Arrival and Peak Abundance of Barn Swallows Hirundo rustica in Three Regions of South Africa in Relation to Climate Indices, Deduced from Bird Atlas Data

Simple Summary

We reveal that the timing of Barn Swallows’ arrival from breeding areas across Eurasia in South Africa on their southward migration is influenced by the large-scale climate indices: Scandinavian Pattern (SCA), North Atlantic Oscillation (NAO), Mediterranean Oscillation (MOI), Southern Oscillation (SOI), and temperature and precipitation in the Mediterranean region and the Sahel. These climate indices operate at Barn Swallow breeding areas before they migrate to the south, and along four main migration flyways used by different migratory populations that meet in three regions of South Africa: Greater Gauteng, Greater Cape Town and Greater Durban. Based on the checklists from the Second Southern African Bird Atlas Project (SABAP2) from 2007 to 2024, our exploratory analysis showed that Annual Anomalies of Barn Swallow arrival timing and abundance in these three regions were related to the climate indices in the areas they visit before arriving in South Africa.

Abstract

For three regions of South Africa (Greater Gauteng, Greater Cape Town and Greater Durban) with the greatest coverage by bird atlas (SABAP2) fieldwork, we related arrival timing and abundance in each region of a long-distance migrant bird, the Barn Swallow, from July 2007 to March 2024. Using monotonic regression, from the atlas data we derived (1) the “annual anomaly of arrivals” from the average multi-year pattern; (2) the “average maximum” reporting rate at the completion of arrivals, in each region. We related these measures of the Barn Swallow timing and abundance in each of the 17 seasons of arrivals (July–January) in each of three region, with the large-scale climate indices, and temperature and precipitation in the Iberian+Apennine Peninsulas, for the Balkan Peninsula, and for the Sahel, averaged for the months when Barn Swallow visits areas between where these climate indices operate, at their breeding grounds in Eurasia, and along four southwards migration routes to South Africa. We used multiple regression modes with no more than two climate indices at a time, out of 84 explanatory variables, allowed by 17 data points (seasons) for each region. Our exploratory analysis indicated that the timing and abundance of Barn Swallow arrivals were related to a selection of these climate indices. The related climate indices varied between three regions in a pattern corresponding with the proportions of Barn Swallows arriving there from different breeding areas along different flyways, derived from an earlier study based on ringing recoveries. The paper shows the potential of the SABAP2 database as an annual monitoring approach, primarily due to the strong fieldwork protocol. We recommend that the project be continued indefinitely.

1. Introduction

Shifts in phenology of spring arrivals of Barn Swallows Hirundo rustica in Europe have been well-documented and related to changes in temperatures and rainfall at their stopover and breeding sites across Europe and northern Africa, e.g., Refs. [1,2,3,4]. The timing of arrivals of long-distance migrants from Africa to Europe has been shown to be shaped by different climate factors that influence migrants at their previous life stages [5,6,7,8,9,10]. However, attempts to relate arrivals of long-distance migrants in their staging grounds in Africa with conditions they experience during their post-breeding migration to the south have been scarce. This paper relates the timing of arrival in South Africa, after southwards migration from Eurasia, of a long-distance migrant, the Barn Swallow, with large-scale climate indices, which reflect environmental conditions along their migration from Europe and Africa, similar to those we used in earlier papers on long-distance migrants.

The Barn Swallow is a common farmland passerine bird breeding across Europe and Asia, of the Least Concern status [11], but its European breeding population shows a moderate decline since the 1990s [12]. Barn Swallows are exclusive insectivores and almost exclusively aerial feeders, which are particularly prone to weather-induced mortality [2,13,14]. Barn Swallows are long–distance migrants, which twice a year cross up to 10,000 km between their breeding grounds in the northern hemisphere, where they stay during April–August and raise up to two broods per season, and non-breeding grounds in sub-Sahelian Africa, including South Africa, where they arrive between October and January and stay till March; the migration to the south takes place during September–October, and migration to the north occurs during March–May [13]. Barn Swallows which stay in the non-breeding season in South Africa are a mixture of populations from the entire breeding range from Western Europe to the western part of Asia, and from these wide areas they migrate to non-breeding grounds in southern Africa using the main migration flyways of birds withing the Eurasian-African bird migration system, i.e., the Western, Central and Eastern Flyways between Europe and Africa, and the West Asia–East Africa Flyway (Figure 1A) [13,15,16,17,18].

The mixture of Barn Swallows arriving during the non-breeding season in South Africa from different parts of their breeding grounds is not uniform, and the proportions in which birds from different regions of Europe and western Asia are represented vary across South Africa [19]. The percentage of Barn Swallows staying in South Africa splits between swallows coming from the western, central and eastern Palearctic (cut lines at 10° E and 60° E; Figure 1A), differed between regions of the country (Figure 1B), and were respectively estimated to be 16:80:4 in Greater Gauteng, 0:46:54 in the Greater Durban area, and 34:35:31 in Greater Cape Town area [19]. Neither do all the birds staying in one area of the non-breeding range come from a single part of the breeding range, as Szép et al. (2006) [20] tried to imply. The connectivity between the breeding grounds and non-breeding grounds, based on ring recoveries, has been estimated for Barn Swallows as moderate [17]. For Barn Swallows breeding in Eastern Europe, the likelihood of staying in southern Africa was estimated at 100%, for those from northwestern Europe at 91.1%, for Northern Europe at 75%, and at 11.8% for Central Europe, with the border between Central and Eastern Europe set along the border between Poland and Ukraine, at c. 20°–30° E [17].

Considering their mixed origin, we expected that the pattern of Barn Swallow arrival and peak abundance in South Africa would reflect a combination of the effects of climate factors they experience across their breeding grounds, and during migration to South Africa from Europe along the Western, Central, Eastern Flyways and along Western Asia–East Africa Flyway, which likely also reaches South Africa (Figure 1A).

Figure 1. Geographical range of the Barn Swallow, its simplified migration routes between Europe and Africa, the study area, and regions where the climate indices operate. (A) Breeding (yellow) and non-breeding (blue) areas of the Barn Swallows, their migration routes between Europe and Africa and the areas influenced by the climate indices. Black lines = cut lines (at 10° E and 60° E) between the Western, Central and Eastern Palaearctic, and the eastern breeding limit (ca 90° E) of Barn Swallows arriving in South Africa, as in [19]. Green arrow = the Western Flyway, grey arrow = the Central Flyway, blue arrow = the Eastern Flyway, and brown arrow = the East Asia-East Africa Flyway, based on ringing recoveries and geolocator tracks [15,16,17,18]. Purple frames = areas of the Iberian + Apennine Peninsula, and the Balkan Peninsula; green frames = regions of the Sahel, from which we used monthly averages of temperature (T) and precipitation (P) for: the Iberian + Apennine Peninsula (TIA, PIA), the Balkan Peninsula (TBL, PBL), the Eastern Sahel (TSE, PSE), and the Western Sahel (TSW, PSW). SCAND = Scandinavian Pattern, NAO = North Atlantic Oscillation Index, MOI = Mediterranean Oscillation Index, IOD = Indian Ocean Dipole, and SOI = Southern Oscillation Index. The symbols of climate indices are as in Table 1. Thick black frame = the region enlarged in Figure 1B. (B) Three study regions in South Africa analysed in this study. Map of the Barn Swallow range after [21], ranges of climate indices after [6], modified.

Figure 1. Geographical range of the Barn Swallow, its simplified migration routes between Europe and Africa, the study area, and regions where the climate indices operate. (A) Breeding (yellow) and non-breeding (blue) areas of the Barn Swallows, their migration routes between Europe and Africa and the areas influenced by the climate indices. Black lines = cut lines (at 10° E and 60° E) between the Western, Central and Eastern Palaearctic, and the eastern breeding limit (ca 90° E) of Barn Swallows arriving in South Africa, as in [19]. Green arrow = the Western Flyway, grey arrow = the Central Flyway, blue arrow = the Eastern Flyway, and brown arrow = the East Asia-East Africa Flyway, based on ringing recoveries and geolocator tracks [15,16,17,18]. Purple frames = areas of the Iberian + Apennine Peninsula, and the Balkan Peninsula; green frames = regions of the Sahel, from which we used monthly averages of temperature (T) and precipitation (P) for: the Iberian + Apennine Peninsula (TIA, PIA), the Balkan Peninsula (TBL, PBL), the Eastern Sahel (TSE, PSE), and the Western Sahel (TSW, PSW). SCAND = Scandinavian Pattern, NAO = North Atlantic Oscillation Index, MOI = Mediterranean Oscillation Index, IOD = Indian Ocean Dipole, and SOI = Southern Oscillation Index. The symbols of climate indices are as in Table 1. Thick black frame = the region enlarged in Figure 1B. (B) Three study regions in South Africa analysed in this study. Map of the Barn Swallow range after [21], ranges of climate indices after [6], modified.

