How extensive is the effect of modern farming on bird communities in a sand dune desert?

Bird community structure and diversity measures in sand dune habitats far from and close to modern farms in Wadi Araba, south-west Jordan, were compared using 52 line transects for breeding birds and habitat variables. A change in the bird community of sand dunes surrounding farming projects was measured to a distance of 1 km, but could neither be related to changes in habitat structure nor to the activity of opportunistic predators (Red Fox) as these did not vary signifi cantly between the two samples. Th e farms included lines of trees and off ered a constant source of water, which attracted a variety of opportunistic species, thus increasing bird diversity and total bird abundances. Th e absence of characteristic ground-dwelling species of open sand dune habitats in the structurally intact sand dunes surrounding farms was likely to be the result of localized, but eff ectively far-reaching habitat modifi cation (farms acting as barriers) and/or competition with some of the opportunistic species, which were common around farms.


Introduction
Th e anthropogenic impact on bird communities of arid ecosystems has been studied only tentatively.Most studies have focused on the eff ects of grazing, large irrigation projects and invasive plant species on ecosystem function and the diversity of ani-mals, including birds (Dean et al. 2002;Desmond 2003;Rahmani and Soni 1997;Whitford 1997).
Th e hyper-arid Wadi Araba in south-western Jordan exhibits a variety of habitat types ranging from sand dunes and alluvial fans to salt marshes and arid acacia savannahs.Although still considered as one of the least populated and spoilt areas of Jordan (Andrews 1995), current expansion of agricultural projects in Wadi Araba is probably interfering with ecosystem function and modifying the habitats of birds at an increasing rate.Th e aim of this study was to determine the impact of intensive agricultural practices on bird communities by comparing the community structure and diversity measures of sand dune habitats immediately surrounding farms with those, which are distant but of the same type.We tested a variety of approaches and analytical methods to determine the best indication of agricultural impact in arid landscapes using quantitative data on birds.

Study area
Wadi Araba, the northern extension of the Great Rift Valley, is a low-lying, fl at strip of land between the mountain range of southwest Jordan and the Negev plateau of southern Israel.It stretches between the Gulf of Aqaba in the south and the Dead Sea in the north, with a maximum width of 30 km.Th e study included the southern and central parts of the Jordanian side of Wadi Araba (c.30°00'N, 35°10'E), 60% of which consists of sand dune desert.Th e climate is generally hot and hyper-arid; the mean annual precipitation is approximately 50 mm, while the annual mean evaporation rate lies around 2.000 mm (National Atlas of Jordan, 1984).Th e occasional rainfalls usually occur during the winter months when temperatures are lower than in summer; temperatures range from 4 °C (absolute minimum) in the winter to 45 °C in the summer, with an annual mean of 24 °C (National Atlas of Jordan, 1984).Local disparities in soil and hydrological conditions cause a variety of vegetation and habitat types (Baierle, 1993).Th e sand dune habitats considered in this study vary from highly mobile sand dunes, nearly devoid of vegetation to more stable and fl at sand dunes, where the surface is occasionally covered with a thin layer of fi ne gravel.Th e vegetation usually consists of a sparse cover dominated by Haloxylon persicum and a substantial annual cover after the rainy season.A few shallow waterbeds intercept the studied sites, which are lined with relatively dense vegetation dominated by H. persicum, and occasionally including the shrubs Calligonum comosum and Retama raetam.Th ere are a few villages, in addition to nomadic Bedouins, but most of the area is unpopulated and considered a closed military zone due to its proximity to the political borders.Nevertheless, grazing by goats and camels of local Bedouins is permitted and widespread all over.Natural herbivores, e.g.Gazella dorcas, occur in small numbers near the borders with Israel.Th e establishment of large projects of intensive, irrigated farming is ongoing in several parts of the study area, which exploit ground water for irrigation and include fenced areas lined with introduced trees as windbreakers.

