The distribution of birds, be they Afro-Palearctic (migrants) or Afro-tropical species (residents), in the transient zone between Sahara and humid forests is far from random. Midwinter/dry season distribution of birds species was found to be strongly associated with rainfall gradient and woody cover (Zwarts et al. 2023b,c). This paper attempts to answer the question of whether the distribution of arboreal birds is primarily determined by the selection of a specific rainfall zone or by the distribution of preferred woody species. First, we show how bird species are distributed across the various rainfall zones. Second, we use the average bird density per woody species and the percent woody cover of the selected woody species to calculate the predicted distribution of all bird species assuming species-specific densities in the woody species are uniform. Comparing the observed distribution to the modelled distribution may show to what degree bird species select specific rainfall zones, independent of tree selection. Third, we investigate in five bird-rich and common woody species with a wide distribution whether the density of different bird species in these tree species varies with rainfall. Fourth, we explore whether (a proxy of) food supply for insectivorous birds varies per rainfall zone.

METHODS

The methods, and their pitfalls, for counting birds and woody plants are extensively described by Zwarts & Bijlsma (2015) and briefly again in Zwarts et al. (2023a,b). In short, between 2011 and 2019 birds were counted in 1901 randomly selected study sites (each 4.5 ha) between 7° and 22°N and between 17°W and 42°E (in total 10 million km²; Figure 3 in Zwarts et al. 2023a) during the dry season (20 November – 10 March). All woody plants were registered separately (species, height, width of canopy) within the study sites. We also noted for every arboreal bird the individual woody plant in which it was detected. This allowed us to convert bird counts in the study sites into bird densities per ha, but also to calculate bird density per ha canopy, separately per woody species. We used a five-point scale to score the opacity of the canopy. The data were collected in random sites, but the analysis of the distribution of birds present in five bird-rich trees in relation to average annual rainfall necessarily also includes some data from non-random sites, because we actively searched for specific trees beyond their normal climate range, thus in relatively dry and humid habitats.

Two potential problems arise when attempting to interpret the distribution of birds in relation to rainfall independent of tree selection. First, woody cover increases with rainfall from few and scattered trees in the arid zone to a denser woody cover in more humid zones. If birds avoid scattered, isolated trees, one may expect, on average, fewer birds in woody plants in the arid zone. Similarly, when the presence of a bird in a tree is affected by other trees in its direct surroundings, it might bias our analysis. No evidence was found, however, that isolated trees were visited less often by woodland birds, nor that the presence of (preferred or non-preferred) woody plants in the vicinity had an impact on bird density in specific woody plants (Zwarts et al. 2018). We therefore disregard woody cover and nearby presence of other tree species in our analysis. Second, bird density in woody plants is defined as birds per ha canopy (measured in a horizontal plane), lumped for all woody plants irrespective of size. This potentially complicates our analysis because woody plants are smaller in the arid zone, not only between, but also within woody species. Thus, when bird density in woody plants increases with height and width, we may expect systematically higher bird densities per woody species in the more humid zone. Hence, we did a separate analysis of bird density in relation to tree height (Supplementary Material 1) with which the impact of the height of woody plants on the bird density could be integrated in the modelled bird distribution.

We made a prediction of bird distribution assuming that their preference for specific woody species would be the same across the entire distributional range, multiplied by tree density to arrive at a density estimate. This was compared to the overall density actually mapped. If bird density in an area is determined by tree availability alone, the estimates should match. Any discrepancies might then be due to other important factors, especially rainfall. The predicted bird distribution was calculated as follows:

To investigate whether bird densities vary per rainfall zone we selected five common woody species rich in birds and present in a variety of rainfall zones from semi-arid (400–600 mm rain/year) to humid (800–1200 mm rain/year). These five are Umbrella Thorn Acacia tortilis, Egyptian Acacia A. nilotica, Red Acacia A. seyal, Winter Thorn Faidherbia albida and Desert Date Balanites aegyptiaca. Since bird density differed for the western and eastern Sahel, we selected trees from a limited region where most data had been collected, i.e. the area between 1°W and 17°W. To reduce the impact of tree size, we excluded woody plants <6 m high in Faidherbia and <4 m high in other acacias and Balanites (Figure S4 in Supplementary Material 1). Faidherbia is a widely distributed tree species across many climate zones. The most northerly were recorded in desert villages in Sudan and Mauritania (17°N, 150–200 mm rain per year), the most southerly in Ethiopia and in a village in the Central African Republic (8°N; Photo 1). Faidherbia was as common in the coastal rice fields at sea-level in Guinea-Bissau (12°N and 1750 mm rain) as in the Ethiopian Highlands (up to 2300 m above sea level).

