Drivers of species richness and abundance of butterflies in coffee–banana agroforests in Uganda

Abstract

A study was conducted in 26 sites on agricultural landscapes in Central Uganda to collect baseline information about important drivers of butterfly richness and abundance. Data were collected for 1 year (2006) using line transects walk-and-counts, fruit-bait traps and handnets sampling methods. A total of 57,439 individuals belonging to 331 species were collected. Totals of 127, 131 and 299 species were recorded in transect counts, banana-bait and handnets, respectively. Of the 57,439 individuals registered, 75%, 19% and 6% were recorded in transect counts, handnet and banana-bait trap, respectively. Butterfly abundance and species richness were significantly (p < 0.05) affected by climatic factors (rainfall, temperature) in previous years (2004 and 2005) and richness and abundance of wild nectaring plants. Butterfly species richness (not the abundance) decreased with land-use intensity (p < 0.05) and was positively related to the cover of semi-natural habitats. Both butterfly species richness and abundance declined sharply with forest distance. Nearby forest remnants and high cover of semi-natural habitats are thus important for conservation of butterflies in coffee–banana agroforestry systems and farmers should be encouraged to protect such resources.

Keywords:

• butterfly diversity
• drivers
• farmlands
• Lepidoptera conservation
• habitats protection
• East Africa

Introduction

Butterflies have been recognized as a useful biodiversity indicator group of tropical land-use systems because they are sensitive and react quickly to subtle changes in environmental and habitat conditions (Kremen 1994; Libert 1994; Brown 1997; Larsen 2008; Özden et al. 2008; Pozo et al. 2008). Due to their short life cycle, narrow niches and relatively low mobility, they are more sensitive to land-cover and land-use changes than long-lived animals. Butterflies are relatively easy to capture, manipulate and identify (Rogo and Odulaja 2001; Fitzherbert et al. 2006; Marin et al. 2009; Nyamweya and Gichuki 2010), which makes them important candidates for monitoring changes in habitat, biodiversity and environmental conditions (Kremen 1992; Howard et al. 2000; Cleary 2004; Cleary and Mooers 2004), including the impact of landscape and habitat management practices and disturbance regimes in terrestrial ecosystems (Stork et al. 2003; Öckinger and Smith 2008).

Butterfly activities are closely controlled by weather and many species are constrained by climate, mostly occupying a small part of the range of their host plants (Fitzherbert et al. 2006; Nyamweya and Gichuki 2010). Butterflies play a significant ecological role in agricultural landscapes. They perform essential ecosystem services (Rogo and Odulaja 2001; Schmidt and Roland 2006), especially in the recycling of nutrients (N, P, K) highly needed by crops. In their larval stage, butterflies feed on leaves of several wild plants found in the agricultural systems and therefore release faeces that contain large amounts of nutrients (Munyuli 2010). In addition, butterflies are food to birds and other predators and are hosts to several parasitoids that suppress crop pests (Summerville et al. 2001; Cardoso et al. 2008, 2009). Consequently, their conservation is essential to sustaining the productivity of natural and agricultural landscapes. Despite their diversity, ubiquity and ecological importance, butterflies remain relatively little studied, particularly with regard to their ecology, behaviour and functional role in farmland habitats (Marchiori and Romanowski 2006; Stireman et al. 2009). In agricultural systems, butterflies are suspected to be important pollinators of wild and cultivated crop species (Munyuli 2010). Roughly, 90% of butterfly species live in the tropics (Boriani et al. 2005; Bonebrake et al. 2010). However, knowledge of butterflies inhabiting farmland habitats is fairly good in Mediterranean regions compared with the sub-Saharan Africa. The relative scarcity of data on tropical butterfly populations hampers the ability to effectively conserve them, particularly as pollinating agents (Bonebrake et al. 2010) in agricultural systems.

Across sub-Saharan Africa, surveys of drivers of butterfly diversity in farmlands are rare in the literature. There are a few published papers quantitatively documenting the diversity of butterflies for some agricultural regions in East Africa. In Uganda, most of the works on butterflies have been carried out mainly in natural areas, forest ecosystems and in protected areas (Tumuhimbise et al. 1998; Howard et al. 2000; Molleman et al. 2006; Tushabe et al. 2006). There exist no published data describing extensively the diversity of butterflies found in agricultural landscapes in Uganda in relationship to climatic, regional, landscape and local drivers. However, such information is important for butterfly biodiversity conservation in the rural landscapes.

