling in Aerial Survey. Wildlife Management, 41, 605–615.

Côté, S. D., Rooney, T. P., Tremblay, J.-P., Dussault, C., & Waller, D. M. (2004). Ecological

Impacts of Deer Overabundance. Annual Review of Ecology, Evolution, and Systematics,

35, 113–147.

Cumming, D. H. M., Fenton, M. B., Rautenbach, I. L., Taylor, R. D., Cumming, G. S.,

Cumming, M. S., Dunlop, J. M., Ford, G. S., Hovorka, M. D., Johnston, D. S., Kalcounis,

M. C., Mahlanga, Z. and Portforts, C. V. (1997). Elephants, woodlands and biodiversity

in miombo woodland in southern Africa. South African Journal of Science, 93, 231–236.

De Beer, Y., & van Aarde, R. J. (2008). Do landscape heterogeneity and water distribution

explain aspects of elephant home range in southern Africa’s arid savannas? Journal of

Arid Environments, 72, 2017–2025.

Epaphras, A. M., Gereta, E., Lejora, I., Ole Meing’ataki, G. E., Ng’umbi, G., Kiwango, Y., …

Mtahiko, M. G. G. (2007). Wildlife water utilization and importance of artificial

184

waterholes during dry season at Ruaha National Park, Tanzania. Wetlands Ecology and

Management, 16, 183–188.

Fahrig, L. (1997). Relative effects of habitat loss and fragmentation on population. The Journal

of Wildlife Management, 61, 603–610.

Fahrig, L. (2003). Effects of habitat fragmentation on biodiversity. Annual Review of Ecology,

Evolution, and Systematics, 34, 487–515.

Field, C., & Ross, I. (1976). The savannah ecology of Kidepo valley National Park. African

Journal of Ecology, 14, 1–15.

Fisher, B., and Christopher, T. (2007). Poverty and biodiversity: Measuring the overlap of

human poverty and the biodiversity hotspots. Ecological Economics, 62, 93–101.

Frost, P. (1996). The ecology of miombo woodlands. In: Campbell, B. (Ed), The Miombo on

Transition: Woodlands and Welfare in Africa. Center for International Forestry Research,

Bogor, Indonesia.

Galanti, V., Preatoni, D., Martinoli, A., Wauters, L. A., & Tosi, G. (2006). Space and habitat use

of the African elephant in the Tarangire–Manyara ecosystem, Tanzania: Implications for

conservation. Mammalian Biology - Zeitschrift Für Säugetierkunde, 71, 99–114.

Gaveau, D. L. A., Epting, J., Lyne, O., Linkie, M., Kumara, I., Kanninen, M., and Leader-

Williams, N. (2009). Evaluating whether protected areas reduce tropical deforestation in

Sumatra. Journal of Biogeography, 36, 2165–2175.

185

Gereta, E., Ole Meing’ataki, G. E., Mduma, S., & Wolanski, E. (2004). The role of wetlands in

wildlife migration in the Tarangire ecosystem, Tanzania. Wetlands Ecology and

Management, 12, 285–299.

Grainger, M., Aarde, R., & Whyte, I. (2005). Landscape heterogeneity and the use of space by

elephants in the Kruger National Park, South Africa. African Journal of Ecology, 43,

369–375.

Grange, S., & Duncan, P. (2006). Bottom-up and top-down processes in African ungulate

communities : resources and predation acting on the relative abundance of zebra and

grazing bovids. Ecography, 6, 899- 907.

Grossmann, E. B., & Mladenoff, D. J. (2007). Open woodland and savanna decline in a mixed-

disturbance landscape (1938 to 1998) in the Northwest Wisconsin (USA) Sand Plain.

Landscape Ecology, 22, 43–55.

Gustafson, E. J. (1998). Quantifying landscape spatial pattern: what the state of the art?

Ecosystems, 1, 143–156.

Hobbs, N. T., Galvin, K. A., Stokes, C. J., Lackett, J. M., Ash, A. J., Boone, R. B., … Thornton,

P. K. (2008). Fragmentation of rangelands: Implications for humans, animals, and

landscapes. Global Environmental Change, 18, 776–785.