We did not attempt to quantify the food supply in trees but used an indirect measure to estimate the abundance of the moth Crypsotidia conifera, an important prey for birds present in Faidherbia. During daytime, the moths were hidden in cracks in the bark. We counted the number of flushed moths after striking the bark three times at breast level (see Figure 16 in Zwarts & Bijlsma 2015), thus providing the fraction of Faidherbia with moths per rainfall zone. In addition, we noted systematically if and when birds recorded during the standard counts were handling a large, easily visible prey, usually a caterpillar or moth. The fraction of birds with large prey was used to show variation in food abundance per rainfall zone.

RESULTS

Observed and predicted distribution of birds

Assuming that during the dry season bird densities vary between woody species (see Table 2 in Zwarts et al. 2023c) but do not vary geographically, we compared observed densities per grid cell (using data from Zwarts et al. 2023b) with predicted distributions of three migratory species (Figure 1). The total number of Western Bonelli's Warblers Phylloscopus bonelli (Figure 1A) observed in the different study sites per 100 mm rainfall zone closely resembled the predicted number (r = +0.91, P < 0.001). Within the range of Bonelli's Warbler, density apparently depended on the distribution of their preferred tree species. The correlation between observed and predicted number is even higher in the Subalpine Warbler Curruca iberiae + subalpina + cantillans (r = +0.94), but we noted that observed numbers were slightly higher than predicted when the average annual rainfall was <300 mm, and lower when rainfall was >300 mm (Figure 1B). Thus, as in Bonelli's Warbler, trees determined the distribution of Subalpine Warbler, apart from a preference for drier habitats in the latter. The same applied for Western Orphean Warbler Curruca hortensis (Figure 1C) which showed an even stronger preference for drier habitats (but correlation between observed and expected was rather small; r = +0.67, P = 0.03). Maps as in Figure 1 were also made for a selection of 5 other migrants and for 5 residents, all species occurring in a wide range of rainfall zones (Supplementary Material 2).

Photo 1.

Winter Thorn Faidherbia albida has a wide distribution in arid, semi-arid and humid zones, the extremes here exemplified by a desert village in Sudan (164 mm rainfall/year, on average; 26 January 2018; left) and a coastal rice field in Guinea-Bissau (1758 mm rainfall/year; 28 November 2014).

When plotting the ratio between observed and predicted bird numbers against annual rainfall, migrants were more common than expected in the rainfall zone <100 and 100–200 mm, but less common when rainfall exceeded 300 mm (Figure 2). The difference between expected and observed was particularly pronounced in Western Orphean Warbler and Lesser Whitethroat Curruca curruca, less so (although significant) in Subalpine Warbler and Western Bonelli's Warbler and still less, and non-significant, in three other migrants (Common Whitethroat Sylvia communis, Common Redstart Phoenicurus phoenicurus, Western Olivaceous Warbler Iduna opaca). In four resident species (Senegal Eremomela Eremomela pusilla, Tawny-Flanked Prinia Prinia subflava, Little Weaver Ploceus luteolus, Green-backed Camaroptera Camaroptera brachyura) the trend was opposite to those of the migrants, with more birds than predicted in the zone with a higher annual rainfall. No significant difference was found in Northern Crombec Sylvietta brachyura and Pygmy Sunbird Hedydipna platura.

Figure 1.

Predicted (◦) and observed distribution () of (A) Western Bonelli's Warbler, (B) Subalpine Warbler and (C) Western Orphean Warbler in the Sahel in winter (20 November – 10 March); same data as in Supplementary Material of Zwarts et al. (2022b): (Figure S10, S24 and S21, respectively). The graph shows observed and predicted total number of birds in all study sites for eleven rainfall zones.

Figure 2.

The ratio of observed and expected numbers of birds as a function of annual rainfall in (A) seven migrants and (B) six residents; original data are given in Figure 1 and in Supplementary Material 2 (Figure S5–S14). The observed numbers deviate significantly from expected in 8 of the 13 bird species and when migrants and residents are taken together (marked *: P < 0.001, using χ² tests).

Distribution of birds in preferred trees

The previous section showed that the distribution of birds and their preferred trees were correlated, but also that bird species differed in their preference for drier (most migrants) or more humid (most residents) zones. When bird species are more abundant in the arid (or humid) zone than expected given the distribution of their preferred trees, a rainfall-related shift in density in those trees might underly the observed pattern. The analysis of whether bird density (per ha canopy) was related to rainfall was done in five tree species rich in birds and found over a wide range of rainfall zones (Photo 1), unlike most other woody species that were either avoided by birds or occurred within a relatively narrow climate zone.

