The Cuito catchment of the Okavango system: a vascular plant checklist for the Angolan headwaters

Abstract This paper aims to provide a baseline for conservation planning by documenting patterns of plant diversity and vegetation in the upper catchment of the Cuito River. 417 species are recorded from this region. Nine of these are species potentially new to science. Ten species are newly recorded from Angola, with an additional species only recorded previously within Angola from the northern enclave of Cabinda. The 108 new provincial records for Moxico clearly indicate the lack of collections from Angola’s largest province. We note the existence of extensive peat deposits in the Cuito river system for the first time and suggest that one of Barbosa’s vegetation types in the area needs to be reassessed.


Introduction
Internationally famous for its wildlife, the Okavango Delta wetland in northern Botswana was the 1000 th World Heritage Site to be designated by UNESCO and is surrounded by desert. The hydrology and ecology of the Delta are dependent entirely on rainfall in the highlands of central Angola, and the flow of water south and east through the Okavango's two principal tributaries, the Cuito and Cubango rivers. The Cubango system has been studied extensively in recent years (Oldeman et al. 2013), but little attention has been paid to biodiversity or conservation of the Cuito drainage.
Central and eastern Angola is overlain by deep Kalahari sands formed from uplifted and reworked deposits of an ancient palaeo-lake. The upper catchment of the Cuito and Cuanavale rivers falls mostly within Moxico Province where the plateau lies at an altitude of around 1500m, and the rivers have cut down to an elevation of around 1350 m. The landscape receives rainfall of approximately 1250 mm a year in the headwater lakes region, dropping to around 750 mm at the southern limits of the core study area which is marked by the Menongue -Longa -Cuito Cuanavale road in Cuando Cubango Province. The rainy season lasts from November to April and soils are highly leached. In consequence, they support very little agriculture (Diniz 1973). Barbosa (1970) assigned the vegetation of the region stretching from just east of Camacupa [General Machado] to Luena [Luso] and south to Longa to vegetation type 17A. This he described as dense, high, mixed (Zambesian and Congolian) miombo woodland with "chanas" or geoxylic-rich grasslands. According to Barbosa, these woodlands comprise Brachystegia species (B. spiciformis Benth. and B. longifolia Benth.) and Julbernardia paniculata (Benth.) Troupin, with some Guibourtia species, Cryptosepalum species and Marquesia species. Around Longa, the vegetation transitions into Barbosa's vegetation type 24, which he describes as a mosaic of savanna, woodland and dry forest with characteristic woody vegetation containing Brachystegia bakeriana Hutch. & Burtt Davy and Burkea africana Hook. White (1977) drew attention to the high rainfall highly leached Kalahari sand system and its peculiar flora in a seminal paper on the underground forests of Africa, extrapolating from his knowledge of similar habitats in north-west Zambia. But detailed surveys of the flora of this region are lacking.
Angola is the least intensively inventoried country in southern Africa for plants (Goyder and Gonçalves in press) -this can be seen graphically in the paucity of plant distribution records for the country (Fig. 1) at the start of the National Geographic Okavango Wilderness Project series of expeditions in 2015. Not only is the whole country under-recorded in terms of plants, but the eastern half of the country has very little geo-referenced specimen data (Marshall et al. 2016, Stropp et al. 2016, Sosef et al. 2017. Early collectors such as the Austrian botanist Friedrich Welwitsch collected along the coast, and along routes into the interior as far as Malange Province in the north and the Huíla Plateau in the south, but no further east (Albuquerque 2008, Goyder and Gonçalves in press). Swiss botanist John Gossweiler collected in all of Angola's provinces over the course of fifty years but spent very little time in central and eastern parts of the country apart from surveys of the Dundo area, Lunda Norte, in 1946 and 1948 funded by the diamond concession DIAMANG (Cavaco 1959, Figueiredo andSmith 2008). In addition, many of Gossweiler's collections are difficult to localise with outdated place names, and duplicates in herbaria accessible to the authors frequently omit locality data altogether. Slowly, as Angola has become more accessible following the end of the civil conflict in 2002, botanical surveys have resumed in areas of high endemism or conservation concern along the western escarpment (Hind and Goyder 2014), but the large eastern provinces of Moxico and Cuando Cubango remain poorly documented.
The only major expedition to study parts of the Cuito catchment botanically was the Kunene-Sambesi Expedition led by Pieter van der Kellen, and that covered only the area either side of the present-day Menongue -Longa -Cuito Cuanavale road. The expedition was in the Cuito system from 17 th December 1899 to around 4 th March 1900, and again between 4-18 th April 1900. The botany of the expedition was written up by Warburg (1903) and summarised by Figueiredo et al. (2009) who included notes on the botanist Hugo Baum and on the itinerary. Collections which form the basis of the many species described by predominantly Berlin-based botanists in Warburg (1903) and by subsequent authors were made from the Longa, Cuiriri and Cuito rivers. The area was revisited by Mendes whose 1959-1960 expedition covered the area between Kuvango [Artur de Paiva], Menongue [Serpa Pinto] and Cuito Cuanavale. Prior to the start of the Okavango Wilderness Project many species were known only from this area, and the surveys offered the chance to see if they occurred more widely.