While several studies (Chay-Hernández et al. 2006; Kivinen et al. 2008; Pickens and Root 2008; Dover and Settele 2009) indicate that climatic, regional and landscape factors are important drivers for butterflies in farmland areas, other studies indicate that local factors (availability of nectaring resources) are of foremost importance in explaining the variation in species richness and total density of butterflies in farmland regions (Pöyry et al. 2009). It is not clear which local, landscape, regional and climatic factors are important predictors of butterfly species richness and abundance in Central Uganda.

The aim of this study was to explore the relationships between butterfly assemblages (species richness and abundance) and local, landscape, regional and climatic drivers.

Materials and methods

Study area and selection of study sites

This study was conducted in 26 different study sites in the banana–coffee system of Lake Victoria Arc, covering several districts of Central Uganda (Munyuli 2009a, 2009b, 2009c, 2011a, 2012; Munyuli et al. 2009). The Lake Victoria Arc is characterized by ferrisoils with high to medium fertility level and receives on average 1000–1800 mm of rains in a bimodal pattern (rainy seasons: March–May, September–November; semi-dry to dry seasons: June–August, December–February) with 22–28°C and 60–75% of mean annual temperature and relative humidity, respectively (Munyuli 2011b). Rainfall amounts and patterns are unpredictable. Several food and cash crops are grown in small-scale monoculture and/or polyculture fields that are integrated into the coffee–banana agroforest production systems where coffee and banana are the main crops (Munyuli 2011c). Rural Central Uganda is a mosaic landscape where ‘islands’ of patches of natural habitats (forest fragments, forest reserves, wetlands, woodlands) are found scattered within agricultural matrices dominated by linear and non-linear features of semi-natural habitats (fallows, hedgerows, grasslands, woodlots, rangelands). All study sites were situated at approximately 790–1000 m above sea level. A total of 26 sites were selected to represent a range of vegetation and habitat types of varying degrees of anthropogenic disturbance and management intensities. Selected study sites were grouped using human population density as a surrogate measure of agricultural intensity (Bolwig et al. 2006). Detailed vegetation, environmental and landscape characteristics of the 26 sites are presented in Munyuli (2011a).

Butterfly sampling

Field sampling of butterflies was conducted in each of the 26 study sites. In each study site, an area of 1 km² was selected and divided into five linear parallel transects 1000 m long and 200 m apart. Butterfly data were collected on one of the five transects per study site, and the selected transect was surveyed during five study site visits in 2006. Butterflies were sampled using three complementary methods (transects walk-and-counts, handnets and fruit-bait traps) universally recommended and extensively used to survey and monitor butterfly populations and communities (Pollard and Yates 1993; Raguso 1993; DeVries and Walla 2001; Kitahara and Sei 2001; Yahner 2001; Fermon et al. 2002; Kitahara 2004; Barlow et al. 2007; Kitahara et al. 2008; van Swaay et al. 2008; Marín et al. 2009; Vu 2009) in terrestrial ecosystems (Lehmann and Kioko 2005) in the tropics. These sampling methods have been applied in Uganda in previous studies (Owen 1971; Coe et al. 1999; Molleman et al. 2006; Akite 2008). Butterfly specimens were identified by consulting literature, nomenclature and coloured plates of butterflies and by using the reference collection available at Makerere University Zoology Museum. To avoid a mix up of the results, specimens collected by each of three sampling methods applied were identified separately. The taxonomic characteristics of butterflies were obtained from standard guides, including (i) Butterflies of Kenya (Larsen 1996), (ii) ‘Butterflies of West Africa’ (Larsen 2006) and (iii) ‘Butterflies of Uganda’ (Carder and Tindimubona 2002). Identification of all voucher specimens was confirmed by a butterfly taxonomist based at Makerere University Zoology Museum. All voucher specimens collected during the present research were deposited at Makerere University Zoology Museum.

Measurement of local, landscape, regional and climatic drivers

Local drivers

Local drivers (richness and abundance of floral nectaring resources) of importance for butterflies were measured following approaches described by Munyuli (2010). These included the percent cover of flowering plants (trees, shrubs, herbs and weeds), the number of nectaring plant species and the percentage cover of cultivated floral resources (proportion of cultivated pollinator-dependent and non-pollinator-dependent crops) per 1 km² area. Data on herbs were collected from 5 m × 5 m (25 m²) quadrats, while those on shrub and trees were collected from 10 m × 50 m (500 m²) quadrats. A total of 2025 m² quadrats and 20,500 m² quadrats were established per transect per study site.

