The Miombo Network

Edited by Desanker, P.V., P.G.H. Frost, C.O. Frost, C.O. Justice, and R.J. Scholes IGBP Report 41 The International Geosphere-Biosphere Programme (IGBP) Stockholm, Sweden

Citation:
Desanker, P.V., P.G.H. Frost, C.O. Frost, C.O. Justice, and R.J. Scholes, (eds.). 1997. The Miombo Network: Framework for a Terrestrial Transect Study of Land-Use and Land-Cover Change in the Miombo Ecosystems of Central Africa, IGBP Report 41, The International Geosphere-Biosphere Programme (IGBP), Stockholm, Sweden, 109 pp.

IGBP Secretariat, Royal Swedish Academy of Sciences, Box 50005, S-104 05, Stockholm, Sweden, http://www.igbp.kva.se

This report describes the strategy for the Miombo Network Initiative, developed at an IGBP intercore-project workshop in Malawi in December 1995 and further refined during the Land Use and Cover Change (LUCC) Open Science Meeting in January, 1996 and through consultation and review by the LUCC Scientific Steering Committee (SSC). The Miombo Network comprises of an international network of researchers working in concert on a ‘community’ research agenda developed to address the critical global change research questions for the miombo woodland ecosystems. The network also addresses capacity building and training needs in the Central, Eastern and Southern Africa (SAF) region, of the Global Change System for Analysis Research and Training (START). The research strategy described here provides the basis for a proposed IGBP Terrestrial Transect study of land cover and land use changes in the miombo ecosystems of Central Africa. It therefore resides administratively within the LUCC programme with linkages to other Programme Elements of the IGBP such as Global Change and Terrestrial Ecosystems (GCTE).

The report provides the framework for research activities aimed at understanding how land use is affecting land cover and associated ecosystem processes; assessing what contribution these changes are making to global change; and predicting what effects global change in turn could have on land use dynamics and ecosystem structure and function. The key issues identified are:

	Patterns, causes and rates of change in land cover in relation to land use
	Consequences of land-use and land-cover changes on regional climate, natural resources, hydrology, carbon storage and trace gas emissions
	Determinants of the distribution of species and ecosystems in miombo
	Fundamental questions of miombo ecosystem structure and function.

Seven core experiments were proposed during the meeting, each addressing a critical objective relating to miombo ecosystems. These were (in no particular order):

	Land use dynamics
	Land cover dynamics
	Ecosystem dynamics and disturbance
	Nutrient limitation in miombo
	Biogeography and distribution of miombo
	Regional primary production and carbon balance of miombo
	Sustainable natural resource management.

In addition a number of integrating activities are proposed, including modelling, developing shared databases, and building regional research capacity. Each core experiment will be undertaken by a consortium of participating institutions and individuals at a network of sites across the miombo region. A Task Team (TT) for the Miombo Network will guide the scientific content and implementation of the projects. The TT will consist of regional and international representatives of participating IGBP and the International Human Dimensions Programme on Global Environmental Change (IHDP) programme elements. It is proposed that a Network Coordinator (NC), appointed by and responsible to the TT, will facilitate the activities of the network. An interim TT has been set up to oversee the launch of the programme and to consolidate the establishment of the Miombo Network

What Is Miombo and Why Is It Important?

Miombo is a vernacular word that has been adopted by ecologists to describe those woodland ecosystems dominated by trees in the genera Brachystegia, Julbernardia and Isoberlinia of the family Fabaceae, subfamily Caesalpinioideae (CSA/CCTA 1960; Wild and Fernandes 1967). Such woodlands extend across about 2.8 million km2 of the southern subhumid tropical zone from Tanzania and Zaire in the north, through Zambia, Malawi and eastern Angola, to Zimbabwe and Mozambique in the south. Their distribution largely coincides with the flat to gently undulating African (early Tertiary) and post-African I (Miocene) planation surfaces that form the Central African plateau. The soils are predominantly infertile and are derived from underlying acid, crystalline rocks of the Basement Complex. These woodlands constitute the largest more-or-less contiguous block of deciduous tropical woodlands and dry forests in the world.

Interspersed within the woodlands are broad, grassy depressions (called dambos or mbuga in the vernacular, see Figure 1(a)). These seasonally waterlogged bottomlands can cover up to 40 % of the landscape in some areas. Dambos are not old river systems, as is often supposed, but are set into the landscape through differential weathering and subsurface removal of material by the lateral flow of groundwater (McFarlane and Whitlow 1990). They are important sites for cultivation and livestock grazing.

The soils of the miombo landscape form a catenary sequence, with well-drained, deeply weathered soils on the higher areas (often with lithosols on the ridge crests in drier areas) giving way to a narrow zone of sandy soils along the footslopes and then to imperfectly- to poorly-drained vertisols in the dambo (Watson 1964; Webster 1965). The dambo clay is largely smectitic and formed in-situ by the concentration of aluminium, silica and bases leached from the interfluves (McFarlane and Whitlow 1990). The basic landscape unit in miombo therefore comprises miombo woodland on the well-drained soils of the interfluve giving way abruptly at the footslope to hydromorphic grasslands on vertisols in the dambo (Figure 1(b)). The sharp transition between woodland and grassland reflects the point where the wet-season water table reaches the surface.

The components of these landscapes are strongly linked by the movement of water, soil and nutrients downslope; by the return of nutrients upslope through animal activity and atmospheric transport; by animal movements and fire; and by patterns of land use. An integrated study of these components is therefore essential. For this reason, the narrow botanical use of the term ‘miombo’ is inappropriate for this programme. In this report the term is used more broadly to encompass not only the woodlands themselves but also the associated land types within the same ecosystem. For the purposes of this programme, therefore, miombo is defined as follows:

Miombo encompasses those southern, Central and East African ecosystems occurring under a hot, seasonally wet climate on soils derived from acid crystalline bedrock, in regions where the woody component of the vegetation is normally dominated by species in the subfamily Caesalpinioideae (family Fabaceae), particularly in the genera Brachystegia, Julbernardia and Isoberlinia.

The area which satisfies this definition is shown in Figure 2. Ecologically similar Isoberlinia woodlands, lacking Brachystegia and Julbernardia, form the Guinea savannahs of West Africa (White 1983). Woodlands dominated by related genera, principally Baikiaea, Pterocarpus, Erythrophleum, Burkea and Cryptosepalum, but differing ecologically in important ways, occur on sandy soils on the fringes of the core miombo region, most notably on Kalahari sands. At a global scale, the ecology of miombo ecosystems is closely related to that of the cerrado and related ecosystems of South America, the dipterocarp woodlands of south-east Asia and the monsoonal tallgrass eucalypt woodlands of northern Australia. Together these ecosystems occupy about 6% of the global land surface.

Miombo ecosystems directly support the livelihoods of about 39 million people in seven Central African countries, including some with among the lowest per capita income and highest per capita population growth rates in the world. A further 15 million people living in towns and cities throughout the region also depend on food, fibre, fuelwood and charcoal produced in miombo. Estimated woody biomass fuel consumption alone amounts to about 48 Tg yr-1, releasing almost 22 Tg C. Other natural and anthropogenic processes, such as wildland burning, clearance of land for cultivation, slash-and-burn agriculture and the cultivation of wetlands, also contribute unquantified amounts of trace gases to the atmosphere, as well as altering the nature of the land cover and hydrological processes. On a more positive note, about 9 % of the region is protected as national parks, safari areas, and wildlife and forest reserves. Moreover, important advances have been made towards developing community-based natural resource management programmes, initially based on the sustainable use of wildlife but increasingly being extended to woodland resources in general (Child 1995, Bradley and McNamara 1993, Dewees 1994).