The paper has three main goals, with the first two being methodological. The first aim is to develop the quantitative tools for enabling application of the database assembled by the Second Southern African Bird Atlas Project (SABAP2) to annual monitoring of bird migration, using daily reporting rate data collected by SABAP2, based on year-round dense coverage of atlas cards, using monotonic regression. The second goal is to demonstrate the value of the approach by applying it to the SABAP2 data for a long-distance migrant, the Barn Swallow, from three regions of South Africa with the most intensive bird atlas fieldwork. The third aim was to explore whether there are relationships between arrival pattern and abundance in each arrival season in the three regions, and if there are any, to measure their strength.

2. Materials and Methods

2.1. Bird Data

There is no bird observatory in southern Africa generating bird ringing data comparable to the long-term monitoring datasets generated across Europe and North America, e.g., [22,23,24]. To do a comparable analysis in South Africa, we therefore used data from the Second Southern African Bird Atlas Project (SABAP2) [25,26,27,28]. We used bird atlas data from the start of SABAP2 in July 2007 up to March 2024; the project is ongoing.

We downloaded the SABAP2 data for the Barn Swallow for the three regions of South Africa for which data volumes were largest (Greater Gauteng, the area between 25–27° S and 27–29° E; Greater Cape Town, 33–35° S and 17–21° E; and Greater Durban, 28°30′–30°30′ S and 29°30–32° E) (Figure 1B). The total number of bird atlas project checklists for the three study areas were 76,958 checklists for Greater Gauteng, 29,159 for Greater Cape Town had, and 25,481 for Greater Durban. For each arrival season we made use of the checklists collected between 1 July and 31 January of the following year, for 17 seasons between 2007/2008 and 2023/2024.

The SABAP2 data consist of comprehensive checklists of all species positively identified within units of space called pentads, 5′ north to south and 5′ east to west; the project runs continuously throughout the year [26,27]. Each checklist was made over a minimum period of two hours, frequently longer; observers were instructed to explore all potential bird habitats within the pentad. Experience demonstrated that this time period was adequate to generate a comprehensive list of species present in the pentad at the time of the fieldwork. SABAP2, and its predecessor SABAP1, have made extensive use in publications of the concept of the reporting rate for species, defined as the number of checklists on which the species was recorded divided by the total number of checklists [25,26,27,28,29]. In this paper, we make use of reporting rates calculated on a daily basis. Because fieldwork is done throughout the year, it is feasible to use the daily reporting rates to describe arrival and departure patterns of migrants, a concept pioneered in 1992 [30]. In this context, Harrison et al. (1997) [31] provided a qualitative discussion of the value of, and caveats to, the use of reporting rates. For example, the reporting rate of a species depends on its conspicuousness; this in turn might vary through the annual cycle because of (a) species effects, such as the acquisition of bright plumage during the breeding season or (b) habitat effects, such as reduced foliage on trees during winter. None of these limitations apply to the reporting rates for Barn Swallows, reporting rates thus provide insights into migration phenology, including that of partial migrants [31].

Overall, there is consensus that, within a species, a relationship exists between bird abundance and reporting rates [26,27,28]. As abundance increases, so do reporting rates. A leading example of evidence for this is provided by the pattern of reporting rates for migrant species such as the Barn Swallow in southern Africa. Reporting rates steadily increased during the arrival period in the austral spring, reached a plateau during the period of residency in midsummer, and decreased during the departure period in autumn—the pattern verified what was already known [30,31,32]. More subtly, reporting rate patterns also confirmed that Western Cattle Egrets (Ardea ibis) were partial migrants in the South African Highveld; winter reporting rates decreased to about half of the summer values [33]. However, Tarboton et al. (1987) [34] considered that 90% of Western Cattle Egrets departed. It is likely that the change in reporting rate is considerably smaller than the change in relative abundance. The First Southern African Bird Atlas Project (SABAP1) made extensive use of the reporting rate concept, justified by the discovery that reporting rates provided a quantitative description which coincided with the qualitative understanding of the phenology of migration [31]. In their discussion of caveats that must be taken into account when interpreting reporting rates, Harrison et al. (1997) [31] stated that reporting rates should be thought of as an “index” of the density of a species. In other words, they considered the relationship to be monotonic; as density increases, so does reporting rate, without a particular mathematical relationship.

The initial development of the concept of reporting rates for bird species was made almost a century ago by Linsdale (1925) [35]. He considered that reporting rates would form a satisfactory index of relative abundance and provided caveats to their interpretation. Temple & Temple [36] used reporting rates in a bird atlas project in the state of Wisconsin, USA. For a selection of species for which count data were available, they used Spearman’s rank correlation to demonstrate a strong positive relationship between the counts and reporting rates. Their choice of Spearman’s rank correlation is an acknowledgement of their grasp that the relationship between abundance and reporting rate was non-linear but monotonically increasing.

For each of the three regions, we computed daily reporting rates for Barn Swallows from 1 July to 31 January the following year. Based on Earlé et al. [32], we anticipate that the daily reporting rates during July will be zero, because, apart from vagrants which failed to migrate, none of the checklists will record Barn Swallow. During the months from August onwards, we anticipate a gradual increase in daily reporting rates during the arrival period [32]. Ultimately, arrival is complete, and we anticipate a period during January, prior to departure on northwards migration, in which the daily reporting rates reach a stable maximum [32]. We calculated daily reporting rates for both the individual seasons and for the period as a whole. There were 17 seasons of data (Season 1: 1 July 2007 to 31 January 2008, …, Season 17: 1 July 2023 to 31 January 2024). This analysis considers two components of the annual reporting rate patterns: (1) the “average maximum” reporting rate reached at the completion of migration; (2) the pattern of the reporting rate plots for the individual years in relation to the 17-year pattern, i.e., the “annual anomaly of arrivals”. (1) provides a quantification of the abundance of Barn Swallows in each year; (2) provides a means to quantify the extent to which the pattern of arrival of migrants in each year was early or late relative to the 17-year pattern.

The “average maximum” for a season was defined as the median of the daily reporting rates between 11 and 31 January (of the following year—e.g., the “average maximum” reached during the 2007/2008 arrival period was taken as the median of the reporting rates for the final three weeks of January 2008). We term this the “achieved midsummer reporting rate”. We chose this period of January for this purpose because this is the month when the least movement takes place; southward migration is complete, and northward migration has not yet begun [32].

Based on the underlying concept we developed [5], we calculated an “annual arrival anomaly” to measure how early or late arrival was in each season in relation to the long-term (2007–2024) average pattern. The concept in [5] was to cumulate the number of birds ringed on a daily basis during the season to estimate the pattern and timing of arrival in that year in relation to the multi-year average. With the reporting rates we use here, the cumulative effect is built into the observation process, because individuals of the species which have already arrived are available to boost the reporting rate; in other words, the reporting rates measure not only the birds which arrived on the day, but also previous arrivals.

However, the reporting rates for individual days are based on sampling, and as a result of sampling variability, the rates do not increase steadily through the arrival period, although the overall pattern is clearly upwards. To overcome this problem, we calculated the monotonic regression of daily reporting rates in each season in each region. One of the first applications of monotonic regression appears to have been in non-metric multidimensional scaling [37,38,39], and the properties of the method and the algorithm are described there. In our context, given reporting rates r1, r2, … r215 for the days from 1 July to 31 January, the monotonic regression is a set of values d1, d2, … d215 having the properties that d1 ≤ d2 ≤ … ≤ d215, and sum(ri–di)2 is minimized. In other words, the monotonic regression creates an increasing set of numbers as close as possible to the erratically but steadily increasing reporting rates during the arrival period. The di do not increase linearly, and in fact large blocks of them are identical numbers.

We computed the monotonic regression of arrival for the overall 17-year period and for each of the 17 seasons. As done in [5], we define the anomaly for a season as the sum of the daily differences between the values of the overall monotonic regression for all 17 seasons and those from the monotonic regression for the season. These differences have a sign, positive or negative, which is retained. If the monotonic regression for the season lies above and to the left of the overall pattern, then the bulk of the differences are negative, the anomaly is negative, and arrival was early in that year. If the differences are positive, then migration was late in relation to the overall multiyear pattern, used as a baseline.

We did not calculate the anomalies over the full 215-day period from 1 July to 31 January. We did the arithmetic for the periods in which arrivals are evident in each region [32] (Gauteng: 6 October to 6 December, Cape Town: 16 October to 11 January, Durban: 12 October to 8 January).

2.2. Climate Variables

2.2.1. Selection of Climate Variables for Analysis

We used the same analytic approach as in [5,6,7]. Given the prior knowledge on migration routes of Barn Swallows (Figure 1A), we selected 13 climate indices (Table 1) which, considering the regions they operate, might influence the environmental conditions Barn Swallows experience prior to and during their migrations south from Eurasia to South Africa.