Landscape drivers

The area size of different land-use types in each 1 km² site was determined using a global positioning system (GPS) or using a tape measure for measurement of the dimensions of small fields (<50 × 50 m). Different land uses were grouped into types. Three landscape variables of high ecological importance for butterfly community studies in agricultural matrices were determined for each site: (i) the percentage of semi-natural habitat per square kilometre area; the term semi-natural habitats included all linear and non-linear semi-natural habitats such as fallows, hedgerows, field margins, grasslands, roadsides, woodlands, woodlots, track sides and stream edges. Because fallows play a particular role (as butterfly reservoirs and as foraging/breeding habitats) in the maintenance of many butterfly species in rural landscapes, the proportion of cover (%) of young fallows per site was calculated alongside the proportion of cover of semi-natural habitats; (ii) the percentage of the land area cultivated; and (iii) the forest distance, that is, the distance from a given study site to the nearest potential pollinators’ source. Distances up to 100 m were measured with a tape measure and distances greater than 100 m were measured with GPS (Garmin International, Olathe, KS, USA; corrected to ± 1 m accuracy with Pathfinder v 2.0).

Regional land-use intensity factors

Different land-use categories were obtained from the Makerere University Geographic Information Service that classifies land use around Lake Victoria into four categories: low, medium, high and very high. Land uses classified as low are areas where three-quarters of the land are maintained uncultivated compared with one-quarter that is under crop/animal production. Land uses classified as medium are managed habitat types where there is an almost equal amount of land cultivated and uncultivated, whereas land uses classified as high are areas dominated by crops or livestock and land uses classified as very high represent large monocultures (estates of tea, sugar, coffee, etc.).

Climatic drivers

Data on regional climatic factors were obtained from meteorological stations located in the study area (Kamenyamigo, Entebbe, Jinja and Kiige meteorological stations). Using raw data from the stations, monthly means for 10 years (1997–2006) for temperature and rainfall (Figure 1) were calculated to detect trends in the rainfall patterns and temperature, since such oscillations may affect the patterns (trends in occurrence and activities) of butterfly species richness and abundance in rural landscapes. Later, other variables (of relevance to the study of patterns of butterfly communities in farmlands) were calculated: (i) the overall mean rains (means/month/10 years); (ii) the overall daily mean minimum and maximum temperatures (mean of 10 years); (iii) the mean monthly rainfall (2004, 2005, 2006); (iv) the mean monthly maximum temperature (2004, 2005, 2006); and (v) the mean monthly minimum temperature (2004, 2005, 2006).

Figure 1. General pattern in temperature and rainfall data in the coffee–banana system, Central Uganda. Note: Data are means of five meteorological stations and means of 10 years: 1997–2006.

Data analysis

Data from the three sources (transect walk-and-counts, banana-bait traps and handnets) were pooled to obtain total butterfly biodiversity per transect per study site per sampling round. Lumping data together helped to minimize variance associated with individual sampling methods (Bossart et al. 2006). Butterfly abundance was calculated as the total number of individuals recorded per transect per study site each sampling day, whereas butterfly species richness was calculated as the total numbers of butterfly species recorded per transect per study site each sampling round. To determine the suite of explanatory variables most closely related to butterfly species richness and abundance measurements, a correlation analysis was conducted. Thus, correlation among dependent variables (butterfly species richness and abundance) and independent variables (climatic, local, landscape and regional variables) characterizing the 26 study sites was tested using Pearson correlation after transformation (log10 and arcsine square root transformation) of the raw data (Munyuli 2009a, 2009b, 2009c; Munyuli et al. 2009). Based on the correlation matrix of all variables measured, only non-collinear independent variables that were significantly (p < 0.05) related to dependent variables were selected for further analyses in simple regression analyses. These also illustrated the trends and magnitude of the effects of independent factors on dependent variables. Scatter plots were used to illustrate the scale dependency of butterflies on the different drivers. For all simple regression models obtained, the coefficient of determinations (R²) were plotted to demonstrate the level of influence of the type of drivers that correlate with the response variable (abundance and species richness of butterflies). The value of the R² was also used to identify variables with the most predictive effects. Thus, for all explanatory variables, the best regression model was selected based on the highest cross-validated R².

To analyse the effects of land-use categories on butterfly abundance and diversity, a general linear model analysis of variance was conducted with butterfly community variables (abundance, species richness) as the dependent variables, and the regional land-use categorical variables (low, medium, high and very high) as fixed factors. The least significant difference (LSD) tests were used as post hoc tests for multiple comparisons of means.