Given the predominance of woodland cover and the areal extent of miombo, the region has the potential to be either a source or a sink for carbon in the global carbon cycle, depending on the land-management practices adopted. The research activities proposed here are designed to provide a better understanding of the processes involved in land-use and land-cover changes, their socio-economic and biophysical contexts, and their global and regional consequences, as an input to the development of policies and programmes aimed at encouraging sustainable patterns of resource use.

Figure 1
The miombo landscape: (a) a plan section showing the interdigitation of woodland (strippled) and dambo grasslands(clear); (b) data on organic carbon, cation exchange capacity (CEC) and total exchangeable bases (TEB) from Webster (1965). The rolling relief is typical of the undulating erosion of the Central African Shield.

Figure 2

Vegatation types
	Mosaic of dry deciduous foreat and secondary grassland, Zambezian
	Wetter Zambezian miombo woodland (Brachystegia, Julbernardia and isoberlinia)
	Drier Zambezian miombo woodland IBrechystegia and Julbernwdie)
	Sudanian woodland wlth abundant Isoberlinia
	Undifferentiated woodland, North and South Zambezian
	Mosaic of Brachystegia bakerana thicket and edaphic grassland

An Overview of Miombo Ecology

Miombo has been portrayed as the archetype of moist-dystrophic savannah ecosystems (Huntley 1982; Frost et al. 1986). The vegetation, landscapes, and ecology of miombo ecosystems have been described by Malaisse (1978a), White (1983), Cole (1986) and Chidumayo (1993), among others. The following overview of miombo ecology is paraphrased from a review by Frost (in press).

The unifying features of miombo ecosystems, as defined in the previous section and elaborated on below, are:

	The upland soils are predominantly alfisols, oxisols and ultisols - infertile, with low cation exchange capacities (2-18 cmol[+] kg-1), low total exchangeable bases (1-15 cmol[+] kg-1), high acidity (pH 4-6 in CaCl2), low available phosphorus (9-18 ppm), and a dominance of low-activity clays and sesquioxides, while the bottomland soils are predominantly gleysols and vertisols.
	The dominant climatic feature is the marked seasonality of rainfall, with more than 95% of the 700-1 400 mm MAP falling during a 5-7 month, hot summer wet season (mean annual temperatures 18-23oC, mean maximum temperatures 24-30oC), which results in marked seasonality of plant production, growth, reproduction, and decay.
	Most of the trees are deciduous but produce new leaves before the rains, drawing on substantial internal reserves.
	The woodlands are dominated by broad-leafed, thornless trees of the subfamily Caesalpinioideae (family Fabaceae), while the grasses mainly comprise species in the family Andropogoneae.
	Woody plants make up more than 95% of total plant biomass in mature woodlands but grasses predominate in dambos.
	The dominant tree genera are all ectomycorrhizal.
	The biomass of, and consumption by, indigenous large herbivorous mammals is generally low.
	Consumption by herbivorous invertebrates is generally low though episodic outbreaks occur every few years.
	Fire is frequent and important, both as an ecological factor and as a management tool.
	Miombo ecosystems have been occupied and used by people for millennia, with slash-and-burn agriculture in the woodlands and cultivation of dambos major traditional forms of land use, though there are currently profound changes occurring in both the kinds and intensity of use.

One of the characteristic features of miombo woodland is its apparent uniformity over large regions. The woodland typically comprises an upper canopy of pagoda- or umbrella-shaped trees, occasionally with a scattered layer of subcanopy trees, and a discontinuous understorey of shrubs and saplings with a patchy herbaceous layer of grasses, forbs and suffrutices. This uniformity is due partly to the remarkably similar physiognomy of the dominant canopy trees, a reflection of their origins in the Caesalpinioideae, and partly to broadly similar environmental conditions across the region.

Differences in structure and species composition occur along rainfall gradients, from the drier fringes of the miombo region to the wetter core. White (1983) divided miombo into wet and dry types. Wet miombo is centred on eastern Angola, northern Zambia, south western Tanzania and central Malawi in areas receiving more than 1000 mm rainfall per year. Canopy height is usually greater than 15 m, reflecting the generally deeper and moister soils which create favourable conditions for growth. The vegetation is floristically rich and includes nearly all of the characteristic miombo species. Dry miombo occurs in Zimbabwe, central Tanzania, and the southern areas of Mozambique, Malawi and Zambia, in areas receiving less than 1000 mm rainfall annually. Canopy height is shorter than 15 m and the vegetation is floristically impoverished. Many of the dominant canopy tree species of the wet miombo are either absent or local in occurrence.

The present-day distribution of miombo reflects its history, principally past climatic changes and past and present human activities, though the scarcity of well preserved and accurately dated pollen profiles from Central Africa currently hampers the detailed reconstruction of the palaeoenvironments and past vegetation of the region (Scott 1984). Sediment cores containing Brachystegia pollen and covering various periods over the past 38 kyr have been recovered from a number of sites across northern Zambia (Lawton 1963; Clark and Van Zinderen Bakker 1964; Livingstone 1971), and from the southern basin of Lake Tanganyika, Tanzania (Vincens 1991), all near the centre of distribution of miombo. Brachystegia pollen has also been recovered from cores taken in the Inyanga highlands, Zimbabwe, at the present upper altitudinal limit of Brachystegia (Tomlinson 1974), and at Dundo, Angola (Van Zinderen Bakker and Clark 1962), close to the current lower altitudinal limit, adjacent to the lowland moist tropical forests of Zaire. In addition, traces of Brachystegia pollen have been found in 1-4 kyr BP sediments from near Naboomspruit and Pretoria, South Africa (Scott 1982, 1983), well beyond the present southern limit of the genus.

Brachystegia appears to have been widespread across Central Africa prior to the Last Glacial Maximum, but at the height of the glaciation, around 19 kyr BP, the vegetation became more open with Compositae, Cliffortia and Cyathea dominant, all elements of the present-day montane flora of Central Africa. This open vegetation was later replaced by Podocarpus, Olea, Myrica and Ericaceae (Scott 1984; Vincens 1991). Interpretations of these data, together with general circulation model (GCM) simulations of palaeoclimates, suggest that at the Last Glacial Maximum around 18 kyr BP the climate was about 3-5 oC cooler, and mean annual precipitation about 250 mm lower, than now (Vincens et al. 1993; Jolly et al. in press). This would have produced widespread change in the distribution of miombo, with the woodland vegetation of the central plateau probably being displaced to lower altitudes as the climate became cooler and drier.

With the amelioration of climate during the Holocene, Brachystegia again became a prominent component of the plateau vegetation. Subsequent expansions and contractions are suggested by changes in pollen abundance in late-Holocene sediments from the Inyanga highlands, where the genus had become well established by 4.7 kyr BP (Tomlinson 1974); and by traces of Brachystegia pollen in 2-4 kyr BP and < 1 kyr BP sediments from near Naboomspruit, South Africa (Scott 1982) and in < 1 kyr BP sediments from Pretoria (Scott 1983). The marked northwards contraction in the geographic range of Brachystegia in the past 1 000 yrs from sites more than 400 km south of its present distribution in Zimbabwe implies either a sudden change in climate during that period, something for which there is no evidence (Scott 1984), or that relatively minor shifts in temperature and moisture regimes during that period adversely affected the population dynamics of Brachystegia at its geographical limits.