Faidherbia attracted 60–100 migrants per ha canopy in rainfall zones with <1000 mm per year, declining to 25–31 migrants/ha in more humid regions. Insectivorous residents were much less common in Faidherbia but showed a similar decline with rainfall as the migrants (from 20 to 6/ha). In all bird species, except two insectivorous residents, the density was significantly related to rainfall (Table S1). However, the trends differed per bird species. In migrants, density declined with rainfall in Subalpine Warbler (from 29 to 0/ha canopy) and in other Curruca species, but peaked in Western Bonelli's Warbler at 800–1000 mm rainfall (with lower numbers in drier and more humid zones). Melodious Warbler Hippolais polyglotta, and especially Willow Warbler Phylloscopus trochilus, were restricted to the most humid zone. Little Weaver was restricted to the drier zones and the only insectivorous resident that was abundant in Faidherbia. The few Pygmy Sunbirds in Faidherbia trees were exclusively observed in the dry zone, the much more numerous Beautiful Sunbird Cinnyris pulchellus only in the humid zones.

In four other tree species, each with a more restricted distribution than Faidherbia, Iberian Chiffchaff, Subalpine Warbler and Western Orphean Warbler reached their highest densities in the driest zones. Opposite trends were found in Western Bonelli's Warbler and Common Whitethroat: increasing densities with rainfall in Balanites, declining densities in Acacia seyal and A. nilotica, and no trend in A. tortilis. Common Redstart and insectivorous residents were not recorded in the arid zone.

Within their preferred tree species Iberian Chiffchaff, Western Orphean and Subalpine Warbler consistently occurred in higher densities than expected in dry habitats, and two other migrants (Willow and Melodious Warbler) in higher-than-expected densities in humid habitats. No such effect was found in the remaining migrants, nor in residents.

Figure 3.

The frequency distribution of the opacity score (0 = bare tree to 4 = dense canopy) in Faidherbia (n = 2237), Balanites (n = 997), Acacia nilotica (n = 311) and A. seyal (n = 1627), separately for three rainfall zones. Selection made of trees ≥6 m high in Faidherbia and ≥4 m in other three species, from December–February.

Suitability of woody plants from the arid zone for birds

Previous work showed that Faidherbia and Acacia seyal host many insectivorous birds unless without leaves or thinly leafed (Zwarts & Bijlsma 2015). Consequently, the average number of birds per rainfall zone present in these tree species would be affected by the fraction of leafless trees. We investigated whether the opacity score (from 0 = bare to 4 = dense canopy; see Figure 3 in Zwarts & Bijlsma 2015) differed per rainfall zone for common and bird-rich tree species. In Faidherbia, we recorded more bare and sparsely leafed trees in the dry zone (<400 mm rain/year; Figure 4), but the opposite in Balanites, A. nilotica and A. seyal. A similar comparison could not be made for A. tortilis due to lack of trees in the zone with >400 mm rain, but opacity of A. tortilis did not differ in the four driest rainfall zones (<100 mm to 300–400 mm). With the exception of Faidherbia, we found no evidence that, during the dry season, in arid conditions (<400 mm rain/year) bird-rich trees were more often bare (Figure 3) and for that reason less attractive to insectivorous birds than in the more humid zone.

Figure 4.

Average number of moths flushed from the trunk of Faidherbia after three strikes, shown for six rainfall zones. The number of Faidherbia trees (≥6 m high) varied for the six categories between 192 and 631. The presence of moths differed significantly per rainfall zone (χ² = 63.6, n = 6, P < 0.001); the percentage of trees without moths increases with rainfall (linear regression analysis: r = +0.91, P <0.01).

Figure 5.

Percent of observed birds during standard counts eating a large prey as a function of annual rainfall. Number of observed birds varies per category between 78 and 925 for Faidherbia and between 100 and 804 for other acacias (excluding species restricted to floodplains: A. kirkii, A. seyal and A. nilotica). The relative frequency at which large prey were taken differs significantly per rainfall zone for Faidherbia (χ² = 44.3, n = 11, P < 0.001), but not for the other acacias (χ² = 13.2, n = 8, P = 0.07).