Hobbs, N., & Gordon, I. (2010). How does landscape heterogeneity shape dynamics of large

herbivore populations? In Owen-Smith, N. (Ed): Dynamics of large herbivores in

changing environments towards appropriate models (p. 141–164). West Sussex, UK:

Willey-Blackwell. 186

Holdo, R. M., Holt, R. D., & Fryxell, J. M. (2009). Grazers, browsers, and fire influence the

extent and spatial pattern of tree cover in the Serengeti. Ecological Applications, 19, 95–

109.

Hopcraft, J. G. C., Sinclair, A. R. E., & Packer, C. (2005). Planning for success: Serengeti lions

seek prey accessibility rather than abundance. Journal of Animal Ecology, 74, 559–566.

Jachman, H. (2002). Comparison of aerial counts with ground counts for large herbivores.

Journal of Applied Ecology, 39, 841–852.

Kalwij, J. M., De Boer, W. F., Mucina, L., Prins, H. H. T., Skarpe, C., & Winterbach, C. (2010).

Tree cover and biomass increase in a southern African savanna despite growing elephant

population. Ecological Applications, 20, 222–233.

Kerley, G. I. H., Landman, M., Kruger M. L. & Owen-Smith, N. (2008). Effects of elephants on

ecosystem biodiversity. In Scholles, R.J. and Mennell, K. G. (Eds.), Elephant

management. A scientific assessment for South Africa. Johannesburg, South Africa. Wits

University Press.

Kie, J. G., Bowyer, R. T., Nicholson, M. C., Boroski, B. B., & Eric, R. (2002). Landscape

heterogeneity at differing scales: Effects on spatial distribution of Mule Deer. Ecology,

83, 530–544.

Kikula, I. S. (1986). The Influence of Fire on the Composition of Miombo Woodland of SW

Tanzania. Oikos, 46, 317–324.

187

Lawton, R. (1978). A study of the Dynamic Ecology of Zambian Vegetation. Journal of

Ecology, 66, 175–198.

Leatherwood, S., Gilbert, J. R., & Chapman, D. G. (1978). An evaluation of some techniques for

aerial censuses of bottle nosed dolphins. Journal of Wildlife Management, 42, 239–250.

Leuthold, W. (1977). Changes in tree populations of Tsavo East Africa National Park, Kenya.

African Journal of Ecology, 15, 61–69.

Loarie, S. R., van Aarde, R. J., & Pimm, S. L. (2009). Elephant seasonal vegetation preferences

across dry and wet savannas. Biological Conservation, 142, 3099–3107.

Manase, E., Gara, J. I., & Wolanski, E. (2010). A review of the water crisis in Tanzania’s

protected areas, with emphasis on the Katuma River-Lake Rukwa ecosystem.

Ecohydrology and Hydrobiology, 10, 153–166.

Mapaure, I., & Moe, S. R. (2009). Changes in the structure and composition of miombo

woodlands mediated by elephants (Loxodonta africana) and fire over a 26-year period in

north-western Zimbabwe. African Journal of Ecology, 47, 175–183.

Mcgarigal, K., & Marks, B. (1995). Spatial pattern analysis program for quantifying landscape

structure. Corvallis: Oregon University State.

McGarigal, K., Cushman, S. A., & Ene, E. 2012. Fragstats v4: Spatial Pattern Analysis Program

for Categorical and Continuous Maps. Computer software program produced by the

authors at the University of Massachusetts, Amherst. Available at the following web site:

http://www.umass.edu/landeco/research/fragstats/fragstats.html.

188

Mfunda, I. M., & Røskaft, E. (2010). Bushmeat hunting in Serengeti, Tanzania : An important

economic activity to local people. International Journal of Biodiversity and

Conservation, 2, 263–272.

Msoffe, F. U., Kifugo, S. C., Said, M. Y., Neselle, M. O., Gardingen, P. van, Reid, R. S., Ogutu,

J. O., Herero, M., and Leeuw, J. De. (2011). Drivers and impacts of land-use change in

the Maasai Steppe of northern Tanzania: an ecological, social and political analysis.

Journal of Land Use Science, 6, 261–281.

Mtui, D. (2014). Evaluating landscape and wildlife changes over time in Tanzania’s protected

areas. PhD Dissertation. University of Hawai‘i at Mānoa, Honolulu.