Within the broad divisions between wet and dry miombo, there is much local variation in composition and structure in relation to soil depth, texture and drainage; fire regime; wildlife impact; and past and present land use. Aboveground woody biomass of old-growth stands in wet miombo averages about 90 Mg DM ha-1 (range 47-144 Mg ha-1), while that of old-growth stands in dry miombo is generally lower, averaging 55 Mg DM ha-1 (range 21-84 Mg DM ha-1) (Malaisse 1978a; Guy 1981; Stromgaard 1985; Chidumayo 1987b, 1988, 1990, 1991; Frost in press). The biomass of disturbed woodlands is lower still. Mean aboveground biomass increases with increasing mean annual rainfall of a site. Recorded mean annual increments of biomass in coppice woodland are 1.2-2.0 Mg ha-1 yr-1 in dry miombo and 2.2-3.4 Mg ha-1 yr-1 in wet miombo, about 4-7 % of aboveground biomass (Chidumayo 1990, 1991, 1993).

Much less is known about the amount of biomass belowground. Miombo species have horizontally and vertically extensive root systems. Maximum recorded lateral distances are 27 m (J. globiflora: Strang 1965) and 15 m (B. longifolia, B. spiciformis and J. paniculata: Savory 1963). The tap roots of these species can exceed 5 m in deep soils. Available estimates of belowground biomass are 7-39 Mg ha-1 (24-53 % of aboveground biomass or 20-35 % of total biomass: Malaisse and Strand 1973; Chidumayo 1993, 1995; Malimbwi et al. 1994).

Woody plants make up 95 - 98 % of the aboveground biomass of undisturbed stands; grasses and herbs comprise the remainder (Martin 1974; Malaisse 1978a; Chidumayo 1993; Frost in press). Aboveground herbaceous biomass in relatively undisturbed, mixed-aged stands of miombo is usually less than 2.0 Mg ha-1 (range of recorded values: 0.1-4.0 Mg ha-1). On average, forbs make up about 30 % of this biomass, though this varies considerably among sites and between years. Herbaceous yield declines exponentially with increasing woody plant basal area, levelling off at a low value above 10 m2 ha-1, the approximate point where the tree canopy cover is complete.

The nutrient content of the foliage of miombo plants is generally low. The nitrogen content of mature leaves of non-nodulated canopy tree species is about 1.9 %, significantly lower than that of the few nodulated, N-fixing species: 2.7%. The average phosphorus level is 0.17 % P. Mature leaves of understorey species have higher average N and P concentrations than non-nodulated canopy species and higher average P concentrations than nodulated species: 3.0 % N and 0.26 % P (Frost in press). Grass nutritional quality is lower than that of woody leaves for much of the year, 1.3-2.2 N % during the early growing season, falling to 0.5-0.8 % during the early dry season. This is below the approximate level required to maintain grazing ungulates, and is a major constraint to livestock production in the region in the absence of supplementary feeding.

Nitrogen, phosphorus and other elements are withdrawn from the leaves of miombo trees as the leaves age (Ernst 1975). The concentrations of N and P in senescing leaves of canopy trees are respectively about 34 % and 23 % lower than those of mature leaves. This withdrawal reduces the quantity of nutrients being recycled and lowers litter quality. Despite this, the N and P content of litterfall is higher than in other tropical moist forests at equivalent levels of N and P flux in litterfall (Vitousek 1984). Estimated annual fluxes at two dry miombo sites amounted to 30-35 kg N ha-1 and 3.4-4.2 kg P ha-1.

The quality of foliage as a resource for herbivores or decomposers is also influenced by the amount of structural carbohydrates and the concentrations of secondary chemical compounds. Lignin levels are surprisingly low, less than 8 % (Jachmann 1989; Mtambanengwe and Kirchmann 1995). The level of total polyphenols is variable, 0-19 %, with non-N fixing species having higher levels on average, 10.2 %, than N-fixing species, 6.8 % (Palo et al. 1993). The difference in polyphenol:N ratios is even greater. Since leaf chemistry is a factor influencing both consumption by herbivores and decomposition, a wider survey of the chemical composition of leaves is needed, both among species and within a species through time.

The generally low nutritional quality of foliage is reflected in the low biomass of both wild and domestic herbivores. On average, the biomass density of indigenous large herbivores in conservation areas in which miombo predominates is only about 20-30 % of that expected at the same mean annual rainfall in those African ecosystems with nutrient-rich soils. In the richer ecosystems large herbivore biomass increases with increasing mean annual rainfall (Coe et al. 1976; Bell 1982), whereas in miombo it declines. Much of the herbivore biomass is made up of large-bodied species such as elephant, buffalo and selective grazers such as Lichtenstein’s hartebeest and sable and roan antelope. Specialist ungulate browsers are rare.

The level of consumption by large herbivores is low (about 1 % yr-1 of available browse: Martin 1974) and is generally less than the amount eaten by invertebrates, mainly lepidopteran larvae (2-4 % yr-1 of annual tree leaf production in normal years, see Frost in press). Consumption by invertebrates is distributed differently in time and space from that by mammals. Invertebrates are active mainly in the wet season and feeding tends to be distributed more uniformly among preferred food plants at a site. There are also marked year-to-year differences in consumption, associated with periodic outbreaks of lepidoptera larvae (Malaisse-Mousset et al. 1970). In contrast, mammals are active throughout the year but, except for elephant, they feed on herbaceous and understorey plants rather than in the canopy. Feeding is also patchily distributed, due both to selectivity for plant species and parts, and to daily and seasonal movements of the animals. Selection of mature leaves by elephant in Malawi was significantly positively correlated with the sugar and mineral content of tree leaves, and negatively correlated with total polyphenol, lignin and steroidal saponin levels (Jachmann and Bell 1985; Jachmann 1989). The controls on invertebrate herbivory are not known, though secondary chemistry is likely also to be involved.

Large areas of woodland can be defoliated during insect outbreaks. This results in a temporary reduction in leaf area, though the plants rapidly produce a new flush of leaves, particularly if defoliation occurs during the early part of the growing season. Insect outbreaks also produce a pulsed flux of nutrients from trees to the litter layer in insect frass which may stimulate microbial decomposition.

Litter quality is important in regulating the rate of litter decay, though moisture availability is perhaps the overriding constraint. Nitrogen mineralisation from litter is positively correlated with the initial N content and negatively correlated with C:N, cellulose:N and lignin:N ratios, among others (Mtambanengwe and Kirchmann 1995). Both litter decay rates and N mineralisation are low during the dry season. At a wet miombo site in Zaire (1270 mm MAP) more than 90 % of leaf litter decayed within a year (Malaisse et al. 1975), compared with only about 60 % at a drier site (885 mm MAP: Swift et al. in prep.). Termites took much of the litter at the wet miombo site.

Organic matter levels in miombo soils are generally low. The average organic carbon content of A-horizon soils from 64 profiles in miombo woodland was 1.4 % (range: 0.3-3.8 %). Organic carbon levels in dambo grasslands are substantially higher: 8.0 %, range 1.1-21.8 %. Low organic matter is a consequence of several factors: a preponderance of low-activity clays; high soil temperatures; the frequent incidence of fire; and the abundance of termites (Trapnell et al. 1976; Jones 1989). Widespread foraging on litter by termites of the family Macrotermitinae, most of which is completely decomposed by cellulase-producing fungi in their mounds, is particularly important (Jones 1989).

The size, density and regularity of tall termitaria is one of the prominent features of miombo landscapes. Tall mounds made by Macrotermes species can occur at densities of up to 5 mounds ha-1, covering about 8 % of the area (Malaisse 1973). At this density, the biomass of Macrotermes species alone may be as much as 46 kg DM ha-1, outweighing other soil fauna groups except humivorous termites. The biomass of the latter has been estimated to be up to 61 kg ha-1 (Goffinet 1976). This high biomass of termites may be responsible for considerable emissions of methane (Zimmerman et al. 1982, but see Seiler et al. 1984). Considerable amounts of methane are likely also to be emitted from seasonally flooded dambo soils, while during the dry season miombo soils generally may be a sink for methane.

