Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Austral Ecology (2010) ••, ••–•• Disturbance, species loss and compensation: Wildfire and grazing effects on the avian community and its food supply in the Serengeti Ecosystem, Tanzania aec_2167 1..10 A. K. NKWABI,1 A. R E. SINCLAIR,1,2* K. L. METZGER,1,2 AND S. A. R. MDUMA1 Serengeti Biodiversity Program, Tanzania Wildlife Research Institute, Arusha, Tanzania; and 2Centre for Biodiversity Research, University of British Columbia,Vancouver, British Columbia, v6t 1z4, Canada (Email: [email protected]) 1 Abstract An important question in biodiversity studies is whether disturbances in ecosystems will cause a net loss of species or whether such losses can be compensated by replacement of other species. We use two natural disturbances, fire and grazing, to examine the response of bird and arthropod communities in grasslands of Serengeti, Tanzania. Both burning and grazing by migrant ungulates take place at the end of the rains in June–July. We documented the communities before disturbance, then 1, 4 and 20 weeks after disturbance on three replicate plots and compared them with three undisturbed plots. Birds were recorded by observation, arthropods from pitfall, tray trap and sweepnet samples. We expected that as the grass biomass was reduced by either disturbance, bird communities would change with concomitant change in arthropod food abundance. Alternatively, bird communities would change not with the absolute amount of food but with the greater accessibility of food as the grass structure changed from long to short grass. Results showed first that both bird species richness and abundance increased after both types of disturbance, but burnt sites showed a greater increase than that for grazed sites. Second, there was a change in bird species composition with disturbance. The functionally equivalent athi short-toed lark (Calandrella athensis) was replaced by the red-capped lark (Calandrella cinerea). Third, the abundance of most groups of arthropods was lower on disturbed sites than those on undisturbed sites, and the reduction of arthropod numbers was greatest on burnt sites. These results imply that bird abundance did not occur through an increase in arthropod abundance but rather through a change in the grass structure making food more accessible; and the higher predation could have caused the lower arthropod abundance. In addition, some bird species replaced others thus functionally compensating for their loss. Key words: arthropods, compensation, ecosystem function, grasslands, Serengeti birds. INTRODUCTION An important question in biodiversity studies is whether disturbances in ecosystems will cause a net loss of species or whether such losses can be compensated by replacement of other species. In this paper we document the impacts of two natural disturbances, fire and grazing, on the diversity and abundance of birds and arthropods in the Serengeti grassland ecosystem. The Serengeti ecosystem supports some 640 species of birds, of which the greatest number are insectivores. Grass fires have dominated African savanna landscapes for hundreds of millennia (Bird & Cali 1998) and plants are fire tolerant. The savanna plant community has been shaped by the frequency of burns (Frost & Robertson 1987). Hence, the vegetation in savanna Africa has become adapted and shaped by fire *Corresponding author. Accepted for publication June 2010. © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia so that it can be considered as a natural disturbance (van Langevelde et al. 2003), and this applies in the Serengeti ecosystem (Norton-Griffiths 1979; Sinclair & Norton-Griffiths 1979; Stronach 1988; Sinclair et al. 2007). Fire is frequent on the Serengeti long grass plains but not on the short grass plains where migratory grazers concentrate in the wet season, preventing the accumulation of combustible fuel loads (Norton-griffiths 1979; McNaughton 1985). The Serengeti supports the largest herds of migrating ungulates in the world. Estimates put wildebeest (Connochaetes taurinus) around 1.3 million, zebra (Equus burchelli) at 250 000 and Thomson’s gazelle (Gazella thomsoni) at 440 000 (Mduma & Hopcraft 2008). During November–May the migrant grazers occupy the short grass plains. Due to the rainfall gradient these short grass areas dry out first, and as they do so in May–June the herds move rapidly westwards through the long grass areas, grazing down the tall grass, before moving into the woodlands. They remain in the long grass areas for a few days only and so create doi:10.1111/j.1442-9993.2010.02167.x 2 A . K . N K WA B I ET AL. a rapid change in grass structure (Sinclair 1977).They do not return until the next wet season 4–6 months later. Both burning and grazing disturbances