Table 1. Climate indices used as exploratory variables in modelling the abundance (“average maximum”) and the timing of arrivals (“annual arrival anomaly”) of Barn Swallows over 2007–2024 in South Africa. The table summarises the main effects of the positive phase of each climate index on weather in the listed regions of Europe and Africa in the given range of months, according to the provided references. Symbols in each first row for an index indicate a significant linear trend over 1982–2021, based on [7,8]: ↓ = decreasing trend, ↑ = increasing trend, – = no trend. The first column provides the same abbreviatIons for the colimate indices as used as symbols in Figure 1. Climate indices were downloaded from the climate services using web links provided next to the name of each index (last access to all datasets on 30 April 2025).

Symbol	Climate Index	Index Description [Main References]	Influenced Region	Winter (Nov–Feb)	Spring (Mar–Apr)	Summer (May–July)	Autumn (Aug–Oct)
SCAND	Scandinavian Pattern Index [40]	Gradient of pressure systems over western Europe and eastern Russia [41,42,43]	N Europe		warm, dry ↓	warm, dry ↓	warm, dry ↓
			W Europe		cool, wet	cool, wet	cool, wet
			C Europe		cool, wet	cool, wet	cool, wet
			E Europe		cool, wet	cool, wet	cool, wet
NAO	North Atlantic Oscillation Index [44]	Difference in the air pressure between Ponta Delgada (Azores) and Reykjavik (SW Iceland) [45,46,47,48,49]	N Europe	warm, wet −	warm, wet −	warm, dry ↓	warm, dry −
			W Europe	warm, wet	warm, wet	hot, dry	hot, dry
			C Europe	warm, dry	warm, dry	hot, dry	hot, dry
			W Mediterranean	warm, dry	warm, dry	cool, wet	cool, wet
			NW Africa	warm, dry	warm, dry
MOI	Mediterranean Oscillation Index [50]	Difference in the air pressure between Algiers (Algeria) and Cairo (Egypt) [51,52,53,54,55]	E Europe		warm, dry	warm, dry
			E Mediterranean		warm, dry	warm, dry
IOD	Indian Ocean Dipole [56]	Difference in the anomalies in sea surface temperature between the western and the eastern Indian Ocean near the equator [57,58,59,60,61]	Middle East	cool, wet ↑	cool, wet ↑	cool, wet ↑	cool, wet ↑
			E Africa	cool, wet ↑	cool, wet ↑	cool, wet ↑	cool, wet ↑
SOI	Southern Oscillation Index [62]	Difference in the air pressure between Tahiti and Darwin (Australia). La Niña = positive SOI; El Niño = negative SOI [63,64]	E Africa (Tanzania)	cool, wet ↑	cool, wet ↑	cool, wet ↑	cool, dry −
			E Africa (Lake Victoria)	cool, dry	cool, dry	cool, dry	cool, dry
			SE Africa	cool, wet	cool, dry	cool, dry	cool, dry
TIA	Monthly average temperature in the Iberian and Apennine Peninsulas [65]	Averaged daily mean temperatures for the area within 35°32′ N–46°29′ N; 10°32′ W–18°07′ E	W Mediteranean	hot ↑	hot ↑	hot ↑	hot ↑
PIA	Monthly average precipitation in the Iberian and Apennine Peninsulas [66]	Averaged daily total precipitation for the area within 35° 32′ N–46° 29′ N; 10° 32′ W–18°07′ E	W Mediterranean	wet	wet	wet	wet
TBL	Monthly mean temperature in the Balkan Peninsula [65]	Averaged daily mean temperatures for the area within 34°43′ N–44°58′ N; 18°51′–26°31′ E	E Mediterranean	hot ↑	hot ↑	hot ↑	hot ↑
PBL	Monthly average precipitation in the Balkan Peninsula [66]	Averaged daily total precipitation for the area within 34°43′ N–44°58′ N; 18°51′–26°31′ E	E Mediterranean	wet	wet	wet	wet
TSW	Monthly average temperature for the Western Sahel [65]	Averaged daily mean temperatures for the area within 15–20° N, 20° W–10° E	WAfrica	hot ↑	hot ↑	hot ↑	hot ↑
PSW	Monthly average precipitation for the Western Sahel [66]	Averaged daily total precipitation for the area within 5–20° N, 20° W–10° E	WAfrica	wet −	wet	wet ↑	wet
TSE	Monthly average temperature for Eastern Sahel [65]	Averaged daily total precipitation for the area within 15–20° N, 10° E–35° E	E Africa	hot ↑	hot↑	hot ↑	hot
PSE	Monthly average precipitation Anomaly for Eastern Sahel [66]	Averaged daily total precipitation for the area within 15–20° N, 10° E–35° E	E Africa	wet −	wet	wet ↑	wet

We used time series for 2007–2024 of monthly averages of these climate indices, downloaded from the relevant websites (Table 1).

Table 1. Climate indices used as exploratory variables in modelling the abundance (“average maximum”) and the timing of arrivals (“annual arrival anomaly”) of Barn Swallows over 2007–2024 in South Africa. The table summarises the main effects of the positive phase of each climate index on weather in the listed regions of Europe and Africa in the given range of months, according to the provided references. Symbols in each first row for an index indicate a significant linear trend over 1982–2021, based on [7,8]: ↓ = decreasing trend, ↑ = increasing trend, – = no trend. The first column provides the same abbreviatIons for the colimate indices as used as symbols in Figure 1. Climate indices were downloaded from the climate services using web links provided next to the name of each index (last access to all datasets on 30 April 2025).

Symbol	Climate Index	Index Description [Main References]	Influenced Region	Winter (Nov–Feb)	Spring (Mar–Apr)	Summer (May–July)	Autumn (Aug–Oct)
SCAND	Scandinavian Pattern Index [40]	Gradient of pressure systems over western Europe and eastern Russia [41,42,43]	N Europe		warm, dry ↓	warm, dry ↓	warm, dry ↓
			W Europe		cool, wet	cool, wet	cool, wet
			C Europe		cool, wet	cool, wet	cool, wet
			E Europe		cool, wet	cool, wet	cool, wet
NAO	North Atlantic Oscillation Index [44]	Difference in the air pressure between Ponta Delgada (Azores) and Reykjavik (SW Iceland) [45,46,47,48,49]	N Europe	warm, wet −	warm, wet −	warm, dry ↓	warm, dry −
			W Europe	warm, wet	warm, wet	hot, dry	hot, dry
			C Europe	warm, dry	warm, dry	hot, dry	hot, dry
			W Mediterranean	warm, dry	warm, dry	cool, wet	cool, wet
			NW Africa	warm, dry	warm, dry
MOI	Mediterranean Oscillation Index [50]	Difference in the air pressure between Algiers (Algeria) and Cairo (Egypt) [51,52,53,54,55]	E Europe		warm, dry	warm, dry
			E Mediterranean		warm, dry	warm, dry
IOD	Indian Ocean Dipole [56]	Difference in the anomalies in sea surface temperature between the western and the eastern Indian Ocean near the equator [57,58,59,60,61]	Middle East	cool, wet ↑	cool, wet ↑	cool, wet ↑	cool, wet ↑
			E Africa	cool, wet ↑	cool, wet ↑	cool, wet ↑	cool, wet ↑
SOI	Southern Oscillation Index [62]	Difference in the air pressure between Tahiti and Darwin (Australia). La Niña = positive SOI; El Niño = negative SOI [63,64]	E Africa (Tanzania)	cool, wet ↑	cool, wet ↑	cool, wet ↑	cool, dry −
			E Africa (Lake Victoria)	cool, dry	cool, dry	cool, dry	cool, dry
			SE Africa	cool, wet	cool, dry	cool, dry	cool, dry
TIA	Monthly average temperature in the Iberian and Apennine Peninsulas [65]	Averaged daily mean temperatures for the area within 35°32′ N–46°29′ N; 10°32′ W–18°07′ E	W Mediteranean	hot ↑	hot ↑	hot ↑	hot ↑
PIA	Monthly average precipitation in the Iberian and Apennine Peninsulas [66]	Averaged daily total precipitation for the area within 35° 32′ N–46° 29′ N; 10° 32′ W–18°07′ E	W Mediterranean	wet	wet	wet	wet
TBL	Monthly mean temperature in the Balkan Peninsula [65]	Averaged daily mean temperatures for the area within 34°43′ N–44°58′ N; 18°51′–26°31′ E	E Mediterranean	hot ↑	hot ↑	hot ↑	hot ↑
PBL	Monthly average precipitation in the Balkan Peninsula [66]	Averaged daily total precipitation for the area within 34°43′ N–44°58′ N; 18°51′–26°31′ E	E Mediterranean	wet	wet	wet	wet
TSW	Monthly average temperature for the Western Sahel [65]	Averaged daily mean temperatures for the area within 15–20° N, 20° W–10° E	WAfrica	hot ↑	hot ↑	hot ↑	hot ↑
PSW	Monthly average precipitation for the Western Sahel [66]	Averaged daily total precipitation for the area within 5–20° N, 20° W–10° E	WAfrica	wet −	wet	wet ↑	wet
TSE	Monthly average temperature for Eastern Sahel [65]	Averaged daily total precipitation for the area within 15–20° N, 10° E–35° E	E Africa	hot ↑	hot↑	hot ↑	hot
PSE	Monthly average precipitation Anomaly for Eastern Sahel [66]	Averaged daily total precipitation for the area within 15–20° N, 10° E–35° E	E Africa	wet −	wet	wet ↑	wet

We used time series for 2007–2024 of monthly averages of these climate indices, downloaded from the relevant websites (Table 1).

