A reconnaissance survey of the vegetation of the North Luangwa National Park, Zambia

A comprehensive survey of the vegetation of the North Luangwa National Park (NLNP) was carried out over a period of two years. The main aims of the survey were to describe the major vegetation communities in the park and to produce a vegetation map of the NLNP Initial differentiation of vegetation units was established by the appearance of the vegetation on aerial photographs Further information was derived from 353 ground plots in which > 20 000 woody plants were identified and measured Thirteen broad vegetation types were recognised in the NLNP Details of their physiognomy, species composition, distrib­ ution, topography and edaphic associations are given.


INTRODUCTION
Until the present study, no detailed vegetation survey of Zambia's North Luangwa National Park (NLNP) had been carried out. Naylor et al. (1973) and Phiri (1989) have described the vegetation of the Luangwa valley, but in both of these surveys, the area covered is too large and the classification too broad to be useful to park man agers. The most detailed study of the region available is that of Astle et al. (1969), who surveyed the South Luangwa National Park, the North Luangwa National Park east of the Muchinga Escarpment, and surrounding areas. Habitat classification in this study was based on physiognomic units recognised in a series of landsystems representing all of the topographic units present in the survey area.
The close relationship between vegetation, climate, landsystems. edaphic factors is well established (Cole 1982;Bell 1984), and the rationale for producing a vegetation-based habitat classification of the NLNP was founded on the premise that vegetation is a readily iden tifiable and measurable facet of habitat (Timberlake et al. 1993). In the long term, the vegetation boundaries desig nated in this study will form part of a geographical infor mation system for the NLNP. together with soil, geolog ical. hydrological, topographical and other information which, combined, will help define habitats for individual species. Justification for studying the vegetation compo nent of habitat as a first step is that a vegetation type is readily disccrnable to managers and park personnel, peo ple who are not necessarily plant ecologists. Descriptions based on vegetation physiognomy (e.g. woodland, wood ed grassland, thicket) and four or five characteristic species should be recognisable to all.
As well as providing a valuable basis for habitat description and monitoring, floristic data are a useful measure of biological diversity. Conservation resources * The Herbarium. Royal Botanic Gardens, Kew. Richmond, Surrey. TW9 3AB. UK MS. received 1997-04-29. are scarce in Zambia, as in most developing countries, and as a signatory to the Convention on Biological Diversity, Zambia is committed to conserving as much of her biological resource as possible. Clearly, a rational approach to conservation needs to be devised in which limited financial and technical resources can be used to maximum benefit. An essential first step is to carry out a biological resource assessment (Article 7a of the Convention) with the aim of identifying areas of high biodiversity for priority protection. Zambia has set aside 63 585 km: (8.5%) of its total land mass as national park and a further 10% is designated as forest reserve. All of these areas require protection, but only a fraction have been subjected to biological inventory'. Without such basic inventory, monitoring (Article 7b). identification of adverse processes (7c) and maintenance or management (7d) of the biodiversity resource is impossible. Due to its great topographical diversity. North Luangwa National Park is potentially one of the most biologically diverse of Zambia's protected areas. The results of the present study will help to confirm this and will form baseline data against which future trends in plant diversity can be mea sured.
This paper describes the major vegetation types pre sent in the NLNP. and presents a vegetation map of the park. A comprehensive checklist of the plants collected during this study is published elsewhere (Smith 1998).

STUDY AREA
The North Luangwa National Park (NLNP) is the most northerly national park in Zambia's Luangwa Valley (Figure 1). It covers an area of 4 636 km; (between 11°25* S to 12c20 S and 31°45' E to 32°40' E) and is situated entirely on the west hank of the Luangwa River, which forms its eastern boundary. In the west, the park boundary incorporates part of the Muchinga Escarpment which constitutes approximately 24% (1 113 knr) of the park s area. The northern boundary of the park is formed by the Lufila River, while in the south, the NLNP is bordered bv the Munvamad/i corridor.  Karoo System (Permian) sedimentary rocks form the dominant strata of the valley floor while the escarpments to the east and west of the Luangwa valley are made up of igneous (e.g. granite) and/or metamorphic (e.g. gneiss and quartzite) rocks (Utting 1976). In the NLNP, the ele vation of the Luangwa River and its adjacent floodplain is ± 600 m. Topographical relief in this region is only a few metres. Further west, but still on the valley floor, the terrain becomes more dissected (relief 5-50 m) and the elevation rises to between 700 and 800 m. In the far western areas of the park the Muchinga Mountains rise 600-700 m above the valley floor, with the highest peaks within the National Park reaching between 1 200 and I 300 m. In this area the terrain is deeply dissected with a relief of 500-600 m.
The valley receives a moderate rainfall of between 700 mm and 900 mm per annum, the wet season extend ing from November through to April. At higher altitudes, on the Muchinga Escarpment and plateau, rainfall is cor respondingly higher [mean for the plateau (Mpika), 1 065 mm].
No comprehensive study of the soils of the Luangwa valley has yet been carried out. However, in their land classification study of the South Luangwa National Park (SLNP), Astle et al. (1969) described and mapped the heterogeneous soil types of the park according to their association with the vegetation. Soil types range from the deep, red sandy loams of the upper escarpment to the alluvial sands and clays of the Luangwa floodplain.
The NLNP forms part of the Zambezian Regional Centre of Endemism (White 1983) and, due to its great topographical diversity, it probably possesses a greater range of habitats than any other park in Zambia. In the higher (800-1 200 m) deeply dissected terrain of the Muchinga Mountains, miombo woodland predominates, while in the lower (600-700 m) nearly flat valley regions, mopane woodland is the dominant vegetation type. Fire (Naylor et al. 1973) and elephant damage (Caughley 1976) have been the most influential factors affecting vegetation in the NLNP in recent years.