have their major effect by altering the grass structure from long grass (80–100 cm) to shorter grass (5–25 cm), and both occur at about the same time of year, the beginning of the dry season. Fire removes effectively all the biomass, which must then re-grow as short grass. Because fires occur at the start of the dry season (the majority occur in July) subsequent rainfall is low and re-growth is in the form of leaves less than 15 cm high with no flowerheads. Grass then remains in this structure until the rains begin in November. Grazing removes some 50% of the biomass leaving a stubble some 10–25 cm high (Sinclair 1977; Sinclair et al. 1985). Thus, fire is a more extreme disturbance than grazing in altering the grass structure. In addition, fire, unlike grazing, can potentially kill arthropods living in the long grass (Evans 1984). Time is required both to re-grow green vegetation and to recolonize the habitat with arthropods. We consider to what extent the community of birds using the Serengeti long grass plains is affected by these two natural disturbances. We examine three aspects: First, how does the richness of bird species change with burning and grazing and to what extent is this affected by disturbances to their arthropod food supply? Second, how does the abundance of birds that use the grass layer change with disturbance? We hypothesized that the richness and abundance of bird species should be altered because the arthropod food supply in the grass layer should decline when the structure is changed from tall grass to short grass. Alternatively, bird species richness and abundance should be altered because the change in grass structure could affect their foraging mode even if food abundance did not change. Third, are there changes in the types of birds in the different grass structure? Richness alone does not indicate whether species replace each other within functional groups. Thus, we examine changes in individual species. MATERIALS AND METHODS Study area The Serengeti-Mara ecosystem (Fig. 1) is an area of some 25 000 km2 on the border of Tanzania and Kenya, East Africa (34° to 36°E, 1° to 3°30′S). The extent of the ecosystem is defined by the movements of the migratory wildebeest.There are two major habitat types, a treeless grassland in the southeast of the system, and savanna dominated by Acacia trees in the west and north. This study is confined to the plains Fig. 1. Map showing the Serengeti National Park, and the long and short grass plains. The sites where replicate treatments (burned, grazed and undisturbed plots) are located are indicated by arrows. doi:10.1111/j.1442-9993.2010.02167.x © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia D I S T U R B A N C E O N AV I A N A N D A RT H R O P O D C O M M U N I T Y habitat, which is divided into two parts, the long grass and the short grass plains (Fig. 1). Soils on the eastern plains are highly saline, alkaline and shallow as a result of their recent volcanic origin. Consequently, the grasslands here are only some 10–15 cm high. The soils become progressively deeper and less alkaline towards the northwest plains and into the woodlands. Rain typically falls in a bimodal pattern, with the long rains during March–May and the short rains in November– December (Norton-Griffiths et al. 1975). There is a rainfall gradient from the dry southeast plains (<500 mm per year) to the wet northwest long grass plains (800 mm) and continuing on to the Kenya border (1200 mm per year) (Sinclair & Norton-Griffiths 1979). The bird fauna Descriptions of the avifauna in the Serengeti ecosystem are given in Sinclair (1978), Folse (1982), Sinclair et al. (2002) and Mwangomo et al. (2007). Although the whole Serengeti ecosystem supports some 640 species of birds (A. Sinclair unpubl. data, 2009), we consider in this paper only those inhabiting the treeless grassland habitat often referred to as the plains. These grasslands, as described above, differ in both structure and plant composition between the southeast short grass swards and the northwest long grass stands. Associated with these plant communities are different associations of bird species. They support four major groups of insectivorous birds: cisticolas, larks, pipits and plovers, in two feeding guilds; (i) those, like some cisticolas (small warbler-like forms), that in tall grass feed among the grass stems above ground; and (ii) those that feed on the ground, which include species from all four groups, in both long and short grass. There are also two main families of seed eating birds, the weavers and estrildines. In all there are 27 families with 78 species documented in Appendix S1 for the different grassland types. Sites and treatment plots The long grass plains comprise a highly uniform grassland