Has been separated in Natalornis, on basis of sexual dimorphism, distinctive plumage (without rufous, no white in tail), nest structure and egg coloration; within present genus, however, these features (apart from nest structure) are variable, suggesting that separate genus is not warranted. DNA data (Dor et al. 2010) indicate sister relationship with H. nigrorufa, despite striking morphological differences. N population, with plumage washed violet, sometimes separated as subspecies lynesi, but differences insignificant. Monotypic.

This intra-African migratory species is threatened by the conversion and degradation of montane grassland habitat and loss of suitable underground nest sites in breeding areas, and degradation of wetland habitats and grassland in non-breeding sites. Over the past century the area of suitable habitat has reduced such that it now has a small population, which as a migratory species is cautiously treated as a single subpopulation. This decline continues at a slow rate overall, although it is rapidly disappearing from its southernmost range, and without conservation action throughout the range is unlikely to cease. The combination of a small, connected population that is declining makes this species best assessed as Endangered.

Hirundo atrocaerulea is an intra-African migrant that breeds in east South Africa, west Eswatini, east Zimbabwe, Malawi, north-east Zambia, south-western Tanzania, west Mozambique, and south-east Democratic Republic of Congo (DR Congo) (Keith et al. 1992, Evans and Bouwman 2010). It is a non-breeding visitor to north-east DR Congo (Lendu Plateau), south Uganda (M. Carswell in litt. 1999) and west Kenya (Keith et al. 1992), and increasingly in adjacent northern Tanzania (Evans et al. 2015).

Most, 74%, of the total breeding population occurs in Zimbabwe, Malawi and southern Tanzania (Evans and Bouwman 2010, Evans et al. 2015, Evans et al. 2016). The species range in South Africa and Eswatini is thought to have contracted by 74% (Evans and Bouwman 2010), where there are now between 35-50 pairs (Evans 2015). In 2019/20 only 19 active nests were located in KwaZulu-Natal, down from 45 in 2003/4 (per Anon. 2021), while only 12 pairs remained in Eswatini in 2012 (Evans et al. 2015). Breeding ceased at both the Blue Swallow Natural Heritage site in South Africa and at Mahamba Mountain in Eswatini between 2005 and 2012 (Evans et al. 2015) and it may no longer breed in Limpopo Province (Mattison 2004, Anon. 2005a, b, Evans et al. 2015). Blue Swallows no longer nest at six of 26 IBAs notified with the species present (Evans et al. 2015). It is uncommon to rare in Zambia, but the occurrence in Mozambique requires further investigation and the 50 pairs estimated across four areas may be an underestimate (Allport et al. 2021). The numbers breeding in south-east DR Congo are also uncertain.

Uganda and Kenya host an estimated 73% of the non-breeding population, with fewer than 200 mature individuals in northeastern DR Congo and a few hundred in northern Tanzania (Evans and Bouwman 2010).

Numbers for each country are assessed and revised below, resulting in an updated estimate based on information from the breeding range of 1,564-3,074 mature individuals, rounded here to 1,500-3,000 mature individuals. The large range reflects the considerable uncertainties within the available data, especially in Zimbabwe and Malawi. Uncertainty over the population in DR Congo is not addressed here due to the lack of data, rather the value based on expert opinion (Evans et al. 2015) is retained unqualified. The total population has previously been estimated at between 1,169 and 1,338 pairs (Evans et al. 2016), equivalent to a population size of 2,338-2,676 mature individuals. 
 
In deriving the revised population estimate here data used was in the form of individuals observed during the breeding season, where it is reasonable to equate numbers observed with mature individuals for this species (Evans 2008, Evans et al. 2016). The number of breeding pairs might be overestimated if the number of adult birds counted during a survey are converted to breeding pairs without taking the potential for cooperative breeding, frequently noted in parts of the range but not others (Evans et al. 2016), into account. However, as helpers are assumed to be capable of breeding should the opportunity arise and the population size is treated as the number of mature individuals, no adjustment to account for cooperative breeding is applied here. While the previous population size was conservatively placed in the band 1,000-2,499 mature individuals given the uncertainty over estimated numbers (BirdLife International 2016), the appraisal of available data from the whole breeding range indicates the minimum value was too low.