Site selection and data collection
The approach employed for this survey was essential ly inductive, with the primary aim being vegetation description rather than analysis (Kent & Coker 1992). As a first step, the vegetation of the park was stratified according to its macroscopic appearance, and subse quently systematic ground survey was used to provide descriptive detail.
Initial differentiation of vegetation units was carried out using aerial photographs and aerial survey. In this procedure, panchromatic aerial photographs of the NLNP (1983, scale 1: 34 000) were used to delineate nat ural vegetation boundaries. Interpretation of the air photo mosaic was carried out in order to define areas homogeneous in tone and texture, corresponding to homogeneous vegetation types. Thus, physiognomic units (woodland, bushland, grassland etc.) were differen tiated and marked on the aerial photographs. Air-photo information was validated and augmented by aerial sur vey. Homogeneous areas of vegetation (and to a certain extent, their composition) were identified by an observer and positions were recorded using a Garmin inboard Geographical Positioning System (GPS). A total of 116 points were recorded in a systematic transect survey and a further 105 points were recorded on other aeroplane and helicopter flights over the park.
In the second pan of the survey, floristic and physiog nomic vegetation data were collected in ground studies. Physiognomic classification followed White (1983) (Table 1). A total of 353 belt transects were located with in the vegetation types defined in the air-photo mosaic. Placement of transects was not random as an effort was made to cover as many vegetation units (homogeneous areas defined on the aerial photographs) as possible. Other considerations were accessibility and avoidance of ecotones and atypical landscape features e.g. termite mounds, tracks, roads (Walker 1976). Transect positions were determined using a portable GPS (Magellan 5000). Standard transects measured 50 x 5 m (= 250 m2), rec tangular plots being chosen for ease of sampling and to maximise species diversity (Brown 1954;Condit et al. 1996). Transects incorporated a minimum of 50 trees/ shrubs or at least 15 specimens of a dominant species. Where these criteria were not met, transects were broad ened or lengthened accordingly (Taylor & Walker 1978). Within each transect, woody plants (> 1 m in height) were identified, counted and measured (diameter at breast height and estimated height). Herbs and grasses were collected and/or noted at all transect sites.
Samples of all plant species recorded were collected, pressed and dried. Identifications were carried out in herbaria in Lusaka. Kitwe, Harare and Kew. Nomenclature follows Flora zambesiaca. Flora o f tropical East Africa and Lebrun & Stork (1991-1997. Voucher specimens are lodged at the Kew. Missouri and Mount Makulu herbaria. A full checklist of the plants collected in the NLNP is pub lished elsewhere (Smith 1998).
In addition to the vegetation survey, a limited soil sur vey of the NLNP was carried out and the resulting soil types were related to the plant communities defined above. In order to include the main vegetation types, soil pits were dug beside a vehicle track which transects the park, from Mano Game Scout Camp (11 °37' S. 32°02' E) on the Muchinga Escarpment in the west, to the Luangwa River (11°51' S, 32°26' E) in the east. Additional pits were dug between 11 °37' S, 32°30' E and 11°33' S, 32°25' E along a cutline parallel to the Lufila River. Soil pits (2 m deep) were dug at ± 2 km intervals within 5-10 m of the track. Soil profiles were measured and pho tographed and a soil sample (200 g) from each horizon was placed in a plastic bag and retained for further analy sis. The displaced soil was returned to the pit and the sur face levelled in an attempt to minimise disturbance to the site. Retained soil samples were measured for colour, texture and pH. Colour was determined using Munsell Soil Colour Charts (1949). Soil texture (stoniness, shape of peds, structure, consistency and roots) was described according to Courtney & Trudgill (1988); sand, silt and clay content of the soil was determined using the hydrometer method (Pramer & Schmidt 1964). Finally, the pH of the soil was recorded using a Kel soil tester (for acidity and soil moisture).

Data analysis and interpretation
Ground study provided detailed information on vege tation structure and composition. Thus, for each vegeta tion unit delineated from the aerial photographs, it was possible to: I, list plant species; 2, calculate tree/shrub density, expressed as n ha 1 (Brown 1954); 3, calculate woody biomass, expressed as t ha'1 using the following formulae developed in similar habitats in Sengwa, Zimbabwe by Guy (1981): Tree biomass (Kg) = 0.0549 x (diameter at breast height)25101 Shrub biomass (Kg) = 1.2102 x (canopy volume)0' 118 4, determine dominant species by calculation of impor tance values (Curtis & McIntosh 1951) for each species according to the formula: Importance value = relative frequency + relative density + relative dominance Ground study data permitted further division of vege tation units according to the floristic criteria above and the environmental information collected from each plot (altitude, soil type, topography).
Vegetation types were described in terms of their typ ical vegetation structure, characteristic woody species, associated grass and herb species, soils, geology, topog raphy, distribution and area covered.