Results

Characteristics of butterfly communities recorded during the surveys

In total, 331 species and 95 genera were registered comprising a total of 57,439 individuals in 26 study sites. Genera with the greatest number of species were Acraea (10.8%), Charaxes (8.9%), Bicyclus (7.6%) and Neptis (5.7%). The most rich tribes were Acraeni (15.5%), Pierini (14.5%), Charaxini (12.7%), Mycalesini (10.9%) and Limenitidini (8.6%). The dominant (>5% of total individuals recorded) butterfly species were Acraea acerata, Bicyclus safitza, Catopsilia florella and Junonia sophia. These dominant and widespread species could be found in all study sites. The most abundant genera were Acraea (17.6%), Junonia (17.4%), Bicyclus (11.1%), Catopsilia (7.9%), Eurema (7.7%), Belenois (6.0%) and Ypthima (5.1%), whereas the most abundant tribes included Junoniini (22.6%), Acraeini (22.4%), Mycalesini (14.2%), Pierini (10.2%), Polyommatini (7.4%) and Satyrini (5.2%). Around 91.2% (331 observed species against 360 expected species) of the species present in the study area were found. The results suggested that a complete inventory of the regional species pool was obtained. Fifteen species were identified as the most ecologically important species (‘characteristic species’) of the coffee–banana system butterfly assemblages, since they occurred with high constancy value (>50%). These were C. florella, J. sophia, B. safitza, A. acerata, Eurema hecabe, Ypthima albida, Zizula hylax, Acraea ventura, Eurema brigitta, Neptidopsis ophione, Junonia eonone, Zizeeria knysna, Cupidopsis cissus, Junonia chorimene and Acrae uvui. They should serve as reliable and efficient indicators of change in conditions in agroecological environments in further surveys. They may also be good candidates for monitoring the state of butterflies of the farmland habitats in East Africa.

Local drivers of butterfly richness and abundance

The results of simple linear and non-linear regressions indicated that mass flowering wild plant resources (nectar flower abundance) were significantly positively related to both species richness (Figure 2A) and abundance (Figure 2B) of butterflies. Overall, mass flowering plants explained 48% and 62% of variations in butterfly species richness and abundance, respectively. Similarly, the number of nectaring plant species (all weeds, herbs, shrubs, trees) was positively related to both species richness (Figure 2C) and abundance (Figure 2D) of butterflies. The number of nectaring plant species accounted for 38% and 22% of variations in butterfly species richness. This result suggests that diverse flowering plant species stimulated attraction of a species-rich flower-visiting butterfly community in the farm landscapes. Unexpectedly, cultivated pollinator-dependent floral resources were not related to butterfly species richness. Similarly, the relationship between cultivated pollinator-dependent crops and butterfly abundance was not significant (p > 0.05), and this indicated that some butterfly species could not visit animal-pollinated crop species. Apparently, butterflies do not benefit much from available mass flowering pollinator-dependent crops. However, cultivated non pollinator-dependent floral resources were significantly negatively related to butterfly species richness (Figure 2E). The percentage of cultivated non-pollinator crops explained 21% of the variation in butterfly species richness. Butterfly abundance was not significantly (p > 0.05) related to the proportion of pollinator-dependent crops, although there was a negative trend. These results indicate that the occurrence of butterflies in the agricultural landscapes of Central Uganda was largely influenced by wild floral resources and by cultivated non-pollinator-dependent crops rather than by cultivated pollinator-dependent floral resources.

Figure 2. Relationship between (i) nectar flower abundance (percentage of mass flowering plants), (ii) the abundance of cultivated floral resources (percentage of cultivated non-pollinator-dependent crops), (iii) the number of flowering plant species (weeds, herbs, trees, shrubs) and the number of butterfly species (A, C, E) and butterfly abundance (B, D, F) per 2 ha transect/study site. Note: Pollinator-dependent crops included all annual, biannual and perennial non-entomophilous crops potentially offering nectar to butterflies.

Landscape drivers of butterfly richness and abundance

Cultivation intensity was significantly negatively related to butterfly species richness (Figure 3A), but not to butterfly abundance (Figure 3B). Proportion of cultivation intensity explained 48% of the variation in butterfly species richness. Similarly, the proportion of semi-natural habitats in a 1 km² area was significantly positively related to butterfly species richness (Figure 3C) and not related to butterfly abundance (Figure 3D). Overall, the proportion of semi-natural habitats accounted for 23% of the variation in butterfly species richness. These results indicated that the increase in amount of semi-natural habitats in the landscape was likely to lead to increased species richness (not the abundance) of butterflies in agricultural matrices of Central Uganda. Forest distance was significantly negatively related to both butterfly abundance (Figure 3F) and butterfly species richness (Figure 3E). Thus, the forest distance explained 69% and 27% of variations in butterfly species richness and abundance, respectively (Figure 3).