Material and methods
The core study area is located to the south of Munhango (Figs 2, 3), and fieldwork was centred initially around the source lakes of the Cuito and Cuanavale rivers (Fig. 4), with excursions radiating from these points to the area south of Tempue and to nearby headwater lakes of other river catchments. In addition, more southerly tributaries such as the Longa (Fig. 5), Luassingua and Cuiriri river valleys were accessed from the Menongue -Cuito Cuanavale road. The darker green area towards the top left of Fig. 2 corresponds with the elevated and dissected plateau covered with moist miombo woodland which formed our core study area.
Botanical surveys were conducted at four different seasons to maximise recording of plant diversity -May-June 2015 (dry season), February-March 2016 (height of the rainy season), October-November 2016 (early rainy season) and April 2018 (late rains/ early dry season). DG took part in all four surveys and focussed principally on the higher rainfall zones of the catchment between the headwaters and the Menongue -Cuito Cuanavale road (Barbosa vegetation type 17A and its transition to vegetation type 24). FG participated in the third of these surveys, and AF focussed on the Longa and Cuiriri river valleys (transition zone between Barbosa 17A and 24 vegetation types), which were the core of Baum's study in 1899 and 1900, and which had proved to be of particular botanical interest in earlier surveys. NB, SB and MJ surveyed the Longa area and the catchment south of the Menongue -Cuito Cuanavale road in June 2015.
Plant diversity was mostly assessed through walk-over surveys of each habitat in turn. But for grasses specifically, plots were set up in February-March 2016 following the methodology of Vorontsova et al. (2016) in order to feed into wider continental assessments of natural and anthropogenic grassland diversity. One plot was set up in undisturbed valley grassland near Tempué, a second in grassland possibly cleared from plateau woodland, but apparently long established, above the Cuito source lake, and the third plot was placed in open miombo woodland on the slope immediately adjacent to the Cuito source lake.
The major vegetation types generally form discrete, readily observable units in different parts of the landscape and were categorised informally.
Herbarium collections were made in sets of four where possible and deposited in two Angolan institutions (the National Biodiversity Institute of the Ministério do Am- biente in Luanda and the Lubango Herbarium (LUBA) at ISCED-Huíla), one in the UK (Royal Botanic Gardens, Kew (K)) and one in South Africa (the SANBI Herbarium in Pretoria (PRE)). Plants covered by CITES regulations (Aloe, succulent Euphorbia, Orchidaceae) were deposited only in Angolan institutions, and identified from photographs. Plants were dried on a frame over a gas burner, using aluminium corrugates to transmit heat and dry air through the press. Collections were identified principally by DG at Kew by reference to the unrivalled tropical African collections and literature held there. Expert opinion was sought from specialists in particular plant groups: Gill Challen -Euphorbiaceae, Phyllanthaceae; Phillip Cribb -Orchidaceae; Iain Darbyshire -Acanthaceae, Linderniaceae, Orobanchaceae; Sebsebe Demissew -Asparagus; Peter Goldblatt -Gladiolus; Nicholas Hind -Compositae; Isabel Larridon -Cyperaceae; Gwylim Lewis -Leguminosae; Mike Lock -Xyridaceae, Zingiberaceae; Inger Nordal -Crinum; Jorge Paiva -Polygala; Alan Paton -Lamiaceae; Sylvia Phillips -Eriocaul-  Roux (2009). On occasion, accepted names diverge from these resources where expert opinion suggests otherwise. Where new country or provincial records are reported, Figueiredo and Smith (2008), recent taxonomic revisions, and searchable online herbarium catalogues (principally Kew (K), the Natural History Museum, London (BM) and the Tropical Institute, Lisbon (LISC)) have been used as the baselines for comparison.
Local usage of plants was documented on 5 th and 9 th March 2016 thanks to the inhabitants of Samenunga village (12°56'00"S, 018°48'54"E) who explained which plants had medicinal properties, and which were used to make items such as fish traps and beehives. Several cultural artefacts were purchased and deposited in the Economic Botany collections at Kew, where some have since been put on public display. Vouchers of the relevant plants were taken for verification at Kew.