community over large areas (about 1500 km-2) dominated by tall (100 cm) grasses such as Themeda triandra, Pennisetum mezianum and Digitaria macroblephora. The migratory herds of wildebeest move through these long grass plains each year at the end of May and their routes and timing have been highly predictable for the past 40 years. Three sites were selected for the treatment plots after the wildebeest had arrived in the long grasslands and after burning commenced, the two treatments starting effectively simultaneously. The selection of sites had to await these events in order to place the treatments. Within a site there were three replicates for grazing and three for burning with paired undisturbed plots (ungrazed and unburnt). Prior to the treatments we set up six pretreament plots in the general vicinity of the sites. The exact locations of the grazed plots were chosen once the herds had moved through, which normally takes a few days. Burns were set by park rangers during the time that the wildebeest migration moved through, and located at our © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia 3 request. The burnt treatment plots were located within these burned areas. The matched undisturbed plots were then placed nearby where neither grazing nor burning had taken place. Each replicate plot was 70 m ¥ 70 m in size (0.5 ha), located using GPS, and marked by flagging tape at the corners of the site. The demarcation defined the boundaries for counting birds. Censusing birds Six plots were surveyed per day (that is all burn and undisturbed or grazed and undisturbed were counted on 1 day). These counts were repeated for 3 days and the numbers for each plot averaged over the 3 days. The number of days was determined from a pilot study where plots were counted for up to 6 days. Since all species present on a plot had been recorded after 3 days this number of days was chosen. A 20-min count period was used in each plot, and the time of day between 0630 and 1030 h in the morning was rotated among the six plots to even out this possible bias. Birds were counted by walking slowly across the plot and back again. Birds were identified by both sight and call, and numbers recorded (Pomeroy & Tengecho 1986; Pomeroy 1992; Bibby et al. 2000). Birds were counted in four time periods. Period 1, pretreatment: Counts were conducted in long grass areas prior to burning and grazing to act as the baseline before treatments were applied. Period 2: Counts were conducted shortly (<1 week) after burning or grazing affected the study plots. Period 3: Plots were monitored 4 weeks after burning or grazing, this time being in the middle of the dry season but after green re-growth had taken place. Period 4: Counts were conducted at the beginning of the subsequent rains when grass began rapid growth (20 weeks after disturbances). This was to examine whether there was a delayed effect of disturbance on use by birds. As part of a long-term study we conducted a vehicle transect through both the long grass (30 km) and short grass plains (25 km) (Fig. 1) in the years 1997–2008. Birds were counted within 50 m of the vehicle twice a year. This provided an index of abundance of species within the communities of the two habitats. Arthropod sampling Methods for the collection of arthropods followed those recommended by New (1998). Ground surface dwelling arthropods – ants, spiders and Coleoptera – were sampled with pitfall traps (8.7 cm diameter). Tray traps (28 cm diameter), painted yellow because they are seen as green by arthropods, attracted flies and wasps. Traps were left for 3 days. Sweep net samples (standard 30 sweeps per sample) captured arthropods in the herb layer, largely Orthoptera and Hemiptera. Two samples from each method were pooled from each plot. Captured arthropods were collected and preserved in alcohol. Arthropod samples were sorted into major groups, counted and voucher specimens were mounted for later identification. doi:10.1111/j.1442-9993.2010.02167.x 4 A . K . N K WA B I ET AL. Data Analysis In this analysis we included all birds that fed on arthropods or seeds in the herb and ground layers. We excluded vertebrate feeders, and aerial insectivores because their activities generally covered a much larger scale than that of our experiments. Species richness was estimated using rarefaction (Krebs 2001). Rarefaction is a statistical method for estimating the number of species in a given sample of individuals. It allows two samples of different sizes, representing different sampling efforts, to be compared. There are two restrictions on the use of this method (Krebs 2001). First, the groups of species must be taxonomically similar. Second, the two groups must be estimated by similar methods. Both of these restrictions were