Mapping Distribution of Butterflies in Central Bobiri Forest Reserve And

Mapping distribution of butterflies in central Bobiri Forest Reserve and investigation of logging and stage of regeneration on butterfly species richness and diversity.

Addae-Wireko Louis March, 2008

Mapping distribution of butterflies in central Bobiri Forest Reserve and investigation of logging and stage of regeneration on butterfly species richness and diversity.

by

Addae-Wireko Louis

Thesis submitted to the International Institute for Geo-information Science and Earth Observation (The Netherlands) and Kwame Nkrumah University of Science and Technology (Ghana) in partial fulfilment of the requirements for the degree of Master of Science in Geo-information Science and Earth Observation, Environmental Systems Analysis and Modelling (ESAM) Specialisation

Thesis Assessment Board Prof. Dr. Ir. Alfred de Gier, Chairman, Degree Assessment Board, ITC (chair) Dr. B. K. Prah (External Examiner) Dr. S.K. Oppong, Internal Examiner, KNUST Ir. Louise M. van Leeuwen, Course coordinator, ITC Dr. William Oduro, First supervisor, KNUST

Supervisors: Dr. Jan de Leeuw (ITC), Dr. William Oduro (KNUST), Mr. J. Quaye Ballard (KNUST)

INTERNATIONAL INSTITUTE FOR GEO-INFORMATION SCIENCE AND EARTH OBSERVATION, ENSCHEDE, THE NETHERLANDS KWAME NKRUMAH UNIVERSITY OF SCIENCE AND TECHNOLOGY, KUMASI, GHANA

Disclaimer

This document describes work undertaken as part of a programme of study at the International Institute for Geo-information Science and Earth Observation and Kwame Nkrumah University of Science and Technology. All views and opinions expressed therein remain the sole responsibility of the author, and do not necessarily represent those of the institutions.

Abstract

The contribution of remote sensing data to the mapping and prediction of invertebrate diversity is poorly investigated and even when done, much of these have focused on pests rather than for conservation purposes. This study sought to map the distribution of selected butterflies in central Bobiri Forest Reserve using Genetic Algorithm for Rule-set Production (GARP) and Maximum Enthropy (Maxent) methods with presence data only. Distributions of six common butterfly species of Ghana were successfully mapped in the study area. Receiver operating characteristic curves for eight (8) distribution maps were outstanding with four being excellent. There was not much difference between the distribution maps generated with all predictor variables considered and those with the best five predictors in terms of AUC. Maxent maps generally had higher AUCs when ROC curves were plotted. Distance to rivers, Enhanced Vegetation Index (EVI), Elevation and Distance to roads were the most relevant predictors of species distribution. Enhanced Vegetation Index performed better than Normalised Difference Vegetation Index (NDVI) and Soil Adjusted Vegetation Index (SAVI) in predicting species distribution. A Kruskal-Wallis test of butterfly diversity indices of three forest classes with different levels of disturbance and at different stages of regeneration revealed significant differences (P<0.0001) in all the three diversity indices used (Fisher’s Alpha, Shannon-Weiner and Simpson’s Inverse diversity) with further post tests (Dunn’s multiple comparison) of forest blocks showing significant differences (p<0.001) for all diversity indices. The recently logged and the not logged blocks were the most similar, with 61 common species and a Morista-Horn Index of 0.951. Although the recently logged (RL) had higher numbers of captures and rarefied abundance of individuals than the not recently logged (NRL) and unlogged (NL), it was also the least diverse with regard to Simpson’s inverse diversity index. Therefore every effort should be made to prevent degradation of the forest which would allow few species most suited for degraded areas and secondary forests to greatly benefit at the expense of specialists found in forests in pristine conditions.

i Acknowledgements

To God most high for making this study possible.

My profound gratitude goes to my supervisors Jan de Leeuw, William Oduro and Jonathan Quaye Ballard. I owe this work to you and it would not have been possible without your time, interest in my work and the constructive criticisms.

I wish to thank the ITC, Wildlife and Range Management Department, FRNR, KNUST and KAAD for their immense support in making the dream of undertaking this course a reality. I am especially thankful to Michael Weir and the coordinators of the GISNATUREM programme for the roles they played.

To all my colleagues, staff of KNUST and Bobiri Forest Reserve, friends, family and all who contributed in diverse ways in making this work a success, I say God richly bless you.

ii Table of contents

1. Introduction ...... 1 1.1. Background ...... 1 1.1.1. Habitats ...... 1 1.1.2. Biological indicators...... 2 1.2. Research Problem ...... 3 1.3. Prior work ...... 4 1.4. Research Aim ...... 4 1.5. Objectives ...... 4 1.6. Research Questions ...... 5 1.7. Hypothesis ...... 5 1.8. Assumptions ...... 6 1.9. Research Approach ...... 6 2. Methods and Materials ...... 9 2.1. Study Area ...... 9 2.1.1. Compartment Designation ...... 10 2.2. Site Selection...... 10 2.2.1. Sampling Design ...... 10 2.3. Data...... 11 2.3.1. Butterfly Sampling...... 11 2.3.2. Habitat Variables...... 13 2.3.3. Topographic Features ...... 14 2.3.4. Distance maps ...... 14 2.3.5. Satellite image processing ...... 15 2.3.6. Modelling of species distribution ...... 15 2.3.7. Selection of model species ...... 15 2.3.8. Development of Models ...... 16 2.3.9. Evaluation of Models ...... 16 3. Results ...... 18 3.1. Intermediate maps for mapping distribution of species...... 18 3.2. Sampling Effort ...... 19 3.3. Distribution of trapped butterfly species among butterfly families...... 20 3.4. Distribution of captured butterflies among forest disturbance classes ...... 21 3.4.1. Butterfly species richness of forest disturbance classes ...... 22 3.4.2. Butterfly diversity of forest disturbance classes ...... 23 3.5. Distribution of captured butterflies among months...... 24 3.6. Species Rank Abundance...... 25 3.7. Modelling of Species Distribution...... 25 3.7.1. Vetting of Predictors...... 26 3.7.2. Results of Models ...... 26 4. Discussion ...... 35 4.1.1. Captures per sampling effort ...... 35 4.2. Spatio-temporal variation of butterfly diversity and species richness ...... 35 4.2.1. Spatial variation of butterfly diversity and species richness ...... 35

iii 4.2.2. Temporal variation of butterfly diversity and species richness ...... 36 4.3. Predictors of Species Distribution ...... 36 4.4. Species Distribution Maps ...... 37 5. Conclusion and Recommendation ...... 39 5.1. Conclusion ...... 39 5.1.1. Implications for conservation ...... 40 5.2. Recommendations ...... 41 References ...... 42 Appendices ...... 48 Appendix A: Checklist of butterfly species caught ...... 48 Appendix B: Input maps of environmental variables ...... 52 Appendix C: Results of statistical analysis ...... 54 Appendix D: Accuracy assessment of unsupervised classification...... 56 Appendix E: Probabilistic predictive maps of Maxent and GARP for the selected species of butterflies. 57 Appendix F: Response curves for best five predictors of maxent model ...... 58

iv List of figures

Figure 1-1 Conceptual framework of Research Approach ...... 7 Figure 2-1 Study area: Bobiri Forest Reserve (BFR) ...... 9 Figure 2-2 Fruit baited trap ...... 11 Figure 3-1 Unsupervised Classification of Landsat 2007 image of study area ...... 18 Figure 3-2 Species accumulation curve showing the relation between cumulative number of species recorded in all 160 traps against length of the sampling period (days)...... 20 Figure 3-3 Number of Individuals and species recorded for the sub-families of the Nymphalidae ...... 20 Figure 3-4 Species Richness estimation for the forest blocks (Rarified: 29 samples) ...... 22 Figure 3-5 Predicted geographical distribution of Bicyclus funebris in the study area ...... 28 Figure 3-6 Predicted geographical distribution of Gnophodes betsimena in the study area ...... 29 Figure 3-7 Predicted geographical distribution of Charaxes protoclea in the study area ...... 30 Figure 3-8 Predicted geographical distribution of Euphaedra phaethusa in the study area ...... 31 Figure 3-9 Predicted geographical distribution of Kallimoides rumia in the study area ...... 32 Figure 3-10 Predicted geographical distribution of Palla ussheri in the study area ...... 33

v List of tables

Table 2-1 Strata composition and Number of trap stations allocated ...... 10 Table 2-2 Derived indices used ...... 13 Table 3-1 Butterfly capture success rate per stratum and sampling effort standardization...... 19 Table 3-2 Total number of butterflies and number of species recorded in the three forest classes ...... 21 Table 3-3 Distribution of abundance of the Nymphalidae captured in the three forest categories over the five subfamilies ...... 21 Table 3-4 Kruskal-Wallis test of significance for diversity indices of the forest blocks ...... 23 Table 3-5 Dunn's Multiple Comparison Post Test of forest blocks per diversity index ...... 23 Table 3-6 Complementarities or similarity Index for forest blocks ...... 24 Table 3-7 Distribution of abundance of the Nymphalidae captured in sampled months from 160 trap stations...... 24 Table 3-8 Most abundant species collected with their distribution in the strata ...... 25 Table 3-9 Distribution and abundance of modelled species ...... 25 Table 3-10 Percentage contribution of environmental layers to maxent models obtained through the Jacknife test for variable importance...... 26 Table 3-11 Receiver Operating Characteristics (ROC) for predictive distribution maps ...... 27

vi List of Abbreviations

ATCOR Atmospheric Correction AUC Area Under the Curve BFR Bobiri Forest Reserve DEM Digital Elevation Model ETM Enhanced Thematic Mapper EVI Enhanced Vegetation Index FORIG Forestry Research Institute of Ghana FPAR Fraction of absorbed Photosynthetically Active Radiation FSD Forest Services Division GARP Genetic Algorithm for Rule-set Production LAI Leaf Area Index Maxent Maximum Entropy MMMeans Michaelis-Menten Mean NDVI Normalised Difference Vegetation Index NL Not Logged NRL Not Recently Logged RL Recently Logged ROC Receiver Operating Characteristic SAVI Soil Adjusted Vegetation Index

vii

MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

1. Introduction

1.1. Background Man has continuously relied on natural resources for survival. Realising that life depends on the availability of natural resources and the possibility of these resources being over- exploited, various measures have been implemented globally by governments, civil and conservation organisations to ensure that resources are sustainably managed by maintaining viable populations in order for renewable natural resources to regenerate (Paloniemi and Tikka, 2008). Other measures include regulation of resource use, the creation of reserves and complete protection of species and portions of natural resources (Gutierrez and Menendez, 2007; Pressey et al., 2004; Tole, 2006).

Information is a requisite in deciding which areas or species are to be reserved for future use or accorded the needed protection based on their uniqueness or exceptional biodiversity value (Gutierrez and Menendez, 2007; Rodrigues and Gaston, 2002; Tole, 2006). Over the years various studies have been conducted to document the kinds, distribution and abundance of plant and animal species for conservation purposes (Burgess et al., 2007; Hall and Swaine, 1981). After establishing protected areas, reliable and timely information on the abundance of species, their distribution and information about their habitats as well as threats are requisites for proper management (de Leeuw et al., 2002). However, even when this information is present, changes resulting from processes of natural selection, resource use, and changes in environmental conditions make further studies necessary (de Leeuw et al., 2002).

1.1.1. Habitats Deforestation has remained a threat biodiversity through causing habitat loss in the light of destruction of several million hectares of forests in the tropics since 1900 (Barlow et al., 2007; Koh, 2007). Habitat destruction has led to extinctions, reduction in species richness and the concentration of the most valuable pieces remaining, ‘hotspots’, in small pockets spread across the earth (Brooks et al., 2002). Despite setting aside 16% of Ghana’s land surface area for maintaining representative samples of natural ecosystems in reserves and protected areas, pressure from agricultural expansion, deforestation, urbanisation, desertification and mining cause the country to lose an estimated 22,000ha (1.3%) annually resulting in an area of intact forest of 6.9% of the country’s total land area as at 2002, from an original cover of 35% and 11.8% by the turn of the century and mid-1970s respectively (Ministry of Environment and Science, 2002). Degradation has not been limited to off-reserves only. Whilst less than 1% of the forest cover remain outside forest reserves, timber extraction as well as selective logging

1 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY. in the reserves have contributed to simplifying the structure of forest habitats and removed ecological niches leading to a reduction in species diversity. (Ministry of Environment and Science, 2002). Efforts have been made to assess and quantify the rapidly degrading resources for the identification of unique areas with much biological diversity remaining for conservation. However, these assessments are laborious and require some degree of expertise to accomplish (Fleishman et al., 2000; Gardner et al., 2008; Hermy and Cornelis, 2000; Kati et al., 2004).