2.2.2. Large-Scale Climate Indices

The large-scale climate indices we used (Table 1), including the Scandinavian Pattern Index (SCAND), the North Atlantic Oscillation Index (NAO), the Mediterranean Oscillation Index (MOI), the Indian Ocean Dipole (IOD) and the Southern Oscillation Index (SOI), are derived by meteorologists as differences in the air pressure or temperatures between two selected weather stations, or areas in the ocean used for IOD [42]. These differences serve as indices that reflect the patterns of atmospheric oscillations, which determine the climate over regions much larger than just the areas between the sites selected to calculate these indices. We used these large-scale climate indices as proxies for environmental conditions over wide areas of Eurasia and Africa [46] because we have only a general knowledge of areas visited by Barn Swallows on their migration to the south, based on ringing recoveries and geolocator studies (Figure 1) [15,16,17,18]. We downloaded these large-scale indices from the databases of the Climate Prediction Center of the National Oceanic and Atmospheric Administration, US Department of Commerce National Weather Service [42] as the monthly values, which we then averaged for the selected months.

2.2.3. Precipitation and Temperature

Precipitation and temperatures in the Sahel and in the Mediterranean region are shaped by overlapping influences of several oscillation patterns, including NAO, MOI and IOD, in a gradient changing from the west to the east (Figure 1A) [53]. Thus, besides these large-scale climate indices, we also chose to use the actual rainfall and temperature conditions, which often interact, for the western and the eastern Sahel, and for the Iberian, Apennine and Balkan Peninsulas at the south of Europe.

For these regions, we used the mean daily air temperature at 2 m above the ground from the ERA5 model, averaged for the areas selected within the rectangles of coordinates (Table 1, Figure 1A). Analogously, we used daily precipitation values from the GPCC model averaged for these areas (Table 1). We downloaded both temperature and precipitation as daily values; these were provided by the World Meteorological Organisation through the Climate Explorer facility [60]. We then averaged these daily values for each month, which we then averaged for the selected months.

For southern Europe, we used temperatures and rainfall averaged for the main peninsulas in the Mediterranean region located on main migration routes of Barn Swallows to the south: the Iberian Peninsula, along the Western Flyway, the Apennine Peninsula on the Central Flyways, and the Balkan Peninsula on the Eastern Flyway (Figure 1A). We combined the Iberian and Apennine Peninsulas as one region, because temperatures in these areas were highly correlated [8]. The temperatures in this combined region in subsequent seasons were much weaker correlated (r < 0.7) with those on the Balkan Peninsula [8], which reduced any bias from using these variables in one regression model [66]. Thus, we used temperature and precipitation in the Balkan Peninsula separately.

For the Western Sahel, we downloaded temperatures and rainfall data for the same range of coordinates (Figure 1A, Table 1) as used in earlier studies to derive the Sahel Precipitation Index (SPI) [5,14,67,68,69,70], to enable comparisons. This region is used as a stopover or final non-breeding destination for Barn Swallows using the Western Flyway (wide green arrow in Figure 1A). The weather stations in the Central Sahel were too scarce to obtain reliable estimates of precipitation and temperature (http://climexp.knmi.nl). Thus, we used their temperature and rainfall anomalies averaged over the Sahel east of 10° E, including central and eastern Sahel, which we call “Eastern Sahel” in this study for convenience. This region includes likely stopover sites for Barn Swallows using the Central and the Eastern Flyways (grey and red wide arrows in Figure 1A, respectively). The main precipitation period across the Sahel is June–October, with the peak in August, and November–March is the cool and dry season [42,70]. Temperatures in the Sahel have increased since 1950, and rainfall in the whole Sahel has increased since the 1980s, followed by “greening” of the Sahel in terms of the vegetation cover [70,71,72].

For analyses, we selected 84 exploratory variables a priori (Appendix A,

Table A1. Climate indices used as explanatory variables in modelling the arrival of Barn Swallows in three regions of South Africa over 2007–2024. The last column provides the source websites from where these climate indices were downloaded (the last access to all datasets on 30 April 2025). The symbols of variables include the abbreviations of months for which the mean monthly values of each climate index were averaged.

No	Symbol of Variable	Climate Index	Source
1	SCAND_MAY_JUL	Scandinavian Pattern Index	[40]
2	SCAND_MAY_AUG
3	SCAND_JUN_JUL
4	SCAND_JUN_AUG
5	SCAND_AUG_OCT
6	SCAND_SEP_OCT
7	NAO_MAY_JUL	North Atlantic Oscillation Index	[44]
8	NAO_MAY_AUG
9	NAO_JUN_AUG
10	NAO_JUN_AUG
11	NAO_AUG_OCT
12	NAO_SEP_OCT
13	MOI_MAY_JUL	Mediterranean Oscillation Index Algiers/Cairo	[50]
14	MOI_MAY_AUG
15	MOI_JUN_JUL
16	MOI_JUN_AUG
17	MOI_AUG_OCT
18	MOI_SEP_OCT
19	IOD_MAY_JUL	Indian Ocean Dipole	[56]
20	IOD_MAY_AUG
21	IOD_JUN_AUG
22	IOD_JUN_JUL
23	IOD_SEP_OCT
24	IOD_AUG_OCT
25	IOD_NOV_FEB
26	SOI_MAY_JUL	Southern Oscillation Index	[62]
27	SOI_MAY_AUG
28	SOI_JUN_AUG
29	SOI_JUN_JUL
30	SOI_SEP_OCT
31	SOI_AUG_OCT
32	SOI_NOV_FEB
33	TIA_MAY_AUG	Monthly average temperature in the Iberian and Apennine Peninsulas (35°32′ N–46°29′ N; 10°32′ W–18°07′ E)	[65]
34	TIA_MAY_JUL
35	TIA_JUN_JUL
36	TIA_JUN_AUG
37	TIA_AUG_OCT
38	TIA_SEP_OCT
39	PIA_MAY_JUL	Monthly average precipitation in the Iberian and Apennine Peninsulas (35°32′ N–46°29′ N; 10°32′ W–18°07′ E)	[66]
40	PIA_MAY_AUG
41	PIA_JUN_JUL
42	PIA_JUN_AUG
43	PIA_AUG_OCT
44	PIA_SEP_OCT
45	TBLK_MAY_JUL	Monthly average temperature in the Balkan Peninsula (34°43′ N–44°58′ N; 18°51’–26°31’ E)	[65]
46	TBLK_MAY_AUG
47	TBLK_JUN_JUL
48	TBLK_JUN_AUG
49	TBLK_AUG_OCT
50	TBLK_SEP_OCT
51	PBLK_MAY_JUL	Monthly average precipitation in the Balkan Peninsula (34°43’ N–44°58’ N; 18°51’–26°31’ E)	[66]
52	PBLK_MAY_AUG
53	PBLK_JUN_JUL
54	PBLK_JUN_AUG
55	PBLK_AUG_OCT
56	PBLK_SEP_OCT
57	TSAHW_MAY_JUL	Monthly average temperature for the Western Sahel (15–20° N, 20° W–10° E)	[65]
58	TSAHW_MAY_AUG
59	TSAHW_JUN_AUG
60	TSAHW_JUN_JUL
61	TSAHW_SEP_OCT
62	TSAHW_AUG_OCT
63	TSAHW_NOV_FEB
64	PSAHW_MAY_JUL	Monthly average precipitation for the Western Sahel (15–20° N, 20° W–10° E)	[66]
65	PSAHW_MAY_AUG
66	PSAHW_JUN_JUL
67	PSAHW_JUN_AUG
68	PSAHW_AUG_OCT
69	PSAHW_SEP_OCT
70	PSAHW_NOV_FEB
71	TSAHE_MAY_JUL	Monthly average temperature for Eastern Sahel (15–20° N, 10° E–35° E)	[65]
72	TSAHE_MAY_AUG
73	TSAHE_JUN_JUL
74	TSAHE_JUN_AUG
75	TSAHE_AUG_OCT
76	TSAHE_SEP_OCT
77	TSAHE_NOV_FEB
78	PSAHE_MAY_JUL	Monthly average precipitation for Eastern Sahel (15–20° N, 10° E–35° E)	[66]
79	PSAHE_MAY_AUG
80	PSAHE_JUN_JUL
81	PSAHE_JUN_AUG
82	PSAHE_AUG_OCT
83	PSAHE_SEP_OCT
84	PSAHE_NOV_FEB

), based on averages of these 13 climate indices over periods of two to four months which made biological sense for Barn Swallows, in terms of the timing of their life stages (Figure 2) and places they visit during their migration to the south (Figure 1A) [14,15,16].