Using moths flushed from the bole as an indicator of food supply for insectivorous birds, 60% of the Faidherbia trees were found to have moths when the annual rainfall was <400 mm; this gradually declined to 25–30% at rainfall >700 mm. Trees with more than 20 moths (maximum 120) flushed were only recorded in areas where rainfall was <700 mm/year. In general, Faidherbia in humid zones carried fewer moths than in drier zones (Figure 4).

While searching for birds in random sites, 3% of the insectivorous birds observed in Faidherbia and 2% of birds in other acacias were recorded as handling a large prey, usually a moth or caterpillar (Figure 9 in Zwarts et al. 2023c). Handling large prey was less often recorded in the more humid zone, in Faidherbia also less often in the most arid part of its distribution area (Figure 5).

Photo 2.

Moths Crypsotidia conifera, hiding in the cracks of the trunk of Faidherbia albida, are flushed to be counted. The caterpillars of C. conifera were locally so abundant that they completely defoliated mature trees, as described by Dunham (1991) for Faidherbia along the Zambesi River. Photos were taken in agroforestry parklands near Djenné, Mali, on 26 January 2016, where all Faidherbia trees were defoliated over an area of at least dozens of km². Western Bonelli's Warbler, and in October–November also Wood Warbler Phylloscopus sibilatrix and Willow Warbler, captured moths from the trunk and from branches in the canopy of Faidherbia by hovering and making sallies.

DISCUSSION

Our data suggest that tree preferences largely explain the distribution of birds within their own distribution range. Take, for instance, the Western Bonelli's Warbler of which 56% of the birds were found in Faidherbia, a tree species common in the semi-arid zone (annual rainfall 400–600 mm), and another 27% in Acacia tortilis (300–400 mm rainfall/year) in which they reached a high density (Table 1 & 2 in Zwarts et al. 2023c). Not surprisingly, most birds were found in the rainfall zones where their preferred tree species were most common (Figure 1A). The density of Western Bonelli's Warbler in the grid cells was highly correlated with the combined woody cover of these two preferred tree species:

(linear regression: r² = 0.61, n = 84, P < 0.0001; raw data in Figure S10 in Zwarts et al. 2023b (bird density) and Figure S2 and S4 in Zwarts et al. 2023c (woody cover of Faidherbia and Acacia tortilis); selection made for all grid cells west of 8°E).

In a multiple regression with woody cover, rainfall and rainfall² using the same data, rainfall and rainfall² appeared to be far from significant (P = 0.519 and P = 0.398, respectively); the total explained variance (r² = 0.61) was the same as in the simple regression (equation 1). In the humid zone (800–1200 mm rainfall/ year) Faidherbia was rare and Bonelli's Warblers were absent in our random study sites (Figure 1A). However, since we searched systematically for Faidherbia trees in the humid zone, we found that in those rare trees Bonelli's were even twice as abundant as in the sub-humid zone (Table S1). The density of Bonelli's Warbler, however, declined in Faidherbia trees in the hyper-humid zone (>1200 mm rainfall/year; Table S1). This suggests that the distribution of the Western Bonelli's Warbler is largely determined by the availability of their preferred tree species, but only to a certain extent given their absence from Faidherbia in the hyper-humid zone.

Willow Warblers and Melodious Warblers showed the opposite trend in Faidherbia (Table S1). These two species were not or rarely recorded in Faidherbia outside the hyper-humid zone where they spend the northern winter. Their distribution was predominantly determined by rainfall and not by the availability of Faidherbia, although this tree species was commonly used in the humid zone, a few hundred km to the north, during migration in October and November (Zwarts et al. 2023c).

The presence of Western Bonelli's Warbler in Faidherbia outside their normal distribution area raises the question of whether other migrants concentrated in acacias and other thorny species in the Sahel would extend their distribution area farther south into the humid zone if their preferred trees were available. The small numbers of Sahelian migrants observed in Ghana, Togo and Benin were indeed recorded in the few available Sahelian trees (Dowsett-Lemaire & Dowsett 2014, 2019). An Acacia nilotica plantation in southern Mali (average annual rainfall 1073 mm, i.e. far beyond the normal climate range of 200–800 mm rainfall for this tree species) was found to attract many Sahelian migrants, albeit in lower densities than in their normal distribution range (Zwarts et al. 2023c). Of the bird species investigated, Curruca warblers showed, within their range of occurrence, the most pronounced preference for trees of the semi-arid and arid zone (100–600 mm rain/year; Figure 2A).

Photo 3.