Consequently, the challenge to measure levels of organisation of biological variation in composition, structure and function has led to the search for relevant biological indicators from which biodiversity can be measured (Leyequien et al., 2007). Measurable indicators have been used in models for describing the environment in which a species has been recorded, identifying other locations where the species may currently reside, and identifying where the species may occur when habitat variables change (Beaumont et al., 2005). Considering the continuous anthropogenic and natural disturbances in the environment and difficulties in species assessment, the most promising approach for such timely assessment, monitoring, prediction to inform conservation of faunal biodiversity appears to be the synergy of remote sensing products and auxiliary data with ecological biodiversity models (Leyequien et al., 2007; Seto et al., 2004). These predictions and assessments however require subsequent validation of results using traditional observation techniques, such as sampling for presence of species, trapping or direct counts to validate predictions (Leyequien et al., 2007).

1.1.2. Biological indicators. Biological indicators, also referred to as ‘bioindicators’ have often been used to assess the state of resources. To qualify as a bioindicator, species should respond to slight ecosystem changes in a predictive manner (Bouyer et al., 2007).

Butterflies belong to the Order Lepidoptera with over 150,000 species (Burnie and Tschinkel, 2005) which is part of the Class Insecta of which more than a million species have been described representing about half the global diversity (Larsen, 2005; Ramos, 2000). Butterflies are the most popular and best known group of all insects (DeVries, 2001; Larsen, 2005) with well known taxonomic ranking and ecologically highly diversified (Bouyer et al., 2007) and well studied due to their being aesthetically nice colour patterns, relative large size, naturally inoffensive, ease of recording and identifying (Devries et al., 1997; Larsen, 2005; Lewis, 2001).

Butterflies have been used in several studies as good indicator species to predict changes in the environment (Thomson et al., 2007; Ulrich and Buszko, 2004; Vanreusel and Van Dyck, 2007). Butterfly species abundance (number of individuals of a species) and species richness (number of species in an area) may change in response to environmental variables such as plant composition and density, humidity, light intensity and degree of disturbance of vegetation thereby making them good indicators of change (Barlow et al., 2007; Bossart et al.,

2 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

2006; Bouyer et al., 2007; Cleary et al., 2005; Lewis, 2001; Nelson, 2007). This is exemplified by Nelson (2007) as he demonstrated butterfly response to light intensity reaching the forest floor as a measure of closeness of canopy or extent of degradation. The relative response by species to changing environmental conditions have resulted in a classification where a species may be termed specialist (restricted only to habitats where their usually narrow requirements of environmental conditions are met) or generalist when they exist under wider range of conditions (Thomas et al., 2004; Vanreusel and Van Dyck, 2007).

Several indices have been developed and used in studies to depict the attributes of vegetation (Goetz et al., 2007; Houborg et al., 2007; Huete, 1988; Mutanga and Skidmore, 2004; Schlerf et al., 2005). Vegetation Indices (VI), as they have been called, have been extracted from combinations of wavelengths to provide information on vegetation attributes or properties through the use of broadband - multispectral sensors (Houborg et al., 2007; Huete et al., 1999; Huete, 1988; Oindo and Skidmore, 2002; Pettorelli et al., 2005; Seto et al., 2004) as well as narrow band hyperspectral images (Kalacska et al., 2007; Schlerf et al., 2005) known to enhance measurement of vegetation cover. The indices have various applications including characterization of different cover types, estimation of vegetation health or stress, measurement of light use efficiency, as well as assessing the extent of degradation or intactness of habitats as a measure of Leaf Area Index (LAI). The traditional remote sensing approach for estimating LAI is based on the combination of a chlorophyll sensitive band (red) and a band located in the high reflectance plateau of vegetation canopies (NIR band) (Houborg et al., 2007). Normalised Difference Vegetation Index (NDVI) (Tucker, 1979) has been widely used for the retrieval of LAI by combining the NIR band and red band reflectances of satellite images. The relationship between the NDVI and vegetation productivity is well established and recent ecological studies have highlighted the relevance of NDVI as an index linking vegetation to animals (Leyequien et al., 2007; Oindo and Skidmore, 2002; Pettorelli et al., 2005; Seto et al., 2004). Nevertheless, NDVI is known to saturate in dense vegetation with high values of LAI (Houborg et al., 2007). Soil Adjusted Vegetation Index (SAVI) (Huete, 1988) and variants of these, proposed as improvements over NDVI, minimise the effect of soil background and suitable to parameterize LAI but are subject to an increase in sensitivity in the atmosphere (Leprieur et al., 1994). However, the Enhanced Vegetation Index (EVI) (Huete et al., 1997) minimises the soil background as well as atmospheric aerosol effects thereby optimising the vegetation signal in high LAI regions.

1.2. Research Problem Compared to other taxonomic groups, the ability of remote sensing data to contribute to the mapping and prediction of occurrence of invertebrate diversity appears to be poorly investigated and even if done, the majority of published studies have concentrated on insects that are considered pests whilst few studies deal with conservation efforts (Leyequien et al., 2007). Bobiri Forest Reserve is renowned for its butterfly species richness having about 456

3 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY. species of Ghana’s known total of 930 species (Larsen et al., 2007). Having been elevated to the status of butterfly sanctuary and aiding in the promotion of eco-tourism, the spatial distribution of butterflies of this reserve has not yet been modelled or mapped. Hence the distribution of species cannot easily be determined to favour specific conservation efforts towards the sustainability of butterflies in the reserve. This is important because subtle changes in the microhabitat affects the distribution, abundance and diversity of butterfly species. It is therefore essential that effects of changes in environmental variables on distribution, abundance and richness of butterfly species be investigated.

1.3. Prior work A review of the history of collection of butterflies in Bobiri Forest Reserve (BFR) by Larsen et al (Larsen et al., 2007) recounts several expeditions in BFR and Butterfly Sanctuary by lepidopterists, tourists, experts and novices alike. Most of these could be considered as the gathering baseline data; thus for the purpose of indicating which species are present or absent in BFR, describing environments of captured species and finding possible factors accounting for the presence. Other studies have concentrated on vertical stratification, butterfly diversity between the natural forest, the Bobiri Aboretum and farmlands outside the reserve and the comparison of diversity in Bobiri Forest Reserve to other reserves and sacred groves (Bossart et al., 2006). Much of the work in Bobiri has however remained published. There are no publications on any attempt aimed at mapping the distribution of butterfly guilds or individual species in this reserve and in Ghana as a whole.

1.4. Research Aim The main aim of this study is to contribute to filling the information gap that exists in terms of species spatial distribution in Bobiri Forest Reserve by mapping the distribution of six common butterflies and assessing the effect of selective logging and stage of regeneration on the richness and diversity of butterflies thereby predicting the consequences of habitat degradation on butterfly species.

1.5. Objectives The following specific objectives were set in order to achieve the research aim. • To map the distribution of six common butterfly species of Bobiri Forest Reserve with widespread, moderate and restricted distributions using GARP and Maxent modelling algorithms. • To compare predictive maps produced by GARP and Maxent through the use of ROC curves. • To compare the mapped butterfly ranges.

4 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

• To compare performance of vegetation indices (EVI, NDVI and SAVI) in determining butterfly species distribution. • To compare butterfly diversity and richness of three classes of forests at different stages of regeneration.

1.6. Research Questions The following research questions posed will be answered by this study. • Are the models of distribution of the six common butterfly species significantly different from each other? • Do GARP and Maxent predict equal species distributions? • Do species abundance, distribution and species richness vary with months? • Does species diversity and richness vary with different stages of regeneration? • Do EVI, NDVI and SAVI correlate with species richness; if so, which is the best? • How well can remote sensing be used to discriminate butterfly species richness of the logged and unlogged forest? • Which of the measured environmental variables strongly determine species distribution?

1.7. Hypothesis The following hypothesis will be tested during the study. Hypothesis 1 Ho: The extent of canopy closure does not influence butterfly diversity present. Ha: The extent of canopy closure influences butterfly diversity present.

Hypothesis 2 Ho: The chosen environmental variables do not predict butterfly diversity better than prediction by chance. Ha: The chosen environmental variables predict butterfly diversity better than prediction by chance.

Hypothesis 3 Ho: Accuracy of model prediction algorithm of GARP does not differ from Maxent. Ha: Accuracy of model predictions algorithm of GARP differs from Maxent.

5 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

1.8. Assumptions In conducting this research, the assumption made was that the successful trapping of butterflies at designated trap stations signify that conditions necessary for the butterfly to thrive in this environment are present and vice versa

1.9. Research Approach The following approach (Figure 1-1) was employed in the study and could be classified under four major stages. 1. Butterfly sampling 2. Selection of model species and extraction of presence data 3. Generation of environmental variables and input layers. 4. Modelling of Species Distribution

6 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Butterfly Sampling Environmental layers

Site Selection Distance to Rivers Roads & Logging Distance Rivers History to Roads

Stratification (NL, NRL, NL) Aspect

Compart- ment Size Slope DEM

Proportional Allocation of trap stations Elevation

Location of Classified Trap Stations Image

Veg. Indices Corrected NDVI,€€ EVI, Landsat Trapping of butterflies SAVI, LAI 2007

Radiation Indices Butterfly Abundance FPAR, Net Rad, & Number of Abs. Solar Rad.) Species

Selection of Model Species Legend

Data

Decision

Habitat Suitability Model Process

Figure 1-1 Conceptual framework of Research Approach

7

MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

2. Methods and Materials

2.1. Study Area The study area, Bobiri Forest Reserve (BFR), is located approximately 30 km east of Kumasi, Ghana and covers an area of 54.6 km2 (Figure 2-1). The reserve is within the moist semi- deciduous south east forest zone (Hall and Swaine, 1981) and has a mean annual precipitation of between 1200 and 1750 mm, and the main dry season occurs between December and March with a shorter dry season in August as rainfall in the south is bimodal. Thus two rainy seasons; March to June and September to November.

Figure 2-1 Study area: Bobiri Forest Reserve (BFR) A: ASTER 2003 Image of BFR depicting extent of degradation around BFR, B: Progress map of BFR showing compartment designation, selected study site and trap stations, C: Location of Bobiri Forest Reserve with respect to Ghana.

Selective timber logging in the forest reserve about 35 years ago has resulted in a canopy 25m to 30m and many openings (Nichols et al., 1998). Human pressure on the ecosystem is evident from the various farms around the reserve, collection of non-timber forest products and its logging history (Figure 2-1a). These anthropogenic influences have seriously degraded other reserves in Ghana(Hawthorne and Abu-Juam, 1995).

9 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

2.1.1. Compartment Designation Bobiri forest reserve is divided into 73 compartments. These compartments have four designations based on their use, namely, research, butterfly sanctuary, strict nature reserve and production forest (Figure 2-1b). These blocks are under two separate managements. The productive forest is managed by the Forest Services Division (FSD) whilst the research, butterfly sanctuary and strict nature reserve designated compartments are managed by the Forestry Research Institute of Ghana (FORIG). Many floral and faunal studies have been conducted in the blocks managed by FORIG. Majority of the studies on butterflies in Bobiri Forest Reserve have taken place in the research and butterfly sanctuary blocks.

2.2. Site Selection. Regardless of the long distance from the point of entry of the reserve and accessibility problems in the wet season this study was conducted in central BFR. The selected site was of much interest because not much is known about butterflies here and it had compartments that have been exposed to different years of selective logging and stages of regeneration as well as those with no records of logging. Eleven (11) of the 73 compartments of BFR (Figure 2-1b and Table 2-1) formed part of the selected site. These compartments were categorised into three strata based on the logging history. The stratified zones comprise a stand with no history of logging (not logged, NL), recently logged (RL) which were logged up to 10 years ago and a third stand that were logged 25 to 45 years ago (not recently logged, NRL).

2.2.1. Sampling Design The proportional stratified random sampling design was used. This design takes advantage of knowledge of a population which is used to increase the precision or usefulness of the sample (Freese, 1984; Hirzel and Guisan, 2002).

Table 2-1 Strata composition and Number of trap stations allocated Traps Actual last year Size Strata Compartment Total (Area) allocated No. of logged (ha) traps Up to 10 years ago 61 2001 82.99 (RL) 63 2007 106.09 189.08 34 30

14 1970 71.36 15 1970 79.45 16 1970 82.45 25 to 40 years ago 19a 1970 71.36 588.24 106 101 (NRL) 65 1975 134.12 66 1976 71.28 67 1975 45.5

13 - 77.23 Not logged (NL) 106.15 20 29 19b - 28.92 Total 10 731.71 883.47 160

10 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

The three stratified zones in the subset study area were used. Table 2-1 summarizes information used in the sampling design. Based on the total size of the compartments that make up each stratum, the trap stations were proportionally allocated.