Results
Approximately 1100 plant collections were made over the course of the four expeditions, with a further 40+ site-based observations recorded.
The principal vegetation types of the core study area are outlined below.

Moist miombo woodlands
Vast swathes of central and eastern Angola are covered in this vegetation. The most common trees we observed were Brachystegia bakeriana, B. longifolia, Cryptosepalum exfoliatum  Brachystegia bakeriana is most common near the outer margins of Cuito miombo woodland, and where the miombo patches are very small, as in the "fairy forests" near the Cuanavale source, these are dominated by this species. More extensive miombo is on the slopes is usually dominated by Julbernardia paniculata, and some plateau miombo (presumably with different soil composition) by Cryptosepalum exfoliatum subsp. pseudotaxus, which can form dense, closed canopy stands of miombo forest rather than woodland. Forest lacks the flammable grass layer that is present in woodland and under Cryptosepalum we frequently observed the presence of a hummock-forming moss not generally found elsewhere. Julbernardia paniculata was seen as the principle nectar source for honey bees during our 2016 surveys.

Swamp forest
We spent a short time in a small patch of swamp forest at the source of the Rio Cuiva (Kwanza drainage). Swamp forest appears to be rare and highly localised in Moxico, unlike in Lunda Norte where extensive formations occur along tributaries of the Kasai River (Congo drainage). The Cuiva swamp forest contained species of Guineo-Congolian affinity such as Zanthoxylum gilletii (De Wild.) P.G.Waterman and Syzygium owariense (P.Beauv.) Benth.

Seasonally burned savannas
These high rainfall grasslands receive 750-1250 mm of rain a year in the upper Cuito catchment, and are on highly leached Kalahari sand. Eastern Angola contains probably 80% of this habitat, which also extends into parts of NW Zambia and western parts of the DR Congo. This habitat is fire-adapted, and is dominated by grasses or by geoxylic suffrutices, plants with large underground woody biomass and seasonal above-ground shoots. Factors governing whether grasses dominated, or geoxylic suffrutices dominated these areas were not clear. Maurin et al. (2014) argue that across Africa, fire is the evolutionary driver of such lifeforms, whereas Finckh et al. (2016) provide convincing evidence that in upland central and eastern Angola, frost also plays a principal role, with cold air pooling in valley bottoms in the winter dry season and "burning" new shoots. Proximity to the water table limits growth of trees also.
The 2016 surveys took us to several sites with significant expanses of natural or little disturbed grasslands. They were particularly extensive near the confluence of the Cuito and Calua rivers downstream of the Cuito source lake, and the equivalent confluence downstream of the Cuanavale source lake. The third notable site was the Tempué valley grasslands. Grassland diversity plots were placed at three sites -one on the plateau above the Cuito source lake, one in the nearby miombo, and one in the Tempué valley grassland. Loudetia species dominated -L. simplex (Nees) C.E.Hubb. in open areas and L. lanata (Stent & J.M.Rattray) C.E.Hubb. in the woodland. Five to seven grass species were found in each plot. Total grass diversity in the upper Cuito-Cuanavale system was 27 species, the majority (18) occurring in open grassland. Grassland diversity appears significantly higher than in the lower altitude plateau grasslands of Lunda Norte, also dominated by Loudetia simplex (Darbyshire et al. 2011(Darbyshire et al. , 2014. Polygala robusta Gürke seems to be associated with diverse natural grassland and could perhaps be considered an indicator of good quality habitat. Another rare species encountered in this environment was the Angolan endemic Blepharis flava Vollesen, known from just eight earlier collections. Both of these species are newly recorded from Moxico. A spectacular blue-flowered Barleria is new to science and was collected at the Cuito-Calua confluence. Also new to science is a geoxylic species of Baphia (Leguminosae), a genus of around 50 species of tree and shrub -the "underground forest" life form had not been recorded in Baphia before. This taxon was only seen in one area of the upper Lungué-Bungo catchment, in plains with a rich flora of geoxylic legume species.
Burkea africana was a common tree in savanna vegetation at the Cuanavale source lake. This was encountered much less frequently in the Cuito source region.
Further south, the upper Longa valley, despite large-scale conversion to rice cultivation, has extensive areas of burned savanna, with some extremely rare species. Orthanthera gossweileri C.Norman was known only from the type, but we recollected it in the Longa valley in March 2016, and at the Cuanavale source in October 2016 extending its range some 200 km to the north.