met because similar groups of birds were compared in the analyses and the data were collected by the same methods. The abundances of ground feeding birds in the different treatments and time periods were averaged across the three replicates for each treatment and time, and standard errors computed. The abundances of above ground arthropods were averaged across the replicates for different methods of capture, treatments and time periods, and standard errors computed. RESULTS Bird species richness Rarefaction estimates for ground feeding birds were used to calculate the probable number of species seen in burned versus undisturbed and grazed versus undisturbed sites on the long grass plains for the three time periods, namely 1, 4 and 20 weeks, following the start of the perturbations (Fig. 2). The rarefaction curves show that immediately after either disturbance there was an increase in bird diversity relative to the undisturbed sites and this persisted throughout the subsequent dry season and early growing season (20 weeks). However, with burning this increase in diversity became less pronounced as time proceeded after the disturbance (Fig. 2a,b,c), while the increase in diversity after grazing became more pronounced with time (Fig. 2d,e,f) burned and undisturbed plots was accounted for by the burning effect where mean abundance of 139 birds per plot on burns was much higher than the 45 on undisturbed plots (t = 5.12, n = 18, P = 0.0001). Although there was a grazing effect (mean abundance grazed 140, undisturbed 98 per site) it was not significant (t = 1.4, n = 18, P = 0.17). Inspection of the data on individual bird species showed that the differences in abundance with treatment were due largely to changes in the larks, cisticolas and plovers (Table 1). Thus, of the top 10 most frequent species in undisturbed plots, only two, rufousnaped lark (Mirafra africana), African quail finch (Ortygospiza atricollis) appear in the top 10 burned plot species, and then at much lower frequency. In burned and grazed plots there were four abundant species that were not recorded at all in undisturbed plots [redcapped lark (Calandrella cinerea), Richard’s pipit (Anthus novaeseelandiae), black-winged plover (Vanellus melanopterus), chestnut-banded sandgrouse (Pterocles exustus)]. We have chosen six species to illustrate the different responses to treatments (Fig. 4). Both the athi short-toed lark (Calandrella somalica) (ground living) and croaking cisticola (Cisticola natalensis) (grass dwelling) had higher abundances in undisturbed plots throughout the period of study. The Fischer’s sparrow-lark (Eremopterix leucopareia) was abundant in undisturbed plots when they were still green (pretreatment) but they moved into both burnt and grazed plots when these appeared while their abundance in undisturbed plots declined progressively through the dry season. Richard’s pipit, crowned plover (Vanellus coronatus) and red-capped lark were not present in pretreatment plots but appeared in treated plots in large numbers when these disturbances occurred while being largely absent in the equivalent undisturbed plots. Some species appeared to replace others when the treatments were imposed. The athi short-toed lark, initially abundant in pretreated sites, dropped in number as soon as burning and grazing occurred (Fig. 5), and only regained abundance on those plots once growth took place at 20 weeks. At the same time the red-capped lark, which was absent before the treatments, increased to high numbers in burned and grazed plots replacing the athi short-toed lark. Abundance of grassland birds Transects in long grass and short grass plains Generally, in the undisturbed plots abundances of birds dropped from time period 1 (pretreatment) to period 2 and subsequently as conditions became drier (Fig. 3). In contrast, bird abundance remained higher on the combined burnt and grazed treatments than the equivalent undisturbed plots, and in all time periods following application of the disturbances (Fig. 3) (t = 3.8, n = 36, P = 0.0006). This difference between doi:10.1111/j.1442-9993.2010.02167.x Data from the twice yearly vehicle transects showed the frequency distribution of the insectivorous and granivorous ground and herb layer birds for the two habitats (Appendix S1). This is our best estimate of these bird communities in the two habitats. We compared the frequency distributions of birds in the undisturbed and treated (burnt, grazed) plots with those © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia D I S T U R B A N C E O N AV I A N A N D A RT H R O P O D C O M M U N I T Y a unburnt burnt Burnt 1 week d Grazed 1 week 5 ungrazed grazed 30 25 25 20 Species Species 30 20 15 10 15 10 5 5 0 0 0 b 200 400 Individuals 600 0 800 e Burnt 4 weeks 25 200 