The species has been treated as occurring in multiple subpopulations on the basis of high natal fidelity (Evans et al. 2015). Three subpopulations were tentatively identified by Evans and Bouwman (2010): 1, those breeding in southeastern Democratic Republic of Congo and suspected to winter on the Lendu Plateau (c. 200 mature individuals); 2, those breeding in southeastern Malawi and southern Tanzania (Livingstone Mountains and Udzungwa Mountains) and wintering in Kenya and SE Uganda; and 3, those breeding in South Africa, Eswatini, Mozambique, northern and southwestern Malawi and southwestern Tanzania and wintering in southwestern Uganda. The latter grouping was separated on the basis of geographic distance by Evans et al. (2015) giving four 'clusters', however individuals from both the South Africa/Eswatini and Malawi/Tanzania breeding populations were suggested to have been found within a very small sample identified in from the same non-breeding area through stable isotope analysis (Wakelin et al. 2011), although that adult birds had been demonstrated to moult solely in non-breeding areas (Earle 1987) appears to undermine this finding (S.W. Evans in litt. 2023). However, there seems little biogeographic explanation for a division between those breeding in different parts of Malawi and Tanzania. While hypothesised to spend the non-breeding season on either side of Lake Victoria (Evans and Bouwman 2010), it may be that individuals wintering here originate from multiple breeding areas, as there is no apparent biogeographical barrier to birds coming from the south and therefore no barrier to mixing of dispersing birds. The apparent shift of non-breeding individuals into the Masai Mara National Park (Evans et al. 2015, eBird 2023) supports the suggestion that there may be movement between sites in the nonbreeding season. Logically this suggests that most of the population is subject to some level of exchange of individuals and therefore operate as a single subpopulation (per IUCN 2022). The separation noted for birds within the Democratic Republic of Congo (breeding in the southeast and occurring on the Lendu Plateau in the non-breeding season) suggests this may be a separate subpopulation. Comprising only around 200 mature individuals (though poorly evaluated), this would represent a maximum of 13% of the global population. However, in the knowledge that the species is fully migratory, capable of significant dispersal and that there is no firm evidence of subpopulation structure within the species, the precautionary position is adopted here and the species is assessed as a single subpopulation.

The largest reported breeding numbers are in Malawi, southern Tanzania and eastern Zimbabwe. In Malawi, most breed within Nyika National Park, where 28 territories were located in 71 km² in 1987 (Holroyd and Quinn, n.d.), with this used to generate a population of between 204-260 pairs for the central plateau above 2,200 m indicating that approaching 300 pairs were probably present (Fishpool and Evans 2001). The same data were used to suggest that given the amount of suitable habitat above 2,100 m at least 300 pairs are likely present (Dowsett-Lemaire 2006). A survey of Nyika National Park in January and February 2004, qualified as short (16 days) and noted to not cover the dry season, recorded 59 individuals but no attempt was made to derive an overall population size (Anon. 2005). It is possible that the specificity of suitable nest-sites for the species may mean that extrapolation from an apparent high density of nests in one area is inappropriate. A precautionary revision would be to halve the minimum value, suggesting a rounded preliminary figure between 100 and 300 pairs present, or 200-600 individuals. Additionally small numbers breed in Misuku Hills, North and South Viphya, Kirk Range and Mount Mulanje, with a value of 40 pairs/80 individuals assigned to sites in Malawi away from Nyika (Evans et al. 2015).

In southern Tanzania, surveys in 2008/9 and 2012 generated a total of 151 individuals observed in an estimated 1,905 km² covered, with a habitat suitability model predicting a total suitable area of between 7,598-17,177 km² and therefore a population between 602 and 1,362 individuals (Evans et al. 2016). There is one potential bias unaccounted in this data: that the transect route was not representative of the modelled suitable habitat area (Evans et al. 2016), however given the wide variance in potentially suitable habitat area used a significant error is unlikely.