Figure 3. Relationship between landscape drivers (cultivation intensity, distance to forest margins and percentage of semi-natural habitats) per square kilometre area and butterfly species richness (A, C, E) and abundance (B, D, F).

Regional drivers of butterfly richness and abundance

There were no significant variations in butterfly abundance among the four land-use intensity categories (p > 0.05, Figure 4B), indicating that butterfly abundance was not very sensitive to the degree of agriculture intensification. However, there were significant effects of the land-use intensity on butterfly species richness (p < 0.05; Figure 4A) that were significantly higher in low to medium land-use intensity categories than in high and very high land-use categories (Figure 4A). Study sites located in the low land-use category had almost double the number of species.

Figure 4. Effects of the regional land-use intensity categories on the species richness (A) and abundance (B) of butterflies collected from farmlands of Central Uganda in 2006. Notes: Means (x ± SE) followed by different letters are significant at p < 0.05 according to the LSD test. GLM, general linear model; ANOVA, analysis of variance.

Climatic drivers of butterfly richness and abundance

There were significant correlations between climatic variables and butterfly community parameters (Table 1). Rainfall of the year 2004 was negatively correlated to both species richness (r = –0.47, p < 0.05, n = 26) and abundance (r = –0.49, p < 0.05, n = 26) of butterflies. Similarly, the mean rainfall of the year 2005 was significantly negatively correlated to both species richness (r = –0.48, p < 0.001, n = 26) and abundance (r = –0.53, p < 0.05, n = 26) of butterflies. However, butterfly abundance and species richness were not significantly (p > 0.05) associated with the mean rainfall of the year 2006 or the overall mean rainfall of 10 years (Table 1). In addition, the mean maximum temperature of 10 years was positively correlated to both species richness (r = 0.51, p < 0.05, n = 26) and abundance (r = 0.48, p < 0.05, n = 26) of butterflies. In addition, the mean butterfly abundance was positively correlated to both the maximum temperature of the year 2004 (r = 0.51, p < 0.05, n = 26) and the maximum temperature of the year 2005 (r = 0.46, p < 0.05, n = 26) (Table 1). These results indicated that current butterfly abundance and/or species richness were likely being affected by past events. In other words, butterfly communities found in farmlands of Central Uganda may be sensitive to climate change. Although rainfall is expected to affect positively the number and cover of wild and cultivated flowering plant species, no significant (p > 0.05) correlations were found between the abundance (and the diversity) of cultivated and wild floral resources and the mean rainfall of 10 years, or the mean rainfall of the year 2004 or 2005 (Table 1).

Discussion

Local, landscape and regional drivers of butterfly abundance and diversity

Among local drivers, the abundance and species richness of wild flowering plants significantly affected butterfly abundance and species richness. Both the number and cover of wild blooming plant species were more important local drivers for butterflies than the abundance of cultivated floral resources in the coffee–banana agroforest systems in Central Uganda. It is likely that cultivated floral resources (pollinator-dependent and non-pollinator-dependent crops) are not the primary food plants for the majority of butterfly species inhabiting farmlands probably because locally grown plants are less rich in nectar, the main floral resource collected by most butterflies foraging in agricultural landscapes. A significant strong positive relationship between butterfly species richness and nectaring plant species richness was found in Central Uganda. This result is consistent with Simonson et al. (2001) who found that butterfly species richness was positively correlated with total vascular plant species richness and native plant species richness in Colorado (USA). Worldwide, positive relationships between butterfly diversity and plant species richness are reported (e.g. Kumar et al. 2009): plant species serve as important food plants for larvae, flowers providing nectar resources for adult butterflies and the physical structure of plants create microhabitats providing shelter to numerous species of butterflies in rural landscapes. Thus, high plant species diversity is accepted to be an indicator of good habitat quality for generalist and specialist butterfly species (Kumar et al. 2009) inhabiting farmland habitats.

In this study, it was observed that butterfly species richness was sensitive to land-use intensity whereas the abundance was not, probably because very few adapted farmland species can easily be represented by a high number of individuals in the farmlands while the majority (>70%) of species may be ‘rare species’ represented by one to two individuals only. While studying the effect of agricultural intensification on the homogenization of Lepidoptera community in Finland, it was observed that the diversity of butterfly and day-active geometrid moth communities found within 134 differently fragmented landscapes decreased with increasing agricultural intensity with the consequence: promotion of habitat generalists in landscapes with moderate to intensive agriculture compared to region with low agriculture intensity (Ekroos et al. 2010).