Randomly generated numbers were picked and these numbers were paced to, beginning from corners of the compartments along the roads. To facilitate the daily collection of specimen for the research period, random points chosen served as the starting point for two transects in the opposite directions. Transects were laid nearly perpendicular to roads (figure 2-1b). Though transects begin from the road, the first trap stations were located about 50m from the road. Two transects were laid parallel to existing roads and 15m to each side to compare the species present. Compartments scheduled to be logged between September and December 2007 were excluded from the samples.

2.3. Data. Data was collected for three months (October 2007 to December 2007). For each month, data was collected for two weeks in two sets of 80 traps each; a total of 160 trap stations located approximately 50m from each other. A typical transect had 10 trap stations. Coordinates of all the trap stations were recorded with a Garmin GPSmap 60CSx which has good satellite reception even under dense canopy.

2.3.1. Butterfly Sampling.

Under-storey fruit-feeding butterflies were captured with fruit baited traps (Figure 2-2) (Daily and Ehrlich, 1995; DeVries, 1987) for the study. The cylindrical trap nets were at least 90cm in height in order to minimise the escape of butterflies once it entered the trap (Barlow et al., 2007). Butterfly traps were hung on a support usually shrubs and was within 0.1 m and 1m from ground level (Bossart et al., 2006). Figure 2-2 Fruit baited trap

Traps were baited with pulverised, fermenting bananas covered in a rubber bucket two days prior to each collection. The first day of the week for collection was used to set the traps. This included the hanging of traps from ropes attached to branches of trees and/or shrubs. Grease was improvised for tangle foot which was applied to the ropes to prevent ants from descending on the traps. The trap stations were then labelled. The prepared bait was mixed with fresh palm wine for maximum attraction of the butterflies and also to speed up the

11 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY. fermentation process. Baits were then put in plastic plates and placed in the trap to complete the procedure. Captured butterflies in the traps were removed about 24 hours from the time the traps were set and for the next five days and placed into labelled glassine envelopes. Species were kept in storage bowls and put in a deep freezer till they were to be identified. Butterflies of West Africa (Larsen, 2005), an identification guide was used for identifying the species.

2.3.1.1. Assessment of butterfly diversity. Species diversity at a local scale of analysis is expressed as “indices that weight both the richness and equitability (evenness of abundance across species) of a sample (Whittaker et al., 2001). Diversity was assessed at different levels. These include alpha diversity - diversity within samples, species communities or patches - and beta-diversity or species turnover which measures the diversity among samples or communities (Beck and Vun Khen, 2007; Summerville et al., 2003; Whittaker et al., 2001). Species diversity was quantified in terms of abundance and species richness. Abundance was calculated as the “total number of individuals of a species” whilst species richness refers to “the total number of species observed” (Goetz et al., 2007) or “the total number of species in a sample” (Whittaker et al., 2001).

The data was pooled at two levels – monthly and per forest disturbance class (stratum) - before diversity indices were calculated in order to reduce the variations that existed in the capture success of the traps (Barlow et al., 2007; Bossart et al., 2006). Since the three forest disturbance classes had different areas of coverage and sampling effort was proportional to the area covered by each class, the abundance of the larger areas were rarefied (Sanders, 1968) by randomly selecting within this large population smaller samples equivalent to the number of the least sampled to enable comparisons of species diversity and richness statistics. The rarefaction, calculation of species diversity indices and shared species indices were done using EstimateS (Colwell, 2005) and EcoSim (Gotelli and Entsminger, 2007) software. The diversity indices generated by EstimateS and EcoSim were subjected to statistical tests in a statistical software (Prism 4 for windows; Graphpad Software Incorporated) to check whether the data has been sampled from a Gaussian population. Since not all the indices passed the normality test, the Kruskal-Wallis test – also called Kruskal-Wallis one-way analysis of variance by ranks – a non parametric test which does not assume that the data has a Gaussian distribution was used to compare the diversity indices (Shannon-Weiner Index, Simpson’s Inverse Diversity Index and Fisher’s Alpha) for the three forest blocks to ascertain the significance of differences between them. Following the Kruskal-Wallis test, the Dunn’s (multiple comparison) post test was done to compare the difference in the sum of ranks between each two blocks per diversity indices.

12 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

2.3.2. Habitat Variables. With the thermal band of the 2007 Landsat image, relative differences in thermal properties of study site were extracted. These included absorbed solar radiation, net radiation and Fraction of absorbed Photosynthetically Active Radiation (FPAR). Two Hobo pro data loggers installed from October to December 2007 and used to collect temperature, dew point and relative humidity (RH) readings for some trap stations in the study area (two at a time) did not vary much so this data was not used in the modelling. Hemispherical photographs taken with a digital fish-eye lens camera at the trap stations were analysed with the GAP Light Analyser software (Frazer et al., 1999) and used for estimating the degree of openness of canopy which was used as the major discriminating factor in the classification of the Landsat 2007 image. The camera was mounted on a tripod, 1 meter from the forest floor for uniformity and levelness. Four attributes of vegetation in the study area were extracted from the Landsat 2007 image; Soil Adjusted Vegetation Index (SAVI), Leaf Area Index (LAI), Enhanced vegetation Index (EVI) and Normalised Difference Vegetation Index (NDVI).

2.3.2.1. Derivation of Vegetation and Radiation Indices Selection of vegetation indices were based on the extent of use, its applicability for the study and their relative advantages. The indices used are summarised in Table 2-2 and followed by the equations used.

Table 2-2 Derived indices used Index Image Used Data Range Software used

Normalised Difference Vegetation Landsat ETM+ -1 to 1 ATCOR, Erdas Imagine Index (NDVI)

Soil Adjusted Vegetation Index (SAVI) Landsat ETM+ 0 to 1000 ATCOR, Erdas Imagine

Enhanced Vegetation Index (EVI) Landsat ETM+ -1 to 1 ENVI 4.3

Leaf Area Index (LAI) Landsat ETM+ 0 to 10000 ATCOR, Erdas Imagine

Fraction of absorbed photosynthetically Landsat ETM+ 0 to 1000 ATCOR, Erdas Imagine active radiation (FPAR)

The formulae used in deriving the indices are shown below.

Normalised Difference Vegetation Index (NDVI) (Tucker, 1979). The Normalised Difference Vegetation Index was calculated from

ρ NIR− ρ RED NDVI = (1) ρ NIR+ ρ RED where ρ NIR and ρ RED are NIR and Red bands

13 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Soil Adjusted Vegetation Index (SAVI) (Huete, 1988)

ρ NIR− ρ RED SAVI=+ (1 L) (2) ρρNIR+ RED + L where ρ NIR and ρ RED are NIR and Red bands and L=0.5

Enhanced Vegetation Index (EVI) (Huete et al., 1997)

⎛⎞ρ NIR− ρ RED EVI= 2.5⎜⎟ (3) ⎝⎠ρρNIR+−67.51 RED ρ BLUE + where ρ NIR and ρ RED are NIR and Red bands

Leaf Area Index (LAI) (Baret and Guyot, 1991)

⎛⎞⎛1 aVI0 − ⎞ LAI=−⎜⎟⎜ ln ⎟ (4) ⎝⎠⎝aa2 1 ⎠ where VI = SAVI, ao=0.82, a1=0.78, a2=0.6

Fraction of absorbed photosynthetically active radiation (FPAR) (Asrar, 1989; Wiegand et al., 1991) This index estimates the fraction of absorbed photosynthetically active radiation and is associated with productivity. FPAR is approximated from a model given by

FPAR=−CA[ 1 exp( − BLAI x ] (5) where C=1, B=0.4, A=1

Net Radiation (Richter, 2004) Expressed as the sum of three radiation components, Net radiation is given by

RRn=+− solar R atm R surface (6) where Rn = Net Radiation, Ratm = radiation emitted from the atmosphere toward the surface,

Rsurface = radiation emitted from the surface into the atmosphere

2.3.3. Topographic Features A digital elevation model (DEM) of the study area was resampled to 15m resolution using the bilinear interpolation method. The resulting image was used for generating slope, aspect and elevation maps for the study area.

2.3.4. Distance maps Distance maps were generated from existing roads and rivers in the study area using ILWIS 3.3.

14 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

2.3.5. Satellite image processing Recent Landsat Enhanced Thematic Mapper (ETM plus) images of the study area acquired on 24th February 2007 images were used. The Landsat ETM+ image was chosen because it has a blue band which is required for deriving the Enhanced Vegetation Index.

2.3.5.1. Corrections The image was radiometrically corrected for haze and atmospheric correction using the ATCOR 3 workstation module of ERDAS Imagine 8.7 since the terrain of the study area is not flat. The radiometrically corrected image was subsequently corrected geometrically through an image to map registration using a topographic sheet with forty distinguishable and nearly permanent features used as ground control points with a Root Mean Square Error of 0.16, an error margin of 2.2m per pixel.

2.3.5.2. Classification of Image An unsupervised classification was run on the Landsat ETM+ image in Erdas Imagine version 9.1. The image was first classified into 30 classes based on site reflectance values. The resulting classes were then aggregated and merged into four canopy density classes (high canopy density, medium canopy density, low canopy density and no canopy cover) based on visual inspection of the image and cross-checking with field data. Results of the hemispherical image analysis were used together with 76 ground truthing points for accuracy assessment and validation of the classified image.

2.3.6. Modelling of species distribution Most of the modelling techniques used to define species-environment relationships rely on the identification of two observation sets: one that identifies locations in which species is present and one in which it is absent (Corsi et al., 2000; Hirzel et al., 2001; Phillips et al., 2006). Due to time constrains and the uncertainties with regard to generating acceptable species absence data from presence data (Clark et al., 1993), Genetic Algorithm for Rule- set Production (GARP) (Stockwell, 1999) and Maximum Entropy (Maxent ver 3.0.4 beta) (Phillips et al., 2006) modelling algorithms that require only presence data will be used for the modelling.

2.3.7. Selection of model species Species selected were based on a number of criteria: • Ecology of the species and its response to forest degradation. • Spread of species in the study area (common, moderately common and few) • Availability of data from butterfly sampling.

15 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

2.3.8. Development of Models The generated environmental layers for input into the models were converted to ASCII format with the same coordinate system and pixel size. An initial model was run by the Maximum Entropy Species Distribution Modelling software (Version 3.0.4 beta) with all the environmental layers for the selected species of butterflies. Thirty percent random points were reserved for testing of the model generated. A maximum of 10000 background points were specified with 500 maximum iterations, a regularization multiplier of 1 and a convergence threshold of 1x10-5. Feature types used included linear, quadratic, threshold and hinge. Response curves were also generated for the environmental layers (Appendix F). Through the Jackknife test, heuristic estimate of the relative contributions of the environmental variables to the maxent model were plotted to show variable importance. Based on the relative contribution of the environmental layers to any given model, the five best predictors for the model were selected and used to re-run the model for each of the species. In the Genetic Algorithm for the Rule-set Production model (GARP; DesktopGarp, cersion 1.1.6) two sets of models were run with all the environmental layers and the top five for each species. The following parameters and settings were used for the runs: Optimization parameters were 20 runs, 0.01 convergence limit and 1000 maximum iterations. The Atomic, Range, Negated Range and Logistic Regression rule types were activated. Seventy percent of the points were used for training. Ten (10) best subsets of models were selected based on a hard omission threshold with 10% omission. These ten subsets were then averaged in ArcGIS 9.2 to create the final probabilistic predictive maps for each species with a range between 0 and 1.

2.3.9. Evaluation of Models The Hawth Analysis Tools extension of ArcGIS 9.2 were used to generate random points used as pseudo-absences (Phillips et al., 2006) to extract an equal number of points for testing the accuracy of the training and testing datasets. Receiver Operating Characteristic (ROC)(Fielding and Bell, 1997) curves were used to measure and compare the accuracy of the predictive models as a single value by estimating the area under a curve (AUC) generated by plotting omission error (sensitivity) on the y-axis against commission error (1-specificity) on the x-axis (Phillips et al., 2006). In the ROC plots, used points are classified into two (“true” or “false”). The fraction of all positive instances classified “true” – also known as Sensitivity or the true positive rate – are plotted against the fraction of all negative instances classified “true” – also known as the false positive rate or 1-specificity – for all possible thresholds (Phillips et al., 2006). Same points were used to extract distribution predictions for the Maxent, GARP, all predictors and the best 5 predictors models so that their performance could be compared.

In all, methods employed included sampling of butterflies, their identification, collation of habitat variables and the integration of all these data in analysis and modelling algorithms for

16 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY mapping of selected species. Few challenges faced included the relatively short time that had to be used for butterfly collection, their identification and confirmation that the species have been rightly identified.

17 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

3. Results

3.1. Intermediate maps for mapping distribution of species. The classified image (Figure 3-1) had an accuracy of 76.32% and an overall Kappa statistic of 0.5693 (Appendix D). Four canopy density classes were distinguished; high (138.533 ha), medium (484.515 ha), low (120.533 ha) and no canopy cover (10.957 ha). The northernmost part of the study area encompasses the nature reserve and some research plots that have not been exposed to logging (NL).