Map production
A preliminary working vegetation map of the NLNP was derived from the 1983 panchromatic aerial photographs. Transparent overlays marked with the vegetation bound aries delineated on the aerial photographs were reduced from a scale of 1 : 34 000 to a scale of 1 : 250 000 on a pho tocopier (Xerox U.K. Ltd). This reduction was then super imposed onto the 1 : 250 000 UTM Ordnance Survey maps of the NLNP (no. SD-36-2 and SC-36-14) and redrawn. This rough map was used as a working image upon which ground survey could be based.
A second more accurate, more detailed and up to date vegetation map (scale 1:100 000) was produced from a partial scene (90 x 90 km) LANDSAT TM satellite image (Band 4, 7th July 1995), georeferenced and geocoded to UTM map projection (Satellite Applications Centre, Pretoria). The image was supplied by SAC as digital data (EOSAT Fast Format) on CD-ROM, which was processed into a photographic image (Hunting Technical Services, UK) upon which vegetation bound aries could be marked by hand. The vegetation types delineated on this image were assigned according to the aerial and ground survey results presented below and the end result is a map incorporating 11 broad vegetation types and in which vegetation mosaics are indicated ( Figure 1). This map was digitised using PC ARC/INFO (Kent Cassells, DICE, UK) and incorporated into the beginnings of a Geographical Information System (GIS) for the NLNP. More detailed vegetation data, as well as information about geology, topography, water relations, plant-animal interactions, etc. can all be incorporated into this database in the future in order to build up a more complete picture of the park ecosystem.

RESULTS AND DISCUSSION
During this study, the vegetation of the NLNP was stratified into six floristic classes, divided into 13 distinct vegetation types. Vegetation type distributions are shown in Figure 1, with the exception of types A2 and F3, the areas of which are too small to be shown at this scale.
The vegetation descriptions below are largely based upon quantitative tree/shrub data from sample sites. The herbaceous component is described from qualitative data and collection material. Further division of vegetation types into subtypes is based on substrate. A detailed description of all vegetation types and subtypes is given below. Key characteristics of woody vegetation types are given in Table 2. A. RIPARIAN FORESTS. WOODLANDS AND THICKET This vegetation class comprises moist forests to woodlands and thicket fringing perennial and seasonal watercourses. At lower altitudes (600-700 m) valley riverine fringe vegetation takes the form of woodland (usually with a well-developed shrub layer), and thicket. It should be noted that even on the valley's perennial rivers (Luangwa, Lufila and Mwaleshi), fringe woodland and thicket is discontinuous, being interspersed with other vegetation types such as riverine grassland (F2) and/or non-riverine vegetation types. This is even more noticeable on seasonal watercourses where fringe river ine trees are usually mingled with trees and shrubs char acteristic of adjacent vegetation types. At higher altitudes (800-1 300 m), on the Muchinga Escarpment, riverine fringe vegetation takes the form of dense evergreen for est which, on the larger watercourses, may extend to adjacent swampy areas. This vegetation type tends to be well developed and distinctive (although narrow) even on the smaller, seasonal watercourses.
Valley riverine woodland (Al), as described in the pre sent study, is readily recognisable in the Luangwa valley surveys of Astle et al. (1969), Naylor et al. (1973) and Phiri (1989) (Table 3). Escarpment riverine forest (A2) however, is not described in detail in any of the above studies but is covered thoroughly by Fanshawe (1971). Vegetation type A2 contains elements of Fanshawe's Riparian forest' (p. 34) and 'Swamp forest' (p. 32).

Al. Valley riverine woodland and thicket
The Luangwa River lies at an altitude of ± 600 m and topographical relief is only a few metres. The large trib utaries of the Luangwa in the NLNP are the Lufila, Mwaleshi and Mulandashi Rivers. These rivers emerge from the Muchinga Escarpment in the west at an altitude of between 700 and 800 m. Topographical relief along these watercourses varies from as much as 5-50 m near the escarpment to only a few metres at the Luangwa con fluences.
Valley riverine woodland and thicket is associated with the rich, recently deposited alluvial soils which lie adjacent to the rivers of the valley floor. This alluvial belt may only be a few metres wide or. along the Luangwa and its major tributaries, may extend many hundreds of metres from the river. Riverine valley soils show consid erable diversity and cannot be assigned to any one soil class. In general, these are deep, stoneless, clearly strati fied soils which vary in texture from sands to clays (see vegetation type F2). Soil pH is usually around neutral but ranges from pH 5.6 to pH 6.9. Soil colour varies from light yellow brown to dark grey.
The fringe woodland of the valley's perennial and sea sonal rivers is generally two-storeyed in structure, with canopy trees reaching 20 m or higher and a well-devel oped shrub layer, which may extend to form areas of bushland or thicket. Characteristic tall trees include Kigelia africana, Diospyros mespiliformis, Trichilia emetica, Lonchocarpus capassa, Colophospermum mopane, Combretum imberbe, Faidherbia albida, Sclerocarya birrea and Tamarindus indica (invariably associated with termitaria). Other less frequent but typi cal large trees include Hreonadia salicina and Khaya nyasica. Commonly occurring small trees and shrubs are Piliostigma thonningii, Ziziphus abyssinica, Oncoba spinosa, Feretia aeruginescens, Flueggea virosa, Anti-desma venosum, Phyllanthus reticulatus. Acacia sieberiana and A. polyacantha subsp. campylacantha. Amongst the thicket-forming shrubs, the genus Combretum is well represented: C. obovatum (on clay), C. fragrans and C. imberbe (as a shrub) are all common. C. fragrans may form areas of bushland in which it is the single dominant species. Other frequent thicket-forming species are Diospyros senensis (on sand), Keetia zanzibarica and Friesodielsia obovata. Climbers in this vegetation type include Jasminum fluminense, Abrus precatorius and Dregea macrantha. The grass layer associated with welldeveloped fringe woodland or thicket is sparse and con fined to shade-loving species such as Panicum maximum, Phyllorachis sagittata and Setaria homonyma. Herbs typical of the stratified soils of the larger rivers of the valley floor include Senna obtusifolia, Indigofera tinctoria, Sida alba, Vemonia glabra and spp.. Duosperma spp., Corchorus spp. and Ocimum spp.
On the major rivers such as the Luangwa. Lufila. Mwaleshi and Mulandashi. vegetation type Al occurs in a substrate-dependent mosaic with the riverine herba ceous habitats described in vegetation type F2.