400 Individuals 600 Grazed 4 weeks 16 14 20 Species Species 12 15 10 10 8 6 4 5 2 0 0 0 c 200 400 Individuals 600 0 f Burnt 20 weeks 100 200 300 Individuals 400 Grazed 20 weeks 18 25 16 14 Species Species 20 15 10 12 10 8 6 4 5 2 0 0 0 200 400 600 Individuals 0 50 100 Individuals Fig. 2. Rarefaction curves for total number of ground feeding bird species on the long grass plains observed at different times in weeks following application of burning and grazing (solid circles) in comparison to undisturbed (unburnt and ungrazed) plots (open triangle). Vertical lines, one S.D. frequencies obtained from the standard vehicle transects through the long grass and short grass plains (Table 2).The Bray-Curtis similarity coefficients show that the untreated long grass communities were most similar to the long grass plains community, and least similar to the short grass plains community. The similarity value between untreated and treated communities falls between these former two values. Moreover, the treated (burnt plus grazed) bird communities were most similar to the short grass plains community. © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia Abundance of arthropods In general, there was greater abundance of arthropod groups in the undisturbed grassland compared with that in both burnt and grazed plots (Appendix S2), and this difference is illustrated in Figure 6 for the combined catch by each method. The only exceptions were that wasps were marginally more abundant in grazed and burnt plots than in undisturbed plots, and beetles were more abundant in grazed plots (Fig. 6, doi:10.1111/j.1442-9993.2010.02167.x 6 A . K . N K WA B I ET AL. Longgrass burnt Longgrass unburnt a Longgrass grazed b Longgrass ungrazed 200 200 160 Number Number 160 120 80 40 120 80 40 0 0 Pre 1 4 20 Pre Weeks 1 4 20 Weeks Fig. 3. Mean abundance of birds per sites of all ground feeding birds in plots at different time periods following burning (a) or grazing (b) compared with abundances on sites prior to these perturbations or at contemporaneous undisturbed sites. Open bar, unburnt or ungrazed; solid bar, burnt or grazed; vertical lines, one S.E. Table 1. Top: The number and proportion of birds recorded for the top 10 species in the combined undisturbed sites, compared with that of the same species in the combined burnt and grazed plots (disturbed). Bottom:The number and proportion of birds recorded for the top 10 species in the disturbed plots compared with that of the same species in the undisturbed plots Top 10 in undisturbed plots Species Zitting cisticola Athi short-toed lark Rufous-naped lark Croaking cisticola African quail finch Fan-tailed widowbird White-winged widowbird Rosy-breasted longclaw Senegal bustard White-tailed bushlark Number undisturbed Proportion undisturbed Number disturbed Proportion disturbed 749 632 271 133 98 79 61 50 46 45 0.308 0.260 0.111 0.055 0.040 0.032 0.025 0.021 0.019 0.018 182 77 182 0 43 0 0 0 9 29 0.067 0.029 0.067 0 0.016 0 0 0 0.003 0.011 Top 10 in treated plots Species Red-capped lark Crowned plover Caspian plover Fischer’s sparrowlark Zitting cisticola Rufous-naped lark Richard’s pipit Black-winged plover Superb starling Capped wheatear Number disturbed 704 301 226 209 182 182 130 125 92 87 Appendix S2). Time trends in catch were not consistent across treatments or trapping method. In both treatments catches were usually higher after 20 weeks, when re-growth following the start of the rains was occurring, compared with the 1- and 4-week samples. doi:10.1111/j.1442-9993.2010.02167.x Proportion disturbed 0.261 0.112 0.084 0.077 0.067 0.067 0.048 0.046 0.034 0.032 Number undisturbed 0 1 0 33 749 271 7 0 19 0 Proportion undisturbed 0 0 0 0.014 0.308 0.111 0.003 0 0.008 0 DISCUSSION We proposed that bird richness and abundance should change with burning or grazing due to a concomitant change in food supply. The results indicated that on the long grass plains there was an increase in bird © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia D I S T U R B A N C E O N AV I A N A N D A RT H R O P O D C O M M U N I T Y Burnt Grazed Control ASTL 25 7 CRCI 3.5 3 20 15 Mean Mean 2.5 10 2 1.5 1 5 0.5 0 0 Pre 1 4 20 Pre 1 4 Period 20 Period RIPI FSLA 9 7 8 6 7 5 Mean Mean 6 5 4 3 4 3 2 2 1 1 0 0 Pre 1 4 Pre 20 1 4 20 Period Period CRPL RCLA 9 35 8 30 7 25 Mean Mean 6 5 4 3 20 15 10 2 5 1 0 0 Pre 1 4 20 Period Pre 1 4 20 Period Fig. 4. Mean abundance per site of individual bird species observed at different time periods in long grass plains. Open bar, undisturbed; solid bar, burnt, and hatched bar, grazed. Vertical lines, one S.E. ASTL, athi short-toed lark; CRCI, croaking cisticola; FSLA, Fischer’s sparrow-lark; RIPI, Richard’s pipit; CRPL, crowned plover; RCLA, red-capped lark. species richness on both