For example, we selected the Scandinavian Pattern Index (SCAND) between March and October, which is the period of Barn Swallows’ stay at the breeding grounds, within the influence of this oscillation pattern, but we did not consider SCAND for November–February, when Barn Swallows are south of the climatic influence of SCAND (Figure 1A). We used varied ranges of months for each climate index, e.g., NAO May–July which would likely influence conditions during Barn Swallow breeding season in Europe, but PSW May–August, as the period preceding Barn Swallows arrival in West Africa in September–October, considering wide inter- and intra-population variation in timing of breeding, migrations, and stay at the final non-breeding grounds, including South Africa (Figure 2). We also included the calendar year to control for the effect of the year [73], considering that some of the climate indices have multi-year trends [7,8], but others show no such trend.

2.3. Statistical Analyses

We used these climate variables as exploratory variables in multiple regression models, in which the midsummer reporting rates or the annual arrival anomalies over 17 seasons (2007–2024) were the subsequent response variables. We did not modify this choice of explanatory variables based on experiences doing the analysis. We did a visual inspection of the plots of the explanatory variables against year, to ensure that there were no gross errors, outliers or eye-catching peculiarities in the scatter diagrams, and to assess which variables showed any multi-year trend. We standardized all explanatory variables and response variables to mean zero and standard deviation one for the regression analyses.

With 17 data points, we limited the maximum number of explanatory variables that could be fitted in one model to be two [74,75]. This aids in avoiding the inclusion of spurious and highly correlated variables in one regression model.

Model selection used a forward stepwise procedure, allowing a maximum of two explanatory variables, starting with a null model, and subsequently adding variables that provide the greatest improvement at each step until a stopping rule is met. We explored the data using a combination of the RSEARCH procedure and the STEP directive in Genstat v. 22 [75]. The inratio and outratio criteria were both set to the default values of 1. Many of the variables were closely correlated [7,8]. Thus, in most cases, there were multiple alternative models with almost identical percentages of variances explained and similar values for the Akaike Information Criterion. We aimed to choose biologically meaningful models, guided by our cumulative eight decades of experience. Because of the forward stepwise approach to variable selection and the human choice between closely equivalent models, the statistical results have to be treated as exploratory rather than as confirmatory. Thus, we make no claims in terms of statistical significance; rather, we interpret our models at face value, tabulating the percentage of variance explained because it provides a simple and intuitive way of evaluating model fit. In a nutshell, we were searching for models with at most two explanatory variables which explained the largest proportion of the variance in each of six contexts: three regions, each with two measures (the Annual Anomalies and the Midsummer reporting rates). By generating scatterplots, we checked whether the stepwise procedure had chosen a model because of a single outlier point. We considered whether the selected explanatory variables were potentially biologically interesting. In choosing models for presentation in the Results, we were sensitive to the issue of multicollinearity, and avoided it using standard approaches, i.e., we avoided including highly correlated variables (at |r| > 0.7) in one model, and we monitored the Variance Inflation Factors in each model, which in none of our models exceeded VIF = 3 [67].

3. Results

Daily reporting rates for Barn Swallows in Greater Gauteng showed a general pattern of increase through the arrival period, but, as a result of sampling variation, there were frequent small decreases in the reporting rate on successive days (Figure 3A). The overall pattern of increase is captured by the monotonic regression of the daily reporting rates; this generates a non-parametric line which closely fits the observed reporting rates (Figure 3A). The overall monotonic regression lines were calculated for Greater Cape Town and Greater Durban (Figure 3B), and also for the individual years in each region in order to calculate the annual anomalies.

The timing of arrival of Barn Swallows in Greater Gauteng and Greater Durban areas was almost identical until about Day 115 (23 October) (Figure 3B). After this date, reporting rates in Greater Durban continued to increase, whereas those in Greater Gauteng started to flatten out to reach an asymptote. The arrival pattern in Greater Cape Town increased more slowly than in other regions, so that the date when the monotonic regression of reporting rate reached 0.10 in Greater Cape Town was 17 days later than in Greater Gauteng and six days later than in Greater Durban (Figure 3A,B). The monotonic regression of reporting rate reached 0.6 in Greater Cape Town, 37 and 39 days after reaching this level in Greater Gauteng and Greater Durban, respectively (Figure 3A,B). In Greater Gauteng, Greater Cape Town and Greater Durban the monotonic regressions of reporting rates reached within 0.03 of their maximum levels (0.71, 0.74 and 0.84) on Day 159 (6 December), Day 194 (10 January) and Day 191 (7 January), respectively (Figure 3B).

The annual midsummer reporting rates in the three regions were calculated as the median of the daily reporting rates for the final three weeks of January (Figure 4). There were significant trends for decreases in Greater Gauteng (r = −0.822, p < 0.001,

Table 2. Correlations between calendar year, the annual midsummer reporting rates (Mid) and the annual anomalies (Anomaly) in the three regions for the Barn Swallow over the period 2007 to 2024. In each case the sample size was 17, so the critical values at the 5% significance value, indicated by * in the table was 0.482, at the 1% level (**) was 0.606, and at the 0.1% level (***) was 0.725.

Variable	Year	Mid Greater Gauteng	Mid Greater Cape Town	Mid Greater Durban	Anomaly Greater Gauteng	Anomaly Greater Cape Town
Mid Greater Gauteng	−0.822 ***
Mid Greater Cape Town	−0.103	0.261
Mid Greater Durban	−0.823 ***	0.677 **	0.082
Anomaly Greater Gauteng	0.038	0.096	0.498 *	−0.132
Anomaly Greater Cape Town	−0.588 *	0.451	0.161	0.568 *	0.120
Anomaly Greater Durban	−0.457	0.232	0.268	0.602 *	0.065	0.552 *

) (Figure 4A) and Greater Durban (r = −0.823, p < 0.001) (Figure 4C), but this is not the case in Greater Cape Town (Figure 4B).

For each region, the annual anomalies were calculated as the areas between the overall annual arrival curve provided by the monotonic regression line (Figure 3B) and the annual arrival curves, each calculated from the monotonic regression line of the reporting rates for the year (Figure 5). The areas were assigned signs so that a positive annual anomaly indicates early arrival (i.e., the monotonic regression line for the year lies mostly to the left of the overall monotonic regression line). The annual anomalies showed no trends in Greater Gauteng (Figure 5A) and Greater Durban (Figure 5C) but showed a trend towards earlier arrival in Greater Cape Town (r = –0.588, p < 0.01) (Figure 5C). The correlations and statistical arguments for the statements in this paragraph are provided in Table 2.

3.1. Greater Gauteng

The Annual Anomaly of Barn Swallow arrival in Greater Gauteng was unrelated to the calendar year (Model 1,

Table 3. Results of regression models for the Annual Anomaly of the arrival of Barn Swallows in Greater Gauteng, 17 years of SABAP2 data (2007 to 2024). The signs in the brackets indicate the signs of the coefficients for the relationships of the listed explanatory variables. % = the proportion of variation the model accounts for; AIC = Akaike Information Criteria; nbm = not biologically meaningful.

Model	%	AIC	Explanatory Variables
1	0.7	92.51	Year (nbm)
2	19.0	89.24	Precipitation Sahel West (May–Aug) (+)
3	11.1	90.73	Temperature Sahel East (Sep–Oct) (–)
4	34.4	86.77	Precipitation Sahal West (May–Aug) (+), Temperature Sahel East (Sep–Oct) (–)
5	28.9	88.06	Precipitation Sahel West (May–Aug) (+), Indian Ocean Dipole (Jun–Jul) (–)
6	27.4	88.40	Precipitation Sahel West (May–Aug) (+), Scandinavian Pattern (Sep–Oct) (+)

, Figure 5A). Temperature and precipitation in the Sahel in periods immediately prior to the migration passage of the swallows were responsible for the largest percentage of explained variance in the Annual Anomaly (examples are Models 2, 3 and 4, Table 3). Precipitation during the rainy season in the Sahel (May to August) was positively related to the Annual Anomaly, so that if rainfall was low, then arrival was early; temperature in the Sahel was negatively related to the Annual Anomaly, so that if temperature was low, then arrival was late.

Models 5 and 6 (Table 3) are presented because they involved explanatory variables from outside the Sahel. Model 5 includes the Indian Ocean Dipole during the period several months prior to migration, and Model 6 includes the Scandinavian Index in the southward departure period.