A2. Escarpm ent riverine forest
This forest is found fringing the rivers and streams of the Muchinga Mountains in the west of the NLNP. Elevation ranges from 800-1 300 m and the terrain is deeply dissected with relief measured in hundreds of metres. The igneous/metamorphic geology of the Mu- Together, vegetation types Bl and B2 cover an area of approximately 180 km:, and seem to occupy a niche on soils of an intermediate nature between the low, poorly drained clays of vegetation type FI and the higher, sandy soils of vegetation type Dl. Mixed alluvial thicket is the character istic thicket vegetation type north of the Mwaleshi River, while Combretum thicket largely occurs to the south of the Mwaleshi. Both types arc dominated by different species of thicket-forming shrubs, but share many species in common. The absence of C. elaeagnoides and C. celastroides in large areas of Mixed alluvial thicket, north of the Mwaleshi River, which led to the distinction being made in the present sur vey, is apparently a local phenomenon because these two species do occur further north, on the Lufila River. The rea son for their absence in the middle of the park is unknown, and requires further study. Astle et al. (1969) make no distinction between Mixed alluvial thicket ( B l) and Combretum thicket (B2), referring to this vegetation type as Thicket on freely draining alluvium'. Their species list for this vegetation type contains the dominant elements of both Bl and B2.

Bl. Mixed alluvial thicket
This vegetation type is found on the alluvial soils associated with the Luangwa River in the east of the park and is the characteristic thicket vegetation type north of the Mwaleshi River. Elevation in this area is between 600 m and 625 m and the terrain undulates slightly with a relief of only a few metres. This vegetation type is fre quently found in mosaic with Chloris-Dactyloctenium-Echinochloa secondary grassland (FI) but occupies higher ground and better drained, sandy soils so that while the surrounding grassland becomes waterlogged during the wet season, the areas of thicket remain com paratively dry.
Soils associated with this vegetation type are freely draining and are typically pale brown to orange, stone-less, sandy clay loams (60-75% sand) of slightly acid to neutral pH (pH 6.5-7.0).
Vegetation type Bl comprises closed and open stands of bushes 2-7 m high with occasional tall trees. Colo phospermum mopane is present to a greater or lesser extent. It is found in clumps on patches of calcareous or sodic clay soil but also as individual, usually tall trees dotted throughout the habitat. Other occasional tall trees include Xeroderris stuhlmannii, Pseudolachnostylis maprouneifolia, Stereospermum kunthianum and Adansonia digitata.

B2. Combretum thicket
This is the characteristic thicket type south of the Mwaleshi River. Elevation is ± 600-620 m and the ter rain is flat with a relief of only a few metres. As with Mixed alluvial thicket (Bl). Combretum thicket is found on alluvial, sandy soils close to the Luangwa and Mwa leshi Rivers.
Vegetation type B2 is comprised of thicket-forming shrubs. 2-7 m in height, which grow in a mosaic of closed and open stands. The grass layer is not well devel oped. Tall trees are rare but species such as Manilkara mochisia and Diospyros quiloensis may occur infre quently. Small trees found in Combretum thicket include Schrebera trichoclada and Combretum collinum subsp. gazense. The dominant shrubs are Combretum elaeag noides, C. celastroides, C. obovatum. Holarrhena pubescens, Vangueria infausta and Markhamia spp. Grasses and herbs associated with this habitat are the same as those in vegetation type Bl.

C MIOMBO W OODLANDS
The miombo woodlands of North Luangwa National Park cover an area of approximately 1 300 knr. This is the dominant vegetation type of the Muchinga Escarp ment and its attendant foothills, Chinshenda. Soma and the Mvumvwc range. The vegetation described under the umbrella of miombo woodlands is physiognomically diverse, ranging from closed woodland to open wood land to scrub woodland. For the purposes of this classifi cation. miombo woodland in the NLNP has been divided into two types, Cl Upper escarpment and C2 Lower escarpment/hill miombo woodland, a division largely based on floristic composition. Further division into subtypes is dependent on substrate.
Miombo woodland is clearly recognised in all three of the previous valley surveys (Table 3). Further de scriptions are available from Trapnell (1953) and from Fanshawe (1971). C l. Brachystegia-Julbernardia-Isoberlinia u p pe r escarpment and plateau m iom bo woodland At elevations over 1 (XX) m, this is the most important and extensive vegetation type of the Muchinga Escarp ment. It can be separated into three distinct vegetation subtypes dictated by substrate. Subtype 1 occurs over most of the upper escarpment, on deep laterite soils. Subtype 2 is rare in the NLNP and is found in isolated patches on shallower plateau soils. Subtype 3 is associat ed with the granite outcrops dotted throughout the escarpment terrain.
Vegetation type Cl (subtype 1) corresponds to Trapnell's E and El types (Trapnell et al. 1950), which occur on the deep soils of the escarpment, as evidenced by the dominance of Brachystegia utilis and B. spiciform is, species which cannot tolerate shallow soils (Fanshawe 1971). The tall trees and comparatively sparse grass layer seen in this vegetation type are proba bly due to the improved drainage and better soils associ ated with the dissected terrain (Cole 1963). In this vege tation type, laterite was more commonly found as nod ules in the B horizon rather than the impermeable layer characteristic of the pediplain plateau soils (Cl subtype 2). Subtype 2 corresponds to Trapnell's P4 type (Trapnell et al. 1950), and Subtype 3 is described by White (1983) as 'Zambezian rupicolous bushland and thicket'. Subtype 1. Upper escarpment miombo woodland: is the dominant vegetation type of the upper Muchinga Escarp ment in the west of the NLNP. and is found at elevations ranging from 1 000 to > 1 300 m over deeply dissected ter rain where relief is measured in hundreds of metres.
Subtype 1 is associated with deep, red, stoneless sandy loams or sandy clays. These soils are slightly acid (pH 6.6-7.0) and usually contain laterite nodules and mica aggregates in the B horizon. Subtype 2. Plateau miombo woodland: occurs in isolat ed patches in the far west of the NLNP. and is associat ed with flat terrain. Plateau soils tend to be shallow, poor in nutrients and humus, slightly acid, and are typically leached, with a laterite or gley horizon near the surface. In contrast to upper escarpment miombo woodland, plateau miombo woodland is of single storey structure and is characterised by stunted Brachystegia-Julbernardia, interspersed with Uapaca, Protea, Faurea and Monotes species. In addition, the shrub and grass layers arc comparatively well developed, with Hyparrhenia and Andropogon spp. predominant. This subtype occurs on the comparatively shallow, infertile soils of the plateau peneplain and over large areas of Zambia appears to be secondary miombo woodland which has been subjected to repeated fires and cultivation (Fan shawe 1971). Subtype 3. Rupicolous miombo woodland: the rocky out crops and granite kopjes of (he Muchinga Escarpment support a distinctive vegetation type and although many of the taxa listed above may occur, additional species such as Brachystegia microphylla, Pterocarpus rotundifolius, Schrebera trichoclada, Kirkia acuminata, Landolphia parvifolia and Tarenna neurophylla are typi cal. Carphalea pubescens is a characteristic subshrub.