Numbers in the eastern highlands of Zimbabwe were given as 620 individuals in Evans et al. (2015). Much of this is comprised of an estimated 200 pairs in Nyanga National Park and an additional 100-200 pairs patchily along the remainder of the highlands (Childes 2001, Mwizabi et al. 2003). However, Childes (2001) reported a total of 400 birds in 1988 having reduced to 271 by 1996, with 71 pairs reported from the Nyanga IBA in 2001 (per Mudereri et al. 2009). A survey was conducted in 2013/14 across the known sites within the country, recording 175 individuals from Nyanga in a surveyed area of 75 km², suggesting that the total number of pairs at this time likely exceeded 71 but does suggest a decline since 1988 (Matsvimbo et al. 2014). Significant loss of grassland across the area is noted as the driver of the decline (Matsvimbo et al. 2014). The 2013/14 survey recorded 34 individuals in a surveyed area of 5 km² in Chimanimani and only 4 individuals in each of Stapleford forest area and the Bvumba Highlands (Matsvimbo et al. 2014). At the latter site the species now appears to be restricted to a single site, with Matsvimbo et al. (2014) stating the future looks bleak for the species in the Bvumba highlands, and the survey also found little potential for larger numbers in the Stapleford forest area. This survey suggests that at a considerably lower minimum number of individuals is required to capture the potential reduction in the Zimbabwe population for this reassessment. A range of 100-200 pairs/200-400 mature individuals is assigned to Nyanga, 50-150 individuals at Chimanimani and with 10-50 mature individuals at the remaining sites, giving a precautionary revised country total of 260-600 individuals.

There is limited habitat in Zambia hence few individuals, 20-30 (Mwizabi et al. 2003), and these are considered contiguous with the Nyika plateau population (Evans et al. 2015). While there is a lack of proper survey data, very few have been observed recently suggesting even fewer, 10-15 may be present (F. Willems in litt. 2024).

In Mozambique the estimate of 50 pairs/100 individuals is considered a minimum estimate and previous reported numbers have been similar (Evans and Bouwman 2010, Little 2013, Allport et al. 2021). These are found in four or five sites, with 10-15 pairs in Catandica and Serra Choa, 25-30 pairs at 'Penhalonga' in Manica, 1-3 pairs at Mt. Tsetserra, likely breeding in the Mozambique side of Chimanimani and possible breeding in the vicinity of Mt. Gorongosa (Allport et al. 2021).

Recent, rapid declines have been recorded in the southern portion of the breeding range, estimated at 54% over 7 years between 2005 and 2012, which would equate to 77 % over 10 years (Evans et al. 2015). The population in South Africa and Eswatini was estimated at around 100 pairs (Monadjem et al. 2006) around 2000, with 84 pairs recorded in South Africa (Anon. 2005b, M. McNamara in litt. 2006) and 106 pairs in 2005 (Evans et al. 2015). By 2012 this figure was 57 pairs (Evans et al. 2015), though elsewhere was thought to potentially be only 40 pairs (McKechnie and Little 2012). Evans (2015) placed the number at between 35-50 pairs, and Evans et al. (2015) reported 39 pairs in South Africa and 12 in Eswatini, hence 102 individuals down from 84 and 22 pairs respectively in 2005. The most recent information suggests there are between 35-40 pairs in South Africa (Anon. 2021), consistent with Evans et al. (2015).

Declines are thought to be occurring in Zimbabwe (Matsvimbo et al. 2014) and also possibly Mozambique (Evans et al. 2015) but this is uncertain given the need for further survey (Allport et al. 2021). In Malawi/Zambia the important population around Nyika is thought stable and there is no evidence for a contraction in breeding range for southwestern Tanzania (Evans et al. 2015). No information on any change in the population breeding in DR Congo is available.

Numbers in non-breeding areas are poorly monitored. The species has ceased wintering around Mabira Forest and apparently also from the Busia Grasslands, attributed to loss of grassland habitat to grazing and agriculture (Ogoma 2013a), although the species was reported to still occur by the local community. However the species now occurs in the Masai Mara in Kenya and Serengeti National Park in Tanzania, apparently a recent change to the non-breeding distribution given the high level of visitation to these protected areas (Evans et al. 2015).