Termite mound soils have higher levels of total N, acid-extractable P and cations than surrounding soils (Watson 1977; Trapnell et al. 1976; Jones 1989), due to both the concentration in, and subsequent decomposition of, organic matter within the mounds (Jones 1989), and the concentration of minerals through the evaporation of soil water in chimneys within the mound (Weir 1973). This produces nutrient-rich patches within an otherwise nutrient-poor landscape. The termitaria, as a result, invariably support distinctively different vegetation from the surrounding woodlands (Malaisse 1978b), and are often the focus of activity for birds and other animals which might otherwise not do so well in such a largely unproductive environment. Termitarium soil is widely used by farmers as an amendment to their fields (Watson 1977).

Miombo is notable among dry tropical woodlands and forests for the number of tree species having ectomycorrhizae (ECM) rather than vesicular-arbuscular mycorrhizae (VAM) associations (H÷gberg 1982, 1992; H÷gberg and Piearce 1986). Most of the dominant tree species including species of Brachystegia, Julbernardia, Isoberlinia (Fabaceae: Caesalpinioideae), Marquesia and Monotes (Dipterocarpaceae) and Uapaca (Euphorbiaceae) have ectomycorrhizae. Fewer tree species in miombo have VAM, and only some of these are nodulated and so potentially fix nitrogen. In contrast, many of the dominant understorey shrubs and herbs are nodulated, particularly in regularly burnt communities. The dominance of ECM tree species in miombo may reflect the advantage that such species have on seasonally-dry infertile soils; the few other tropical ecosystems dominated by ECM tree species also occur on infertile soils (H÷gberg 1982). Ecto-mycorrhizae may be particularly important in enabling plants to take up nutrients direct from the litter. The mycorrhizae depend on carbon supplied by the host plant but the cost of maintenance to the plants has not been measured in miombo. It could be substantial.

Dry season fires are a regular and frequent feature of miombo ecosystems (Trapnell 1959), occurring at most sites at 1-3 yearly intervals. Fires in miombo probably constitute the single largest area burned in the world, around 1 million km2 yr-1 (Scholes et al. 1996). Fire-return intervals depend on fuel accumulation rates, both at the site and in the surrounding vegetation, and on proximity to potential sources of ignition. Fires in the miombo region are an important source of trace gas emissions, methane, carbon monoxide and nitrous oxide (Scholes et al. 1996, Andreae et al. 1996). Most fires probably originate accidentally from people preparing land for cultivation, collecting honey, or making charcoal (Chidumayo 1995). Deliberate burning is done by hunters and livestock owners, to provide a green flush that attracts the animals. Fire is also used to clear areas around homesteads and alongside paths between settlements.

Fires in miombo are fuelled largely by grass and fine woody material, and their impact on plants depends on their intensity and timing in relation to plant phenology. Late dry season fires are more intense and destructive than fires occurring in the early dry season when fuel moisture contents are relatively high and ambient temperatures and windspeeds are low. Fuel load is largely a function of time since the last fire, the previous season’s rainfall (which determines grass production), the amount of grazing and the extent of woody plant cover. Fires tend to be more frequent and intense in areas of low woodland cover and high mean annual rainfall, where grass production is high but where grass quality and therefore grazing pressure is low. Emission factors for trace emissions from these fires can vary by a factor of 6 to 8, depending on grass:woodland ratio (Ward et al. 1996).

Four functional groups of miombo plants are recognised in relation to their degree of tolerance to fire (Trapnell 1959; Lawton 1978): fire-intolerant, such as the forest species; fire-tender (including most of the dominant canopy species when young); and semi-tolerant and fire-tolerant species. Changes in fire frequency therefore change vegetation structure and composition. Frequent late dry season fires eventually transform woodland into open, tall grass savannah with only isolated, fire-tolerant canopy trees and scattered understorey trees and shrubs. In contrast, woody plants are favoured by both early burning and complete protection. Since grass biomass declines sharply as tree cover increases, a period without fire would lead to gradual tree canopy closure, the suppression of grass growth, and lower fuel loads. Lower fuel loads mean less intense fires, less damage to woody plants, uninterrupted woody regrowth and continued canopy closure.

Fire interacts with large herbivores in affecting vegetation structure (Guy 1989). Elephants are well known for their habit of ringbarking, breaking, pushing over and uprooting trees and shrubs (Thomson 1975). This transforms relatively dense woodlands into more open wooded grasslands with scattered tall trees, resprouting tree stumps, and a dense layer of low growing shrubs. The increased grass production leads to higher dry season fuel loads, more frequent and intense fires and further suppression of woody plants.

The dominant trend in regenerating miombo in the absence of frequent hot fires or other intense disturbances is therefore towards the development of woodland (Strang 1974). The trees are difficult to eradicate, most of the woody plants resprout vigorously from surviving stems and rootstocks. Only where the plants are thoroughly uprooted during disturbance are marked changes in composition likely. Both the dispersability and longevity of the seeds of Brachystegia, Julbernardia and most other Caesalpinioideae are low so that, where the trees are eliminated, recovery to the original state, at best, will be slow.

The dynamics of miombo woodlands have been interpreted largely in terms of a single-state equilibrium model of succession to a regional climax vegetation - dense woodland in drier regions and semi-evergreen or evergreen forest in wetter areas - with fire and disturbance by people or wildlife being the main agents of disturbance (Freson et al. 1974; Strang 1974; Lawton 1978). For example, Freson et al. (1974) propose a three-stage regressive series (dense dry forest-open miombo woodland-savannah) induced by the combination of wood cutting and fire. The view that miombo is a sub-climax formation arises partly from observations of miombo juxtaposed with patches of evergreen forest where topography and soils seem superficially to be the same (e.g., Lawton 1978). White (1983), however, notes that where evergreen forest occurs alongside wet miombo there is a transition to deeper soils, suggesting an edaphic control.

More recently, there have been attempts to describe miombo dynamics in terms of multi-state models in which, following abandonment of cultivated fields, or the combined impacts of elephants and fire, there is a transition either to open woodland dominated by Combretum spp. or to grassland (Stromgaard 1986; Starfield et al. 1993). Fire is considered to be the driving force behind the transitions. A longer than normal fire-free period is needed to return the vegetation to Brachystegia-dominated woodland.

Few studies of vegetation change in miombo have followed changes at a site over time. Most interpretations have been based instead on contemporaneous studies of the vegetation at a number of sites presumed to differ only in fire regime (Trapnell 1959; Lawton 1978; Chidumayo 1988) or time since abandonment of cultivation (Stromgaard 1986). Short-term studies in any event are unlikely to produce many insights, given that most of the tree species, including all the canopy dominants, are relatively long-lived (lifespans 60-120 years). Simulation models of plant community dynamics (e.g., Desanker and Prentice 1994), supported by long-term monitoring of vegetation change at a range of sites, will be needed to predict vegetation change under existing and changed conditions.

Land Uses in Miombo

Human population densities over much of the miombo region are relatively low, averaging about 15 persons km-2 (range: 7-8 persons km-2 in Angola and Zambia to over 110 persons km-2 in Malawi), generally lower than those found in climatically similar areas elsewhere in Africa. Population growth rates are high, 2.5-3.5 % per annum, and although some of the increase in population is being absorbed by urbanisation (bringing its own problems), the expanding populations mean greater demands on the land and its resources and changes in the kinds and intensities of land use.