Wetland
Wetlands tend not to be very diverse botanically, nor to have local endemics. They are however, poorly sampled in Angola.
The extensive peaty wetlands of the Cuito have a much more diverse flora than the rather limited equivalent on the Cubango, which is a much faster flowing river running through a rocky valley. Clump-or tussock-forming plants such as Eriocaulaceae and Xyridaceae are common, while plants such as Droseraceae and Lentibulariaceae are able to supplement the limited nutrients available to other plants by trapping and digesting insects or aquatic invertebrates. Sedges (Cyperaceae) are present but are not as common as preliminary palynological records might suggest (unpublished preliminary results).
The headwater lakes of the Cuito system support a wider range of open water aquatics (true water lilies (Nymphaeaceae) and other aquatics such as Nymphoides and Brasenia) than is present on the Cubango. One unusual aquatic plant encountered in the fast-flowing upper Longa river was Mayaca baumii Gürke (Mayacaceae), a nearendemic and the only old-world representative of this otherwise entirely neotropical family.
Robust river-margin plants include Gardenia imperialis K.Schum. (Rubiaceae) are present throughout both river catchments, while plants such as Tacazzea rosmarinifolia Oliv. (Apocynaceae) with rheophyic adaptations and requiring a rocky footing are found only on the Cubango. Many wetland species have their known distributions extended dramatically. Genlisea angolensis R.D.Good, for example, was formerly known in Angola from just Cuando Cubango and from one collection in the DR Congo -collections in both 2015 and 2016 demonstrate this species occurs throughout the catchment of the Cuito and Cuanavale rivers (Goyder 2016). Wetland species of Polygala and Eriocaulaceae show similar distributions. The photographic record of Crinum binghamii Nordal & Kwembeya from just N of Cuito Cuanavale demonstrates this also, as it was formerly known only from western Zambia Kwembeya 2004, Zimudzi et al. 2008). While extending the known distributions, the new limits reflect the high rainfall, low nutrient Kalahari sand ecology.
The source lakes generally have deep accumulations of unconsolidated peat at their margins. We measured these to a depth of at least five metres at the Cuito source lake. The valleys also have more consolidated peat deposits. Such deposits are rare in tropical Africa. Reiley and Page (2016), in a recent volume on tropical peatland, state that the only significant peat deposits in Angola are on the lower Cuanza River 50 km from Luanda. The upper Cuito and Cuanavale lakes and wetlands seem to have been overlooked, despite reference in the same volume to peaty deposits in the Okavango Delta in Botswana. Analysis of peat cores from these deposits in ongoing at the University of the Witwatersrand in South Africa -pollen trapped in different strata has the potential to shed light on changes in vegetation in the region over thousands of years.