Mundia, N., & Murayama, Y. (2009). Analysis of land use/cover changes and animal population

dynamics in a wildlife sanctuary in East Africa. Remote Sensing, 1, 952–970.

Mwalyosi, R. (1992). Land-use changes and degradation in south-west Masaai land, Tanzania.

Environmental Conservation, 19, 145–152.

Nagendra, H., Pareeth, S., Sharma, B., Schweik, C. M., & Adhikari, K. R. (2007). Forest

fragmentation and regrowth in an institutional mosaic of community, government and

private ownership in Nepal. Landscape Ecology, 23, 41–54.

Newmark, W. D. (2008). Isolation of African protected areas. Frontiers in Ecology and the

Environment, 6, 321–328.

Norton-Griffiths, M. (1978). Counting animals. (2nd Edition) Handbook no. 1. Techniques in

African wildlife ecology. African Wildlife Foundation, Nairobi.

189

Ogutu, J. O., & Owen-Smith, N. (2003). ENSO, rainfall and temperature influences on extreme

population declines among African savanna ungulates. Ecology Letter, 6, 412–419.

Ogutu, J. O., & Owen-Smith, N. (2005). Oscillations in large mammal populations: are they

related to predation or rainfall? African Journal of Ecology, 43, 332–339.

Osborn, F. V. (2004). Seasonal variation of feeding patterns and food selection by crop-raiding

elephants in Zimbabwe. African Journal of Ecology, 42 322–327.

Ott, T. (2007). Landscape heterogeneity as a determinant of range utilization by African

elephants (Loxodonta africana) in mesic savannas. Pretoria, South Africa: University of

Pretoria.

Ottichilo, W. K., Leeuw, J. De, & Prins, H. H. T. (2001). Population trends of resident

wildebeest [ Connochaetes taurinus hecki (Neumann)] and factors influencing them in

the Masai Mara ecosystem, Kenya. Biological Conservation, 97, 271–282.

Owen-Smith, N. (1988). Megaherbivores: The Influence of very large body size ecology,

Cambridge. UK: Cambridge University Press.

Owen-Smith, N. (2004). Functional heterogeneity in resources within landscapes and herbivore

population dynamics. Landscape Ecology, 19, 761–771.

Owen-Smith, N., Kerley, G. I. H., Page, B., Slotow, R., & van Aarde, R. J. (2006). A scientific

perspective on the management of elephants in the Kruger national park and elsewhere.

South African Journal of Science, 102, 389–395.

190

Owen-Smith, N., Mason, D. R., & Ogutu, J. O. (2005). Correlates of survival rates for 10

African ungulate populations: density, rainfall and predation. Journal of Animal Ecology,

74, 774–788.

Pelkey, N. W., Stoner, C. J., & Caro, T. M. (2003). Assessing habitat protection regimes in

Tanzania using AVHRR NDVI composites : comparisons at different spatial and

temporal scales. International Journal of Remote Sensing, 24, 2533–2558.

Pellew, R. A. (1984). Food Consumption and Energy Budget of the Giraffe. Journal of Applied

Ecology, 21, 141–159.

Pienaar, U. D. V. (1963). The Large Mammals of Kruger National Park: Their distribution and

Present day status. Koedoe, 6, 1–37.

Prins, H. H. T. (1987). Nature conservation as an integral part of optimal land use in East Africa:

The case of the Masai ecosystem of Northern Tanzania. Biological Conservation, 40,

141–161.

Rabe, M. J. Rosenstock, S. S., & deVos, J. C. (2002). Review of Big-Game Survey Methods

Used by Wildlife Agencies of the Western United States. Wildlife Society Bulletin, 30,

46–52.

Riginos, C., & Grace, J. B. (2008). Savanna tree density, herbivores, and the herbaceous

community: bottom-up vs. top-down effects. Ecology, 89, 2228–2238.

Samuel M. (1981). Correction of visibility bias in aerial surveys where animals occurs in groups.

Journal of Wildlife Management, 45, 993–997.

191

Sánchez-Azofeifa, A. G., Daily, G. C., Pfaff, A. S. P., & Busch, C. (2003). Integrity and isolation

of Costa Rica’s national parks and biological reserves: examining the dynamics of land-

cover change. Biological Conservation, 109, 123–135.