burned and grazed areas compared with those on the undisturbed plots. Insectivorous birds formed the majority of ground-living birds in these Serengeti habitats.Thus, arthropod food abundance was taken as the measure of food supply for these birds. In general, most groups of arthropods declined in abundance in both burning and grazing treatments in contrast to the increase in bird richness and abundance, a result contrary to our expectations. Thus, these higher abundances of birds were not due to higher absolute food resources. The lower food resources could have resulted either from the damaging effects of the disturbances or from the higher depredations of the more abundant insectivorous birds. An alternative explanation is that bird species richness and abundance changed with these disturbances © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia due to changes in grass structure that affected foraging mode, but not necessarily food abundance. Because our results showed that food abundance declined as both bird richness and abundance increased they imply that birds were taking advantage of the change in grass structure, were able to access food easier and so reduce arthropod abundance. In essence, this result implies that there was a change in ecosystem function by increasing the top-down predation effect on arthropods due to a change in habitat structure. Two other factors could contribute to both higher bird species richness and higher abundance in burned and grazed areas on the long grass plains compared with those on the undisturbed plots. First, although grass height was lower in burned and grazed areas, doi:10.1111/j.1442-9993.2010.02167.x 8 A . K . N K WA B I ET AL. 35 8 30 7 6 5 20 4 15 3 10 Mean ASTL Mean RCL A 25 2 5 1 0 0 0 1 2 3 4 5 Period Fig. 5. Mean abundance per plot of red-capped lark (RCLA solid circles) and athi short-toed lark (ASTL open triangles) on burned areas of the long grass plains. Time 1, pre-treatment; time 2, 1 week; time 3, 4 weeks; time 4, 20 weeks after burning. Table 2. The Bray-Curtis coefficients of similarity for the bird communities on the long grass and short grass plains recorded on the vehicle transects, and on the long grass undisturbed plots, and the burnt and grazed plots Bray-Curtis coefficients Undisturbed versus long grass transect Undisturbed versus short grass transect Undisturbed versus burnt + grazed Burnt + grazed versus long grass transect Burnt + grazed versus short grass transect Coefficient 0.384 0.092 0.243 0.329 0.546 habitat heterogeneity through a mosaic of different heights may actually have been greater. Second, many bird species in these areas are nomadic and this habit enables them to locate recently burnt or grazed patches. Influxes of species, such as red-capped larks, black-winged plovers and Richard’s pipits appeared on burned sites, while these and Caspian plovers appeared on grazed sites. Their nearest locations were on the short grass plains, which occur some 60 km to the southeast. It is possible that the change in grass structure from tall to short grasslands could be mimicking the creation of patches of short grass plains within the long grass habitat. Such patches may be attracting the avifauna of the short grass plains. Although we did not conduct site counts on the short grass plains, we have the 11 years of data from transect counts conducted in both the long and short grass plains. These showed that the bird communities of the disturbed sites were more similar to that on the short grass plains than that on the undisturbed long grass plains. Thus, the perturbations of burning and grazing changed the bird community towards that seen on the short grass plains. doi:10.1111/j.1442-9993.2010.02167.x We also examined whether individual species might replace each other ecologically, showing opposite trends in abundance with treatment. The results showed that at least in the long grass habitats some lark species with similar ground living insectivorous food habits replaced others after treatments were imposed (Fig. 5). Therefore, these results show that species can replace each other functionally following disturbances such as burning and grazing. Few other studies have examined the effects of burning and grazing on bird populations in Africa (Parr & Chown 2003). Ogada et al. (2008) reported that the removal of large herbivores from African savanna plots increased bird diversity. The effects of the large mammals were through the removal of tree canopy, which reduced the number of bird species. There were little effect of the mammals on the grass layer. O’Reilly et al. (2006) have documented the effects of experimental fires on African savanna birds in Kenya. They recorded that burning increased bird diversity relative to controls but