Plant diversity
417 species of vascular plant were recorded from the core study area of the high-rainfall upper Cuito and Cuanavale drainage system. The Checklist was compiled principally from our own collections from the high-rainfall zone, but with some additional collection made by Hugo Baum in the transition zone to the south. The majority of Baum's collections from the Cuito drainage system were, however, made in Barbosa's drier vegetation type 24 even further to the south and are not included in this checklist. Note that Baum's specimens citing Longa as the locality refer to the river, not to the village currently known as Longa, which is at the southern limit of our core study area, nor to Baixo Longa 100 km to the S, and outside the core study area. A further point of confusion is Warburg's (1903) map showing the route of the Kunene-Sambesi Expedition places "Hadjon Longa" close to the confluence of the Longa and Cuito rivers even further south in the region of the present-day village of Nankova.
We report nine species from the core study area which are potentially new to science (Table 1). Ten species are newly recorded for Angola with an additional species which had only been recorded within Angola from the northerly enclave of Cabinda.
Orthochilus is a new generic record for the country (Table 2). But it is the new provincial records that give the clearest indication of how poorly studied the core project area has been to date -we recorded ten new records for Bié Province, ten for Cuando Cubango, and 108 for Moxico -the largest province in Angola. Botanically, the pyrophytic grassland zone between the marsh and the miombo woodland contains most of the new and interesting species. Over 40 underground forest species (whose nearest relatives are forest trees or shrubs) were recorded from this zone and as part of the ground flora of neighbouring miombo woodlands. They include Napoleonaea gossweileri Baker f. (Lecythidaceae), Trichilia quadrivalvis C.DC (Meliaceae), and an undescribed species of Baphia (Leguminosae). The Baphia was flowering profusely at ground level in the upper Lungué-Bungo catchment, where it occurred in an assemblage of other underground forest species. Baphia is a genus of 50 species of trees and shrubs in Africa and Madagascar -this is the first record of a pyrophytic underground forest species in the genus, and it appears to be a species new to science. The diversity of rubber-producing Apocynaceae species in the grassland zone was not-ed -Landolphia lanceolata (K.Schum.) Pichon, L. thollonii Dewèvre, Chamaeclitandra henriquesiana (Hallier f.) Pichon, and Raphionacme michelii De Wild. were common elements and have been used as sources of natural rubber in earlier times. Other much rarer species of Apocynaceae were also recorded from this zone, including Orthanthera gossweileri C.Norman, which we found at the source of the Cuanavale river, 200 km north of its earlier known distribution. The new species of Baphia will be described separately, along with a more detailed discussion of the geoxylic suffrutex flora of the region.
Thirty-nine legume species were recorded from the upper catchment of the Cuito Cuanavale system and were found in both open and woodland habitats. Most of the miombo trees belong to this family, but there were many herbs also. Other significant elements of the flora include Rubiaceae (26 spp.), Apocynaceae (19 spp.), Lamiaceae (20 spp.) and the genus Polygala (Polygalaceae) with 14 species recorded -each habitat had its own suite of Polygala species. Monocot diversity was also substantial, with 31 grass species recorded, 17 orchids -mostly in the marsh and grassland zones -and seven species of Gladiolus (Iridaceae).
A flame lily species, Gloriosa sessiliflora Nordal & M.G.Bingham, was recorded from Angola for the second and third times ever, by the headwater team and the Longa/Cuiriri team respectively. It was described from similar marshland habitats in western Zambia in 1998.