There is considerable diversity in the ways people use miombo (Trapnell and Clothier 1937; Allan 1965; Puzo 1978). Apart from commercial farming (tobacco, maize, soya) and land set aside for wildlife conservation and tourism (national parks, safari areas and game reserves etc.) the main form of land use is small-scale sedentary and shifting cultivation of maize, cassava, small grains and pulses. Where present, dambos are used for growing cereals and root crops during the wet season, and for growing vegetables and grazing livestock during the dry season. The interfluves are used for cultivating rain-fed crops, wet-season grazing, and as the source of wood for fuel and construction, as well as of various non-timber forest products (NTFPs). Apart from the drier areas of Zimbabwe and Tanzania, there are relatively few livestock, many fewer than people would like to have. The reasons are complex; low forage quality and animal diseases are major proximate constraints, themselves reflections of broader influences of climate, geology and soils (Bell 1982).

Many households depend on natural products derived from the surrounding woodlands and grasslands. These products include fuelwood, charcoal, timber, thatching grass, fibre, fruit, mushrooms, honey, edible caterpillars and other animals, organic inputs for fields, fodder for livestock, and many other items. At least 50 edible plant species are recorded from miombo (Malaisse 1978a). A number of cultural practices, largely unique to the miombo region in their extent, have developed in response to some of the prominent ecological features of miombo ecosystems. For example, the widespread traditions of bee-keeping, gathering mushrooms, and collecting edible caterpillars reflect, respectively, the dominance of prolific flowering entomophilous trees (Clauss 1992); the dominance of tree species with ectomycorrhizal associations, many of which produce edible mushrooms (Pegler and Piearce 1980); and periodic eruptions of some lepidoptera populations (Malaisse-Mousset et al. 1970).

The farmers use miombo landscapes in a fine-grained manner, exploiting localised patches of higher fertility and avoiding areas of particularly poor soils (Carter and Murwira 1995). Because of the low fertility of most of the soils, few areas, other than fertile alluvium in river valleys, can be cultivated continuously without a decline in fertility. Organic amendments (leaf litter, manure), termitarium soil, ash from burnt brush piles and, where available and affordable, inorganic fertilizers, are used to counter low and declining soil fertility (Swift et al. 1989; Carter and Murwira 1995). Shifting cultivation is carried out where there is enough land, though this is changing as population pressures increase (Chidumayo 1987a; Stromgaard 1989).

Slash-and-burn (ash-fertilisation) agriculture in particular is widely practised in miombo. The practice appears well adapted to the generally infertile soils of miombo, particularly in the wetter regions where woody plant biomass is high and cut trees regenerate rapidly through resprouting. The best-known of these is the large-circle chitemene system practised by the Bemba people of northern Zambia and southeastern Zaire (Richards 1939; Allan 1965; Puzo 1978). In this system the foliage and outer branches of trees are cut from an outfield area and stacked in an infield area where they are left to dry before being burnt just prior to the annual rains. The outfield area is about 5-8 times larger than the infield area, which averages about 0.3 ha (range 0.02-1.03 ha). Traditionally, the trees are not stumped but are allowed to resprout before again being harvested about 25 years later (Chidumayo 1987a; Stromgaard 1988).

The plots to be chopped are selected on the amount of wood available for burning (Puzo 1978). By concentrating the brush into large piles, rather than spreading them over a wider area, farmers obtain a deeper ash bed and a greater fertilising effect through the release of plant nutrients and increases in soil pH (Stromgaard 1984; Mapiki 1988; Araki 1992). A deeper ash bed reduces the amount of soil tillage required, an advantage to farmers who have no livestock, while the greater fertility contributes to larger crop yields. This means that the plots to be cultivated can be smaller and more manageable. More brush also means a hotter fire, more complete combustion of the plant material, and the suppression of weeds during the first few years of cultivation. The land is cultivated until the added fertility is exhausted or until the regenerating woody vegetation and weeds make continued cultivation unproductive. The plot is then abandoned to a long fallow period and new fields are opened up (Puzo, 1978).

Various other forms of shifting cultivation occur within the region. Some are variants of the chitemene system described above, others involve short-rotation fallows. On inherently infertile soils, a long fallow period is necessary for replenishment of both the vegetation and soil nutrients. With the ongoing expansion of the human population in the region, leading to a reduction in the area of uncultivated land, fallow periods are becoming shorter and shifting cultivation is gradually being replaced by more permanent agriculture (Lawton 1982; Chidumayo 1987a; Stromgaard 1989). As the duration of the fallow declines, trees are felled rather than lopped to produce sufficient biomass to burn and fertilize the soil, reducing subsequent fallows to a fire-maintained wooded grassland called chipya. Eventually, most of the potentially arable land is cultivated more-or-less permanently. In the absence of more intensive soil fertility management, this is likely to result in a gradual and long-lasting decline in fertility.

Shifting cultivation is often viewed as an early stage in agricultural development, conditioned by low population densities (and therefore plenty of uncultivated land and no incentive to develop more intensive methods of farming) and limited technology. In miombo, this argument is extended to cover the impact of tsetse fly, mainly Glossina morsitans, the vector of the protozoan blood parasites, Trypanosoma spp., that cause trypanosomiasis (‘sleeping sickness’ in humans; nagana among domestic livestock). The threat posed by trypanosomiasis is thought to prevent people from keeping domestic livestock, particularly cattle and donkeys, leading to shortages of animal draught power and manure for fertilising the fields; inefficient hand-cultivation of small plots; rapid exhaustion of the soil; and early abandonment in favour of virgin woodland elsewhere.

At best, this is a gross oversimplification of the complexities involved in the development and practice of shifting agriculture. Richards (1985) has cautioned against viewing shifting cultivation as a single, undifferentiated system of farming, distinct from permanent cultivation, for example. Instead, it should be seen to constitute a range of actions, among a wider array of land-use practices, each used according to particular ecological and economic circumstances.

Given the low nutrient status and high acidity of the soils it is perhaps more appropriate to view chitemene as well-adapted strategy of ash-fertilisation that makes use of the inherent capacity of the plants to resprout when cut, and which allows people with limited resources to overcome the constraints of their environment through capitalising on the nutrients stored in the vegetation.

The presence of tsetse fly undoubtedly has affected human ecology in miombo, but not necessarily in the way and to the extent commonly assumed.

The presence of tsetse fly and the threat of trypanosomiasis seems to have determined pre-colonial settlement patterns and densities. For example, in south-eastern Tanzania, people lived in small isolated settlements, minimising the risk of infection by human trypanosomiasis (Matzke 1979). People responded to outbreaks of sleeping sickness by abandoning the affected areas and dispersing to new settlements at lower densities. The alternative response, widely promoted by colonial authorities, was to consolidate people into large enough concentrations to clear and keep open the surrounding bush thereby destroying tsetse fly habitat in a sufficient radius around the settlement and associated agricultural lands (Matzke 1979). This was possible in drier areas but has proved difficult to sustain in wetter regions where tree regrowth rates are high and the soils too infertile to be able to support the local density of people needed to keep the vegetation open (Matzke 1979). The presence of tsetse fly still governs aspects of agricultural, veterinary and development policies in the region. Attempts to rid miombo of tsetse fly through bush-clearing, eradication of wildlife, ground- and aerial-spraying of insecticides, and the use of insecticide-impregnated, odour-baited ‘targets’, have all had, and continue to have, an impact on land use and land cover across large areas of miombo.