Figure 3-1 Unsupervised Classification of Landsat 2007 image of study area

Generally, high canopy density classes are found along the banks of the rivers whilst the low canopy density and no canopy were near roads (Figure 3-1). The most degraded areas revealed by this classification is the southern portions of the study area which has current

18 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY logging activity (RL). The NRL block which comprise of the fairly degraded NW-SE belt towards the northern part of the study area and the central portions which have a sizeable amount of the high and medium canopy density.

Fourteen environmental layers used as input into the model algorithms for the generation of species distribution maps are shown in Appendix B.

3.2. Sampling Effort Table 3-1 shows the butterfly capture success rate and sampling effort standardization for the three forest strata. The recently logged forests had the highest number of captures per trap per day (1.08) while the not-logged forest recorded the lowest (0.82). The mean capture per effort was 0.97.

Table 3-1 Butterfly capture success rate per stratum and sampling effort standardization. Size Sampling Sampling Total Captures per Strata Traps (N) (ha) days (d) effort (N x d) captures trap per day NL 106.15 29 15 435 358 0.82 NRL 588.24 101 15 1515 1480 0.98 RL 189.08 30 15 450 484 1.08 Total 883.47 160 45 2400 2322 0.97 NL= Not Logged, NRL= Not Recently Logged, RL= Recently Logged.

The species accumulation curve for the study does not approach an asymptote (Figure 3-2). More species remain to be discovered with increasing sampling effort in the areas study area. Total species richness estimated by the second-order Jackknife estimator (Burnham and Overton, 1978; Burnham and Overton, 1979; Smith and van Belle, 1984) in EstimateS (Colwell, 2005) for mean runs is 170 species.

19 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

120

100

80

60 Species Observed

40 Cummulative number of Species

20

0 123456789101112131415 Sampling Effort (Days)

Figure 3-2 Species accumulation curve showing the relation between cumulative number of species recorded in all 160 traps against length of the sampling period (days).

3.3. Distribution of trapped butterfly species among butterfly families. The 2322 trapped butterflies belonged to three (3) families namely Nymphalidae, Pieridae and Lycaenidae with 99% belonging to the. This is not surprising as the traps targeted capturing of fruit feeding butterflies (Nymphalidae).

1800 60 1600 50 1400 1200 40 1000 30 800 600 20

400 Number of Species Number of Individuals 10 200 0 0 Biblidinae Heliconiinae Nymphalinae Charaxinae Limenitidinae Satyrinae Abundance 2 3 26 179 563 1546 No. of Species 2 1 4 205232 Subfamily

Figure 3-3 Number of Individuals and species recorded for the sub-families of the Nymphalidae

20 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

The abundance vis-à-vis number of species in the six (6) subfamilies under Nymphalidae is summarized in Figure 3-3. Satyrinae was the most abundant group with 1546 individuals (66.67%) recorded (Figure 3-3) representing 5 genera with the least abundant being Biblinidae (0.08%). The most species rich subfamily was Limenitidae with 52 species (46.85%) from eleven (11) genera with the least species rich being subfamily Heliconiinae (Figure 3-3).

3.4. Distribution of captured butterflies among forest disturbance classes One butterfly each was recorded for the Pieridae and Lycaenidae in the RL and NRL stratified forest disturbance classes respectively (Table 3-2). From the recently logged forest block (RL) the recorded individual, Leptosia hybrida (family Pieridae) is noted for seeking flowers fervently, congregating on forest paths with flowers (Larsen, 2005). From the total of 114 species recorded, 111 belonged to the Nymphalidae family (Table 3-2).

Table 3-2 Total number of butterflies and number of species recorded in the three forest classes Family Forest Disturbance Classes NL NRL RL Total Nymphalidae 364 (69) 1458 (96) 497 (69) 2319 (111) Pieridae 0011 Lycaenidae 0 1 0 1 Unidentified 0 1 0 1 Total 364 1460 498 2322 (114) ( ) no. of species, NL= Not Logged, NRL= Not Recently Logged, RL= Recently Logged.

Individuals from all the six (6) subfamilies of the Nymphalidae were recorded for the NRL and RL whilst NL had five (5) subfamilies represented (Table 3-3).

Table 3-3 Distribution of abundance of the Nymphalidae captured in the three forest categories over the five subfamilies Forest Disturbance Classes TAXON NL NRL RL Total Biblidinae (2) 0 [0.00] 1 [0.01] 1 [0.03] 2 Charaxinae (2) 47 [1.62] 110 [1.09] 22 [0.73] 179 Heliconiinae (1) 1 [0.03] 2 [0.02] 0 [0.00] 3 Limenitidinae (11) 110 [3.79] 361 [3.57] 92 [3.07] 563 Nymphalinae (4) 8 [0.28] 13 [0.13] 5 [0.17] 26 Satyrinae (5) 198 [6.83] 971 [9.61] 377 [12.57] 1546 Total 364 [12.55] 1460 [14.46] 498 [16.60] 2319 ( ) no. of genera represented, [ ] average number per trap, NL= Not Logged, NRL= Not Recently Logged, RL= Recently Logged. Taxa: FAMILY, Subfamily

21 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Although 971 members of the Satyrinae were captured in the NRL, the highest average per trap (12.57) was recorded in the recently logged (RL) with the least in the NL (6.83). The NL block however had the highest averages for Charaxinae, Heliconiinae, Limenitidinae and Nymphalinae (Table 3-3).

3.4.1. Butterfly species richness of forest disturbance classes Rarefied total species richness estimated (downscaled to 29 samples to permit comparison) for the blocks by the second order Jackknife estimator for the NL, NRL and RL forest disturbance classes were 105, 106 and 109 respectively.

The Michaelis-Menten [MMMeans] and Bootstrap richness estimators predicted lower total species richness for all the forest classes but the results from these two estimators were much comparable (Figure 3-4). The RL block had the highest number of species and individuals recorded. NL recorded the least number of individuals (358) in 29 samples but tied with RL on the number of species observed (70).

500 450 400

350

300 NL 250 NRL Count 200 RL 150 100 50 0 Number of Sobs (Mao Jack2 bootstrap MMMeans Individuals Tau) NL 358 70 105 81 89 NRL 425 65 106 77 81 RL 468 70 109 82 86 Spe ci e s Ri chne ss Estim a tors

Figure 3-4 Species Richness estimation for the forest blocks (Rarified: 29 samples) Sobs (Mao Tau)=Species Observed, Jack2= 2nd order Jackknife Estimator, bootstrap=bootstrap estimate, MMMeans=Michaelis-Menten richness estimator.

The NRL had the lowest number of observed species (65) in the 29 samples. The number of singletons (species that were recorded only once) and doubletons in all strata were almost the same 24 to 27 and 12 to 14 respectively (Appendix A). The singletons were twice as much as the doubletons in all the different strata.

22 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

3.4.2. Butterfly diversity of forest disturbance classes Fisher’s alpha diversity (diversity within sampled forest blocks) was highest in the recently logged forest (25.18) with the least being the not logged block (Table 3-4). Visual examination of the mean Shannon diversity index showed no clear differences between the three forest disturbance classes but the not logged (NL) block had the highest diversity with regard to the Simpson’s Inverse diversity index (16.5) with the lowest being the recently logged forest (14.21). When subjected to a normality test (Appendix C), not all the indices passed so a non- parametric test that does not assume that the data is sampled from a Gaussian distribution was employed. The Kruskal-Wallis test of variance by ranks revealed significant differences (P<0.0001) in all the three diversity indices for the forest blocks (Table 3-4). Due to the small P values observed (P<0.0001), it is concluded that differences in overall medians at p<0.05 are not due to coincidence but differences within each forest disturbance class.

Table 3-4 Kruskal-Wallis test of significance for diversity indices of the forest blocks Shannon-Weiner Simpson's Inverse Fisher's Parameter Index Index Alpha NL 3.4 16.5 24.29 Forest Block means NRL 3.43 14.36 25.18 RL 3.44 14.21 25.41 Kruskal-Wallis test P value P<0.0001 P<0.0001 P<0.0001 P value summary *** *** *** Do the medians vary signif. (P < 0.05) Yes Yes Yes Number of groups 3 3 3 Kruskal-Wallis statistic 49.63 106.4 87.95 ***= Extremely significant

Further post-tests (Dunn’s Multiple Comparison) revealed significant differences between each pair of forest blocks compared (NL vs NRL; NL vs RL and NRL vs RL) for all the diversity indices (Table 3-5). The difference in rank sum was very significant (P<0.01) for the NL and NRL in terms of the Shannon-Weiner index and extremely significant (P<0.001) for comparisons of all other indices.

Table 3-5 Dunn's Multiple Comparison Post Test of forest blocks per diversity index Indices Blocks compared Difference in rank sum P value Summary NL vs NRL 29.96 P < 0.01 ** Shannon-Weiner NL vs RL 82.66 P < 0.001 *** NRL vs RL 52.7 P < 0.001 ***

NL vs NRL 66.84 P < 0.001 *** Simpson's Inverse NL vs RL 124.3 P < 0.001 *** NRL vs RL 57.46 P < 0.001 ***

NL vs NRL 70.86 P < 0.001 *** Fisher's Alpha NL vs RL 110.8 P < 0.001 *** NRL vs RL 39.91 P < 0.001 *** ** = Very significant, ***= Extremely significant

23 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Table 3-6 shows the extent of overlap of the species observed when forest blocks were compared two at a time. The NL and RL had the highest (0.951) Morista-Horn Index as they shared 61 of the respectively 70 and 71 species observed in their samples. The forest blocks with the lowest similarity was the NRL and RL block which shared only 49 of the 99 and 71 species observed in their samples. The NL shared the most species with the NRL however the Morista-Horn Index was not the highest because the latter had about 30% of its species not shared with the former.

Table 3-6 Complementarities or similarity Index for forest blocks Number of Species Shared Species Morisita-Horn First Sample Second Sample First Sample Second Sample Observed Index NL NRL 70 99 64 0.896 NRL RL 99 71 49 0.825 NL RL 70 71 61 0.951

3.5. Distribution of captured butterflies among months. December recorded about 49% of the total number (2322) of butterflies captured in the three-month study. Increases in the number of butterflies were noted from October through November to December (Table 3-7). There was a three-fold increase in the numbers of the subfamily Satyrinae captured in December as compared to what was captured in October. This was accounted for by Bicyclus funebris recorded 0, 19 and 451 individuals in October, November and December respectively (Appendix A).

Table 3-7 Distribution of abundance of the Nymphalidae captured in sampled months from 160 trap stations. Months TAXON Oct Nov Dec Total NYMPHALIDAE Biblidinae (2) 1 0 1 2 Charaxinae (2) 55 86 38 179 Heliconiinae (1) 0 0 3 3 Limenitidinae (11) 156 197 210 563 Nymphalinae (4) 8 8 10 26 Satyrinae (5) 298 376 872 1546 Sub total 518 667 1134 2319 ( ) no. of genera represented

Individuals of the subfamily Nymphalinae and perhaps could be said to be the most stable as they did not vary much in the months sampled (Table 3-7). The Nymphalinae subfamily recorded just two more individuals in December whilst number of individuals recorded was same for the other months.

24 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

3.6. Species Rank Abundance. The top ten most abundant species sampled constituting 60% of the total number of species had at least 50 individuals trapped (Table 3-8). Bicyclus funebris was the most abundant with 20% of the total individuals, followed by Bicyclus sangmelinae and Gnophodes betsimena with 10.3% and 8.9% respectively. There were thirty-one (31) singletons (species with just one member collected) and seven (7) doubletons (appendix A). Thirty-nine species were found to be unique because they were recorded in only one forest disturbance class. The 100 least abundant species only contributed to about 29% of the total number of species.

Table 3-8 Most abundant species collected with their distribution in the strata Forest Disturbance Class No. of traps Species Total NL NRL RL seen in Bicyclus funebris 46 313 111 470 130 Bicyclus sangmelinae 27 177 35 239 91 Gnophodes betsimena 22 119 65 206 103 Aterica galene 17 64 9 90 59 Bicyclus xeneas 14 46 24 84 59 Gnophodes chelys 12 46 14 72 52 Euphaedra medon 14 44 8 66 46 Charaxes protoclea 19 33 7 59 38 Euphaedra phaethusa 13 40 6 59 41 Bicyclus sandace 2 32 16 50 38

3.7. Modelling of Species Distribution. Species were ranked in terms of spread (number of traps seen in). Six (6) species were selected from various portions of the ranking for modelling (Table 3-9). These comprise two (2) each of common (greater than 100 traps), moderately common (between 50 and 60 traps) and the few (in less than 20 traps).