The midsummer peak reporting rate in Greater Gauteng decreased significantly through time (

Table 4. Results of selected regression models for the midsummer peak reporting rates for Barn Swallows in Greater Gauteng, 17 years of SABAP2 data (2007 to 2024). The signs in the brackets indicate the signs of the coefficients for the relationships of the listed explanatory variables. % = the proportion of variation the model accounts for; AIC = Akaike Information Criteria.

Model	%	AIC	Explanatory Variables
1	65.3	29.53	Year (–)
2	70.9	27.54	Year (–), Precipitation Sahel East (Sep–Oct) (+)

, Figure 4A).

Model 1 (Table 4) notes that linear regression explained 65.3% of the relationship between calendar year and the Midsummer reporting rates. The regression coefficient for the explanatory variable Year, when not standardised, was −0.0138 (SE = 0.00246, t₁₅ = 5.59, p < 0.001), suggesting an annual decrease in reporting rate, when expressed as a percentage, of 1.38% per year, or 23% over the 17-year study period. The climate variable which best improved the fit of the regression model involved precipitation in the Sahel (Model 2, Table 4), but the regression coefficient was not statistically significant, and the Akaike Information Criterion decreased, indicating that this model was a poorer fit than the one with only calendar year.

3.2. Greater Cape Town

Because the Annual Anomaly for the arrival of Barn Swallows in Greater Cape Town was strongly correlated with calendar year (Table 2, Figure 5B), that was an appropriate starting point for modelling this anomaly.

Model 1 (

Table 5. Results of regression models for the Annual Anomaly of the arrival of Barn Swallows in Greater Cape Town area, 17 years of SABAP2 data (2007 to 2024). The signs in the brackets indicate the signs of the coefficients for the relationships of the listed explanatory variables. % = the proportion of variation the model accounts for; AIC = Akaike Information Criteria; nbm = not biologically meaningful.

Model	%	AIC	Explanatory Variables
1	0.7	92.51	Year (nbm)
2	19.0	89.24	Precipitation Sahel West (May–Aug) (+)
3	11.1	90.73	Temperature Sahel East (Sep–Oct) (–)
4	34.4	86.77	Precipitation Sahal West (May–Aug) (+), Temperature Sahel East (Sep-Oct) (–)
5	28.9	88.06	Precipitation Sahel West (May–Aug) (+), Indian Ocean Dipole (Jun–Jul) (–)
6	27.4	88.40	Precipitation Sahel West (May–Aug) (+), Scandinavian Pattern (Sep–Oct) (+)

) indicates that 45.8% of the variance of the Annual Anomaly was accounted for by year. However, the stepwise regression, set to select two explanatory variables, excluded year as an explanatory variable, and the two best models used three climate variables: Temperature Sahel East (Jun–Jul), Temperature Iberia-Apennines (Aug–Oct), and precipitation Balkans (Sep–Oct) (Models 5 and 6, Table 5). The four climate variables that most improved the percentage variance accounted for are given as Models 2 to 5 (Table 5). Models 2, 3 and 4 provide the results when each of these three variables is the only explanatory variable.

Because of the large number of potential explanatory variables, there are other similar models, because many of the variables were correlated. The striking result in this modelling exercise is the fact that, in spite of the strong trend in the Annual Anomaly through time, Year was not selected for any of the best models.

There was no trend in the midsummer peak reporting rates for Barn Swallows in Greater Cape Town (Model 1,

Table 6. Results of selected regression models for the Midsummer peak reporting rates for Barn Swallows in Greater Cape Town area, 17 years of SABAP2 data (2007 to 2024). The signs in the brackets indicate the signs of the coefficients for the relationships of the listed explanatory variables. % = the proportion of variation the model accounts for; AIC = Akaike Information Criteria; nbm = not biologically meaningful.

Model	%	AIC	Explanatory Variables
1	0.2	49.09	Year (nbm)
2	26.6	43.88	Southern Oscillation Index (Nov–Feb) (–)
3	12.7	46.83	North Atlantic Oscillation Index (May–Jul) (+)
4	35.4	42.61	Southern Oscillation Index (Nov–Feb) (–), North Atlantic Oscillation Index (May–Jul) (+)

, Figure 4B).

The variable which was most successful in explaining the Midsummer peak reporting rates was the Southern Oscillation Index for the coinciding midsummer period November-February; the percentage of variance explained was 26.6% (Model 2, Table 6). The forward stepwise regression analysis suggested that the combination of this variable and the North Atlantic Oscillation for May to July was the two best explanatory variables for this analysis, explaining 30.4% of the variance (Model 4, Table 6). The May-July period coincides with the immediately preceding breeding season in Eurasia. Many similar models can be obtained by varying the periods of months over which the climate variables are averaged, but the models presented here are representative.

3.3. Greater Durban

The best available model to describe the annual anomaly of the arrival of Barn Swallows in Greater Durban (Figure 5C) was Model 2, which included the Scandinavian index during the preceding breeding season, and precipitation in the Iberia-Apennines region during southward migration (

Table 7. Results of regression models for the Annual Anomaly of the arrival of Barn Swallows in Greater Durban, 17 years of SABAP2 data (2007 to 2024). The signs in the brackets indicate the signs of the coefficients for the relationships of the listed explanatory variables. % = the proportion of variation the model accounts for; AIC = Akaike Information Criteria.

Model	%	AIC	Explanatory Variables
1	21.9	95.30	Year (–)
2	56.8	86.71	Scandinavian Pattern (May–Aug) (+), Precipitation Iberia-Apennines (Aug–Oct) (+)
3	39.6	91.19	Scandinavian (May–Aug) (+)
4	11.6	97.28	Temperature Iberian+Apennine Peninsulas (Aug–Oct) (+)
5	52.7	88.17	Scandinavian (May–Aug) (+), Temperature Sahel East (Sep–Oct) (+)
6	49.6	89.17	Scandinavian (May–Aug) (+), Mediterranean Oscillation Index (Aug–Oct) (–)
7	50.1	89.02	Precipitation Sahel East (Sep–Oct) (+), Indian Ocean Dipole (Jun–Jul) (–)
8	55.6	87.17	Precipitation Sahel East (Sep–Oct) (+), Mediterranean Oscillation Index (May–Aug) (+)

). Comparing Models 3 and 4 with Model 2, it is clear that the Scandinavian index is dominant. There were multiple alternatives (Models 5 to 8) with similar explanatory power to Model 1 (Table 7).

The Midsummer peak reporting rate in Greater Durban showed a decrease through time and explained 64.4% of the variability of this variable (Figure 4C,

Table 8. Results of selected regression models for the Midsummer peak reporting rates for Barn Swallows in Greater Durban, 17 years of SABAP2 data (2007 to 2024). The signs in the brackets indicate the signs of the coefficients for the relationships of the listed explanatory variables. % = the proportion of variation the model accounts for; AIC = Akaike Information Criteria.

Model	%	AIC	Explanatory Variables
1	64.4	28.22	Year (–)
2	75.2	23.32	Year (–), Mediterranean Oscillation Index (May-Jul) (+)
3	74.0	24.12	Year (–), Temperature Balkans (May–Aug) (+)
4	63.0	29.72	Mediterranean Osc, Index (May–Jul) (+), Scandinavian Pattern (May–Aug) (+)

). When the model was given the option of adding a climate variable to the Year, the two best variables were the Mediterranean Oscillation Index (May to July) and Temperature in the Balkans (May to August), which increased the percentage variance explained to 75.2% and 74.9%, respectively (Table 8). When year was excluded as a potential explanatory variable, and the forward selection was allowed to select two climate variables, the model was almost as good as year alone, explaining 63.0% of the variance in the Midsummer peak reporting rate. The two variables were the Mediterranean Oscillation Index (May to July) and the Scandinavian Index (May to August) (Table 8). From the chosen months, these are variables that operate during the breeding season and hence are related to breeding conditions.

4. Discussion

The striking result of this research is that the inter-year variation in the timing of arrival of Barn Swallows in three regions of South Africa is influenced by sets of climate variables operating over the previous year during breeding and on southward migration. Besides this, the particular climate variables make sense in relation to the breeding areas and migration routes taken to the three regions. This discussion is structured around the classes of climate variables (precipitation, temperature, Scandinavian Pattern, and other climate indices) rather than the three regions themselves.

The explanatory variables used in this study were selected prior to any analyses being undertaken. Although the models provided insights into what might have been statistically more significant results than those presented here, we did not modify the original set of climate explanatory variables. What is striking is that, for each of the six response variables we modelled, there are plausible climate variables that “explain” substantial proportions of the variance.

The selected explanatory variables include a broad set of climate variables that span Europe and Africa. The shortness of the time period (17 years) precludes us from including more than two explanatory variables in each model [75,76,77,78]. We revealed that, as with northward migration of other long-distance migrants [7], the southward migration of the Barn Swallow is influenced by the major climate variables, which operate at different parts of the wide breeding range of the species and along its Western, Central or Eastern migration routes to the south. Similarly, in the Great Lakes region, USA, the timing of Barn Swallows’ post-breeding roosting, which is a first step of their migration to the south, was shaped by conditions, such as temperatures and precipitation, at their distant breeding and stopover sites [79].