Based on an analysis of the change in occupied sites over the full historical span of records, an annual rate of population reduction was derived of between 0.36% overall (1850-2005) and 0.64% for South African and Eswatini if the start of declines is set to 1890 to coincide with the onset of commercial forestry expansion (Evans and Bouwman 2010). Loss of the suitable habitat to agriculture and forestry is continuing today (Monadjem et al. 2006, Evans 2015, Evans et al. 2015). Based on a linear model of reduction the population is suspected to be declining at between 4-6 % over ten years (Evans and Bouwman 2010), while a population viability model predicted an ongoing rate of 8 % decline (Evans et al. 2015). In conjunction with the rapid rate of reduction noted in the south of the breeding area these analyses infer that there is an overall continuing decline in the population.

This species breeds in montane grassland (Keith et al. 1992, Berruti et al. 1995, Holroyd and Quinn undated), preferring high rainfall, undulating areas such as mist belt grasslands of South Africa (Evans 2000), found from 850 m elevation (Anon. 2021). It favours grassland with interspersed with drainage lines in gullies or valleys, and also frequents wetland systems such as pans and small dams (Anon. 2005b). The species requires such wetlands among grassland patches for foraging (Anon. 2005b). In Eswatini, its breeding sites have been restricted to grasslands above 1,200 m, with most above 1,300 m (Monadjem et al. 2006).

Prior to its northwards migration, it may congregate akin to Barn Swallow, Hirundo rustica, giving twittering calls (Meikle 2012). 

In its non-breeding range it favours open grassland, often with bushes and trees (Keith et al. 1992, Ndang'ang'a 2004, Ndang'ang'a 2007) and marshy areas (Ndang'ang'a 2007).

Its nest is usually attached to the roof or side of a hole in the ground (relatively free of vegetation) (Turner and Rose 1989, Keith et al. 1992, Johnson 1996). Nest-sites in South Africa and Eswatini include underground sinkholes or disused aardvark Orycteropus afer burrows (Mattison 2003, Wakelin 2004, Monadjem et al. 2006), as well as artificial and artificially altered holes (Mattison 2003, Monadjem et al. 2006) such as old mining and prospecting shafts (Meikle 2010). Birds use the same nest-site for up to 30 years (Keith et al. 1992, S.L. Childes in litt. 1999, 2000), as long as the cavity remains stable (Wakelin 2004), although it is not known whether these are the same birds returning (J. Wakelin in litt. 2007). The nest consists of a cup of mud, mixed with fine grass and rootlets, and lined with feathers, and is repaired each year (Wakelin 2004). The species has been recorded breeding cooperatively in some parts of its range (facultative cooperative breeder) (Wakelin 2004, S.W. Evans in litt. 2007). However, it is not known whether helpers are related to the breeding pair (Wakelin 2004). Nesting occurs between October and March (Monadjem et al. 2006). A successful pair may produce two broods per breeding season (Wakelin 2004, Monadjem et al. 2006), with three clutches fledged from one nest in KwaZulu-Natal (J. Wakelin in litt. 2007). It lays two or three eggs, normally three; incubation lasts about 15 days and the chicks fledge after about 22 days (Wakelin 2004). The total number of active nests seems to be positively correlated with rainfall prior to breeding, probably through its influence on the abundance of insects, which the species feeds on (Monadjem et al. 2006). The species often forages in dense fog. They are well adapted to these conditions, having the most water repellent plumage of any terrestrial bird species (Rijke et al. 2000, Kylin et al. 2011). Despite this adaptation, fledging success is lower in years with more days with persistent dense fog (Evans and Bouwman 2010).

The species' nest-site fidelity (Wakelin 2004) exposes the population to the effects of habitat transformation in breeding areas, and similar fidelity to non-breeding areas is suspected. Grassland in many parts of its range is being lost to afforestation (Keith et al. 1992, Wakelin 2004), intense human settlement, cultivation (especially sugarcane [Evans 2000]), intensive livestock-farming, intense grazing, intensive grass-burning (Nasirwa and Njoroge 1997, S.W. Evans in litt. 2007, J. Wakelin in litt. 2007), invasion by non-native trees and bracken (S.L. Childes in litt. 1999, 2000) and (potentially) small-scale mining (S.W. Evans verbally 1998). More than 60% of the South African Grassland Biome Habitat has already been irreversibly transformed (Evans 2000). At five study sites in KwaZulu-Natal, 55% of the suitable grassland habitat was transformed between 1980 and 2000, and the number of active nests also halved (Wakelin 2004, M. McNamara in litt. 2006). At these sites, in addition to afforestation which doubled in area over the 20 year period and also accounted for half of the total land-use of the study area, grassland was also altered by the encroachment of woody vegetation, as a result of a lack of fuelwood harvesting by displaced communities and changes in fire regimes and timber plantation management (Wakelin 2004, J. Wakelin in litt. 2007). The ongoing rate of grassland loss has been estimated at 6% annually in KwaZulu-Natal, with a further 23% of grassland area lost since 2008 (Anon. 2021). In Mpumalanga, 61% of suitable habitat in the North-eastern Mountain Sourveld has been transformed (M. McNamara in litt. 2006). In Eswatini, 42% of potentially suitable grassland above 1,300 m has been transformed or heavily degraded since the 1950s (Monadjem et al. 2006). Suitable breeding habitat in this country has been lost mostly to afforestation for exotic timber plantations, and to a lesser extent urbanisation (Monadjem et al. 2006).