Table 3-9 Distribution and abundance of modelled species Spread (No. of trap Species Abundance Distribution Status stations recorded) Bicyclus funebris 130 470 Widespread Common Gnophodes betsimena 103 206 Widespread Common Euphaedra phaethusa 41 59 Moderate Moderately common Charaxes protoclea 38 59 Moderate Moderately common Palla ussheri 18 22 Restricted few Kallimoides rumia 17 20 Restricted few

25 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

3.7.1. Vetting of Predictors. Table 3-10 shows the percentage contribution of selected predictors for mapping the distribution of the species when all the environmental variables were used.

Table 3-10 Percentage contribution of environmental layers to maxent models obtained through the Jacknife test for variable importance. Bicyclu Gnophode s Charaxes Euphaedra s Kallimoides Palla Variable funebris protoclea phaethusa betsimena rumia ussheri Distance to rivers 4.96 10.93 28.19 3.98 1.27 16.02 EVI 35.90 23.06 29.42 45.92 7.86 0.00 Elevation 17.57 5.08 1.66 5.12 1.29 3.58 Distance to Roads 25.13 10.35 10.14 33.47 0.88 0.00 NDVI 0.54 1.86 0.52 0.28 85.95 47.48 SAVI 2.84 1.80 12.46 2.18 0.00 12.13 ASPECT 1.53 1.84 7.07 1.67 0.03 2.60 Classified Image 0.04 0.01 0.26 0.33 2.07 2.64 LAI 2.23 40.90 1.54 0.00 0.62 3.23

Net Radiation 4.48 0.67 0.00 0.00 0.01 0.00 Absorbed solar Radiation 3.32 0.88 4.56 6.27 0.03 2.64 Fraction of Photosynthetically 0.94 1.14 3.92 0.63 0.00 9.68 Active Radiation Slope 0.52 1.46 0.25 0.15 0.00 0.00 Percentages in bold contributed significantly to the maxent model.

Distance to rivers was an important factor in the distribution maps of all the six species of butterflies (Table 3-10). Enhanced Vegetation Index (EVI) and Elevation were significant contributors to the distribution models of five species, with distance to road contributing to four models. The remaining predictors contributed significantly (among the top five predictors) to at least one species distribution map with the exception of the slope layer which did not contribute significantly to any. The range of values for the slope was from 0 to 13 (Appendix B).

3.7.2. Results of Models Distribution maps of Maxent and GARP returned from the modelling process were continuous probability maps with a range of 0 to 100 and 0 to 10 respectively where pixels with lower probabilities represented areas with low habitat suitability (thus areas less likely for the species to be found there). Table 3-11 shows the predictive ability of the distribution maps (models) as a measure of Receiver Operating Characteristic (ROC) given by the Area Under the Curve (AUC) obtained by plotting 1-specificity (number of wrongly classified

26 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY points) on the x-axis and sensitivity (fraction of wrongly classified points) on the y-axis. The following scale was adopted. AUC = 0.5, no discrimination; 0.7 0.9, outstanding.

Table 3-11 Receiver Operating Characteristics (ROC) for predictive distribution maps All Predictors Best 5 Predictors Pseudo-random Pseudo-random Species Maxent Model AUC* absences AUC** Maxent Model AUC* absences AUC** Training Validation Maxent GARP Training Validation Maxent GARP Bicyclus funebris 0.92 0.83 0.86 0.56 0.91 0.83 0.84 0.62 Gnophodes betsimena 0.89 0.84 0.73 0.63 0.88 0.82 0.8 0.7 Euphaedra phaethusa 0.92 0.88 0.85 0.76 0.9 0.87 0.8 0.84 Charaxes protoclea 0.92 0.82 0.9 0.67 0.9 0.83 0.85 0.85 Kallimoides rumia 0.86 0.91 0.68 0.5 0.88 0.80 0.71 0.74 Palla ussheri 0.92 0.83 0.85 0.74 0.94 0.75 0.9 0.81 * Area Under the ROC Curve (AUC) generated by Maxent during model run. ** AUC generated using pseudo- random absence. Training = AUC generated by training dataset (70% of sample points). Validation = AUC generated by testing dataset (30% of sample points)

Models of distribution generated by Maxent from at least 10000 background points produced their own accuracies at two levels (Table 3-11) thus using the 70% training dataset and subsequently the 30% testing dataset. Eight (8) of these models developed based on 70% of the sample used as training points on the scale of Hosmer and Lemeshow (Hosmer and Lemeshow, 2000) were outstanding (AUC>9.0) whilst the remaining 4 were excellent (0.8

27 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

a b

c d Figure 3-5 Predicted geographical distribution of Bicyclus funebris in the study area a = Maxent: All predictors b= GARP: All predictors c=Maxent: Top 5 predictors d=GARP: Top 5 predictors.

Distribution maps for Bicyclus funebris (Figure 3-5a, b, c and d) generally had low probabilities of occurrence (habitat suitability) on roads which was the second most important environmental variable contributing to the gains in AUC. There was very high probability of occurrence within 50 meters from the roads, decreasing with increasing distance from the roads up to about 500 metres, but increasing again from a distance of 900m to 1400m from the roads (see response curves in appendix F). The nature reserve towards the north of the study area had high values for EVI (appendix B), the most important predictor to the

28 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY distribution of Bicyclus funebris (Table 3-10) so it is not surprising that this area presents a high probability of occurrence for the species.

a b

c d Figure 3-6 Predicted geographical distribution of Gnophodes betsimena in the study area a = Maxent: All predictors b= GARP: All predictors c=Maxent: Top 5 predictors d=GARP: Top 5 predictors.

Since the distribution map for Gnophodes betsimena predicted by pseudo-random absences for all predictor sets (Figure 3-6b) did not return an acceptable AUC (0.63), it is not discussed. Acceptable distribution maps for Gnophodes betsimena (Figure 3-6a, c and d) generally had low probability of occurrence on roads with high probabilities along river courses themselves but with lower probabilities with increasing distance from rivers. Quite similarly to the

29 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY. distribution of Bicyclus funebris, EVI and distance from roads were the most important predictors of their distribution therefore they showed to some extent similar probabilities for occurrence or absence. The predictive map of GARP for the best 5 predictors however had very low probability of occurrence more 350m from roads.

a b

c d Figure 3-7 Predicted geographical distribution of Charaxes protoclea in the study area a = Maxent: All predictors b= GARP: All predictors c=Maxent: Top 5 predictors d=GARP: Top 5 predictors.

Figure 3-7d is excluded from the description of species distribution because pseudo-random absence AUC for was not acceptable. Distribution maps of Charaxes protoclea (Figure 3-7a, b and c) show high affinity for rivers with highest occurrence probabilities within 200m then gradually declining in probability of occurrence with increasing distance from rivers. The

30 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY opposite is true for roads as increasing distance from existing roads rather favour occurrence of the species. Though not affected much by elevation, occurrence probability rapidly declines with increasing elevation from 230m (See Appendix F for response curves)

a b

c d Figure 3-8 Predicted geographical distribution of Euphaedra phaethusa in the study area a = Maxent: All predictors b=Maxent: Top 5 predictors c= GARP: All predictors d=GARP: Top 5 predictors.

Habitat suitability for Euphaedra phaethusa (Figure 3-8a, b, c and d) generally increased with increasing EVI values. This restricts the species to areas with high to medium canopy density (Figure 3-1) and limited suitability for areas with low and no canopy cover. There is low occurrence probability on rivers courses but higher probabilities near roads though not on the roads themselves.

31 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

a b

c d Figure 3-9 Predicted geographical distribution of Kallimoides rumia in the study area a = Maxent: All predictors b= GARP: All predictors c=Maxent: Top 5 predictors d=GARP: Top 5 predictors.

The distribution of Kallimoides rumia (Figure 3-9a, b, c and d) is restricted to high and to some extent medium canopy density areas with higher affinity for areas close to rivers. There is a steady increase in occurrence probability with increasing values of NDVI and EVI (see Appendix F for response curves). Low canopy density areas in the recently logged (RL) and roads present the least suitable conditions for their presence. This is depicted clearly by the models of GARP (Figure 3-9 b and d).

32 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

a b

c d Figure 3-10 Predicted geographical distribution of Palla ussheri in the study area a = Maxent: All predictors b= GARP: All predictors c=Maxent: Top 5 predictors d=GARP: Top 5 predictors.

The distribution of Palla ussheri (Figure 3-10a, b, c and d) is much restricted to some few meters from rivers (up to 250m) as probability of occurrence steadily declines with increasing distance from the rivers; inhabiting areas with high canopy density and clearly excluded from areas with low or no canopies at all. NL and NRL presents suitable habitat for this species.

In summary, the distribution maps for all the six species followed two general trends (Appendix E). Bicyclus funebris, Gnophodes betsimena and Euphaedra phaethusa had higher probability of distribution near existing roads and lower probabilities along river courses whilst the opposite was true for Charaxes protoclea, Kallimoides rumia and Palla ussheri

33

MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

4. Discussion

4.1.1. Captures per sampling effort Almost two-thirds of the total estimated species (170) in the study site was recorded. The successful capture rate (0.97) in the selected site was very low as compared to earlier research by Bossart et. al. (2006) who recorded 4.7 in the butterfly sanctuary. Since there was no difference in the sampling methodology used and the target butterflies, this could possibly be due to the differences in species richness between the two sites (wholly protected butterfly sanctuary and production forest) and the duration of the sampling as a five-year of study by DeVries and Walla (2001) in another tropical region reveals that short term sampling may underestimate the species diversity of the an area as month of sampling and even year of a study vary greatly. This not withstanding, since the same treatments or sampling procedure was used in the three forest disturbance classes, results are comparable and representative of the blocks in the study area and decisions from this will be valid.

4.2. Spatio-temporal variation of butterfly diversity and species richness Among other factors as cost, time and expertise needed for sampling, the mobility of faunal species especially migrants further complicates the assessment of species occurrence, diversity and richness (Kati et al., 2004; Leyequien et al., 2007). Variations observed are be categorised into two; spatial and temporal variations.

4.2.1. Spatial variation of butterfly diversity and species richness Comparison of the three blocks at different stages of regeneration revealed significant differences in butterfly diversity indices (Table 3-4) and species richness estimators. The RL had the highest average number of individuals per trap (16.60) and rarefied number of species (Figure 3-4) followed by the NRL and then NL (Table 3-3). This result confirms earlier assertions that extent of degradation affects the diversity of species also referred to in the ‘the intermediate disturbance theory’. In this theory environments with intermediate disturbances have been found to contain higher species numbers than those that are not; or very much disturbed (Basset et al., 1998; Beck and Vun Khen, 2007; Dumbrell and Hill, 2005; Fuller et al., 2007; Hamer et al., 1997; Koh, 2007; Rodriguez et al., 1998; Willott et al., 2000). RL had the highest Fisher Alpha diversity (diversity within samples). This is probably because the recently logged (RL) presents patches of degraded and primary forests attract more species than a homogenous block which has not been logged or has been extremely degraded (Willott et al., 2000). Although rarefied species observed for the Recently Logged (RL) equalled that of the Not Logged (NL) (Figure 3-4), diversity was higher for the not logged (NL) in terms of

35 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Simpson’s inverse diversity which takes into account the evenness of species captured per block (Bossart et al., 2006; Colwell, 2005). The NL also had the highest averages per trap for four of the six butterfly subfamilies namely Charaxinae, Heliconiinae, Limenitidinae and Nymphalinae whilst the RL topped in the average number of species per trap for the Satyrinae and Biblidinae. It is amply evident from the above that species diversity indices and richness vary in space with regard to stage of regeneration of a forest block. It is however surprising that the NL and RL were most similar in terms of the Morista-Horn Index which is related to the number of species shared by the two blocks(Table 3-6). The NRL and RL had the least similarity.

4.2.2. Temporal variation of butterfly diversity and species richness Abundance of butterflies varied with months (Table 3-7) although number of species did not. Members of the subfamily Satyrinae and Limenitidinae had increases from October to December (Table 3-7). October and November fall in the minor rainy season in southern Ghana and accounts for the disparity in species captured within these two months and December. Number of individuals of Bicyclus funebris and Gnophodes betsimena captured in December, a dry season, increased from the numbers recorded in October and November (Appendix A). A similar finding was made by Fermon (2002) in south east of Cote d’Ivoire where Bicyclus funebris and Gnophodes betsimena were the most abundant species in the dry season with the former not being captured in the wet season. It must however be noted that though number of individuals captured varied with months, no trend can be deduced due to the relatively short duration of the study (3 months) as tropical butterfly diversity is affected by seasonal variation (DeVries and Walla, 2001; Fermon, 2002).