C2. Julbemardia-Brachystegia lower escarpment and bill m iom bo woodland and scrub woodland
This woodland covers much of the lower Muchinga Escarpment and its attendant foothills, C hinshenda. Soma and the Mvumvwe range. It is also found on the upper valley floor where it may intergrade with vegeta tion types D1 and D2.
Vegetation type C2 can be separated into two subtypes according to substrate and vegetation physiognomy. Subtype 1 is miombo scrub woodland, associated with the thin, eroded, stony soils of the hill slopes. Subtype 2 is open woodland, found on the deeper soils of the inter fluves and Hatter sites.
Vegetation type C2, subtype 1 (scrub woodland) is recognised by Astle et al. (1969) who refer to it as 'Miombo scrub on shallow soils'. This type is also described by White (1983), in a direct reference to the Luangwa valley (p. 99). Fanshawe (1971), Naylor et al. (1973) and Phiri (1989) all refer to this distinctive form of miombo as 'Scrub miombo woodland'. Vegetation type C2, subtype 2 (woodland) is designated 'Miombo wood land on deep soil' by Astle et al. (1969). Phiri (1989) cites Brachystegia boehmii, B. bussei, B. manga and Julbernardia globiflora as common taxa in this habitat. Subtype 1. Brachystegia stipulata-Julbernardia globi flora miombo scrub woodland: occurs at elevations rang ing from 700-1 000 m and is the most extensive vegeta tion type of the lower Muchinga Escarpment, where it occurs in mosaic with subtype 2. It also intergrades with vegetation types Dl and D2 on the upper valley floor.
Subtype 1 occurs on the shallow, slightly acid (pH 6.6-6.9) light grey to yellowish brown, generally stony sandy clay loam soils (lithosols and shallow fersiallitic soils) associated with the hill slopes of the lower escarp ment and foothills.  Wild & Barbosa (1967). White (1983) refers to it as 'North Zambezian undifferentiated woodland', comprising miombo associates but not Brachystegia/Julbernardia species. Fanshawe (1971) groups this vegetation type together with the miombo woodlands of the valley floor, but refers to it as 'Erythrophleum woodland'. The classifi cation of the present survey is largely based on floristic composition and our ground survey data strongly indicates that vegetation type Dl does not fall into the miombo class. Brachystegia/Julbernardia species do occur in the ecotones where vegetation type C2 intergrades with Combretum-Temiinalia woodland, but elsewhere they are never dominant and over large areas of this woodland, par ticularly in the east, they do not occur at all. Astle et al. (1969) andPhiri (1989) refer to this vegetation type as 'Terminalia sericea-Erythrophleum africanum woodland savanna' and 'Erythrophleum woodland' respectively (Table 3). This nomenclature can be related to our results in that E. africanum is an important component of this vegetation type. However, in the NLNP at least. Combretum/ Terminalia species are more frequent, more abundant and account for more woody biomass than E. africanum. Like us. Naylor et al. (1973) refer to this veg etation type as *Combretum-Temiinalia woodland'. Astle et al. (1969) classify vegetation type D2. Combretum-Temimalia-Diospyros wooded grassland, under Miombo scrub on shallow soils', although they do make a topographic distinction between this habitat (3:1 and 5:1) and vegetation type C2 subtype 1 (4:2 and 8:1). In Astle's updated landsystem/vegetation map (Astle 1989) a further land facet is added (landsystem 7: facet 6) in which this vegetation type is described as: 'semi-deciduous scrub land with local variation in species composition. A mopane and Terminalia stenostachya association on the flatter sites. Brachystegia stipulata. Combretum apiculatum, Julbernardia globiflora association in areas of greater relief.' This is a landsystem description incorporating the floristic mosaic described below, of w hich vegetation type D2 is the non-miombo component.