Smuts, G. L. (1975). Home range sizes for Burchell’s Zebra Equus burchellii antiquorum from

the Kruger National Park. Koedoe, 18, 139–146.

Steinhorst, R. K. & Samuel M. D. (1989). Sightability adjustment methods for aerial surveys of

wildlife populations. Biometrics, 45, 415–425.

Stoner, C., Caro, T., Mduma, S., Mlingwa, C., Sabuni, G., Borner, M., & Schelten, C. (2007).

Changes in large herbivore populations across large areas of Tanzania. African Journal of

Ecology, 45, 202–215.

TANAPA (Tanzania National Parks). (no date). Tarangire general management plan, Arusha,

Tanzania: Tarangire National Park.

TAWIRI (Tanzania Wildlife Research Institute). (2010a) Aerial Census in the Katavi-Rukwa

ecosystem in dry Season 2009. Internal Report. Njiro, Arusha.

TAWIRI (Tanzania Wildlife Research Institute). (2010b) Aerial Census in the Ruaha-Rungwa

ecosystem in dry Season 2009. Internal Report. Njiro, Arusha.

Teren, G., & Owen-smith, N. (2010). Elephants and riparian woodland changes in the Linyanti.

Pachyderm, 47, 18–25.

Turner, M., Gardner, R., & O'neill, A. (2001). Landscape ecology: In theory and practices.

United States of America: Springer.

192

Valeix, M., Fritz, H., Chamaillé-Jammes, S., Bourgarel, M., & Murindagomo, F. (2008).

Fluctuations in abundance of large herbivore populations: insights into the influence of

dry season rainfall and elephant numbers from long-term data. Animal Conservation, 11,

391–400.

Valeix, M., Fritz, H., Sabatier, R., Murindagomo, F., Cumming, D., & Duncan, P. (2011).

Elephant-induced structural changes in the vegetation and habitat selection by large

herbivores in an African savanna. Biological Conservation, 144, 902–912.

Vijver, C. A. D. M. Van De, Foley, C. A., & Olff, H. (1999). Changes in the woody component

of an East African savanna during 25. Journal of Tropical Ecology, 15, 545–564.

Voeten, M. M., & Prins, H. H. T. (1999). Resource partitioning between sympatric wild and

domestic herbivores in the Tarangire region of Tanzania. Oecologia, 120, 287–294.

Voeten, M. M., Vijver, C. A. D. M. Van De, Olff, H. & Langevelde, F. Van. (2009). Possible

causes of decreasing migratory ungulate populations in an East African savannah after

restrictions in their seasonal movements. African Journal of Ecology, 48, 169–179.

Wilcove, D., McLellan, C., Dobson, A. (1986). Habitat fragmentation in the temperate zone. In

Conservation Biology: The science of scarcity and diversity (p. 584). Sunderland,

Massachusetts: Sinauer Associates, INC. Publishers.

193

Chapter 5. Conclusions and Recommendations

1.0 Introduction The goal of this dissertation was to evaluate changes in landscapes and wildlife populations in the Tanzania’s wildlife protected areas (National Parks and Game

Reserves), from the early 1980s to the early 2010s. Three broad objectives were addressed in the process of reaching this goal which were to: i) investigate the extent to which land cover change has occurred from the 1980s to the 2010s, inside and outside the protected areas, ii) to examine the extent to which wildlife abundance has changed from the 1980s to the 2010s, inside and outside the protected areas, and the direction and rate of change for each species or group, and iii) to determine the influence of land cover spatial components on wildlife species abundance, inside and outside the protected areas.

Specific research questions were developed for each objective and were addressed in chapters 2, 3 and 4, respectively.

2.0 Limitations Before proceeding with a summary of the results, conclusion and recommendations from this dissertation, the shortcomings of this project will be discussed. The major limitation was the small sample sizes in the regions investigated and the mismatch of data used for comparison between variables. Selection of the study sites was limited by availability, accessibility and suitability of data. The study sites selected for this project include Katavi-Rukwa, Ruaha-Rungwa and Tarangire and vary by objective.