4.3. Predictors of Species Distribution From the fourteen (14) environmental layers used as predictors of species distribution, Distance to rivers was an important factor in the distribution maps of all the six species of butterflies (Table 3-10). Enhanced Vegetation Index (EVI) and Elevation were significant contributors to the distribution models of five species, with distance to roads contributing to four distribution maps. The remaining predictors contributed significantly (among the top five) to at least one species distribution map with the exception of the slope layer which did not contribute significantly to any. Probably this could be due to its range of values from 0 to 13 which shows not much variation (Appendix B).

Environmental variables used were good predictors of species distribution as species distribution maps generated were generally outstanding or excellent and predicted better than prediction by chance. But among these predictors, distance to roads was the most relevant predictor being the best predictor for three species and the second best for two (Table 3-10). Enhanced Vegetation Index performed better than Normalised Difference Vegetation Index (NDVI) and Soil Adjusted Vegetation Index (SAVI) in predicting species distribution. This

36 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

result supports proposals of Huete et al. (1999) that EVI is an improvement over the NDVI as it minimises the soil background as well as atmospheric aerosol effects thereby optimising the vegetation signal in high Leaf Area Indices. EVI good predictor for the distribution maps of Bicyclus funebris, Charaxes protoclea, Euphaedra phaethusa, Gnophodes betsimena and Kallimoides rumia; placing second to NDVI only in maps of Palla ussheri and Kallimoides rumia. As NDVI and SAVI contributed significantly to two species distribution maps each, not much can be said about which one is better.

4.4. Species Distribution Maps This study presents the first successful attempt to map the distribution of a species in Bobiri Forest Reserve as well as giving general trends of choice of habitats for species relative to environmental predictors. The distributions of six species have been mapped with a great degree of success using presence data only with GARP and Maxent Modelling algorithms. The accuracy of model algorithms of GARP and Maxent differed with the latter giving better predictions due the fact that its outputs were a continuous probability from 0 to 100 distinguishing well between areas with marginal probability of species occurrence and highly suitable habitats whilst the former only classified habitats as being suitable or not. A visual comparison of the resulting GARP and Maxent distribution maps for the species should be interpreted with caution as locations judged highly suitable for GARP may only be marginally suitable when compared with the Maxent model (Phillips et al., 2006). It is however anticipated that as accuracies of both modelling algorithms were acceptable, these maps can be interpreted without problems.

Generally, maps of all predictors and the best five (Appendix E) similar with the exception of the distribution map of Gnophodes betsimena produced with the best 5 predictors in GARP. Distribution maps of all the selected species could be categorised into two, based on physical features which to some extent determined their distribution. Bicyclus funebris, Gnophodes betsimena and Euphaedra phaethusa generally had higher probability of occurrence with respect to distance from roads whilst Charaxes protoclea, Kallimoides rumia and Palla ussheri showed higher probability of occurrence with respect to distance from rivers (Appendix E). General description given by the Larsen (2005) for each of the species whose probability of occurrence varies with distance from roads are in agreement with the distribution maps as Bicyclus funebris, Gnophodes betsimena and Euphaedra phaethusa have been known to be “essentially forest butterflies centred on drier forests …; fairly common forest butterfly also colonising gallery forest and woodlands …; found in all forest types, tolerating habitat disturbance relatively well …” respectively. Though the distribution map (Figure 3-7) of Charaxes protoclea fits Larsen’s (2005) description thus “do not colonise highly disturbed areas forests”, the description of Kallimoides rumia and Palla ussheri to be “especially found along tracks, forest edges, small clearings but not found in deep forest…; frequently found in strongly degraded forests…” respectively do not fit exactly the distribution maps developed (Figures 3-9 and 3-

37 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

10). Further studies on the ecology of these species with respect to habitat variables such as roads, distance to river, etc is however necessary to ascertain the veracity of the statements with respect to the new information produced by the distribution maps of the species.

The top-most part of the study area (NL) had high probability of occurrence for all species. The opposite is true for the southern-most part (RL) of the study area which was generally had lower probability of occurrence for almost all species. This pattern could be attributed to the fact that the northern-most part of the study area presented better habitat suitability ranges which were significant in predicting species distribution (river, high EVI, LAI, SAVI and NDVI values) whilst same could not be said for the southern-most part of the study area.

38 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

5. Conclusion and Recommendation

5.1. Conclusion In conclusion, the following research questions were answered by the results and methods employed.

1 Does species abundance, distribution and species richness vary with month? Species richness, abundance and distribution vary with month. Some key examples were Bicyclus funebris, Gnophodes betsimena and Melanitis leda which increased in number of individuals captured from October, to December whilst Gnophodes chelys, Bicyclus sangemelinae, Charaxes cynthia and Charaxes bipunctatus had a reduction in number of individuals captured from October to December (Appendix A). The total number of individuals captured in October (518), November (667) and 1134 in December (Table 3-4) provides ample proof of variation of abundance with month but not much can be said about any seasonal trend since due to the short duration for data collection.

2 Does species diversity and richness vary with different levels of forest disturbance? Species diversity and richness vary with levels of forest disturbance as there were significant differences in Fisher’s alpha diversity, Shannon-Weiner Index and Simpson’s Inverse Diversity Index for the three blocks of forest disturbance classes when subjected to an analysis of variance test. The not logged (NL) had the highest Simpson’s inverse diversity with the recently logged (RL) having the least whilst the latter had the highest Fisher’s alpha which the former having the least.

3 Are the distribution maps of common butterfly species significantly different from each other? Predicted species distribution maps were significantly different from each other as no two species distribution maps had the same probabilities of habitat suitability across the study area and species had mostly different set of environmental predictors that predicted its distribution best (Table 3-9 and Appendix E).

4 Do EVI, NDVI and SAVI correlate with species richness; if so, which is the best? The vegetation indices (EVI, NDVI and SAVI) correlated with species richness as each of these was at least a major or relevant predictor in the model of a species’ distribution which were all better than predictions by chance or random predictions. However, EVI was better as it was a relevant predictor for five of the six species modelled with NDVI and SAVI being relevant for only two out of the six species modelled (Table 3-9).

39 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

5 Which of the measured environmental variables strongly determine species distribution? The top four (4) predictors include distance from rivers (relevant in all 6 species distribution maps), EVI and Elevation (relevant in 5 of the 6 species distribution maps) and distance from road (relevant in 4 of the 6 species distribution maps). EVI was however a very important predictor and performed creditably well being the best predictor for three species distribution maps and second best predictor in two.

6 How well can remote sensing be used to discriminate butterfly species richness of the logged and unlogged forest? With the exception of the distance from roads and distance from rivers predictors, all the other environmental layers used for predicting habitat suitability were products of remote sensing. As very high prediction results were achieved by distribution maps of species known to inhabit degraded areas and vice versa, it can be concluded that remote sensing can distinguish between logged and unlogged forests pretty well through the use of these species as bioindicators. Lack of discrimination between the logged and unlogged forests by the predictors would have led to distribution maps with poorer accuracies AUC<0.7 where models would be unacceptable or show no discrimination.

All three null hypotheses stated were found to be false. From the results of the study the following are true: 1. The extent of canopy closure or degradation measured through indices such as LAI, EVI, NDVI and SAVI influences the butterfly diversity. 2. The chosen environmental variables predicted butterfly diversity better than prediction by chance. The outstanding (AUC>0.9) and excellent distribution maps (0.8

5.1.1. Implications for conservation Although it is evident that the recently logged (RL) had higher numbers of captures and rarefied abundance of individuals than the not recently logged (NRL) and unlogged (NL), it was also obvious that it was the least diverse with regard to Simpson’s inverse diversity index – which integrates species richness and evenness as a measure of overall site heterogeneity – even though its number of species observed equalled that of the unlogged. Under the current system of selective logging in the studied production zones of Bobiri Forest Reserve (BFR),

40 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

which is aimed at reducing the amount of degradation and reduction of species richness that would have been caused by unrestricted extraction of timber by choice, care must be taken not to over exploit the resources available thereby changing the present primary forest to other habitat types by opening up the canopy. It is however heart-warming realising that only one individual of Hamanumida daedalus and other butterflies of purely disturbed areas which are not found in forests of good condition was collected during the study. It must however be noted that even though there may not be extinctions if parts of this forest were completely degraded, thoughts expressed by Larsen (2005) could become true where few species more suited for degraded areas and secondary forests greatly benefit at the expense of specialists found in forest in pristine conditions thereby altering the composition of species in the reserve. Every effort must therefore be made to keep the few remaining forest blocks and protected areas intact to prevent habitat loss thereby presenting viable habitats for maintaining the diversity of species that exist in them.

5.2. Recommendations The three-month sampling period for this study was short and this has the tendency of underestimating the species diversity of the study area. The research focused on trapping only fruit-feeding understorey butterflies which only represent a fraction of the total butterfly diversity in the sampled area. In a bid to discover the actual species richness of the sampled area, a more comprehensive study which considers vertical stratification and the capture of butterflies not lured by baited-traps should be considered. A further study of the plots used in this study and those used by Bossart et al. (2006) is recommended in order to answer questions on the differences in capture rates of butterflies in the production block of BFR and the butterfly sanctuary. Due to the seasonal variation of some tropical butterflies, it is also recommended that an extended study with data for at least three wet and dry season could be undertaken to ascertain if species in the study area were truly seasonal and to provide relevant maps for each season.

41 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

References

Asrar, G. (Editor), 1989. Theory and applications of optical remote sensing. J. Wiley & Sons, New York.

Baret, F. and Guyot, G., 1991. Potential limits of vegetation indices for LAI and APAR assessment. Remote Sensing of Environment, 35: 161-173.

Barlow, J.O.S., Overal, W.L., Araujo, I.S., Gardner, T.A. and Peres, C.A., 2007. The value of primary, secondary and plantation forests for fruit-feeding butterflies in the Brazilian Amazon. Journal of Applied Ecology, 44(5): 1001-1012.

Basset, Y., Novotny, V., Miller, S.E. and Springate, N.D., 1998. Assessing the impact of forest disturbance on tropical invertebrates: some comments. Journal of Applied Ecology, 35(3): 461-466.

Beaumont, L.J., Hughes, L. and Poulsen, M., 2005. Predicting species distributions: use of climatic parameters in BIOCLIM and its impact on predictions of species' current and future distributions. Ecological Modelling, 186(2): 251-270.

Beck, J.A.N. and Vun Khen, C., 2007. Beta-diversity of geometrid moths from northern Borneo: effects of habitat, time and space. Journal of Animal Ecology, 76(2): 230-237.

Bossart, J.L., Opuni-Frimpong, E., Kuudaar, S. and Nkrumah, E., 2006. Richness, abundance, and complementarity of fruit-feeding butterfly species in relict sacred forests and forest reserves of Ghana. Biodiversity and Conservation, 15: 333-359.

Bouyer, J., Sana, Y., Samandoulgou, Y., Cesar, J., Guerrini, L., Kabore-Zoungrana, C. and Dulieu, D., 2007. Identification of ecological indicators for monitoring ecosystem health in the trans- boundary W Regional park: A pilot study. Biological Conservation, 138(1-2): 73-88.

Brooks, T.M., Mittermeier, R.A., Mittermeier, C.G., da Fonseca, G.A.B., Rylands, A.B., Konstant, W.R., Flick, P., Pilgrim, J., Oldfield, S., Magin, G. and Hilton-Taylor, C., 2002. Habitat Loss and Extinction in the Hotspots of Biodiversity. Conservation Biology, 16(4): 909-923.

Burgess, N.D., Butynski, T.M., Cordeiro, N.J., Doggart, N.H., Fjeldsa, J., Howell, K.M., Kilahama, F.B., Loader, S.P., Lovett, J.C., Mbilinyi, B., Menegon, M., Moyer, D.C., Nashanda, E., Perkin, A., Rovero, F., Stanley, W.T. and Stuart, S.N., 2007. The biological importance of the Eastern Arc Mountains of Tanzania and Kenya. Biological Conservation, 134(2): 209-231.

Burnham, K.P. and Overton, W.S., 1978. Estimation of the size of a closed population when capture probabilities vary among animals. Biometrika, 60: 623-633.

Burnham, K.P. and Overton, W.S., 1979. Robust estimation of population size when capture probabilities vary among animals. Ecology, 60: 927-936.

Burnie, D. and Tschinkel, W.R., 2005. "Insect", Microsoft Encarta 2006 [CD]. Microsoft Corporation, Redmond, WA:.

42 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

Clark, J.D., Dunn, J.E. and G., S.K., 1993. A multivariate model of female black bear habitat use for geographic information systems. . Journal of Wildlife Management, 57(3): 519-526.

Cleary, D.F.R., Boyle, T.J.B., Setyawati, T. and Menken, S.B.J., 2005. The impact of logging on the abundance, species richness and community composition of butterfly guilds in Borneo. Journal of Applied Entomology, 129(1): 52-59.