Field methods
Quantitative data on birds was collected using the line transect method (Bibby et al. 2000) in the period from March to May 2004, which covered the major part of the breeding season.Th irty-seven line transects were randomly selected in sand dune habitats at distances ranging from 0.5 to 8 km from the nearest farms.Each line transect was 1 km in length, with a belt of 100 m on each side.In addition, 15 line transects were carried out in the same habitat type close to farms.In these cases, the observers walked parallel to and at a distance of 80-100 m from the farm fence, thus additionally covering the birds seen at the border of the farms.Th e individual farms covered areas ranging from 0.5-5 km 2 ,and most had a narrow rectangular shape, with a length exceeding 1 km, thus allowing us to carry out full transects parallel to their borders.Line transects were carried out from sunrise to 10:00 am; the duration of each transect ranged from 25-35 minutes, during which all of the birds within the 100-m belts were recorded by two observers walking parallel, at a distance of 20 m.We excluded in the analysis all bird species that do not breed in the study area, e.g.migrants and birds resident and breeding in diff erent habitat types in neighbouring areas, which occasionally utilised the study area for feeding, as well as birds detected beyond 100 m and those just fl ying over the transects.Habitat variables were quantifi ed at three points (at the two ends and at the middle) along each line transect to a range of 50 m before being averaged.Th is allowed a general description and comparison of the habitats of the two groups of samples (far and close).Additionally, the activity of mammalian predators (mainly the opportunistic red fox Vulpes vulpes) was determined by counting the number of trails crossing the sandy surface of line transects.

Data analysis
Analysis conducted at the community level included the construction of dendrograms of the samples (line transects), using group-average clustering from Bray-Curtis similarities on log-transformed bird abundances (Clarke andWarwick 2001, Khoury andAl-Shamlih 2006)).Diversity measures were calculated for all samples (transects) in each group (far and close) and included species richness (S = number of species), Shannon-Wiener index (H) and evenness (J).Th e calculations were according to the following formulas: Where s is the number of species present in the sample; Pi is the proportion of species i in the sample of s species; Hmax = ln s.
Th e diversity measures and total number of individuals of the two samples (Far and close) were tested for signifi cant diff erences using the non-parametric Mann-Whitney U-test, as they departed from normality.Analysis of variance (ANOVA) was applied to examine diff erences in the habitat variables between the two samples.Th e impact of agriculture on bird assemblages was further analysed by comparing the total number of individuals belonging to bird species originally native to Wadi Araba and referred to as "naturally occurring" with the number of "opportunistic"/"agricophilic" birds, which benefi ted from agricultural expansion as indicated by their range expansion into the general area in previous decades (Table 1; Andrews 1995;Shirihai 1996).Other species, which are locally expanding their range due to agricultural developments, were included in the fi rst group, as they are known to belong to the original avifauna of natural habitats in Wadi Araba (Andrews 1995).Th ese are referred to in Table 1 as agriculture-enhanced. Trumpeter Finch, for example originally occurs in natural habitats in Wadi Araba where a mix of sand and stones or rocks cover the ground, however this species has apparently benefi ted locally from agriculture (Shirihai 1996); Desert fi nches are classifi ed as opportunistic because they invaded Wadi Araba consequent to agricultural developments, i.e. they did not previously breed in Wadi Araba.We used the t-test to compare the number of "naturally occurring" and "opportunistic" species in each sample group (far and close).

Results
A comparison of the habitat variables and red fox activity in the sand dune habitats far from farm (hereafter referred to as F-transects) with those close to farms (C-transects) showed that the two groups did not differ significantly in habitat structure or in the activity of red foxes (ANOVA, p > 0.05 for all habitat variables, Table 2).However, the borders of the farms, which included fences and planted trees, were excluded from this analysis.This lead to the conclusion, that the impact of agriculture on the immediate surroundings is due to indirect factors produced by farming practices rather than a direct modification of habitat structure.Although it is likely that farming activities cause a general increase in the abundance of red foxes in the whole study area, there was no significant difference in its activity and assumed effect on birds when F-transects and C-transects were compared (p > 0.05).
Th e C-transects had signifi cantly higher Shannon indices and species richness than F-transects (U-Test, p < 0.001 in both cases, Table 2), although characteristic species of sandy habitats were completely absent in the C-transects.Th ese were replaced in the surrounding sandy habitats and at the edges of farms by a number of "opportunistic" species.In addition, the total avian abundance was signifi cantly higher around and at the edges of farms when compared to the relatively unaff ected sand dune habitats, or F-transects (U-test, p < 0.001).
Table 2. Th e means (± S.D.) of habitat variables, diversity indices, total number of individuals of all birds and number of individuals belonging to "naturally occurring" and "opportunistic" bird species in the transects far from farms (F-transects) and those close to farms (C-transects).