Although Landsat satellite images were provided free of charge by USGS, a large number of images were affected by clouds, and even for the cloud-free images, a 194

complete set of images for one site (park) acquired during the same time period was difficult to find, a necessary condition to consider when analyzing remote sensing data

(Jensen 2005, 2007). The resolution (30 m) of the satellite data (Landsat TM and ETM+) used in the analysis did not distinguish savanna from cropland/cultivated land. Therefore where savannah is reported to have increased, it could mean an increase in cultivated land.

Accessibility of animal density data from Tanzania Wildlife Research Institute

(TAWIRI) was also problematic, reducing the sample size. The sample size for some of species examined did not meet requirement for statistical analysis, which meant that some species were grouped by type. Such aggregation may reduce the accuracy of the results because species are known to respond differently to changes on landscapes

(Valeix et al. 2011).

The mismatch in dates for wildlife count data and land cover data was another limitation. The earliest wildlife count data available for Katavi was from 1991, which was related to land cover data from 1984. A gap as long as seven years between the two datasets would likely limit the accuracy of the results. Likewise, rainfall data from the

1980s for Katavi was unavailable for comparison with the land cover data of 1984 and the wildlife data of 1991. Therefore, the rainfall component was not included in any of the analyses for Katavi presented in this dissertation. However, information about temporal change in the amount of greenness (NDVI) (Pelkey et al. 2000, 2003), which is a proxy for rainfall, is available for all national parks in Tanzania and for comparison with the present results. Other potentially confounding variables, availability of water

(Gereta et al. 2004, Epaphras et al. 2007), predation by carnivores (Ogutu & Owen-

195

Smith, 2005), and illegal hunting (Caro 1999), were not included in the analyses. Some of these factors have been found to affect wildlife populations, such as for water in

Tarangire (Gereta et al. 2004) and Ruaha (Epaphras et al. 2007), and illegal hunting in

Katavi (Caro 1999, 2008 and 2011).

Despite these limitations, the findings of this dissertation are useful to help manage the respective protected areas.

3.0 Summary In Chapter 2, the extent to which land cover change has occurred from the 1980s to the 2010’s inside Katavi (KNP) and Tarangire (TNP) National Parks in Tanzania, and within their 5 km buffers was investigated. Land cover maps from the 1980s, 1990s, and

2010s for the two parks from Landsat TM and ETM+ satellite images were derived, following procedures described in Swain & Davis (1978), Richards & Jia (2006), and

Jensen (2005). The maps were produced with an overall accuracy of 86.8% and Kappa

Coefficient (K-hat) = 0.85, where producer and user accuracies range between 74 to

100% and 69 to 100%, respectively. These land cover maps set a baseline for the future monitoring of land cover changes on the respective parks. Previous land cover maps for

Tanzania have been produced by the Global Network Land Cover Project (GNLP)

(http://www.africover.org/), from Landsat TM imagery of 1997, though these maps had not been assessed for accuracy (Africover personal communication). Moreover, the

GNLP used a different classification system which was not suitable for this work.

Therefore, the present maps have been assessed for accuracy and are useful for monitoring land cover changes in the respective areas.

196

From the land cover maps generated from the satellite imagery, changes between time periods (i.e. 1980s vs. 1990s and 1990s vs. 2010s) were detected using post classification comparison technique (Singh 1989, Macleod & Congalton 1998, Lu et al.

2003, Coppin et al. 2004), and land cover matrices using cross tabulation technique to explore the land cover dynamics were computed (Macleod & Congalton 1998, Pérez-

Hugalde et al. 2011). The expectation was that the area covered by savannah and woody savannah would increase over time inside and outside of the parks, via a succession of shrubland, grassland, swamps and bare or less vegetated land, following increased efforts to protect wildlife habitats and species populations through implementation of the 1998 national wildlife policy, specifically the establishment of the wildlife management areas

(WMAs). Contrary to expectations, degradation of grassland and woodlands into bare or less vegetated land in TNP was found and as expected, savannah and woody savannah in

KNP increased due to recruitment of open shrubland.

Conversions of woodlands and grasslands into farmlands have been reported inside and outside TNP since 1980s and up to early 2000s (Mwalyosi 1992, Bruner et al.

2001, Msoffe et al. 2011). The present assessment reveals a continuing degradation of types of land cover inside and outside protected areas. The TNP results suggest the implementation of the wildlife policy does not appear to be effective in preventing degradation of wildlife habitats. The policy may appear to be effective in the KNP, although this is likely due to the extension of the park in 1998 (Figure 2.1).