Willow Warblers and Pied Flycatchers spend the northern winter in a region that receives 1000–1500 mm rain/year, but there is no escaping from the dry season (November–May) with its withered vegetation and leafless trees, as shown for the same spot in southern Senegal (14.16°W, 12.74°N; average annual rainfall 1160 mm) in the middle of the rainy season (August 1993) and at the end of the dry season (May 1984; photos: G. Gray Tappan, U.S. Geological Survey, EROS Center, USA).

The distribution of the investigated bird species is determined by the availability of their preferred tree species, with a subsidiary role of rainfall of which the importance differs per bird species. These preferences do not explain why some migratory arboreal bird species spend the northern winter in arid zones when most residents and many migrants are exclusively found in more humid zones (Figures 1 and 2). Several alternative explanations, not necessarily mutually exclusive, can be formulated:

Avoidance of competition has been suggested as the explanatory variable (Rabøl 1987, Leisler 1992), but field data – even when collected on a large-scale – without experiments and without detailed information on habitat choice, food availability and use of seasonal food resources are not suitable to test this hypothesis (review by Salewski & Jones 2006, but see Powell et al. 2021). Nonetheless, interference competition is likely to occur in the Sahel because agonistic interactions are common. Olivaceous Warblers and Curruca species, mostly confined to the semi-arid and arid zone, are the most intolerant species and chase all other bird species. Such aggressive behaviour is rare among residents (Salewski et al. 2007; Zwarts et al. 2023d). The selection of humid habitats by residents may have to do with the presence of aggressive migrants in the arid zone, but apart from circumstantial evidence this is uncertain.

Avoiding parasites: For migrants it might be important to avoid the tropical humid zone and its associated high risk of exposure to avian parasites and their insect vectors (Waldenström et al. 2002, Piersma & van der Velde 2012). In fact, tropical rainforest is largely devoid of Palearctic migrants (Newton 1996).

Lower predation risk: Arboreal birds in arid zones face very few – and mostly aerial at that – avian predators, whereas the raptor guild in the humid, wooded zone is more varied with a higher density; predation risk must vary accordingly from almost nil to substantial (Thiollay 1998, Bijlsma 2001, Figure 4 in Zwarts et al. 2018).

Cover from, and insects in, foliage: In four of the five bird-rich tree species, trees are not better leafed in the humid zone than the same trees in the arid zones (Figure 3). In fact, many trees in the humid zone – just like the ones in the semi-arid and arid zone except Faidherbia – shed their leaves in the dry season (Photo 3), and have a near-similar phenology of flowering and fruiting (Mahamane et al. 2007).

Food abundance: In the most important woody species, Faidherbia, moths (Figure 4), and by default caterpillars, can be abundant. Thorny trees in general – the dominant woody plants in the Sahel – are richer in herbivorous insects than non-thorny trees, the latter in the humid zones replacing thorny trees (Colgan et al. 2015, Zwarts et al. 2015).

Prey size: Birds residing in Faidherbia and other acacias in the arid zone more often take large prey (Figure 5), conducive to a high food intake rate.

Wintering in an arid environment has advantages for arboreal migrant birds, but it is also a risky strategy because the annual rainfall in the drier zone is more erratic, and thus less predictable, than in the humid zones (Hiernaux & Le Houérou 2006; Figure 8 and 11 in Zwarts et al. 2009). As a consequence, herbivorous insects may be unreliable as a food resource because in dry years woody plants shed their leaves earlier or – when retaining foliage in the dry season – have fewer leaves during droughts (Bille 1974, Poupon 1980). Many woody plant species in semi-arid and arid climate zones drop their leaves in the dry season anyway. Even if the average food supply in the arid zone is higher than farther south, the mortality risk of birds, averaged over a series of years, would be higher during prolonged droughts, such as between 1969 and 1992. Many migratory birds are not restricted to the arid zone, however. Several species (e.g. Willow and Wood Warbler, Spotted Flycatcher Muscicapa striata and European Pied Flycatcher Ficedula hypoleuca) are common in the Sahel during passage in the early dry season (October–November) but spend the rest of their wintering period farther south (Zwarts et al. 2023e). They may profit from the high food supply in savannah trees for some weeks during migration periods without having to cope with the harsh conditions there during the dry season in drought years. However, adverse effects of droughts spill over into vegetation zones south of the Sahel. Leaving the Sahel in late autumn does not equate with escaping from severe drought (Photo 3, see also Figure 11 in Zwarts et al. 2009, Nicholson et al. 2018), even when considering the likelihood that adverse effects of droughts are less spectacular for birds wintering in humid zones compared to those staying semi-arid and arid zones.