Checklist of Serengeti Ecosystem Grasses

Abstract We present the first taxonomic checklist of the Poaceae species of the Serengeti, Tanzania. A review of the literature and herbarium specimens recorded 200 species of grasses, in line with similar studies in other parts of East Africa. The checklist is supported by a total of 939 herbarium collections. Full georeferenced collection data is made available alongside a summary checklist in pdf format. More than a quarter of the species are known from a single collection highlighting the need for further research, especially concerning the rare species and their distribution.


Introduction
The Serengeti Ecosystem Region (abbreviated as SER) is an area of 25,000 km , found south of the Tanzania and Kenya border between 2° and 4° S, defined by the movement of migratory wildebeest (McNaughton 1983). The Serengeti National Park (established in 1940, 14,700 km ) and the Ngorongoro Conservation Area (established in 1959, 8,300 km ) form the majority of the SER. Both are also designated by UNESCO as World Heritage Sites and Biosphere Reserves. There are a further five smaller reserves surrounding the Serengeti National Park; Maswa in the south; Grumeti, Ikoma, and Ikorongo in the east and the Loliondo Game Controlled Area in the west (World Bank 1993). Over the border in Kenya and contiguous to the Serengeti National Park is the Masai Mara National Reserve (World Bank 1993).

Vegetation
The SER is part of White's Somalia-Masia vegetation type (White 1983). The Serengeti plains which range from 1600-1800 m are characterised by "short" and "tall" grasslands and woody savanna with rocky outcrops called "kopje" (Anderson and Talbot 1965, Lind and Morrison 1974, Herlocker 1975. The short grasslands in the east are dominated by Sporobolus and Kyllinga species whilst in the west the tall grasslands are dominated by Pennisetum, Andropogon and Themeda species (Lind and Morrison 1974). Intermediate grasslands of Cynodon and Sporobolus are found between the short and tall grassland areas. Seventeen different grassland communities have been characterised (McNaughton 1983) : six short grassland communities, three tall and eight medium. The distribution and growth of grass species in the Serengeti plains depends on a number of factors including soil depth and texture, salt concentration, degree of wind erodability, rainfall and grazing pressure Talbot 1965Lind andMorrison 1974).
The Ngorongoro Conservation Area is dominated by a volcano formed 5-7 million years ago with a large central caldera formed 1-2 million years ago (Anderson and Herlocker 1973). The crater reaches over 3350 m high but the crater floor is lower at 1737 m (Anderson and Herlocker 1973). The main vegetation types are grasslands, swamps, montane heath and woodland (Herlocker and Dirschl 1972). Similar to the Serengeti plains, the distribution of the Ngorongoro grasslands are influenced by soils, climate, grazing pressures and fire (Anderson and Herlocker 1973).

Species
The grass flora of Tanzania is documented through the Flora of Tropical East Africa (Clayton et al. 1970, Clayton et al. 1974, Clayton and Renvoize 1982. Grassland ecology in the SER has been well studied (for example Talbot 1965, McNaughton 1983) but these studies have focused on community structure and distribution not on species composition. There are also vegetation maps for the Serengeti woody species (Herlocker 1975) and the Ngorongoro Conservation Area (Herlocker and Dirschl 1972) but full species checklists are not included. Therefore there is a requirement for a checklist of grass species to support ecological studies on the flora and fauna of the SER.

Materials and methods
To compile the checklist specimen data were gathered from literature and herbaria sources. Specimens from the Tanzania Serengeti Ecosystem region were databased at K, NHT and at the Serengeti Wildlife Research Centre's herbaria in 2013 and 2014. Specimens from AAU, MO and US were downloaded from GBIF on 20 May 2014 (GBIF 2014). Specimen data was obtained from EA herbarium on 3 September 2014 from available digitised records. All specimens were databased using BRAHMS version 7.5 (Botanical Research and Herbarium Management System) software (University of Oxford 2014). Additional specimen literature citations were added from The Flora of Tropical East Africa (Clayton et al. 1970, Clayton et al. 1974, Clayton and Renvoize 1982. All specimens where possible were georeferenced using the FTEA gazetteer (Polhill 1988) and the Serengeti official map (Hopcraft 2013). Species names, authorities, synonymy and global distribution were checked with the World Checklist of Selected Plant Families (Clayton et al. 2014).
Data was uploaded from BRAHMS following the Darwin Core Formats. The specimen barcode was set as the catalogue number when possible or if a barcode was not available a unique code from the BRAHMS Serengeti database was used. Each herbarium specimen duplicate is listed separately in the checklist. If the record was a literature citation only the institution code was set as "?" as it was unknown at which herbaria the specimen is located. Only commonly used synonyms were compiled and cited in the nomenclature section, full synonymy is available at the World Checklist of Selected Plant Families (Clayton et al. 2014). Places of publication for each species name are available in the checklist in the supplementary file.

Species
In total 939 herbarium specimens, housed in six herbaria were digitised for the SER. These represented 200 species from 77 genera. The most frequently collected species was Sporobolus ioclados (Trin.) Nees with 30 specimens (Table 1). Twenty species are known from ten or more herbarium specimens. However 54 species in the SER are only known from one herbarium record. The most commonly collected genus was Sporobolus, followed by Panicum and Pennisetum (Table 2). Most grass species recorded from the SER have a widespread distribution across Tropical Africa with only one Tanzanian endemic species recorded, Apochiton burtii C.E.Hubb. One species has not been recorded in mainland Africa before (Bothriochloa pertusa (L.) A.Camus), and this could be a new introduction or a possible misidentification and needs further investigation. We have not been able to verify reported records of Aristida pilgeri Henrad and Digitaria milanjiana (Rendle) Stapf and these species are not included in the checklist. The top ten commonly collected species in the Serengeti Ecosystem Region Table 2.

Number of specimens
The top ten commonly collected genera in the Serengeti Ecosystem Region Table 3.

Collecting intensity across SER
Specimens have been collected across the SER (Fig. 1) but there has been a collecting bias towards certain areas. For example 124 herbarium specimens have been collected in the area around the Serengeti Wildlife Research Centre and 103 specimens from the Ngorongoro Crater. The five smaller reserves (Maswa, Grumeti, Ikoma, Ikorongo and Loliondo) have few collections compared to the Serengeti National Park and the Ngorongoro Conservation Area.

Discussion
The SER has a comparable number of grass species to other Tanzania  There is also a similar low level of endemicity with only one Tanzanian endemic species found in the SER, one Tanzanian endemic in the Selous Game Reserve and no endemic species in the Mkomazi National Park. In this checklist over a quarter of species are only known from one herbarium specimen. Further research into these species is required, to determine whether they are genuinely rare or merely undercollected. The digitisation of grasses from East African herbaria, Kew and using GBIF records allowed us to find records of these undercollected species: if the checklist had been produced with only East African or only European herbarium specimens then some of these species would have been missed.
The checklist allows researchers to see which species have been commonly collected by botanists but there may be some bias. There is a strong bias in the places that botanists have collected with high numbers of specimens from near the Serengeti Wildlife Research

Distribution of grass specimens within the Serengeti Ecosystem Region
Institute and at the Ngorongoro Crater but there also could be a bias towards collection of the most attractive or well known species.