D l. Combretum-Terminalia w oodland
Vegetation type Dl is characterised by fire-tolerant, sandy soil species (e.g. Terminalia sericea) and bears a close resemblance to other high grass-woodland vegeta tion types such as the Burkea-Erythrophleum woodlands of western Zambia and the Chipya woodlands of the Bangweulu region (Trapnell et al. 1950). The common herbaceous indicators of lake basin Chipya. Aframomum alboviolaceum, Smilax anceps and Pteridium aquilinum are not a feature of this habitat, but like Chipya woodland (Lawton 1978), it is probable that vegetation type Dl is maintained by the fierce dry season fires which sweep through the NLNP every year. Combretum-Terminalia woodland is found in close association with thicket (B1/B2), both vegetation types occupying a belt of deep, sandy soil running parallel to the Luangwa River. The relationship between thicket and woodland is not clear, but it is possible that the two vegetation types are deter mined by edaphic factors. In certain (western) areas of the park, this woodland intergrades with hill miombo woodland (C2). It covers large areas of the valley floor (180 km2 in addition to 313 km: in mosaic with vegeta tion type C2) over elevations ranging from 650 m to 700 m. It is generally associated with flat terrain where relief is only up to 10 m.
Soils associated with this vegetation type are deep, light brown to orange and arc mildly acidic to neutral (pH 6.4-7.0). These soils are characterised by a very high sand content (70-90%) and are probably colluvial (or old alluvial) and derived from Karoo sandstone.
Combretum-Terminalia woodland takes the form of open 1-or 2-storeyed deciduous woodland. Canopy spe cies may be up to 20 m tall. The grass layer is tall and well developed. The tall trees found in this habitat arc dominated by Terminalia sericea (up to and above 15 m high), Pseudolachnostylis maprouneifolia, Pericopsis angolensis, Burkea africana, Erythrophleum africanum and Amblygonocarpus andongensis. Common small trees and shrubs are Combretum molle, C. collinum subsp. gazense, C. zeyheri, Bridelia catluirtica, Crossopteryx febrifuga and Bapliia massaiensis. On the deep soils of this type, the grass layer is well developed with both tall and short grasses. In the NLNP. vegetation type D2 occurs in mosaic with mopane (E1/E2) and hill miombo (C2) throughout its range (see map. Figure 1). The Combretum apicula-tum/Terminalia stenostachya/Diospyros kirkii associa tion (vegetation type D2) is the most extensive vegeta tion type in this mosaic, as it occurs on the shallow, stony fersiallitic soils which cover the gentler slopes and flatter regions of the upper valley floor. Hill miombo (C2 subtype 2) occurs in isolated patches on the deeper soils of the ridge tops, while scrub miombo (C2 subtype 1) occurs on the thin rocky soils of the steeper slopes. Mopane woodland and scrub woodland (El and E2) occur on sodic or calcareous patches of soil dotted throughout the upper valley floor. This grassland, in mosaic with vegetation types El. E2 and C2. covers an area of approximately 1200 km2 in the NLNP and occurs on the upper valley floor and in the foothills of the Muchinga Mountains. Throughout its range this vegeta tion type is associated with gently sloping terrain with relief measured in tens of metres.
The soils associated with D2 wooded grassland are stoneless to very stony soils, over siltstone and grits, which may or may not be covered with a surface mantle of quartzoze stones. They are highly variable in texture but tend to be shallow, grey to reddish brown and mod erately acid (pH 5.4-6.6).
Combretum-Terminalia-Diospyros wooded grassland is defined by a well-developed grass/herb layer scattered with small trees and shrubs covering 10% to 40% of ihe surface. It is heterogeneous in composition and form, grading into woodland in some areas and pure grassland in others. Dominant small trees are Diospyros kirkii, Terminalia stenostacliya, T. stuhlmannii, Combretum apiculatum, C. fragrans, C. zeyheri, Crossopteryx febri fuga and Pseudolachnostylis maprouneifolia. Shrub species include Bauhinia petersiana. Acacia liockii, A. gerrardii and Ximenia caffra subsp. caffra. On the shal low. stony soils of this habitat, the grass layer is well developed and dominated by medium to tall, coarse grasses. Trapnell believes that this vegetation mosaic is the result of sheet erosion. He suggests that the colluvial soils of the escarpment were more extensive in the past, covering much of the upper valley floor in the NLNP. Evidence for this is to be found outside the park, north of the Lufila river, where miombo woodland covers a large area of the upper valley floor (Trapnell et al. 1950;Edmonds 1976) and is found on soils of a colluvial nature containing laterite nodules. It is postulated that, in the NLNP and further south, much of this escarpment soil has been washed away, leaving a dissected, undulat ing terrain, the slopes of which are characterised by a thin, stony soil cover. This niche is occupied by vegeta tion type D2 in the areas of gentler relief and miombo scrub woodland (C2 subtype 1) on the steeper slopes. Hill miombo woodland (C2, subtype 2) occupies the ridge tops, where remnants of the deep escarpment soil cover remain. The patches of sodic or calcareous soil, associated with the mopane vegetation (E1/E2) of the upper valley floor, are almost certainly derived from ancient, and in some cases recent, termite activity (Trapnell et al. 1976).

E. C O L O P H O SP E R M U M MOPANE WOODLAND AND SCRUB W OODLAND
Colophospermum mopane is the single dominant tree species in this vegetation class. In the NLNP. mopane may grow as a tall tree of up to 15 m, or it may take the form of a multistemmed, stunted shrub < 3 m tall. The tall form is typical of Colophospermum mopane wood land (El) and the shrub form is characteristic of Colo phospermum mopane scrub woodland (E2). These two vegetation types may occur in discrete areas or they may grow together in mosaic. The difference between the two types is largely the physiognomy of C. mopane, which can be related to browsing damage and substrate as described below. Together, the two mopane habitats cover an area of approximately 600 knr in the NLNP.