In chapter 3, wildlife aerial census data from 1991 to 2012 was analyzed to understand temporal changes in: i) population abundances of six species or groups of large herbivores, and ii) their area of occupancy, inside and within 10 km buffer of

197

Katavi-Rukwa, Ruaha-Rungwa, and Tarangire protected areas in Tanzania. Here the questions to be answered were : 1) Has the population density of large herbivores and area they occupy, in Katavi-Rukwa, Ruaha-Rungwa and Tarangire protected areas in

Tanzania, changed between 1991 and 2012; 2) If change has occurred, was the degree of change, among species and area they occupy, consistent across protected areas; 3) Were the patterns of change similar inside and outside of the protected areas; and, 4) Have the protected areas been effective in achieving the goal of protecting species diversity? If the densities of large herbivores and/or the areas occupied by large herbivores are increasing over time or have remain unchanged, then the protected areas would be considered as effective. Analysis of animal density data that were obtained from Tanzania Wildlife

Research Institute from 1991 to 2012 determined that the population trends for three of the six large herbivore species or groups, including giraffe, medium antelopes and zebra, and the size of areas occupied by giraffe and zebra, have decreased over time inside and outside the three protected areas. The population trends for buffalo and elephant remained static inside and outside the three protected areas, and small antelopes decreased only in Tarangire. Rainfall in the three regions remained unchanged for about two decades. The shrinkage in the area occupied by buffalo, giraffe and zebra is probably the reason for the observed declines. This shrinkage may have intensified resource competition among species or groups and caused food shortages for specialist feeders such as giraffe and zebra. The contraction of wildlife habitat and concurrent declines of giraffe, zebra and medium antelopes, suggest the wildlife policy has not achieved the goal of conservation. Revision of the existing policy is recommended to help protected areas achieve their intended goal.

198

In chapter 4, first, the temporal changes in the landscape metrics (area, number and size of patches, and distance between suitable patches) of the types of land cover utilized by large herbivores, from 1980s to 2010s, were assessed. Second, the influence of changes in those metrics for each type of land cover on densities of three species of large herbivores: elephant, giraffe and zebra was evaluated. The influence of rainfall on each of the land cover spatial components, and on densities of the three large herbivores was also evaluated because rainfall is also commonly known to have a major effect on the population trends of the large herbivores.

The results showed evidence of loss in the amounts of five out of eight types land cover utilized by wildlife in KNP and TNP, between the 1980s and the 2010s. In both parks decreases in either area and/or, number of patches and/or increase in distance between patches of closed shrubland, open shrubland and grassland were found. In TNP alone, an increase in area covered by bare or less vegetated land, and the distance between patches of woody savannah was also found which indicated a loss in the amount of woodlands and grassland, probably degraded into bare or less vegetated land as observed in objective 1. As expected, areas and/or number of patches of open shrubland, savannah, woody savannah and swamps positively influenced the densities of elephant, zebra and giraffe, whereas the density of giraffe was negatively influenced by area of grassland and bare or less vegetated land. These relationships between land cover components and densities of large herbivores suggest that reduction or homogenization of these types of land cover can have profound consequences for species population viability (Owen-Smith 2004). The observed losses in the amounts of grassland, closed shrubland, and woody savannah in TNP, and grassland, open and closed shrubland in 199

KNP are of great conservation concern in maintaining species populations. Park management should take action to stop activity that causes reduction or total loss of land cover, particularly human encroachment cultivation or mismanagement of fires.

4.0 Synthesis

This dissertation reports evidence of degradation and loss in three (KNP) and five

(TNP) out of eight types of land covers utilized by wildlife, from the 1980s to the 2010s.

These changes have resulted in shrinkage of area occupied by buffalo, giraffe and zebra during the dry season, and consequently the abundances of giraffe, medium antelopes and zebra declined considerably. Degradation and loss of land cover have reduced the seasonal forage (wet-dry season) for these species, limiting their movements, with the probable consequence of degrading their own food through overgrazing and over browsing, hence creating shortage of food for less competitive species. Reducing the extent to which species can interact with resources that vary over time and space can have profound consequences for population viability, even for abundant species (Owen-

Smith 2004). In order to ensure long-term viability of species, especially such selective feeders as giraffe, management action is needed to stop human encroachment inside and on the surrounding buffer zones of protected areas.