there were no significant effects on either richness or total abundance. Thus, numbers were distributed more evenly across the same species, unlike our study that found a clear increase in both richness and abundance with both burning and grazing. Jansen et al. (1999) found that the richness of grassland birds and the density of the red-winged francolin (Francolinus levaillantii) were negatively correlated with grazing intensity and annual burning, whereas grey-winged francolin (Francolinus africanus) density was positively correlated with grazing intensity in South Africa. A comparison of grazed and ungrazed sites in southeast Arizona showed an increase in some bird species on grazed lands that were habitat generalists, but it also demonstrated declines in some grassland sparrow species like the Botteri’s sparrow and Cassin’s sparrows in the same area (Bock & Bock 1993). Woinarski (1990) found in northern Australia that short-term effects of fire involved a substantial increase in both diversity and density of birds. Thus, our results are consistent with some other studies, namely increases in richness and abundance with burning and grazing, but not all results are consistent with this conclusion. It is likely that higher frequencies of burning and grazing may reverse the effects we found, and so variation in the severity of disturbance needs further investigation. Our study also throws light on the possible mechanisms producing the changes in the bird communities. CONCLUSIONS This study investigated how the richness and abundance of bird species changed with two types of disturbance to the grass layer, burning and grazing, and examined whether there was a concomitant change in © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia D I S T U R B A N C E O N AV I A N A N D A RT H R O P O D C O M M U N I T Y 40 45 40 35 35 30 30 Number Number Grazed Ungrazed Burnt Pit grazed Unburnt Pit burnt 25 20 15 25 20 15 10 10 5 5 0 0 Pre 1 4 Pre 20 1 Weeks 20 Tray trap grazed 70 180 60 160 140 Number 50 Number 4 Weeks Tray trap burnt 40 30 120 100 80 60 20 40 10 20 0 0 Pre 1 4 20 Pre 1 Weeks 4 20 Weeks Sweep burnt Sweep grazed 45 45 40 40 35 35 30 30 Number Number 9 25 20 25 20 15 15 10 10 5 5 0 0 Pre 1 4 20 Weeks Pre 1 4 20 Weeks Fig. 6. Mean abundance per sample for arthropods caught by pitfall traps (top, ants, spiders and Coleoptera), tray traps (middle, Diptera, wasps and bees) and sweepnets (bottom, Hemiptera and Orthoptera) in plots at different time periods following burning (left three figures) or grazing (right three figures) compared with abundances on plots prior to these perturbations (Pre) or at contemporaneous undisturbed sites. Open bar, undisturbed; and solid bar, burnt or grazed. Vertical lines, one S.E. the arthropod food for the largely insectivorous bird community. We found that both bird diversity and abundance increased with both disturbances. In contrast, the abundance of arthropods declined with both disturbances.Thus, the bird communities appear to be changing in response to changes in the structure of their habitat providing different conditions for new species to enter, and enabling them to access arthropod food and lowering its abundance. This means there may be a change in ecosystem function through a change in the top-down predation cascade. The disturbances created a grassland structure that resembled the short grass plains, and it was the bird species normally residing in the short grass plains that appeared on the disturbed sites. © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia Species that prefer long grass declined in number with the onset of disturbance. However, some were replaced by functional equivalents. Thus, the redcapped lark and the athi short-toed lark, which are functionally equivalent, replaced each other when the grassland changed from long to short grass in both treatments.This replacement compensated for the loss of some species due to the disturbance and contributed to maintaining diversity. ACKNOWLEDGEMENTS We thank the directors of Tanzania National Parks and the Tanzania Wildlife Research Institute for permission doi:10.1111/j.1442-9993.2010.02167.x 10 A . K . N K WA B I ET AL. to work in the park.We thank especially John Mchetto, Stephen Makacha and Joseph Masoy for their assistance in the fieldwork.This work was supported by the Canadian Natural Sciences and Engineering Research Council grant to ARES. All research contained in this study complied with the laws of Tanzania. REFERENCES Bibby C. J., Burgess N. D., Hill D. A. & Mustoe S. (2000) Bird Census Techniques. Academic Press, London. Bird M. I. & Cali J. A. (1998) A million-year record of fire in sub-Saharan Africa. Nature 394, 767–9. Bock C. E. & Bock J. H. (1993) Cover of perennial grasses in southeastern Arizona in relation to livestock grazing. Conserv. Biol. 7, 371–7. Evans E. (1984) Fire as a natural disturbance to grasshopper assemblages of tallgrass prairie. Oikos 43, 9–16. Folse L. J. (1982) An analysis of avifauna-resource relationships on the Serengeti Plains. Ecol. Monogr. 