Sample clusters and bird assemblages
Th e 52 line transects produced 18 species of breeding birds (Table 1), some of which were present in both samples (far and close).In the dendrogram based on Bray-Curtis similarity coeffi cients, two large clusters emerged, separating at a similarity coeffi cient of 10%.Th e fi rst cluster (Fig. 1, left) included 81% of the F-transects, which diverged to smaller subgroups at various similarity levels above 30%.Th e subdivisions were probably the result of local variations in the sand dune habitat aff ecting bird species composition.Th e second cluster (Fig. 1, right) included all C-transects, which also diverged to several subgroups.However, one subgroup (extreme right, Fig. 1) was formed by 88% of the C-transects, which was separated from other transects at a similarity level of 55%.A few F-transects included "opportunistic" species like the Crested Lark and were located within a distance of 1 km to farms (except F3, Fig. 1), thus they appeared in the dendrogram closer to the C-transects.Th is indicated a gradient of effects on bird assemblages in sand dune habitats ranging to a distance of around 1 Km from the borders of farms.Ground-dwelling species, Hoopoe Lark and Desert Wheatear, which co-occurred locally with the Bar-tailed Lark, Spotted Sandgrouse and other, less frequent species were described as agriculture-sensitive species (Table 1) and were completely absent around farms to a distance of more than 200m (Table 3).C-transects were clustered together according to the opportunistic and agriculture-enhanced species (Table 1) that have recently colonised the general area or have locally expanded their distribution subsequent to agricultural development.Except for the Crested Lark, all of these species are known to use shrubs and trees.Some, especially Collared Dove and Little Green Bee-eater also occurred in a number of F-transects, but usually at lower frequency and abundances.

Discussion
Th e cluster analysis comparing the line transects in sandy habitats close to farms and far from farms confi rmed the impact of intensive farming on bird assemblages.In general, transects with similar bird assemblages clustered together and a major split was observed between most F-and C-transects.A few F-transects were similar to Ctransects, indicating a gradient of agricultural eff ects to a distance of 1 km into the surrounding sand dune habitats.Most of these transects contained a relatively dense shrub cover, thus supporting larger numbers of a few species also occurring along the edges of farms, e.g.Collared Dove and Little Green Bee-eater, and lower numbers of grounddwelling species of the open sand dune habitat.Th e bird community of the sand dunes in Wadi Araba is split into two main clusters (Khoury and Al-Shamlih 2006): the fi rst included agriculture-sensitive species, which inhabit unaff ected sand dune habitats and avoided structurally intact sand dunes surrounding farms, whereas the second cluster includes species which benefi t from intensive, irrigated farming activities.
Intensive farming in Wadi Araba is interfering in the natural processes of the arid ecosystem, as indicated by the local decline of characteristic desert species.Such species were found to be sensitive to intensive agriculture, although farming activities artificially increased productivity, including food resources and water availability that, on the other hand, signifi cantly augmented the number of opportunistic species and total avian abundance in the C-transects.
On the Jordanian side of Wadi Araba, modern farms still cover a small proportion of the general area, although they are currently increasing in number and area.Being fenced and secluded, these farms did not directly change the habitat structure of the surroundings.Nonetheless, a complete absence of characteristic ground-dwelling species in the structurally intact sand dunes was notable to a distance of several hundred meters around farms.Knick and Rotenberry (1995) have previously shown that many characteristic birds of open habitats (core species) avoid marginal sites, which are close to the edges of their natural habitats.Th ere was no evidence for a localized eff ect Table 3. Th e means (±SE) of abundances of two common sensitive species (HL= Hoopoe lark, DW=Desert wheatear) and two opportunistic species (CD=Collared dove, CL=Crested lark) at varying distances from farms: C: 0-0.2 km; F1: 0.5-2.5;F2: 2.7-8km. of the opportunistic, predatory red fox on bird assemblages, as the activity of red foxes was approximately the same over the whole area.Th e absence of the typical desert species was probably because they avoid vertical structures (tree lines) forming barriers, which limit daily movements and obscure vision important for display, effi cient foraging and predator avoidance.Another possible eff ect is competition with opportunistic species, particularly the Crested Lark, which was found in F-transects, mainly within a distance of 2.5 km from the borders of the nearest farm.Th ere was no further evidence for any other eff ects, e.g. the use of pesticides, which may however aff ect opportunistic species more severely, as they would be directly exposed (Mclaughin and Mineau 1995).

Figure 1 .
Figure 1.Dendrogram of F-and C-transects, using group-average clustering from Bray-Curtis similarities on log-transformed bird abundances.Similarity coeffi cient in percent.
* p < 0.05, ** p < 0.01, ** p < 0.001; U-test for two samples (HL, DW) and Kurskal-Walis test for multiple samples (CD, CL).Th reshold distance represents the distance of the nearest transect to farms, in which sensitive species were recorded.