Table 1. Cross-correlation matrix showing naïve multiple correlations of Lepidoptera (butterfly) variables with environmental, local and landscape variables

	B	C	D	E	F	H	I	J	K	L	M	N	O	P	Q	R	S	T	U	V	W	X
A	0.74**	−0.52*	−0.44*	0.46*	0.71**	−0.68**	−0.74**	−0.70**	−0.50*			0.64**							0.62**	0.57**
B	−	−0.49*	−0.57**	0.77**	0.66**	−0.94**	−0.95**	−0.97**	−0.46*										0.69**	0.49*		0.44*
C		−		−0.59**		0.52*	0.57**		0.61**	0.47*
D			−			0.57**	0.52*	0.55**	0.69**	0.78**									−0.50*
E				−		−0.78**	−0.79**	−0.72**											0.52*
F					−	−0.64**	−0.72**	−0.58**	−0.46*										0.45*
H						−	0.93**	0.91**	0.83**	0.52*									−0.58**
I							−	0.87**	0.47*										−0.67**	−0.46*		−0.45*
J								−											−0.67**	−0.49*
K									−	0.83**		−0.50*	0.51*								−0.47*	−0.48*
L										−			0.48*		0.51*	0.46*					−0.49*	−0.53*
M												−0.84**		0.84**	−0.76**	−0.96**	−0.85**	−0.75**		−0.79**	0.76**	0.67**
N												−		−0.69**	0.69**	0.85**	0.67**	0.56**		0.64**	−0.55**	−0.58**
O													−	−0.56**			0.64**	0.58**		0.43*	−0.77**
P														−	−0.93**	−0.95**	−0.93**	−0.52*		−0.47*	0.94**	0.93**
Q															−	0.90**	0.89**				−0.78**	−0.98**
R																−	0.93**	0.61**		0.63**	−0.84**	−0.84**
S																	−	0.60**		0.54**	−0.97**	−0.88**
T																		−	0.83**	0.97**	−0.64**
U																			−	0.87**
V																				−	−0.54**
W																					−	0.80**
X																						−

Notes: Different levels of significance of Spearman rank of correlation coefficients: *p < 0.05; **p < 0.001, otherwise not significant when no value is given. A, human population density; B, cultivation intensity; C, flowering plant species richness; D, density of plant species (abundance of floral resources); E, cultivated with floral resources (pollinator-dependent crops); F, cultivated without resources (pollinator non-dependent crops); H, forest distance; I, percentage of semi-natural habitats/site; J, percentage of young fallows per site; K, mean butterfly species richness; L, mean butterfly abundance; M, overall mean rainfall (10 years); N, overall daily mean maximum temperature (mean 10 years); O, overall daily mean minimum temperature (mean 10 years); P, mean monthly rainfall (2006); Q, mean monthly maximum temperature (2004); R, mean monthly maximum temperature (2005); S, mean monthly maximum temperature (2006); T, mean monthly minimum temperature (2004); U, mean monthly minimum temperature (2005); V, mean monthly minimum temperature (2006); W, mean monthly rainfall (2004); and X, mean monthly rainfall (2005).

The results from Central Uganda revealed that forest distance was the most significant landscape driver followed by the proportion of semi-natural habitats. Both the abundance and species richness of butterflies declined linearly with increasing forest distance. This result is consistent with observations recorded in India where a study on adult butterfly communities visiting coffee plantations around a protected area in the Western Ghats found that both butterfly species composition and species richness declined with increasing forest distance (Dolia et al. 2008) up to 2000 m. Distance of up to 1700 m from mature forests did not have any significant effect on butterfly species richness in a mosaic landscape in Central Sulawesi in Indonesia (Veddeler et al. 2005), whereas in Central Uganda, butterfly species richness declined sharply up to a distance of 2000 m. The fact that forest distance was negatively related to butterfly species richness and abundance suggests that surrounding natural forests of agricultural matrices remain an irreplaceable habitat for several butterfly species belonging to different functional groups (e.g. breeding strategy) and life history traits in Central Uganda. In addition, this observation suggests that the ability of different agricultural land uses (human-dominated landscapes) to support Lepidoptera communities can be enhanced if tree plantations/crop fields are established in proximity to surrounding primary forests (Hawes et al. 2009). Obviously, a contiguous distribution of large secondary forest patches with many host trees is very important to conserve butterfly species in fragmented agricultural landscapes (Kobayashi et al. 2009). The observation made in Central Uganda about the effect of forest fragments is consistent with the conclusion of Koh (2008) who clearly demonstrated that natural forests serve as an important population sources for butterfly species occurring in nearby agricultural matrices. In terms of habitat use, butterflies were found more in forest margins, forest fallows, wetlands, swampy habitats, woodlots, woodlands, grasslands and in adjacent fields (Munyuli 2011a) as compared with fields that were located far from such natural and semi-natural habitats. Forest is a primary factor dictating the occurrence of butterflies in agricultural landscapes of Central Uganda.