Land use in miombo therefore reflects a range of adaptations by farmers to the intrinsic constraints and opportunities of climate, landscapes, soils, vegetation, animals, pests and disease. Many of the adaptations, however, have developed under different socio-economic and demographic circumstances from those prevailing today, and are becoming increasingly difficult to sustain as populations increase and the demand for agricultural land intensifies. Population growth, however, is not the only, nor necessarily the main, cause of pressure on natural resources. Superimposed on a largely inherently infertile environment are inequalities in land distribution; unsecured tenure; deteriorating economic circumstances; lack of credit; technical advances in disease-control and agricultural production; commoditisation of agriculture; declining terms of trade among less developed countries on world trade markets; increasing commercialisation of natural products; and frequent droughts.

Individually and interactively, these factors are affecting the patterns of land-use and productivity. Large areas of woodland have been, and continue to be, modified or transformed by people. The changes include reductions in tree density due to frequent harvesting for fuelwood, charcoal production, construction material, and slash for burning and fertilising fields; declines in woodland cover due to frequent burning; and conversion to permanently cultivated land or, to a much lesser extent, plantations of fast-growing non-miombo trees. These changes in land cover and ecological functioning have long-term socio-economic impacts (such as reductions in the availability of NTFPs, less forage for livestock, and fewer inputs to sustain the fertility of increasing areas of arable land), as well as a range of environmental impacts affecting ecological functioning, carbon storage, trace gas emissions, hydrology and regional climate (Justice et al. 1994 ). This programme is aimed at understanding these impacts and their consequences.

At the Miombo Intercore Project Meeting in Zomba, key issues for a miombo global change research initiative were identified, these were formulated into series of research questions and hypotheses which are presented below.

What are the Current Directions and Rates of Change in Land Use and Land Cover?

Human activities are central to the current dynamics of miombo ecosystems. The extensive and intensive use of the soils and vegetation by agrarian communities has undoubtedly shaped, and continues to shape, much of the present miombo landscape. The details, however, have still to be fully revealed and understood. Much of the available data are descriptive rather than analytical; collected in different ways, making comparisons difficult; short-term, so that lagged feedbacks are not apparent; and focused either at the micro-scale, with extreme emphasis on local detail, or at a regional or larger scale, where the data are so aggregated and generalised that the underlying processes are difficult to discern.

There is therefore an urgent need to link the findings of site-, time-, and circumstance-specific case studies with the broader findings of regional studies of land-use and land-cover change. A more systematic approach is required that aims to understand which details of human circumstances, knowledge, perceptions, motives, actions and responses, so apparent at a local scale, emerge as pattern and influence processes at a larger scale. Moreover, because the circumstances are not static, the focus of research must shift from descriptions of pattern to studies of processes.

	What are the present patterns of land use in miombo, and what are their historical, socio-economic and environmental determinants?
	To what extent, in what ways, and at what rates are the patterns of land use changing, and why?
	What are the external driving forces and internal modifiers of the changes in land use and land cover?
	What are the effects on land use of current macroeconomic policies, and of sectoral policies on agriculture, forestry and conservation?
	Can the location and rates of future land use change be predicted?
	How are these changes altering the attributes of different land cover types and their spatial patterns (sizes, shapes, contiguity and connectivity)?

One hypothesis is that traditional land use practices are governed by the constraints of infertile soils, inadequate draught power (due to a shortage of livestock because of poor-quality grazing and the prevalence of animal diseases), and limited markets for produce. These constraints are currently being alleviated through a wider use of fertilizers and cattle feed supplements, the control of disease, and improvements in infrastructure, coupled with a rapidly-growing population and political stabilization in southern Africa. The prediction is that these developments will lead in turn to massive changes in land use and land cover throughout the miombo region in the coming decades. The process is already observable in countries such as Malawi and Zimbabwe. The rapid conversion of large areas of cerrado in Brazil to crop lands during the 1980s illustrates the speed with which the process can occur once constraints are removed.

Other important issues relate to the human dimensions of land use and resource allocation in this important and sensitive region. Transnational and internal demographic factors, most notably migration, must be considered part of the research. In addition, broad issues which relate land use practices to cultural attributes on the one hand and market forces on the other must be considered in terms of understanding the current and future trajectories of land use and cover change in Miombo. For instance one hypothesis is that strong culturally-endowed land use practices limit peasant involvement in market economies; that African peasants deliberately refuse to be captured by the market economy. Finally the tenent that land use intensification cannot take place in this region must be examined critically.

What are the Consequences of Land-Use and Land-Cover Change?

Changes in land use and land cover in miombo potentially affect a wide range of socio-economic and environmental processes. Reductions in the area of uncultivated woodlands may reduce the availability of fuel wood, construction material, and non-timber forest products; diminish the area of communal grazing land; and adversely affect the free ecological services provided by trees. Such impacts may fall disproportionately on the poorest members of society, worsening their socio-economic circumstances. Reductions in tree canopy cover can adversely affect local, and possibly regional, climates through effects on albedo, air temperatures and relative humidity (Malaisse 1978a), while the reduced leaf area means lower evapotrans-piration rates, greater recharge of groundwater, and changes in hydrological functioning downstream (McFarlane and Whitlow 1990). Conversion to permanently cropped land, in the absence of intensive soil management, can lead to reductions in soil organic matter, nutrient depletion, and soil erosion. In all cases, changes in the kinds and rates of emission of trace gases can be expected.

The land surface, and in particular vegetation, regulates the partitioning of radiant energy into different components. This partitioning is an important boundary condition for the climate system at all scales. Changes in land surface properties will result in changes in energy partitioning, and hence in climate. The land surface is also important for climate as it is a major potential source of radiatively reflective particulates and radiatively active trace gases such as CO2 and methane. Complex feedbacks exist between the land surface and climate which operate through both the hydrological and carbon cycles.

Processes that influence surface albedo are reviewed by Henderson-Sellers and Wilson (1983). In general, human-induced land-use changes tend to increase albedo, resulting in more energy being reflected back into the atmosphere. It has been suggested that this contributes to warming of the upper troposphere, which in turn leads to greater atmospheric stability and less convective rainfall (Shukla et al. 1990). Gornitz (1985, 1987) reconstructed land use changes in West Africa from historical records, censuses, atlases and descriptive reports and calculated changes in albedo due to alterations in land use. He concluded that evidence for apparent desiccation or ‘desert creep’ (‘desertification’) may be attributed, in large part, to adverse changes in soil and stream hydrology caused by anthropogenic disruption of the vegetation cover (Gornitz 1985). There are many counteracting factors to take into account when predicting how land-use changes might affect climate. Therefore, it is important to document the details of land cover and reconstruct its dynamics to analyze the effects of changes on the fluxes of energy and water, and how these in turn will affect climate.

	What are the phenological characteristics, physical features (plant height, density, cover, leaf area and distribution) and surface attributes (albedo, surface roughness, canopy conductance) of different land cover types in miombo?
	In what ways do these features influence soil-vegetation-atmosphere interactions?
	What is the impact of widespread, frequent fires on surface albedo, energy and water balance, and surface roughness?
	What effects do interannual climate variability have on land use, soil-
	vegetation-atmospheric interactions and the frequency of fire?

The headwaters of several major southern and Central African river systems lie in the miombo region. These rivers, on which many people depend, include the Zambezi, Kafue, Lualaba, Okavango, Ruaha, Rufiji, Rovuma, Save and Zaire. Changes in land cover are likely to have a variety of impacts on the total amount of water reaching these rivers, the timing of discharge, and on water quality (especially sediment and nutrient contents). For example, reductions in tree canopy cover on interfluves has been shown to increase annual streamflow in the upper Kafue River (Mumeka 1986), while augmented groundwater flow has been implicated in faster rates of subaerial erosion of dambos (McFarlane and Whitlow 1990).