It is a similar situation in the non-breeding areas, with severe loss of grassland habitat even within protected areas used by the species (Ndang'ang'a 2007). The absence of wintering individuals from the Busia Grasslands (Kenya) in 2013 (Ogoma 2013a) despite presence in 2007 is ascribed to further grazing and conversion of grassland to sugarcane and maize, after having already lost over 60% of their grassland cover since the mid-1990s (Ndang'ang'a 2007), when they were surrounded by intensive agriculture (Ndang'ang'a 2004, Ndang'ang'a 2007), and grazed by livestock (Ndang'ang'a 2007). The remaining grasslands are further severely and immediately threatened by fragmentation, agriculture and overgrazing (Ndang'ang'a 2004). In addition people in Busia commonly trapped swallows for food, which may have included H. atrocaerulea (Ndang'ang'a 2007, P.K. Ndang'ang'a in litt. 2007). Similarly, loss of grasslands outside Ruma National Park (Kenya) is severe, with almost every available area cultivated, heavily grazed or settled (Ndang'ang'a 2007), although within the park two areas of suitable habitat are well-protected and continue to be used (Ogoma 2013b). With much of the non-breeding area in unprotected habitat, estimated at 47% in 2010 (Evans and Bouwman 2010) the scenarios described above are likely to affect a large proportion of the population and to have significant negative effects on individual condition and survival, though this is unstudied.

Additionally, drainage of wetlands is likely to affect both breeding success (Anon. 2005b) and may impact individual condition across the rest of the annual cycle.

Breeding birds are known to abandon their nests in reaction to human disturbance from excessive tourism (Mattison 2003) and land-owners (Wakelin 2004), though the extent to which this affects the population is uncertain now many nests are protected. 
Interviews with local people on the Nyika Plateau indicated that birds were used for medicinal purposes and sometimes food (Anon. 2005). There are no data on the scale or potential impact of this threat.

At the Blue Swallow Natural Heritage Site (BSNHS), South Africa, nesting success was reported to decline during years with increased fog (Evans and Bouwman 2000). Blue Swallows are well-adapted to flying in dense fog and have the most water repellent plumage of any terrestrial bird species (Rijke et al. 2000). Kylin et al. (2011) hypothesise that elevated concentrations of sulphonates in the fog, caused by a combination of increased air pollution from the Johannesburg area and volatile organic compounds (VOCs) released from exotic trees, may have increased the wettability of the blue swallow plumage and lowered nesting success. No breeding has taken place at BSNHS in recent years (Evans et al. 2015), but the final driver of the loss is uncertain.

Modelling of potential availability of suitable habitat under climate change scenarios indicates that habitat area will reduce over the next century (Mudereri et al. 2020), such that a continuing decline in the area of occupancy (AOO) can be inferred even in the absence of the ongoing destruction of grassland habitat outlined above.

Species has not been recorded in trade (Donald et al. 2024) but is the focus of ecotourism at some sites within the breeding range. People in Busia commonly trapped swallows for food, which may have included H. atrocaerulea (Ndang'ang'a 2007, P.K. Ndang'ang'a in litt. 2007). Interviews with local people on the Nyika Plateau indicated that birds were used for medicinal purposes and sometimes food (Anon. 2005). However, there are no data on the scale or potential impact of this threat.