Colwell, R.K., 2005. EstimateS: Statistical estimation of species richness and shared species from samples. Version 7.5 User's Guide and application published at http://purl.oclc.org/estimates.

Corsi, F., de Leeuw, J. and Skidmore, A., 2000. Modeling Species Distribution with GIS. In: Boitani, L. and Fuller, T. K. (Eds) [2000] Research Techniques in animal ecology: controversies and consequences. Columbia University Press, New York, 389-434 pp.

Daily, G.C. and Ehrlich, P.R., 1995. Preservation of biodiversity in small rainforest patches: rapid evaluation using butterfly trapping, . Biodiversity and Conservation, 4: 35-55. de Leeuw, J., Ottichilo, W.K., Toxopeus, A.G. and Prins, H.H.T., 2002. Application of remote sensing and geographic information systems in wildlife mapping and modelling. In: Skidmore, A. K. (Ed) [2002], Environmental modelling with GIS and remote sensing, Taylor & Francis, London, 268 pp.

DeVries, P.J., 1987. The butterflies of Costa Rica and their natural history, Papilionidae, Pieridae and Nymphalidae vol. 1. Princeton University Press, Princeton

DeVries, P.J., 2001. Butterflies. Encyclopedia of Biodiversity, 1. Academic Press.

Devries, P.J., Murray, D. and Lande, R., 1997. Species diversity in vertical, horizontal, and temporal dimensions of a fruit-feeding butterfly community in an Ecuadorian rainforest. Biological Journal of the Linnean Society, 62(3): 343-364.

DeVries, P.J. and Walla, T.R., 2001. Species diversity and community structure in neotropical fruit-feeding butterflies. Biological Journal of the Linnean Society, 74: 1 - 15.

Dumbrell, A.J. and Hill, J.K., 2005. Impacts of selective logging on canopy and ground assemblages of tropical forest butterflies: Implications for sampling. Biological Conservation, 125(1): 123-131.

Fermon, H., 2002. The butterfly community of a managed West African rainforest: patterns of habitat specificity, diversity, stratification and movement, Georg-August-Universität Göttingen.

Fielding, A.H. and Bell, J.F., 1997. A review of methods for the assessment of prediction errors in conservation presence/absence models. Environmental Conservation, 24(1): 38-49.

Fleishman, E., Fay, J.P. and Murphy, D.D., 2000. Upsides and downsides: contrasting topographic gradients in species richness and associated scenarios for climate change. Journal of Biogeography, 27(5): 1209-1219.

Frazer, G.W., Canham, C.D. and Lertzman, K.P., 1999. Gap Light Analyzer (GLA), Version 2.0: Imaging software to extract canopy structure and gap light transmission indices from true- colour fisheye photographs, users manual and program documentation. Simon Fraser

43 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

University, Burnaby, British Columbia, and the Institute of Ecosystem Studies, Millbrook, New York.

Freese, F., 1984. Statistics for Land Managers. Paeony Press, Edinburgh, Scotland.

Fuller, T., Sanchez-Cordero, V., Illoldi-Rangel, P., Linaje, M. and Sarkar, S., 2007. The cost of postponing biodiversity conservation in Mexico. Biological Conservation, 134(4): 593-600.

Gardner, T.A., Barlow, J., Araujo, I.S., Avila-Pires, T.C., Bonaldo, A.B., Costa, J.E., Esposito, M.C., Ferreira, L.V., Hawes, J., Hernandez, M.I.M., Hoogmoed, M.S., Leite, R.N., Lo-Man- Hung, N.F., Malcolm, J.R., Martins, M.B., Mestre, L.A.M., Miranda-Santos, R., Overal, W.L., Parry, L., Peters, S.L., Ribeiro-Junior, M.A., da Silva, M.N.F., da Silva Motta, C. and Peres, C.A., 2008. The cost-effectiveness of biodiversity surveys in tropical forests. Ecology Letters, 11(2): 139-150.

Goetz, S., Steinberg, D., Dubayah, R. and Blair, B., 2007. Laser remote sensing of canopy habitat heterogeneity as a predictor of bird species richness in an eastern temperate forest, USA. Remote Sensing of Environment, 108(3): 254-263.

Gotelli, N.J. and Entsminger, G.L., 2007. EcoSim: Null models software for ecology. Acquired Intelligence Inc. & Kesey-Bear. , Jericho, .

Gutierrez, D. and Menendez, R., 2007. Regional hotspots of butterfly diversity in a protected area: Are they indicators of unique assemblages and areas with more species of conservation concern? Acta Oecologica, 32(3): 301-311.

Hall, J.B. and Swaine, M.D., 1981. Distribution and ecology of vascular plants in a tropical rain forest: forest vegetation in ghana. W. Junk, The Hague, 383 pp.

Hamer, K.C., Hill, J.K., Lace, L.A. and Langan, A.M., 1997. Ecological and biogeographical effects of forest disturbance on tropical butterflies of Sumba, Indonesia. Journal of Biogeography, 24(1): 67-75.

Hawthorne, W.D. and Abu-Juam, M., 1995. Forest Protection in Ghana with Particular Reference to Vegetation and Plant Species. . . IUCN, Gland, Switzerland.

Hermy, M. and Cornelis, J., 2000. Towards a monitoring method and a number of multifaceted and hierarchical biodiversity indicators for urban and suburban parks. Landscape and Urban Planning, 49(3-4): 149-162.

Hirzel, A. and Guisan, A., 2002. Which is the optimal sampling strategy for habitat suitability modelling. Ecological Modelling, 157(2-3): 331-341.

Hirzel, A.H., Helfer, V. and Metral, F., 2001. Assessing habitat-suitability models with a virtual species. Ecological Modelling, 145(2-3): 111-121.

Hosmer, D.W. and Lemeshow, S., 2000. Applied logistic regression, 2nd Edn. Wiley, New York.

Houborg, R., Soegaard, H. and Boegh, E., 2007. Combining vegetation index and model inversion methods for the extraction of key vegetation biophysical parameters using Terra and Aqua MODIS reflectance data. Remote Sensing of Environment, 106(1): 39-58.

44 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

Huete, A., Justice, C. and van Leeuwen, W., 1999. MODIS vegetation index (MOD13). Algorithm Theoritical Basis Document (ATBD). NASA.

Huete, A.R., 1988. A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment, 25(3): 295-309.

Huete, A.R., Liu, H., Batchily, K. and van Leeuwen, W., 1997. A Comparison of Vegetation Indices Over a Global Set of TM Images for EOS-MODIS. Remote Sensing of Environment, 59(3): 440-451.

Kalacska, M., Sanchez-Azofeifa, G.A., Rivard, B., Caelli, T., White, H.P. and Calvo-Alvarado, J.C., 2007. Ecological fingerprinting of ecosystem succession: Estimating secondary tropical dry forest structure and diversity using imaging spectroscopy. Remote Sensing of Environment, 108(1): 82-96.

Kati, V., Devillers, P., Dufrene, M., Legakis, A., Vokou, D. and Lebrun, P., 2004. Testing the Value of Six Taxonomic Groups as Biodiversity Indicators at a Local Scale. Conservation Biology, 18(3): 667-675.

Koh, L.P., 2007. Impacts of land use change on South-east Asian forest butterflies: a review. Journal of Applied Ecology, 44(4): 703-713.

Larsen, T.B., 2005. Butterflies of West Africa. Apollo Books, Stenstrup.

Larsen, T.B., Aduse-Poku, K., Boersma, H., Sáfián, S. and Baker, J., 2007. Bobiri Butterfly Sanctuary in Ghana - Discovering its butterflies (with a checklist of the 930 butterflies of Ghana). Metamorphosis, 18(3): 87-126.

Leprieur, C., Verstraete, M.M. and Pinty, B., 1994. Evaluation of the performance of various vegetation indices to retrieve vegetation cover from AVHRR data. Remote Sensing Reviews, 10: 265-284.

Lewis, O.T., 2001. Effect of Experimental Selective Logging on Tropical Butterflies. Conservation Biology, 15(2): 389-400.

Leyequien, E., Verrelst, J., Slot, M., Schaepman-Strub, G., Heitkonig, I.M.A. and Skidmore, A., 2007. Capturing the fugitive: Applying remote sensing to terrestrial animal distribution and diversity. International Journal of Applied Earth Observation and Geoinformation, 9(1): 1-20.

Ministry of Environment and Science, 2002. National Biodiversity Strategy for Ghana.

Mutanga, O. and Skidmore, A.K., 2004. Narrow band vegetation indices overcome the saturation problem in biomass estimation. International Journal of Remote Sensing, 25(19): 3999 - 4014.

Nelson, S.M., 2007. Butterflies (Papilionoidea and Hesperioidea) as potential ecological indicators of riparian quality in the semi-arid western United States. Ecological Indicators, 7(2): 469-480.

Nichols, D.J., Wagner, M.R., Agyeman, V.K., Paul, B. and Cobbinah, J.R., 1998. Influence of artificial gaps in tropical forest on survival, growth, and Phytolyma lata attack on Milicia excelsa. Forest Ecology and Management, 110(1-3): 353-362.

Oindo, B.O. and Skidmore, A.K., 2002. Interannual variability of NDVI and species richness in Kenya. International Journal of Remote Sensing, 23(2): 285 - 298.

45 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Paloniemi, R. and Tikka, P.M., 2008. Ecological and social aspects of biodiversity conservation on private lands. Environmental Science & Policy, (2008) In Press, Corrected Proof.

Pettorelli, N., Vik, J.O., Mysterud, A., Gaillard, J.-M., Tucker, C.J. and Stenseth, N.C., 2005. Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends in Ecology & Evolution, 20(9): 503-510.

Phillips, S.J., Anderson, R.P. and Schapire, R.E., 2006. Maximum entropy modeling of species geographic distributions. Ecological Modelling, 190(3-4): 231-259.

Pressey, R.L., Watts, M.E. and Barrett, T.W., 2004. Is maximizing protection the same as minimizing loss? Efficiency and retention as alternative measures of the effectiveness of proposed reserves. Ecology Letters, 7(11): 1035-1046.

Raaijmakers, J.G.W., 1987. Statistical analysis of the Michaelis-Menten equation. Biometrics, 43: 793-803.

Ramos, F.A., 2000. Nymphalid butterfly communities in an amazonian forest fragment. . Journal of Research in Lepidoptera, 35: 29-41.

Richter, R., 2004. ATCOR User Manual. GEOSYSTEMS GmbH, Germering.

Rodrigues, A.S.L. and Gaston, K.J., 2002. Optimisation in reserve selection procedures--why not? Biological Conservation, 107(1): 123-129.

Rodriguez, J.P., Pearson, D.L. and Barrera, R.R., 1998. A test for the adequacy of bioindicator taxa: Are tiger beetles (Coleoptera: Cicindelidae) appropriate indicators for monitoring the degradation of tropical forests in Venezuela? Biological Conservation, 83(1): 69-76.

Sanders, H.L., 1968. Marine benthic diversity: a comparative study. American Naturalist, 102: 243-282.

Schlerf, M., Atzberger, C. and Hill, J., 2005. Remote sensing of forest biophysical variables using HyMap imaging spectrometer data. Remote Sensing of Environment, 95(2): 177-194.

Seto, K.C., Fleishman, E., Fay, J.P. and Betrus, C.J., 2004. Linking spatial patterns of bird and butterfly species richness with Landsat TM derived NDVI. International Journal of Remote Sensing, 25(20): 4309 - 4324.

Smith, E.P. and van Belle, G., 1984. Nonparametric estimation of species richness. Biometrics, 40: 119-129.

Stockwell, D., 1999. The GARP modelling system: problems and solutions to automated spatial prediction. International Journal of Geographical Information Science, 13(2): 143 - 158.

Summerville, K.S., Boulware, M.J., Veech, J.A. and Crist, T.O., 2003. Spatial Variation in Species Diversity and Composition of Forest Lepidoptera in Eastern Deciduous Forests of North America. Conservation Biology, 17(4): 1045-1057.

Thomas, C.D., Hanski, I., Ilkka, H. and Oscar, E.G., 2004. Metapopulation Dynamics in Changing Environments: Butterfly Responses to Habitat and Climate Change, Ecology, Genetics and Evolution of Metapopulations. Academic Press, Burlington, pp. 489-514.

46 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY

Thomson, J.R., Fleishman, E., Nally, R.M. and Dobkin, D.S., 2007. Comparison of predictor sets for species richness and the number of rare species of butterflies and birds. Journal of Biogeography, 34(1): 90-101.

Tole, L., 2006. Choosing reserve sites probabilistically: A Colombian Amazon case study. Ecological Modelling, 194(4): 344-356.

Tucker, C.J., 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment, 8(2): 127-150.

Ulrich, W. and Buszko, J., 2004. Habitat reduction and patterns of species loss. Basic and Applied Ecology, 5(3): 231-240.