	In what ways, and to what extent, do changes in land use and land cover affect local, landscape and regional hydrological processes?
	What is the relationship between land use and mass transport of the products of interfluve erosion, and what is the impact on water storage capacity downstream?
	What effects will these changes have on the availability and quality of both surface and subsurface water resources?
	What is the impact of more intensive land use of wetlands, particularly dambos?
	In what ways might changes in climate, including changes in rainfall variability, affect patterns of land use, water storage, and water quality?

The wide extent of miombo means that land use change in the region could have globally-significant impacts. Miombo landscapes are the main source of the cloud of tropospheric ozone which drifts off Africa every September (Fishman et al. 1991), and are the source of about 10% of the global carbon monoxide and biomass-burning aerosol particle budgets (Scholes et al. 1996a). Detailed analyses of tropospheric ozone over southern Africa are given in Thompson et al. (1996a and 1996b). The miombo woodlands are also significant sources of nitric oxide from biogenic processes in the soil; methane from the dambo wetlands, ruminants and termites; and non-methane hydrocarbons from the vegetation.

Mature miombo woodlands contain about 95-136 Mg C ha-1 in the vegetation (~40 %) and soil (~60 %), depending on climate. Conversion of these woodlands to short-duration crop agriculture would release large amounts of carbon dioxide to the atmosphere, perhaps as much as 50 Mg C ha-1, or about 14 Pg C if the entire miombo region were to be converted. This is a potentially large contribution in relation to the current annual global flux due to land use change of about 1.6 Pg C yr-1 (IPCC 1996). On the other hand, if miombo woodlands were managed to maximise carbon storage, a substantial quantity of carbon could be sequestered in biomass, soils and woodland products.

	What are the major trace-gas fluxes between miombo ecosystems and the atmosphere, and how do these differ among the various land-cover types?
	What are the controls on these fluxes?
	What effects do changes in land use and land cover have on the flux of trace gases, both during conversion and afterwards?
	What are the types and amounts and geographic distribution of trace-gas emissions from different kinds of biomass fires (wildfires, slash-and-burn, hearth fires, charcoal fires), and what is their interannual variability?
	What are the residual (post-fire) effects of wildfires on the exchange of trace gases among soils, vegetation, and the atmosphere?

The principle natural resources derived from miombo woodlands are not reflected in the formal economy. They consist of fuelwood and charcoal, essential for household energy for millions of people; fruit and other foods - honey, edible fungi, edible caterpillars, rodents and other small animals; timber for house construction; thatching grass; and wood and fibre for making household implements, baskets and other handicrafts. Rainfall in the miombo area is variable, resulting in periodic food shortages. On these occasions, the availability of wild foods, as well as other natural products that can be harvested and sold or exchanged for food, can be crucial to survival. These products also supplement peoples’ livelihoods and diets during high-production years.

	What are the present trends in the use of natural resources of miombo?
	What impacts are present policies on land use (agriculture, forestry, conservation), agricultural credit, pricing and marketing, and economic structural adjustment, having on the resource-use practices at local and regional scales? What would be the potential impact of major changes in these policies?
	What is the impact of commercialization on the patterns of resource use?
	What are the key ecological processes sustaining production in miombo ecosystems and how are these being affected by land use change?
	In what ways are changes in land use affecting the kinds and availability of natural products?
	To what extent are these changes affecting the patterns and intensities of use of these resources?

What Controls the Distribution of Miombo Species and Ecosystems?

The ability to understand the course of ecological change and to predict its longer-term consequences depends greatly on the level of understanding of how a particular ecosystem functions and what controls the distribution and abundance of key species. Miombo forms one of the major ecosystem types in Africa, occurring over a wide climatic range and a somewhat narrower range of soils, yet still displaying a relatively sharp transition to other major formations. Any Dynamic Global Vegetation Model (DGVM) must be able to predict the current extent of miombo if it is to be useful in predicting the consequences of future climate change. Existing DGVMs have been developed from a north-temperate perspective and are largely untested for major tropical formations. Similarly there is a need for such models to incorporate land use (e.g. fire) as an integrated ecosystem process.

A working hypothesis is that the distribution of miombo ecosystems is controlled by four factors: warm temperatures; moderate rainfall; a marked dry season; and infertile soils. The present biophysical boundaries of miombo are considered to be determined by climatic factors and soil fertility as follows:

	The upper altitudinal limit and possibly the southern geographical limit of miombo is controlled by low temperature (~-4 oC absolute mean minimum temperature: Ernst 1971).
	There is a lower limit to the length of the dry season (~4 months) below which either the trees are evergreen and form a closed canopy, precluding fire, or the soil is permanently saturated and supports hydromorphic grasslands and swamp forest.
	There is a lower limit to water availability (700-1500 mm yr-1 and a minimum of 90-100 days growth opportunity annually) below which trees belonging to the Caesalpinioideae are replaced by more arid-adapted groups.
	There is an upper limit to soil fertility, primarily in terms of P availability and possibly pH, above which miombo species are out-competed by faster growing species characteristic of Acacia-dominated savannahs.

The distribution of the key miombo species is likely to be controlled by a similar combination of climatic and edaphic factors.

What Controls Primary Production in Miombo Ecosystems?

The working hypothesis is that primary production in miombo ecosystems is broadly related to rainfall, but is everywhere strongly constrained by nutrient availability, particularly nitrogen and phosphorus. The relative importance of these two elements, and the extent to which they interact in their effects on miombo production, is not fully known. Because primary production underpins both the natural resource potential of miombo and its potential for absorbing or emitting greenhouse gases, the question of the controls on primary production is of fundamental importance.

The structure and composition of miombo woodlands, and hence its production potential, is also affected by disturbances such as fire and the felling of trees by people or elephants. These disturbances in turn affect the availability of water and nutrients, often positively. The nature of the interactions between climate, disturbance, structure, composition and function need to be explored more systematically, not least because they probably hold the key to understanding how miombo species and systems will respond to changes in global climate.

The proposed research on Miombo is partitioned into seven core experiments and a set of integrating activities that cut across the core experiments. The order in which they are listed does not reflect any particular priority. The approach taken begins with analyses of the character and dynamics of land use and cover change, which is then tied to its ecological effects. In essence, the approach described below first defines the causes of land use and cover change, and then its consequences in ecological terms.

It is recognized that there will also be significant social, institutional and economic consequences to changing land use patterns in the region, and this component will be developed at a later point through close collaboration with the appropriate Programme Elements of the IHDP, such as its emerging Institutions Project. However, since the program structure of the IHDP is in its formative stages, this report focuses the discussion of consequences of LUCC in biophysical terms.

The first two Core Experiments define an approach which is parallel to Focus 1 and Focus 2 of the LUCC Science Plan (IGBP Report 35/HDP Report 7). The approach articulated in this document heavily emphasizes Focus 2 of LUCC, which is oriented around direct observations and empirical models. This approach, which could be called "pattern to process", begins with analyses of the spatial and temporal characteristics of LUCC in the region, and then, through case studies, attempts to define the underlying processes at various scales which give rise to the patterns observed. Further refinement of the science plan is expected, and in doing so the Core research activities would expand the work of Focus 1 through more detailed case studies.

Core Experiment 1: Land Use Dynamics

Case studies, along with spatial analyses from direct observation (Core Experiment 2) can be used to build sets of micro- and meso-level models of land-use and land-cover change through time. The goals of such an approach are both description and explanation in the form of useful causal understanding (Clark, Jones and Holling 1979).

The aim of this experiment is to develop and test hypotheses regarding the controls on land use in the miombo region. The complexity of the problem deepens with the understanding that three dimensions of drivers - socio-economic, biophysical, and land management (proximate causes) - are relevant to land-use/cover change, two of which involve adaptive agents and systems that respond to and sometimes anticipate changes in the other spheres. Global change modellers have only recently begun to deal with this problem with regard even to biotic feedbacks (Baskin 1993). An approach that integrates the three spheres, therefore, is not only necessary, it is forward looking.