In this study, both the abundance and diversity of butterflies increased with increasing amount of semi-natural habitats per square kilometre. Similarly, in southeastern Sweden (Jonason et al. 2010), butterfly species richness (not the abundance) was found to be positively related to increasing tree cover in the farm landscape. Apart from land use and farm management methods or farming practices (Weibull and Östman 2003), vegetation structure, quality of the matrix surrounding an agricultural habitat (Binzenhöfer et al. 2008; Summerville et al. 2008), diversity and types of habitats (Dessuy and de Morris 2007; Ngai et al. 2008; Kumar et al. 2009), landscape heterogeneity and habitat connectivity (Davis et al. 2007) are important factors determining occurrence, movements, population dynamics, seasonality, persistence and long-term survival of Lepidoptera faunal communities in the agricultural landscapes (Dennis 2003; Greza et al. 2004; Chay-Hernández et al. 2006; Kivinen et al. 2008; Öckinger and Smith 2008; Pickens and Root 2008; Stasek et al. 2008; Dover and Settele 2009; Brückmann et al. 2010). Generally, agricultural matrices that are more resembling a nearby forest patch maintain higher butterfly diversity than matrices with lesser shade cover (Summerville et al. 2001; Kitahara and Watanabe 2003; Weibull and Östman 2003; Boriani et al. 2005; Aviron et al. 2007; Ohwaki et al. 2007, 2008; Barlow et al. 2008; Bergman et al. 2008; van Halder et al. 2008; Marín et al. 2009).

Climatic drivers of butterfly richness and abundance

It is generally accepted that climate factors regulate most insect species’ life cycles, including butterflies (Sparks and Yates 1997; Roy and Sparks 2000; Watt and McFarlane 2002). Climatic variables may therefore be very important in determining emergency (van Asch and Visser 2007; Saastamoinen and Hanski 2008), foraging, reproduction and breeding activities of many butterfly species in sub-Saharan Africa. In fact, seasonal flight activity, as well as species diversity and overall abundance vary greatly between sampling rounds (Munyuli 2010) in a study site. It is well established that temperature is a key factor increasing daily activities (flight, foraging movements) of butterflies, whereas rainfall affects indirectly butterfly through enhancing the availability of nectaring resources. Most butterfly species need resources like host plants (for larvae), nectar plants (for adults) and sites for resting, and both natural/semi-natural habitats and nectar plant resources are also affected by rainfall pattern (Öckinger et al. 2009; Rossi and van Halder 2010).

In this study, the abundance and species occurrence of butterflies were significantly related to rainfall and mean maximum temperature in previous years (2004, 2005), not the year of study (2006). The mean precipitation of the past 10 years was not significantly associated with any measure of richness and abundance of butterflies, but the mean maximum temperature in the same period was. The fact that butterfly community variables were related to mean maximum temperature and rainfall in previous years rather than current has also been found in temperate regions (e.g. Pollard 1988). In relation to rainfall, results obtained from Uganda coincide at least partially with results from Malaysia, where it was found (Brehm 2005) that monthly rainfall variation correlated with moth populations in previous months of the year, weather factors and plant phenology in the lowland dipterocarp forest. The intensity of tree flowering in the previous months was an important environmental factor that correlated positively with the numbers of species and individuals of moths that emerged in any subsequent month (Intachat et al. 2001). High rainfall in the previous 3 months led to an increase in moth abundance in the following month (perhaps by stimulating an increase in fresh plant material), whereas high rainfall and relative humidity thereafter served to decrease abundance, possibly by encouraging the spread and activity of pathogens that impacted on early life-stage survivorship (Intachat et al. 2001). It was concluded that the diversity of geometrid moths correlated more with weather parameters than with tree phenology (Intachat et al. 2001) in tropical regions, and important weather parameters that influenced moth abundance include monthly rainfall, relative humidity and minimum temperature in previous months. In addition, it is generally agreed that weather variables can have long-term effects on population fluctuations of different butterfly taxa (Sei-Woong 2003). Weather variables can help in detecting potential butterfly population changes following changes (Sei-Woong 2003) in climatic factors (changes in temperature regimes, rainfall patterns, relative humidity) that are believed to affect negatively/positively species richness, abundance and distribution of butterflies regionally as well as locally (Sei-Woong 2003; Kivinen et al. 2007). Although in Uganda the study was conducted on butterflies in agricultural systems, while in Malaysia the study was conducted on moths, the two studies were able to show that Lepidoptera biodiversity can be related to previous events (historical events) in weather factors.