Vanreusel, W. and Van Dyck, H., 2007. When functional habitat does not match vegetation types: A resource-based approach to map butterfly habitat. Biological Conservation, 135(2): 202-211.

Whittaker, R.J., Willis, K.J. and Field, R., 2001. Scale and species richness: towards a general, hierarchical theory of species diversity. Journal of Biogeography, 28(4): 453-470.

Wiegand, C.L., Richardson, A.J., Escobar, D.E. and Gerbermann, A.H., 1991. Vegetation indices in crop assessments. Remote Sensing of Environment, 35: 105-119.

Willott, S.J., Lim, D.C., Compton, S.G. and Sutton, S.L., 2000. Effects of Selective Logging on the Butterflies of a Bornean Rainforest. Conservation Biology, 14(4): 1055-1065.

47 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Appendices

Appendix A: Checklist of butterfly species caught. Months Forest Blocks TOTAL FAMILY Subfamily Tribe Genus Species Oct Nov Dec NL NRL RL Individuals PIERIDAE Pierinae Pierini Leptosia Leptosia hybrida 1 0 0 0 0 1 1 LYCAENIDAE Polyommatinae Polyommatini Chilades Chilades eleusis 0010 1 0 1 NYMPHALIDAE Biblidinae Eurytelini Ariadne Ariadne albifascia 1 0 0 0 0 1 1 Eurytela Eurytela hiarbas 0 0 1 0 1 0 1 Charaxinae Charaxini Charaxes Charaxes bipunctatus 6 2 1 0 7 2 9 Charaxes boueti 0 0 1 0 1 0 1 Charaxes brutus 1 0 0 0 1 0 1 Charaxes cedreatis 0 1 0 0 1 0 1 Charaxes cynthia 9 6 3 5 13 0 18 Charaxes etheocles 1 1 1 1 2 0 3 Charaxes eupale 1000 1 0 1 Charaxes fulvescens 1 4 10 1 11 3 15 Charaxes numenes 4 3 1 4 3 1 8 Charaxes pleione 2 0 4 1 5 0 6 Charaxes protoclea 9 43 7 19 33 7 59 Charaxes tiridates 5 1 1 0 6 1 7 Charaxes zingha 1 1 1 2 1 0 3 Pallini Palla Palla decius 0322 2 1 5 Palla publius 0 3 1 1 1 2 4 Palla ussheri 10 10 2 5 15 2 22 Palla violinitens 3 3 0 2 2 2 6

48 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Months Forest Blocks TOTAL FAMILY Subfamily Tribe Genus Species Oct Nov Dec NL NRL RL Individuals NYMPHALIDAE Heliconiinae Vigrantini Lachnoptera Lachnoptera anticlia 0 0 3 1 2 0 3 Limenitidinae Adoliadini Aterica Aterica galene 24 34 32 17 64 9 90 Adoliadini Bebearia Bebearia absolon 2 6 3 4 6 1 11 Bebearia cocalia 6 1 7 2 8 4 14 Bebearia lucayensis 0 2 3 1 4 0 5 Bebearia mandinga 1001 0 0 1 Bebearia mardania 11 7 7 8 13 4 25 Bebearia paludicola 2443 4 3 10 Bebearia phantasina 2 1 2 1 4 0 5 Bebearia sophus 6 19 21 7 27 12 46 Bebearia tentyris 8 15 8 5 21 5 31 Bebearia zonara 1 3 3 1 5 1 7 Adoliadini Catuna Catuna crithea 1 1 0 0 1 1 2 Euphaedra Euphaedra ceres 11 3 3 1 13 3 17 Euphaedra edwardsii 0120 1 2 3 Euphaedra eupalus 4 4 1 2 4 3 9 Euphaedra harpalyce 5 8 15 3 22 3 28 Euphaedra inanum 0 1 0 1 0 0 1 Euphaedra janetta 1 5 5 3 6 2 11 Euphaedra medon 18 16 32 14 44 8 66 Euphaedra perseis 0 3 0 0 1 2 3 Euphaedra phaethusa 12 22 25 13 40 6 59 Euphaedra splendens 0 0 1 0 1 0 1 Euphaedra themis 3860125 17

49 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Months Forest Blocks TOTAL FAMILY Subfamily Tribe Genus Species Oct Nov Dec NL NRL RL Individuals NYMPHALIDAE Limenitidinae Adoliadini Euriphene Euriphene amicia 1010 2 0 2 Euriphene aridatha 2 1 0 0 2 1 3 Euriphene barombina 3 1 0 1 2 1 4 Euriphene gambiae 4 1 0 1 3 1 5 Euriphene simplex 1 0 2 1 1 1 3 Euryphura Euryphura chalcis 0 0 1 0 1 0 1 Hamanumida Hamanumida daedalus 0 1 0 0 1 0 1 Limenitidini Cymothoe Cymothoe egesta 0 1 2 2 1 0 3 Cymothoe fumana 0 1 0 0 1 0 1 Cymothoe mabillei 0010 0 1 1 Harma Harma theobene 1 0 0 0 0 1 1 Neptis Neptis nysiades 0 0 1 1 0 0 1 Pseudacraea Pseudacraea lucretia 0 1 0 0 1 0 1 Satyrinae Elymniini Bicyclus Bicyclus abnomis 9 7 21 1 24 12 37 Bicyclus angulosa 0 9 1 1 2 7 10 Bicyclus funebris 0 19 451 46 313 111 470 Bicyclus istaris 2120 5 0 5 Bicyclus madetes 8 8 5 3 16 2 21 Bicyclus martius 6 19 12 7 20 10 37 Bicyclus procora 3 0 0 0 3 0 3 Bicyclus safitza 0 1 6 0 4 3 7 Bicyclus sandace 15 16 19 2 32 16 50 Bicyclus sangmelinae 126 71 42 27 177 35 239 Bicyclus taenias 1 2 2 2 2 1 5 Bicyclus vulgaris 1 3 13 2 11 4 17 Bicyclus xeneas 17 47 20 14 46 24 84

50 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Bicyclus zinebi 1 5 0 1 2 3 6

Months Forest Blocks TOTAL FAMILY Subfamily Tribe Genus Species Oct Nov Dec NL NRL RL Individuals NYMPHALIDAE Satyrinae Elymniini Elymniopsis Elymniopsis bammakoo 5 2 3 4 6 0 10 Hallelesis Hallelesis halyma 1 0 0 0 1 0 1 Melanitini Gnophodes Gnophodes betsimena 27 58 121 22 119 65 206 Gnophodes chelys 33 30 9 12 46 14 72 Melanitis Melanitis leda 4 11 26 4 26 11 41 Melanitis libya 0 1 3 2 1 1 4 Nymphalinae Nymphalini Antanartia Antanartia delius 0120 1 2 3 Junoniini Hypolimnas Hypolimnas salmacis 0 0 2 1 1 0 2 Salamis Salamis cacta 1 0 0 1 0 0 1 ?? Kallimoides Kallimoides rumia 7 7 6 6 11 3 20 ?? Status uncertain.

51 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Appendix B: Input maps of environmental variables.

Elevation Slope Aspect

Absorbed Solar Radiation Net Solar Radiation Fraction of Photosynthetically Active Radiation

EVI NDVI SAVI

52 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Classified Image of study area Leaf Area Index

Distance from Rivers Distance from Roads

53 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Appendix C: Results of statistical analysis

Normality Test for Indices per forest disturbance block Shannon-Weiner Index Simpson Index Fisher's Alpha NL NRL RL NL NRL RL NL NRL RL Number of values 29 101 30 29 101 30 29 101 30 Mean 3.338 3.2483.037 24.02 13.51 12.13 26.02 22.38 20.58 Std. Deviation 0.3395 0.2027 0.253 1.035 0.6974 0.7196 0.6079 1.723 2.028 Std. Error 0.06304 0.02017 0.0462 0.1923 0.06939 0.1314 0.1129 0.1715 0.3702

Normality Test KS distance 0.2386 0.3068 0.2336 0.3313 0.2502 0.248 0.2655 0.211 0.1392 P value 0.0737 P<0.0001 0.0757 0.0034 P<0.0001 0.0499 0.0335 0.0002 P > 0.10 Passed normality Yes No Yes No No No No No Yes test (alpha=0.05)? P value summary ns *** ns ** *** * * *** ns

Coefficient of variation 10.17% 6.24%8.33% 4.31% 5.16% 5.93% 2.34% 7.70% 9.85% ns = Not significant * = Significant ** = Very significant ***= Extremely significant

Kruskal-Wallis test: Shannon-Weiner Index Parameter Value P value P<0.0001 Exact or approximate P value? Gaussian Approximation P value summary *** Do the medians vary signif. (P < 0.05) Yes Number of groups 3 Kruskal-Wallis statistic 49.63

Dunn's Multiple Comparison Test Difference in rank sum P value Summary NL Shannon vs NRL Shannon 29.96 P < 0.01 ** NL Shannon vs RL Shannon 82.66 P < 0.001 *** NRL Shannon vs RL Shannon 52.7 P < 0.001 *** ** = Very significant ***= Extremely significant

54 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Kruskal-Wallis test: Simpson's Inverse Diversity Parameter Value P value P<0.0001 Exact or approximate P value? Gaussian Approximation P value summary *** Do the medians vary signif. (P < 0.05) Yes Number of groups 3 Kruskal-Wallis statistic 106.4

Dunn's Multiple Comparison Test Difference in rank sum P value Summary NL Simpson vs NRL Simpson 66.84 P < 0.001 *** NL Simpson vs RL Simpson 124.3 P < 0.001 *** NRL Simpson vs RL Simpson 57.46 P < 0.001 *** ***= Extremely significant

Kruskal-Wallis test: Fisher's Alpha Parameter Value P value P<0.0001 Gaussian Exact or approximate P value? Approximation P value summary *** Do the medians vary signif. (P < 0.05) Yes Number of groups 3 Kruskal-Wallis statistic 87.95

Dunn's Multiple Comparison Test Difference in rank sum P value Summary NL Alpha vs NRL Alpha 70.86 P < 0.001 *** NL Alpha vs RL Alpha Mean 110.8 P < 0.001 *** NRL Alpha vs RL Alpha Mean 39.91 P < 0.001 *** ***= Extremely significant

55 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Appendix D: Accuracy assessment of unsupervised classification. Unsupervised Classification - Error Matrix Reference Data Classified Data Class 1 Class 2 Class 3 Class 4 Class 1 8 0 0 4 Class 2 0 6 0 0 Class 3 2 0 4 2 Class 4 8 0 2 40 Column Total 18 6 6 46

Unsupervised Classification Accuracy Totals Reference Classified Number Producers Users Class Name Totals Totals Correct Accuracy Accuracy Class 1 18 12 8 44.44% 66.67% Class 2 6 6 6 100.00% 100.00% Class 3 6 8 4 66.67% 50.00% Class 4 46 50 40 86.96% 80.00% Totals 76 76 58

Overall Classification Accuracy = 76.32%

Kappa (K^) Statistics

Overall Kappa Statistics = 0.5693

Conditional Kappa for each Category. Class Name Kappa Class 1 0.5632 Class 2 1 Class 3 0.4571 Class 4 0.4933

56 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Appendix E: Probabilistic predictive maps of Maxent and GARP for the selected species of butterflies.

Maxent Maxent GARP Garp (All predictors) (Top 5 predictors) (All predictors) (Top 5 predictors)

Bicyclus funebris

Gnophodes betsimena

Euphaedra phaethusa

Charaxes protoclea

Kallimoides rumia

Palla ussheri

57 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Appendix F: Response curves for best five predictors of maxent model

Bicyclus funebris Charaxes protoclea Euphaedra phaethusa Gnophodes betsimena

Omission Omission Omission Omission

ROC ROC ROC ROC

Elevation Elevation Aspect Aspect

Distance to rivers Distance to River Distance to River Distance to River

Distance to road Distance to Road Distance to road Distance to Road

Enhanced Veg. Index Enhanced Veg. Index Enhanced Veg. Index Enhanced Veg. Index

Net Radiation Leaf Area Index Soil Adjusted Veg. Ind. Soil Adjusted Veg. Ind.

58 MAPPING DISTRIBUTION OF BUTTERFLIES IN CENTRAL BOBIRI FOREST RESERVE AND INVESTIGATION OF LOGGING AND STAGE OF REGENERATION ON BUTTERFLY SPECIES RICHNESS AND DIVERSITY.

Kallimoides Rumia (KU) Palla Ussheri (PU) Jackknife AUC

Omission Omission Bicyclus funebris

ROC ROC Charaxes protoclea

Classified Image Elevation Euphaedra phaethusa

Elevation FPAR Gnophodes betsimena

Distance to River Distance to River Kallimoides rumia

Enhanced Veg. Index Soil Adjusted Veg. Index Palla ussheri

NDVI NDVI

59