The approach will consider the multiple ways in which land cover and human activity interact. It can change as a result of: (i) independent changes in biophysical drivers (e.g., climate and atmospheric change); (ii) direct alteration (e.g., deforestation); (iii) indirectly mediated through the biophysical processes (e.g., groundwater withdrawal leading to a lowered water table and to reduced stream flow and altered vegetation); or (iv) a more complex chain of human activity, which feeds back to human activity, which then directly alters land cover (e.g., human introduction of rinderpest, change in wild herbivore population, advance of woodland with tsetse, leading ultimately to mechanised clearing, see Sinclair 1979).

There are two steps to disaggregating this problem of multiple factors. The first is to divide forcing functions into social and biophysical drivers and proximal drivers. The second is to analyse each set of drivers in terms of spatial and temporal scale in an attempt to resolve at least some of the complexity and controversy regarding multiple driving forces. This step draws on Holling’s argument that the dynamic behaviour of many ecological systems can be explained by a few driving forces operating at characteristic frequencies and spatial scales - the extended keystone hypothesis (Holling 1992).

Available data (mostly descriptive) will be collated and new data collected. These data will be used to discern underlying processes of land-use change at the regionl scale. Land-use change models will be constructed and will include factors such as biophysical suitability of broad areas for various land uses; trends in regional and global economics and trade; and regional demographic trends. Results from this experiment will be used along with land cover change information from the Land Cover Dynamics Experiment to develop an understanding of how land cover changes are driven by land-use changes.

The kinds of questions which will be emphasized relate to the spatial and temporal dynamics of land use and cover change, in particular investigating what controls the interannual vaiability of land use changes as well as its spatial pattern over time. The research will consider such questions as: (a) Are the inter-annual dynamics and rates of land conversion and abandonment to secondary forest significantly different than the decadal mean trends; and (b) Through the integration of socioeconomic and satellite data and the development of dynamic LUCC models, can we improve our understanding of the dynamics of LUCC and the various controls (proximate and distant determinants) on rates of conversion and regrowth and land use transition sequences?

Activity 1: Microscale Land Use Case Studies

Case studies will document important factors for major current land-use systems such as shifting cultivation, slash and burn systems, modern village scale agriculture, cash-economy agriculture, and government land use (forest reserves, national parks, plantations). Studies will also document patterns of land-use in areas constrained by major pests and disease, such as the tsetse fly and sleeping sickness. Some of these data exist in the region, and will be collated and evaluated.

Task 1: Case Studies of Specific Land-Use Systems, Communities and Cultures

Activity 2: Historical Land-Use Patterns

The purpose of this activity is to build datasets for the development and testing of models. These will be reconstructed from written records for the period 5000 to 20 years ago. Recent data will rely heavily on remote sensing products and regional/global data sets developed by various agencies such as FAO, CIESEN, World Bank, and national archives.

Task 1	Reconstruction of Land Cover and Land-Use over the Last 300 years
Task 2	Development of Spatial Database of Land Use and Socio-Economic Drivers
Task 3	Archaeological and Historical Studies of Land-Use Patterns

Activity 3: Regional Land-Use Change Modelling

The current distribution of land cover, recent trends in land use change when combined with existing socio-economic data bases and trends can be used to model potential changes of land cover and land-use change under different scenarios of population, economy and climate change. In the miombo region there are several relatively small scale studies (e.g., Scudder 1982; Kalipeni et al. 1988) which can form the basic platform for creating a model for the entire region. Scudder has studied one area over a 30+year period, while Kalipeni et al. have documented a case study of the rural communities supporting a major city. Other case studies exist. Information like these is invaluable in looking at human dimensions of land cover change. The main purpose is to build a predictive model of land use change that will capture our current understanding of land use systems in the region. The model will be used to develop scenarios for the immediate future given population and socio-economic projections.

The causes of land cover conversion have often been attributed to such things as population growth, road building, and a host of other similar monotonically varying. Although population and infrastructure are important long term determinants of general trends over decades, interannual variability is not tightly linked to such parameters. Econometric variables such as variations in price structures and local market demands are likely to be better predictors of deforestation and abandonment rates over short time cycles. It is possible to couple satellite-based LUCC datasets to socio-economic data to build diagnostic models of the LUCC process.

Such models could provide: (a) an improved understanding of the factors which control variations in LUCC rates annually; (b) an improved understanding of the factors which determine the rate of secondary fallow forest clearing; and (c) an improved understanding of the factors which control the balance between deforestation (the bringing of land into production) and abandonment (the taking of land out of production).

It is possible to develop models at two levels. The first are models based at Case Study sites and focused on establishing the critical village-level variables, dynamics and proximate determinants. The second level of model analysis could be developed at the scale of countries or the region. The Case Studies can be used to calibrate the regional model and provide a basis and location for validation of regional results.

Task 1	Develop Multiscale Model of Land-Use Dynamics
Task 2	Develop Scenarios of Land Use and Land Cover for Next 30 years

Core Experiment 2: Land Cover Dynamics

Activity 1: Land Cover Change Mapping and Analysis

Recent developments through the NASA Landsat Pathfinder Program have shown that large regions of tropical forest can be mapped using high resolution satellite data to provide the basis for carbon balance studies and natural resource management (Skole and Tucker 1993). Time series satellite data over the last twenty years are used to derive the location, extent and rates of deforestation. Similar base maps are now needed for the miombo region. Due to the strong seasonality and inter-annual variability of vegetation of the miombo system there is the need to augment the Pathfinder approach to ensure selection of the optimum period for discrimination of the desired land cover types and to enable interannual comparisons. This proposed activity would provide a pilot study to evaluate the synergistic use of time-series coarse resolution and temporal samples of high resolution data. The activity would also need to match the components of the miombo system that can be mapped using these complimentary data types with the vegetation classes required by the miombo scientists. During this pilot study period it will be important to ascertain the current status of land cover mapping within the miombo region and identify any recent developments using satellite data to map vegetation and vegetation change. For example, the current status and plans of the United Nations/Food and Agricultural Organization (UN/FAO) Africover Project, World Bank REIMP project, the FAO 2000 Tropical Forest Assessment, the Zimap/SPOT national image mapping project are complementary projects already identified which will involve various national forestry and mapping agencies within the region. Close coordination is desirable with such on-going land cover mapping projects.

Research and development is underway to develop new and improved techniques for surface parameterisation using existing and planned satellite remote sensing systems e.g., Running et al. (1995). There will be a considerable advantage to having the IGBP Transects provide a test bed for developing and testing these algorithms. For example, involvement of miombo scientists in the evaluation of these resulting data sets would help ensure the applicability of the methods to the miombo environment and an early access to improved data sets. It is quite feasible that a series of satellite product validation campaigns could be organized within the region underpinned by a strong ground measurement program. Within NASA’s Earth Observing System (EOS) there is currently discussion of a potential regional miombo validation campaign following the launch of the EOS AM platform.

Task 1

High Resolution Satellite Mapping of the Miombo Region

Following the pilot studies an implementation plan will be developed for a comprehensive mapping of the entire Miombo region using up-to-date high resolution imagery. Emphasis will be given to having in-country expertise closely involved in the image interpretation. The interpreted image data planes could be entered into a Geographic Information System (GIS), mosaiced, registered and mapped to a coordinate system appropriate for the region, following the Landsat Pathfinder approach (Skole and Tucker 1993). The main product will be a spatially explicit data base for mid 1990’s.