Discussion
Miombo woodland is generally regarded as Zambesian floristically. Nevertheless, we encountered a small but significant element of moist-miombo species with Guineo-Congolian affinities. These include several species of Apocynaceae, Uvaria angolensis Welw. ex Oliv. in the Annonaceae, Paropsia brazzaeana Baill. in the Passifloraceae and Englerina gabonensis (Engl.) Balle in the Loranthaceae. The small patch of swamp forest at the head of the Río Cuiva is also Guineo-Congolian in affinity with Syzygium owariense (Myrtaceae) and Zanthoxylum gilletii (Rutaceae) widely distributed in the Congo Basin and West Africa. Phylogenetically, Crinum binghamii (Amaryllidaceae), a wetland species, is closer to Congolian members of the genus than to Zambesian species (Nordal and Kwembeya 2004).
Cape elements in the flora were restricted to savanna or grassland habitats, sometimes where rocky substrate was encountered locally. Protea, Cliffortia and Erica are three genera with predominantly Cape affinities and species radiations.
Floristic links outside of Africa are demonstrated by a couple of wetland taxa. Mayaca baumii (Mayacaceae) is the only African species of an otherwise entirely neotropical genus and family. Mesanthemum glabrum Kimpouni (Eriocaulaceae) is allied phylogenetically to a species from Ecuador (Larridon pers. comm., unpublished work in progress). These distributions probably reflect historic transatlantic dispersal events involving birds.
Human population in the region is low, and the few villages we passed are far apart. The low-nutrient landscape does not support much agriculture. Nevertheless, one or two villages grew a diverse range of crops, and neighbouring miombo woodland was cleared for shifting maize and cassava cultivation. Habitat conversion is local but increasing in what is otherwise a remarkably intact ecosystem. Major human impact on the vegetation of the Cuito-Cuanavale system was only really apparent around the town of Cuito Cuanavale, and the section of the Longa valley affected by the large-scale rice project, although many of the grasslands are burned more frequently that they would be without human presence. Also, timber in the upper Lungué-Bungo valley is increasingly targeted as this area is closer to the provincial capital Luena than the rest of the core project area.
Many native plant species are used as medicines or for construction. It is mostly the most common species that are used.
The most frequent miombo tree, Julbernardia paniculata, is not only the principle source of nectar for honey bees but is also the preferred tree for the construction of beehives (Fig. 6). A cylinder of bark is removed from the tree (killing the individual), stapled together with stakes made from another legume timber (Bobgunnia madagascariensis), and tied together with stringy underbark from a third (Julbernardia paniculata, Brachystegia bakeriana or Cryptosepalum exfoliatum subsp. pseudotaxus). Internal bracing hoops come from flexible young shoots of either Diplorhynchus condylocarpon (Müll.Arg.) Pichon or Baphia massaiensis subsp. obovata, the permanent wooden cap at one end is made from Parinari curatellifolia Planch. ex Benth., while the removable cap at the other end is of woodland grasses, mostly Loudetia spp. Such traditional methods of construction are destructive but sustainable when population levels are low. However, harvesting of honey and production of beehives is becoming an industry, with some villages boasting of 300+ hives in active use.
Large fish traps were constructed from saplings of Englerophytum magalismontanum, tied together with fine bark string as above. The small fishtrap was constructed from the grass Loudetia densispica (Rendle) C.E.Hubb.
Locally made bark canoes were present in most lakes and major watercourses we visited (Fig. 7). These were generally made from bark of the legume tree Erythrophleum africanum and stitched together as above with strips of stringy underbark from Brachystegia or Cryptosepalum spp.
It was noted that local people have a detailed understanding of plants with active biological properties in their immediate environment and know how to use these to treat a variety of conditions in the absence of accessible western medicine.

Conclusions
Over 1100 plant collections were made during the course of the National Geographic Okavango Wilderness Project, the majority from the core project area of the upper Cuito and Cuanavale river catchments. These form the basis of what is undoubtedly the most detailed specimen-based assessment of the vegetation and plant diversity of this region.
The flora of the upper Cuito and Cuanavale system is diverse and endemism is high, although the latter has not been analysed in detail for this study. New records extend the known geographic range of many species 200 km to the north, to the headwaters of the Cuito and Cuanavale rivers. They also underline the need for further surveys in Moxico Province where 108 new provincial records were reported, and provide evidence that the absence of plant records for eastern Angola revealed on the GBIF data map of southern Africa is real, and not a data artefact. All four Protea species collected in Moxico had never been recorded there before. Barbosa (1970)'s vegetation type 17A needs to be critically reconsidered in the light of our findings in this area -we observed a fundamental change in composition of the miombo woodlands east of Cuemba once we moved onto the deep white sands, where several woody species drop out -no Isoberlinia angolensis (Benth.) Hoyle & Brenan was seen east of this point, and Brachystegia spiciformis occurred exceptionally rarely. Both Burkea africana and Brachystegia bakeriana are significant elements of the landscape in the headwater lakes region, not just in the transition zone around Longa. We saw no Marquesia species in the headwaters zone, but Monotes is common. Baphia massaiensis subsp. obovata, more commonly associated with dry Baikiaea-dominated woodland, was a common element of the miombo right up into the headwater region.
We also highlight the existence of extensive peat deposits in the Cuito river system. These are not as extensive as those recently reported from the Congo Basin (Dargie et al. 2017), but must be significant in terms of carbon storage nevertheless.

Checklist
An annotated checklist of the upper Cuito & Cuanavale drainage system -the flora of high rainfall (annual precipitation more than c. 750 mm), highly leached Kalahari sand deposits from the headwaters to c. 15°S, based prinicipally on 2015, 2016 and 2018 field surveys (Barbosa vegetation type 17A and transition to vegetation type 24).