Differences in the effects of various climate variables in the timing of arrivals and abundance of Barn Swallows between the three regions of South Africa correspond well with the pattern of proportions of the populations arriving in the country from different parts of its breeding range, based on ringing recoveries [19].

4.1. Effects of Climate Factors on the Timing of Barn Swallow Arrivals in South Africa

In all three regions, the three best-fitted regression models indicated the relationship between the timing of Barn Swallow arrivals in South Africa and climate factors that operate along the Western flyway (i.e., temperatures or precipitation in the Western Sahel), and those along the Western and Central flyways (i.e., temperatures and precipitation in the Iberian and Apennine Peninsula), and those along the Eastern flyway (i.e., temperatures or precipitation in the Balkans, the Eastern Sahel, MOI, IOD). This is in line with the results from ringing recoveries showing that Barn Swallows arrive in all regions of South Africa from wide breeding grounds using all these three migration flyways between Europe and Africa [15,16,17], and most likely also more eastern flyways from breeding areas east of about 60° E. The dataset is too limited for any further conclusions from the strengths of the relationships with different climate factors. But we expect that with a longer time-series available, the likely differences in proportions of birds using different flyways arriving in the three regions of South Africa might be manifested in the patterns of relationships of the arrival timing to different climate indices.

4.1.1. Effects of Precipitation

Wet conditions during the austral summer in Africa bring an abundance of insects [80,81,82], which provide good conditions for insectivorous migrants, including Barn Swallows, to fuel and moult feathers efficiently [83,84,85,86,87]. High rainfall and low temperatures at the African winter quarters, which are related to high Normalized Difference Vegetation Index (NDVI), which reflects high primary production [88,89], facilitated early arrival of Barn Swallows to their breeding colonies in Italy, likely through a positive effect on their condition [2]. After winters with good rainfall in western Africa, Barn Swallows arrived in Spain early in the spring [3]. On the other hand, drought had a negative effect on the body mass and moult rate of Barn Swallows using roosts in Botswana, likely through reduced food availability, while moderate to frequent precipitation benefitted their good condition and fast moult [14,90,91]. The survival of Barn Swallows was reduced when precipitation in the Sahel and South Africa was low, especially during October–March [2,4,14,92].

The positive relationships we found between the arrival anomalies and the rainfall in the Western or Eastern Sahel between May and October indicate that after low rainfall in these areas, Barn Swallow arrived in South Africa early. These results suggest that if these birds get to the Sahel when it is dry, hence inhospitable in terms of food availability, they keep moving south and thus arrive in South Africa early. If the opposite happens, i.e., if it is wet when Barn Swallows arrive in the Sahel, they do not rush through this region, using food abundance, and thus their arrival in South Africa is late. A similar explanation might be valid for precipitation on the Iberian-Apennine Peninsula and on the Balkans in the Mediterranean region.

4.1.2. Effects of Temperature

Temperature might influence Barn Swallows in two ways: indirectly, through the influence on the availability of their insect food, and directly, through the effect on their thermoregulation [2]. At their breeding grounds, flying insect abundance was high with high temperatures, but not correlated with rainfall [2,93,94,95], and high temperatures during the breeding period advanced egg laying by females [2]. On the opposite, cold, especially combined with heavy rains, might have a negative effect on Barn Swallows during migration, because of the low abundance of flying insects in such conditions [96]. Cold spells, which occur in spring (March–May) over the Sahara, might have a fatal effect on Barn Swallows, causing a shortage of food when the birds are emaciated after long passage [14]. On the other hand, Barn Swallows can also get overheated during hot conditions in Africa, especially during intensive aerial foraging or migratory flight.

The Arrival Anomaly was negatively related to the temperatures in the Eastern Sahel in June–July or September–October, depending on the region of South Africa. Thus, if the temperature was low (negative anomaly) just before or during Barn Swallows’ stay in this part of the Sahel, their arrival in South Africa was late (positive arrival anomaly). However, June–October in the Eastern Sahel are generally hot (ca 25–35 °C) and humid [97], and temperatures milder than usual might facilitate Barn Swallow feeding in the Sahel, and make them stay there longer, and thus arrive in South Africa relatively late in such seasons.

4.1.3. The Effect of the Scandinavian Pattern Mediterranean Oscillation Index on Barn Swallow Arrivals in Greater Gauteng

The positive relationship we found between the Scandinavian Pattern (SCAND) in May-August or September–October, and the Annual Anomaly of Barn Swallow arrivals of Barn Swallows in Greater Gauteng and Greater Durban indicated that with the positive SCAND, their arrivals were later than average. The positive SCAND is related to warm and dry conditions in Northern Europe, i.e., the breeding grounds for a considerable proportion of the Barn Swallows arriving in these two regions [19]. Such favourable breeding conditions can increase the breeding success of Barn Swallows, which can raise two broods [13]. In such a case, the increased proportion of young from second clutches, migrating to the south later than those from first broods, and also delayed migration of adults that had stayed at the north with the second broods longer than in less successful breeding years, might shift the overall arrival timing of the species later than usual. Interestingly, the relationship to the Scandinavian Pattern has not been pronounced for Barn Swallows arriving in the Greater Cape Town area. This might be related to the lower proportion of the Barn Swallows arriving in this region from the Central Palaearctic breeding grounds, including Scandinavia, than in the two other regions of South Africa [19].

4.1.4. The Effect of the Northern Oscillation Index on Barn Swallows’ Arrivals in South Africa

Interestingly, the only relationship of NAO with Barn Swallow arrivals in South Africa occurred for the midsummer peak reporting rates in the Greater Cape Town area, and indicated that with positive summer NAO (May-Jul) (+), Barn Swallows arrived more abundantly in this region. The positive summer NAO is related to warm and dry summers in north-western Europe [47], which is the breeding area for the greater proportion of Barn Swallows arriving in the Cape Town region than in the other two regions of South Africa [19]. We suggest the same mechanism of these late arrivals as for the Scandinavian Pattern, i.e., in years with such favourable breeding conditions, an increased number of young from second broods [13] can cause their more abundant arrival in the Cape Town region than in other years. Interestingly, the pattern of the influence of summer NAO on Barn Swallows arriving in the Cape Town area, and that of summer SCAND, operating more to the east in Europe, on those arriving in the two other regions of South Africa, corresponds with the increasing proportion of Barn Swallows of the more eastern breeding origin towards more eastern regions of south Africa [19].

4.1.5. The Effect of the Mediterranean Oscillation Index on Barn Swallows’ Arrivals in South Africa

Our results showed that the MOI1 in May–August was positively related to both the timing of arrivals (Annual Anomaly) and the abundance of Barn Swallows (Midsummer peak reporting rates) in the Greater Durban area, which indicated that with positive MOI1 in summer, Barn Swallows arrived in this area late and more abundantly than in other years. Positive summer MOI1 brings mild and wet conditions in the eastern Mediterranean region [52], which might benefit Barn Swallows during the early phase of their migration, before crossing the Mediterranean Sea. Such favourable conditions might cause them to stay in the Mediterranean region longer than usual, to utilise food resources and build fitness for farther migration south, and also promote their survival, hence increased abundance.

4.1.6. The Effect of the Indian Ocean Dipole in June–July on Barn Swallows’ Arrivals in Greater Gauteng and Greater Durban Areas

We revealed the negative relationships of the Annual Anomaly of the arrivals of Barn Swallows in two regions of South Africa: Greater Gauteng and Greater Durban, with the IOD in June-July, which indicated that with a positive IOD in these months, Barn Swallows arrived in these areas later than average. The positive IOD in June-July might cause catastrophically high rainfall in East Africa, which might last until August-September [98]. Such conditions might be unfavourable for Barn Swallows, as insects avoid flying in heavy precipitation. We suggest that in such years, Barn Swallows move through East Africa quickly and thus arrive in South Africa early, which is especially pronounced in these two regions, which likely support a considerable proportion of Barn Swallows using the Eastern Flyway [19].

4.1.7. The Effect of the Southern Oscillation Index in November-February on Barn Swallows in Greater Cape Town Area

Midsummer peak reporting rates for Barn Swallows in Greater Cape Town were negatively related to the SOI in November–February, which indicated that they arrived in this region more abundantly than usual with the negative SOI. A negative SOI, indicating El Niño event, often leads to wetter conditions in the austral summer in the Cape Town region than in other years. Such conditions might be favourable and thus attract more Barn Swallows from the neighbouring dry areas as the Karoo than usual.

4.2. Effect of the Year

Where calendar year was the dominant explanatory variable, this is likely to be a result of our lack of climate variables with the many-year trends, which would “overtake” the effect of the year in the models, from some parts of the breeding range and along the migration flyways. For example, we had only one climate variable (SCAND) reflecting conditions in the eastern Palaearctic, and none for western Asia up to 90° E, the breeding grounds of a substantial proportion of the Barn Swallows that migrate to South Africa [19].