Conservation Actions Underway
CMS Appendix I and II. In 1986, an international working group was established for the conservation and recovery of the species throughout its range (Anon. 2005b). The group has helped to prevent afforestation of the species' grasslands and wetlands in five key areas, successfully lobbied for grasslands in eight areas to be rehabilitated, and succeeded in blocking prospecting and mining in two key areas in South Africa (M. McNamara per Anon. 2005b). A workshop was held in South Africa in June 2002, followed by the publication of an international action plan (Evans et al. 2002). A South African working group has been formed to coordinate and encourage conservation of the species (Evans 2000, Holroyd and Quinn undated). In Malawi, Nyika National Park and the Mulanje area are considered secure from agricultural encroachment in the long term, owing to the cold climate and poor soils (F. Dowsett-Lemaire in litt. 1997). In Zimbabwe, c. 200 pairs breed within Nyanga National Park (S.L. Childes in litt. 1999, 2000), and less than 50 pairs breed in Chimanimani National Park (S.L. Childes in litt. 1999, 2000).  The majority of the few known pairs in Eswatini breed in Malolotja Nature Reserve (Parker 1994, Monadjem et al. 2006), with two pairs in recent years on Sibebe Mountain, which has been the focus of a community-based conservation project (Monadjem et al. 2006). In Mpumalanga and KwaZulu-Natal, artificial nest cavities have been dug (Wakelin 2004, Anon. 2005b, S.W. Evans in litt. 2007) to encourage the species into previously unoccupied areas where there are no natural nest sites (Anon. 2005b), and some remain in use (Wakelin 2004). However, artificial nests are less warm than natural nests, and the possible impacts of this for breeding in this species are, as of yet, still uncertain (Wakelin et al. 2013). Of the 44 artificial nest sites created in KwaZulu-Natal, 18 have been used once or more, but 26 remain unused (Anon. 2005b). Research is underway in South Africa to investigate the minimum size of grassland required by breeding pairs through the use of radio-tracking and wing-marking studies (Wakelin 2004). The species' breeding success in South Africa is monitored anually (Mattison 2003, S.W. Evans in litt. 2007). In South Africa, a national count (Mattison 2004, Arnott 2004, Anon. 2005a) and breeding survey in KwaZulu-Natal (Mattison 2003, Arnott 2004, Anon. 2021) and Mpumalanga (S.W. Evans in litt. 2007) are carried out annually in October-November. Environmental education programmes are ongoing in the regions of South Africa where the species breeds (M. McNamara in litt. 2006). The Ruma National Park is a stronghold for the species in Kenya (Ndang'ang'a 2007, Ogoma 2013b).

Conservation Actions Proposed
Identify key wintering sites and conserve them (L. Bennun in litt. 1999). Survey and monitor breeding population size and trend at less well-studied sites (M. Dyer in litt. 1999, S.L. Childes in litt. 1999, 2000). Control and remove non-native plants at breeding sites (S.L. Childes in litt. 1999, 2000). Assess effects of grassland fires on spread of non-native plants. Prevent further loss of grassland habitat in Kenya (Ndang'ang'a 2004). Survey the species' entire range in Eswatini to obtain data on its breeding population (Monadjem et al. 2006). In Eswatini, increase the area of protected grassland above 1,300 m (Monadjem et al. 2006). Consider the burning of grassland patches to prevent shrub and tree encroachment and encourage swarms of insects for the species to feed on (Ndang'ang'a 2007). Initiate a conservation strategy for the Busia Grasslands that can incorporate development (Ndang'ang'a 2007). Raise public awareness of the species' plight and importance of its habitat (M. McNamara in litt. 2006). Prevent any further significant loss of suitable habitat in South Africa (M. McNamara in litt. 2006). Reduce disturbance by tourists and land-owners. Study habitat use in its wintering range (S.W. Evans in litt. 2007). Determine the characteristics of drainage lines used for foraging (S.W. Evans in litt. 2007). Identify migration routes used between breeding and wintering grounds (S.W. Evans in litt. 2007). Study breeding biology in greater detail, including the effects of differences in conditions within natural and artificial nest holes (S.W. Evans in litt. 2007). Study the species' demography and longevity (S.W. Evans in litt. 2007). Study the genetics and taxonomy of the species (S.W. Evans in litt. 2007).