Studies show that land cover degradation in and around Katavi and Tarangire protected areas was already occurring prior or at the time of establishment of the wildlife policy (Mwalyosi 1992, Bruner et al. 2001, Msofe et al. 2011). Similarly, previous reports show that declines on densities of large mammals across protected areas have been occurring prior to establishment of wildlife policy (Stoner et al. 2007, Caro 1999),

200

and continued to occur in Katavi after the policy implementation (Caro 2008, 2011).

While the levels of species abundance were expected to have either increased or remained at the levels of 1998 (i.e. policy implementation), this has unfortunately not been the case, even after a decade of implementation of the wildlife policy. This dissertation reports that degradation and loss of wildlife habitat and consequent declines in abundances of large herbivores continue to occur, inside and outside protected areas.

On this basis, we conclude that the three protected areas (Katavi, Tarangire and Ruaha) appear to be not effective in protecting and maintaining populations of species and their habitats, despite the increased emphasis on wildlife protection through establishment and implementation of the wildlife policy.

5.0 Recommendations

One of the objectives of the Tanzania’s National Wildlife Policy of 1998 was to encourage communities living adjacent protected areas to establish WMAs on communal land, from which they would directly benefits from wildlife resource through tourism enterprises, while preventing degradation and loss of wildlife habitats (MNRT 2007).

Economic benefits from WMAs to local communities adjacent protected areas in

Tanzania are reported to increase (Sulle et al. 2011, USAID 2013) while wildlife populations continue to decline. In early 1990s, similar approaches to WMAs were implemented in Namibia, Zimbabwe and Zambia and are reported to be successful in achieving both the economic development and conservation objectives (Shaw & Platts

2004, Boudreaux, 2010). In Namibia for example, contrast to Tanzania’s WMAs, the wildlife species populations are increasing inside and outside protected areas (Boudreaux,

201

2010). Degradation of wildlife habitats and shrinkage of distributional area of wildlife species reported in this dissertation suggests that the objective of wildlife protection has not been achieved. In order to avoid continual loss of habitats and species population, a revision of the existing policy is recommended, to come up with techniques that would help protected areas achieve their intended goal.

The resolution of Landsat satellite images (30 m) used in the analysis of land cover maps did not distinguish savannah from cultivated land. Since the latter appears to threaten the viability of protected areas in preserving species populations, wherever feasible and given availability of funding, finer scale maps such as ASTER (15 m resolution) or SPOT-4 Vegetation, should be employed to enable quantification of cultivated lands inside and outside the protected areas. If funding is limited, assessment can be done using the same Landsat images that are provided free of charge in combination with high resolution photos, such as Geoeye from Google Earth photos.

Areas observed with cultivation inside and at the buffer zones of KNP and TNP

(S1 and S2) should be closely monitored, to avoid further enchroachment inside the parks. Park ecologists can do this by observing high resolution photos on Google Earth and, as much as funding and manpower allows, conducting ground monitoring. Together with the wildlife census, TAWIRI does record human activties within wildlife ecosystems. However, these data are collected at an interval of two to three years and it may not be helpful for park managers to rely upon this information.

The analyses perfomed in this dissertation did not include a number of variables that might have provided stronger judgment on the effectiveness of the protected areas.

202

These include water availability, predation, poaching, fire and herbivory. Although some of these variables cannot be controlled by man e.g rainfall, the others can be to an extent.

A more comprehensive analysis could include include these variables, to provide a better conclusion about the innefectiveness of the parks at preventing the decline of wildlife populations.

203

6.0 References

Boudreaux, K. C. (2010). Community Conservation in Namibia: Devolution as a tool for

the legal empowerment of the poor. Working paper no. 10 - 64. Mercatus Center,

George Mason University.

Bruner, A. G., Gullison, R. E., Rice, R. E., & da Fonseca, G. (2001). Effectiveness of

parks in protecting tropical biodiversity. Science, 291, 125–8.

Caro, T. (2008). Decline of large mammals in the Katavi-Rukwa ecosystem of western

Tanzania. African Zoology, 43, 99–116.