4.3. Different Patterns of Influence of Climate Indices Between the Three Regions of South Africa

Interestingly, only the temperatures in the Eastern, but not in the Western Sahel, were related to the Barn Swallows’ arrival timing in South Africa, and this relationship was stronger for Greater Gauteng and Greater Cape Town than for Greater Durban areas, which might indicate that Barn Swallows arrive in the first two regions largely along the central and eastern flyways. The relationship of arrivals in Greater Gauteng and Greater Cape Town regions might indicate the origin of a substantial proportion of Barn Swallows arriving in these areas, which corresponds with a 0.75 transition probability between N Europe and South Africa by ringing recoveries [17]. But such a relationship was not prominent for the Greater Durban area, and we suggest that the possible explanation might be that the birds from even more east might prevail in this area.

Burman et al. (2016) [99] estimated that 80% of the Barn Swallows migrating to Greater Gauteng breed between 10° E and 60° E, effectively the eastern part of Europe. The most likely migration routes of these population to southern Africa would pass through the Apennine Peninsula (the Central Flyway), or the Balkans and the Middle East (the Eastern Flyway), across the Sahara to central or eastern Sahel, and farther east following the Nile River, and then to the south along the Great Rift Valley east of the tropical forests of central west Africa, towards the eastern part of southern Africa, including the Greater Gauteng region.

The peak of the rainy season in the Sahel is from June to September. Our results appear to show that, for Barn Swallows migrating to the Greater Gauteng region of southern Africa, if the rains in the Sahel are good, this speeds their passage through the region, resulting in early arrival (a negative Annual Anomaly).

4.4. Effects of Climate Factors on the Reporting Rates of Barn Swallow in South Africa

None of the explanatory variables provided climatic insights into the 23% decrease in midsummer reporting rates in Greater Gauteng during the study period. One likely explanation for this is that the set of climate variables we considered did not deal with the area in which the impact occurred, either on the breeding grounds (e.g., in the eastern half of Europe, mainly Russia, and western Asia), or on the migration route (e.g., the Middle East).

Juveniles arrive in southern Africa after the first southwards migration later than adults [100]. One potential explanation of earlier arrival and lower achieved maximum reporting rates in some seasons (e.g., Greater Cape Town in 2012) is thus a failed breeding season by adult Barn Swallows.

4.5. The Expected Effect of Climate Change on the Migration Patterns of Barn Swallows Between Eurasia and South Africa

The population size of Barn Swallow breeding in Europe has shown a moderate decline since the 1990s [12]. This decline has been attributed to changes in agricultural practices, including increased use of pesticides and a decrease in livestock farming, and the related decline of their aerial insect prey, on which Barn Swallows depend at all stages of their migratory life [100,101,102,103]. Breeding success and adult survival rates of migratory birds feeding on similar aerial insectivores tend to be lower in years with high NAO values and dry conditions at their non-breeding grounds [104,105,106,107]. High summer NAO is related to hot and dry conditions in northwestern Europe and the western Sahel. Such conditions might reduce the survival of juveniles due to their heat stress, as reported during heat waves [108]. Hot and dry weather might limit insect abundance at stopover and staging sites in arid areas of Europe and Africa [14,83], reducing the survival of Barn Swallows during the non-breeding season. Thus, increasing temperatures across Africa and Europe, with more frequent extreme heat waves [42], caused by climate change, do not bode well for Barn Swallows. In addition, two El Niño events, in a short sequence, in 2015/2016 and 2023/2024, reflected with extreme negative SOI, which caused the following summers at both hemispheres to be the hottest on record [108], and brought catastrophic drought to South Africa in 2016 [59], likely have not benefited Barn Swallows. However, analysis of ringing recoveries of the species showed that Barn Swallows visiting the region of Africa south of ca 9° S, including South Africa, have shortened their migration distance over 1912–2008 [109]. If this tendency continues, we can expect fewer Barn Swallows reaching the southernmost destination in South Africa, especially those prone to drought, as the Greater Gauteng area. However, IOD in November-February has also increased since the 1980s and was mostly positive in the last five years, which has been related to wet conditions in East Africa [6,110]. Climate change also involves increased precipitation and “greening” in the Sahel since the 1980s, which has a positive impact on primary productivity and insect abundance [71,72]. Thus, an increased precipitation when Barn Swallows visit West and East Africa, combined with possible shortening of their migration routes into Africa [109], which should reduce costs and risks of migration and enable them to opportunistically use suitable sites, should at least partly counteract the other negative effects of climate change.

4.6. Second Southern African Bird Atlas Project (SABAP2)

This analysis made use of bird atlas data. It uses the concept of reporting rates pioneered over the period 1982–1986 by the bird atlas of the Western Cape [111] and adopted by the First Southern African Bird Atlas Project (SABAP1) [29,30]. We extend and develop the methods used in [31] to describe bird migration by using the bird atlas reporting rates on an annual basis.

The major difference between SABAP1 and SABAP2 was one of scale; the grid cells in SABAP1 were 15′ × 15′, whereas for SABAP2 they were 5′ × 5′. However, the main reason why we were not able to include data from SABAP1 in this analysis was that, once “sufficient” data had been collected in grid cells, the project coordinators encouraged atlasers to operate in the areas with little data to increase overall coverage; a by-product of this was a reduction in data volumes in the latter years of SABAP1 in and around the big urban centres, and precisely the three regions this paper focused on (LGU pers. obs; he regrets being one of the project coordinators responsible for this decision).

At its initiation in 2007, the Second Southern African Bird Atlas Project (SABAP2) was envisaged as a short-term project to produce updated bird distribution maps. It has developed into a long-term project. This paper is the first to demonstrate the value of the protocol adopted by SABAP2 as an annual monitoring project for a migratory bird species. Overall, the monitoring value of SABAP2 can be greatly enhanced by the adoption of systematic data collection. This has been achieved by the citizen scientists in one region of South Africa, who adopted an annual coverage strategy [112,113].

4.7. Study Limitations

There are two sets of limitations to this study; one relates to the data and the other to the statistical methods. It needs to be kept in mind that reporting rates, a concept largely devised by the bird atlas projects in Africa (see Section 4.6), are not linearly related to abundance but are nevertheless monotonically related, in the sense that when abundance increases, so do the reporting rates. Where reporting rates are based on small samples of checklists, they are unreliable [31]. This is the reason why we confined our analysis to the areas of South Africa where there were extremely large sample sizes.

Compared with our other papers in this series [5,6,7,8,10], which all dealt with northward migration, our sample size is small: 17 years vs. 34 to 42. This places limitations on the number of explanatory variables that can be fitted in models; as described above, we were sensitive to the issue of overfitting [75]. As a by-product of this, we emphasise again that our results should be viewed as exploratory and not confirmatory. However, this is the first time an analysis like this has been attempted, considering southward migration. We believe that the results are so striking that publication is justified at this time, even if they are preliminary. In addition, this paper helps justify the long-term continuation of the Second Southern African Bird Atlas Project, so that papers building on this one will be based on longer time series of data.

We did attempt to increase the number of years for which data were available by including the observations from the First Southern African Bird Atlas Project (SABAP1) [30,31]. That ought to have increased the number of years of data by five (1987 to 1991). Unfortunately, that project took the view that it was developing a “snapshot” of bird distribution for the region, and once areas had “enough” data, they were “closed” for further submission of checklists [31]. That decision impacted the three regions considered in this paper. In contrast to the “snapshot” approach of SABAP1, the Second Southern African Bird Atlas Project (SABAP2) embraced the paradigm of generating a video of bird distributions [26]. This is appropriate given a background of climate change, so that the mistake made in SABAP1, of closing areas to further fieldwork, will not be repeated. In reality, the most valuable data are collected in the areas which already have vast quantities of data, so that change can more easily be detected.

5. Conclusions

The take-home message of this paper is that the migration of Barn Swallows to the south and their arrival at their southernmost non-breeding destinations are influenced by many of the major patterns of atmospheric circulation, temperature and precipitation across the species’ breeding range and its migration routes. We presented the results of the first explanatory analysis of such relationships, within the limitations of the datasets we had. Once a longer time-series on arrivals of this species, and those of other long-distance migrants, in southern Africa becomes available, a more formal analysis considering more climate variables at a time becomes feasible, to further explain which environmental conditions and in which locations might shape that and other long-distance migrants’ timing, numbers and survival at the southern end of their journeys.

The paper demonstrates the potential of the Second Southern African Bird Atlas Project as a long-term annual monitoring project for birds. The particular strengths are its strong protocol, which enables the analysis undertaken here, and the fact that, unlike most monitoring projects, it enables monitoring to be done both in space and time. We recommend its indefinite continuation of the SABAP2 project, which with time, will provide more insight into changes in birds’ distribution and timing and their response to climate change in South Africa.