El. Colophospermum mopane woodland
This woodland occurs on the alluvial soils associated with the Luangwa River and its tributaries in the east of the park. The terrain is flat, but due to the sparse herba ceous layer and the impermeable nature of the soil, mopane woodland is usually dissected with drainage channels and erosion gullies.
Mopane woodland soils typically consist of a shallow (sandy loam) A horizon over an impermeable B horizon of brown or grey, cracking, slightly acid (pH 6.0-6.5) clay. These soils are poorly drained and as a result, are waterlogged during the rainy season. Continual sheet erosion has the effect of removing the topsoil, and the roots of mopane trees in this habitat are frequently exposed or undermined.
In this vegetation type, Colophospermum mopane grows as the single dominant species in an open two-storeyed woodland comprising a canopy layer of mature trees (10-15 m tall) and an understorey of trees in various stages of development. Trees and shrubs associated with C. mopane are comparatively few, mainly species found in the thicket habitats B 1 and B2. Other associated species are Afzelia quanzensis, Balanites aegyptiaca, and Ximenia americana. The herbaceous component is dependent on substrate. On soils with a sandy A horizon, the herbaceous layer is sparse and largely composed of grasses, particularly those associ ated with vegetation type FI (e.g. Urocliloa mossambicensis, Chloris spp., Dactyloctenium spp. The tall 'cathedral mopane' woodland associated with deep alluvial soils east of the Luangwa. outside the NLNP. is not common in the park. Instead, this twostoreyed form of woodland has arisen due to browsing pressure, which prevents recruitment into taller size classes (Caughley 1976). The most influential browsers in mopane woodland are elephants, which tend to browse destructively, pollarding the trees as they feed (Anderson & Walker 1974;Caughley 1976;Lewis 1991;Styles 1993). However, in the past ten years the elephant popu lation of the Luangwa valley has been severely reduced due to ivory poaching (Leader-Williams et al. 1990) and if elephants are responsible for the maintenance of this scrub mopane, it is to be expected that without their browsing pressure, scrub mopane trees w ill grow' into tall trees and set seed. In the mopane plots laid down in this study, individual tree heights, extent of damage and seed status were recorded. Measurements in 1993. and1994 suggest that, in many cases, height recruitment is occur ring. Further visits to these plots will be necessary to confirm this trend.

E2. Colophospermum mopane scrub woodland
This woodland occurs in discrete patches throughout all the habitats of the valley floor forming distinct islands of vegetation, which are clearly visible as white patches on aerial photographs. They appear to arise due to local soil conditions. The soils arc typically compacted pinkish grey to light grey sandy silt loams over an impermeable calcareous or sodic clay loam B horizon. The xerophytic conditions created by the impermeability of the soil, and the relative alkalinity of the B horizon produce a hostile environment for herbaceous plant species. The resulting paucity of herbaceous ground cover exacerbates the erosion prob lems in this habitat and. as with vegetation type E l. scrub mopane soils are dissected by numerous drainage chan nels and erosion ditches.
There are two major factors which cause the shrub growth form of Colophospermum mopane in the NLNP: browser damage (see above) and/or soil conditions. Where scrub mopane trees are intimately associated with tall mopane trees, as in vegetation type El on the alluvial soils in the east of the park, stunted growth appears to be largely due to browser-pollarding. Elsewhere, both near the river and throughout the rest of the park, compara tively large, discrete areas of scrub woodland occur. This is a distinct vegetation type in which edaphic factors are more important than browsing pressure in influencing vegetation physiognomy (Dye & Walker 1980).

F. GRASSLANDS
The grasslands of the NLNP. defined as areas of herbaceous vegetation with less than 10% woody vege tation cover, are all associated with water. Vegetation Type FI is seasonally waterlogged, while types F2 and F3 are associated with the rivers and dambos of the val ley and escarpment respectively.
Vegetation type FI is referred to by Phiri (1989) as 'Floodplain grassland' with Echinochloa colona cited as the dominant species. Astle et al. (1969) describe this vegetation type (landsystem 1; facet 7: photo 4) as 'Short Echinochloa grassland with Combretum obovatum shrubs on dark cracking clays'.
Vegetation type F3 is not dealt with by any of the val ley surveys, as it is an upper escarpment and plateau habitat. However. Vesey-Fitzgerald (1963) gives a detailed account of this vegetation type which he calls 'Headwater valley grasslands (dambos)'. Louiletia sim plex and Hyparrhenia spp. are named as characteristic species. White (1983) and Fanshawe (1971) also describe this vegetation type, referring to it as Dambo grassland' and Bush-group grassland' respectively.

FI. Chloris-Dactyloctenium-Echinochltm secondary grassland
This type takes the form of short, annual grassland punctuated with occasional clumps of Combretum obo vatum thicket. It is found on degraded mopane woodland and, as a result, is often scattered with the skeletons of dead mopane trees. Vegetation type FI covers a large area of the NLNP (approximately 335 km2) and is found on recently deposited alluvial soils adjacent to the Luangwa River. The terrain is flat and low lying, and is seasonally waterlogged during the rainy season.
Soils associated with this habitat tend to be shallow, poorly drained, light grey, compacted neutral sandy clays or sandy loams. This grassland (FI) is clearly part of a dynamic suc cession. The extensive areas of FI grassland present in the NLNP today were not recorded by Astle in 1965. At that time, these areas were largely covered by mopane woodland. It is probable that this succession has arisen due to extensive browsing damage in mopane wood lands, concomitant with the well documented increase in elephant numbers recorded in the valley during the 1970's (Caughley 1976). Additional factors such as waterlogging and fire may have been involved in main taining the grassland state. Recent observations made by the author suggest that mopane seedlings are starting to re-invade the FI grasslands in certain areas of the park. This reversion to mopane woodland would be consistent with the hypothesis that these grasslands are created and maintained by elephants because, as stated above, ele phant numbers in the Luangwa valley have declined drastically in the past ten years (Leader-Williams et al. 1990). Furthermore, recent anti-poaching efforts have greatly reduced the incidence of man-made fires in the park. The mechanics of this succession need to be under stood because this grassland is a particularly productive habitat, which aerial survey has shown is utilised by large numbers of grazers (NLCP census 1994). If these grasslands revert back to mopane woodlands, an impor tant grazing habitat will be lost from the NLNP. Further research is needed to investigate the grassland-mopane succession, and the factors which influence it.