52, 111–27. Frost P. G. H. & Robertson F. (1987) The ecological effects of fire in savannas. In: Determinants of Tropical Savannas (ed. B. H. Walker) pp. 93–140. ICSU Press, Miami. Jansen R., Little R. M., Crowe T. M. & Ikanda D. (1999) Implications of grazing and burning of grasslands on the sustainable use of francolins (Francolinus spp.) and on overall bird conservation in the highlands of Mpumalanga province, South Africa. Biodivers. Conserv. 8, 587–602. Krebs C. J. (2001) Ecological Methodology. Harper and Row, New York. van Langevelde F., van de Vijver C. A. D. M., Kumar L. et al. (2003) Effects of fire and herbivory on the stability of savanna ecosystems. Ecology. 84, 337–50. McNaughton S. J. (1985) Ecology of a grazing system: The Serengeti. Ecol. Monogr. 55, 259–94. Mduma S. A. R. & Hopcraft J. G. C. (2008) The main herbivorous mammals and crocodile in the Greater Serengeti Ecosystem – Appendix. In: Serengeti III: Human Impacts on Ecosystem Dynamics (eds A. R. E. Sinclair, C. Packer, S. A. R. Mduma & J. M. Fryxell) Chicago University Press, Chicago. Mwangomo E. A., Hardesty L. H., Sinclair A. R. E., Mduma S. A. R. & Metzger K. (2007) The association of two endemic birds in the Serengeti ecosystem: Habitat selection, diet and group formation of the rufous-tailed weaver, Fischer’s lovebird and associated species. Afr. J. Ecol. 46, 267–75. New T. R. (1998) Invertebrate Surveys for Conservation. Oxford University Press, Oxford. Norton-Griffiths M. (1979) The influence of grazing, browsing, and fire on vegetation dynamics of the Serengeti. In: Serengeti: Dynamics of an Ecosystem (eds A. R. E. Sinclair & M. Norton-Griffiths) University of Chicago Press, Chicago. Norton-Griffiths M., Herlocker D. & Pennycuick L. (1975) The patterns of rainfall in the Serengeti ecosystem, Tanzania. East Afr.Wildl. J. 13, 347–74. doi:10.1111/j.1442-9993.2010.02167.x O’Reilly L., Ogada D., Palmer T. M. & Keesing F. (2006) Effects of fire on bird diversity and abundance in an East African savanna. Afr. J. Ecol. 44, 165–70. Ogada D. L., Gadd M. E., Ostfeld R. S., Young T. P. & Keesing F. (2008) Impacts of large herbivorous mammals on bird diversity and abundance in an African savanna. Oecologia 156, 387–97. Parr C. L. & Chown S. L. (2003) Burning issues for conservation: A critique of faunal fire research in Southern Africa. Austral Ecol. 28, 384–95. Pomeroy D. E. (1992) Counting Birds. African Wildlife Foundation, Nairobi. Pomeroy D. E. & Tengecho B. (1986) A method of analysing bird distributions. Afr. J. Ecol. 24, 243–53. Sinclair A. R. E. (1977) The African Buffalo. University of Chicago Press, Chicago. Sinclair A. R. E. (1978) Factors affecting the food supply and breeding season of resident birds and movements of palaearctic migrants in a tropical African savannah. Ibis 120, 480–97. Sinclair A. R. E., Dublin H. & Borner M. (1985) Population regulation of Serengeti Wildebeest: A test of the food hypothesis. Oecologia 65, 266–8. Sinclair A. R. E., Mduma S. A. R. & Arcese P. (2002) Protected areas as biodiversity benchmarks for human impacts: Agriculture and the Serengeti avifauna. Proc. of the Roy. Soc. of London 269, 2401–5. Sinclair A. R. E., Mduma S. A. R., Hopcraft J. G. C., Fryxell J. M., Hilborn R. & Thirgood S. (2007) Long-term ecosystem dynamics in the Serengeti: Lessons for conservation. Conserv. Biol. 21, 580–90. Sinclair A. R. E. & Norton-Griffiths M. (1979) Serengeti – Dynamics of An Ecosystem. University of Chicago Press, Chicago. Stronach N. R. H. (1988) Grass Fires in Serengeti National Park, Tanzania: Characteristics, Behavior and Some Effects on Young Trees. University of Cambridge, Cambridge. Woinarski J. C. Z. (1990) Effects of fire on the bird communities of tropical woodlands and open forests in northern Australia. Aust. J. Ecol. 15, 1–22. SUPPORTING INFORMATION Additional Supporting Information may be found in the online version of this article: Appendix S1. Family, Latin name and common name of bird species identified in long grass (transects), short grass (transects) and undisturbed (this study) habitats. Appendix S2. The abundance per sample and standard errors (in parenthesis) of the main arthropod groups collected at different time periods in the burnt and grazed plots compared to those collected in the undisturbed plots (unburnt and ungrazed). Pre = pretreatment, time 1 = 1 week, time 4 = 4 weeks, time 20 = 20 weeks after the treatments. © 2010 The Authors Journal compilation © 2010 Ecological Society of Australia