Caro, T. (2011). On the merits and feasibility of wildlife monitoring for conservation: a

case study from Katavi National Park, Tanzania. African Journal of Ecology, 49,

320–331.

Caro, T. (1999). Densities of mammals in martially protected areas: The Katavi

ecosystem of western Tanzania. Journal of Applied Ecology, 36, 205–217.

Coppin, P., Jonckheere, I., Nackaerts, K., & Muys, B. (2004). Review article digital

change detection methods in ecosystem monitoring. International Journal of

Remote Sensing, 25, 1565–1596.

Epaphras, A. M., Gereta, E., Lejora, I., Ole Meing’ataki, G. E., Ng’umbi, G., Kiwango,

Y., & Mtahiko, M. G. G. (2007). Wildlife water utilization and importance of

artificial waterholes during dry season at Ruaha National Park, Tanzania.

Wetlands Ecology and Management, 16, 183–188.

Gereta, E., Ole Meing’ataki, G. E., Mduma, S., & Wolanski, E. (2004). The role of

wetlands in wildlife migration in the Tarangire ecosystem, Tanzania. Wetlands

Ecology and Management, 12, 285–299. 204

Jensen, J. (2005). Introductory digital image processing: A remote sensing perspective

(3rd ed.). Upper Saddle River: Pearson Prentice Hall.

Jensen, J. (2007). Remote sensing of the environment: An earth resource perspective.

Upper Saddle River, NJ: Pearson Prentice Hall.

Lu, D., Mausel, P., Brondizio, E., & Moran, E. (2003). Change detection techniques.

International Journal of Remote Sensing, 25, 2365–2401.

Macleod, R. D., & Congalton, R. G. (1998). A Quantitative comparison of change-

detection algorithms for monitoring eelgrass from remotely sensed data.

Photogrammetric Engineering and Remote Sensing, 64, 207–216.

Msoffe, F. U., Kifugo, S. C., Said, M. Y., Neselle, M. O., Van Gardingen, P., Reid, R. S.,

& de Leeuw, J. (2011). Drivers and impacts of land-use change in the Maasai

Steppe of northern Tanzania: an ecological, social and political analysis. Journal

of Land Use Science, 6, 261–281.

Mwalyosi, R.B.B. (1992). Land-use changes and degradation in south-west Masaai land,

Tanzania. Environmental Conservation, 19, 145–152.

Ogutu, J. O., & Owen-Smith, N. (2003). ENSO, rainfall and temperature influences on

extreme population declines among African savanna ungulates. Ecology Letter, 6,

412–419.

Ogutu, J. O., & Owen-Smith, N. (2005). Oscillations in large mammal populations: are

they related to predation or rainfall? African Journal of Ecology, 43, 332–339.

Owen-Smith, N. (1988). Megaherbivores: The influence of very large body size ecology,

Cambridge. UK: Cambridge University Press.

205

Pelkey, N. W., Stoner, C. J., & Caro, T. M. (2000). Vegetation in Tanzania: assessing

long term trends and effects of protection using satellite imagery. Biological

Conservation, 94, 297–309.

Pelkey, N. W., Stoner, C. J., & Caro, T. M. (2003). Assessing habitat protection regimes

in Tanzania using AVHRR NDVI composites: comparisons at different spatial

and temporal scales. International Journal of Remote Sensing, 24, 2533–2558.

Pérez-Hugalde, C., Romero-Calcerrada, R., Delgado-Pérez, P., & Novillo, C. J. (2011).

Understanding land cover change in a special protection area in central Spain

through the enhanced land cover transition matrix and a related new approach.

Journal of Environmental Management, 92, 1128–37.

Richards J. A. (1998). Remote Sensing Digital Image Analysis. Berlin: Springer.

Shaw, J. & Platts, L. (2004). Building the CAMPFIRE paradigm. Property and

Environment Research Center Report (PERC). Montana, USA.

Singh, A. (1989). Review article: digital change detection techniques using remotely

sensed data. International Journal of Remote Sensing, 10, 989–1003.

Swain, P. H., & Davis, S. M. (1978). Remote Sensing: the quantitative approach. West

Lafayette Ind., USA: Purdue University.

USAID. (2013). Tanzania wildlife management areas evaluation. United States Agency

for International Development. Report.

206