F2. Valley riverine grasslands
These grasslands are associated with the larger rivers of the valley floor. The Mwaleshi, Mulandashi and Luangwa Rivers all have extensive floodplains within their meander belts as well as numerous attendant drain age channels, oxbow lagoons and dambos.
Various soils are associated with this vegetation type, but all are based on the recently deposited alluvium of the large rivers. All tend to be deep and stratified, with soil textures ranging from the well-drained sandy soils of the sandbars to the cracking black clays of the floodplains. Each substrate supports a distinctive grass and herb component. Subtype 3. Aquatic associations: water grass associa tions are found on the seasonally waterlogged clays of the Luangwa River's oxbow lagoons and dambos. In these areas, which remain under water for most of the rainy season, water-loving grasses such as Oryza barthii, Echinochloa colona, Sporobolus pyramidalis and Setaria spp. dominate. Common sedges in this habitat include Cyperus esculentus, C. articulatus. C. distans and Kyllinga alba. Characteristic water-associated herbs arc Polygonum setulosum, Lindernia oliveriana, Sphenoclea zeylanica and Heliotropium spp. When water remains in the lagoons and dambos, the aquatic water weed Pistia stratiotes is characteristic. Around the peripheries of lagoons and dambos as they dry' out, typical herbaceous species are Portulaca oleracea, Ludwigia stolonifera, Hibiscus articulatus, Alternanthera sessilis, Mimosa pigra and Sphaeranthus spp.

F3. Loudetia simplex-Hyparrhenia d am bo grassland
On the Muchinga Escarpment and hills of the NLNP (Chinshenda. Mvumvwe and Soma) the rivers and streams do not have a well-developed meander belt and, as a result, the herbaceous riverside vegetation is less well defined than in the valley. However, the numerous dambos and drainage channels associated with these watercourses do have a characteristic grass and herb component.
Dambo soils are poorly drained and compacted. They are typically leached illuvial soils, black, dark grey or dark brown in colour, and acid (pH 5-6). The dominant grasses and herbs found on the dambos and streams of the lower escarpment and hills are similar to those found in the upper escarpment. However, a number of water-associated grasses and herbs from the valley may also occur. Grasses include Brachiaria brizantha, Digitaria milanjiana, Echinochloa colona. E. pyramidalis, Setaria pumila, S. sphacelata, Sporobolus pyramidalis, Themeda triandra and Urochloa mossambicensis. Subshrubs and herbs include Urena lobata, Senna occidentalis. Aeschynomene mimosifolia, Ageratum conyzoides, Acmella caulirhiza, Tragia lasiophylla. Alternanthera sessilis, Ludwigia stolonifera and Polygonum setulosum.

Vegetation mosaics
The representation of vegetation as a series of discrete types, based on communities or associations is. to some extent, artificial in that observed vegetation associations are almost invariably part of a continuum or occur in mosaic with other communities (Craig 1983). This is a problem of scale. The larger the scale of the map to be produced, the smaller the area of association that can be represented. For practical purposes however, some vege tation communities will always be too small to map sep arately. This applies in areas of mosaic or in cases where vegetation communities occupy a microhabitat. The veg etation associated with termite mounds, for example, is distinctive (Fanshawe 1968) and may be important to the ecology of an area, yet as a type, it is often too dispersed to be mapped separately. In the present study, clear mosaics have been indicated on the map (Figure 1). Elsewhere, dominant associations, not necessarily pure associations, are depicted (see below).

CONCLUSIONS
Vegetation monitoring on any scale requires baseline data from which deviations, or trends, can be measured. Changes in woody cover over large areas, for example, can be discerned from aerial photographs and satellite images. In the case of the NLNP, mopane woodland for example, can be mapped from historic aerial photographs (1952, 1965 and 1983) and compared with the 1995 LANDS ATderived digital data used in the present study (Figure 1).
Although aerial perspectives are useful records of large-scale trends, they provide little information about cause of change. Of much more use in this respect is ground study information about vegetation structure and species composition. In the present study, 353 relocat able plots were laid down and the precise data recorded in each of these plots provides the basis for monitoring vegetation changes in the future.
Finally, it is important to stress that this study presents a coarse stratification of the vegetation of the NLNP, use ful for measuring changes in the park's vegetation at this scale. Although these vegetation units are fairly homoge neous, they still, in some cases, represent more than one plant community. A more detailed study of plant associa tions in the park is needed if we are to begin to compre hend the processes involved in maintaining plant commu nities. At the species level, it is the local combination of abiotic and biotic factors which make up a niche that is important. If we are to understand the autecology of species, or even the ecology of ecotypes, we need to understand the combinations of habitat components which make up the niche. The great potential of geographical information systems is that the detailed local picture can gradually be built up. Spatial information about geology, soil, topography, water relations, fire regimes, animalplant interactions and numerous other factors can be col lected over time and added to the picture. It is only through this process that we can understand enough about ecosystems to effectively manage them.