Practically, weather factors affect the production of floral resources in current and previous years. High production of flowers in the previous wet months of the year may guarantee the availability of food for different butterfly taxa (larval and adult stages), and this in turn may tend to increase larval survivorship and therefore the abundance of adult butterflies. In some cases, the abundance of adult butterflies can increase with high abundance and diversity of flowering plants in the current and not in previous years. An increase in the number of host trees flowering would also mean an increased number of oviposition sites for the butterflies (particularly forest specialist species).

High rainfall in the months of the current year may deter butterflies from flying, thus resulting in a low catch in traps. High rainfall in the previous months of the year may also decrease the survivorship of the larvae, resulting in lower adult butterfly catch in the current month. During very rainy periods, larvae are more likely to be washed off their host plants. Combination of a wetter period and a high relative humidity in the months before adult emergence may also increase activities of disease microorganisms during that period, resulting in higher risk of larval mortality from disease and therefore reduction in adult abundance in the current month. Rainfall in the current month may also encourage disease at the pupal stage and cause mortality if heavy during the vulnerable eclosion period. High rainfall, on the other hand, can encourage blooming of wild and cultivated plants in most tropical agroecosystems for the benefit of butterflies (Munyuli 2010) and other Lepidoptera. Heavy blossoms in the previous months would encourage adult and larval survivorship and therefore increase the number of emergence adult butterflies in the following months of the following year.

In summary, if the annual and seasonal changes in rainfall and temperature continue to oscillate and if rainfall continue diminishing while temperature continue rising in Uganda (Munyuli 2010) this may, combined with already existing negative impacts of environmental degradation, lead to a high loss/decline in butterfly biodiversity while increasing the vulnerability of ecosystems relying on these keystone species to function properly.

Conclusion

Findings of this study indicated that the abundance and species diversity of butterflies in the coffee–banana agroforestry systems of Central Uganda were affected by local, landscape, regional and climate drivers. Plant diversity was the most influential local factor, whereas forest distance and cultivation intensity were the most influential landscape drivers. The presence of natural habitats (e.g. forest reserves, wetlands) and semi-natural habitats in the vicinity of agricultural matrices was related to high diversity of Lepidoptera. It was also found that butterfly communities were affected more by previous climatic events than current ones. Landscape managers should advocate for the maintenance of high-quality habitat matrix in agricultural mosaic farm landscapes to protect native butterfly faunal diversity and reduce their vulnerability to further environmental/climatic changes in the region. This could be supported by policies that enhance protection and restoration of forest fragments and promotion of interventions that enable farmers to adopt butterfly friendly conservation and farming practices, such as agroforestry systems with multipurpose tree species, to sustain butterfly communities and services in rural landscapes of Uganda.

To provide incentives to the conservation of butterfly fauna in agricultural landscapes in Uganda, further research is needed to understand the behavioural ecology of butterflies and to determine the pollination efficiency (Martins and Johnson 2009) of different butterfly species (as well as for other Lepidoptera taxa). Such studies will help in highlighting the importance of butterflies (and moths) as a group providing pollination services of high economic importance to both wild and cultivated plants in Uganda and in East and Central Africa.

Acknowledgements

I am very grateful to the Darwin Initiative (Defra, UK; project reference: 14-032; project title: Conserving biodiversity in modernized farm landscapes in Uganda) for funding this study. I am also very grateful to project leaders Dr Juliet Vickery (RSPB-Cambridge University, UK), Dr Phil Atkinson (British Trust for Ornithology, UK), Prof. Derek Pomeroy (Makerere University, Uganda), scientific supervisors Prof. Simon Potts (University of Reading, UK) and Prof. Philip Nyeko (Makerere University, Uganda). I am also grateful to Dr Bwinja M and to Mr Maurice Mutabazi (research assistant) for his assistance in the field. I am grateful to Akite Perpetra who helped during identification of butterfly to species level at Makerere University Zoology Museum. I am also grateful to anonymous reviewers of this journal for constructive criticisms and comments on earlier versions of this article.