Survival, regeneration and leaf biomass changes in woody plants following spring burns in Burkea africana—Ochna pulchra Savanna*

Effects o f tw o intensities o f spring burn on various aspects o f w oody plants o f a Burkea africana—Ochna pulchra Savanna after one growth season are given. Mortality o f w oody plants was very low with, for example, that o f in­ dividuals o f Ochna pulchra being between 1 and 5% . Som e species where the above-ground parts were often burned away com pletely, as in Grewia flavescens, no mortality o f individuals occurred. Basal regeneration shoot mass was found to depend parabolically on plant height while the ratio o f leaf to twig m ass in basal shoot regeneration varied inversely with plant height in Ochna pulchra. The ability o f Ochna pulchra plants to produce new basal shoots appeared to not only depend on size o f the plant but also on the number o f basal shoots present prior to the fire. In live Ochna pulchra plants basal regeneration shoot biomass per individual was found to increase exponentially with greater reduction in canopy leaf biom ass. This relation was also affected by possible direct heat effects. Basal shoot regeneration mass was found to vary greatly with species and varied from 0,7 g/individual for Dichapetalum cymosum to 285,6 g /individual for Euclea natalensis. There was a clear tendency for non-suffrutex shrub species to have greater mean basal regeneration shoot mass per plant than that o f most tree species. There was a compensatory effect in Ochna pulchra between number and size o f basal regeneration shoots. Standing dead w oody plant in­ dividuals (before the burn) were either felled by fire or apparently unaffected by fire and there was no selectivity by species. Results o f the present study are generally supported by other work on the effects o f fire in savanna and som e other vegetation types.


IN T R O D U C T IO N
Fire has probably occurred regularly in the savan nas of southern Africa, initiated first by agencies such as lightning and later increasingly by man. The structure of savanna is probably related to this incidence of fire. In the past fire, once started, could spread over vast areas of savanna but more recently fire is often contained by artificial fire breaks such as roads. Fire is used as a veld management tool for both livestock farming and wildlife conservation. The two main objectives of burning savanna for live stock farming are to remove moribund grass and to decrease the woody plant element. In wildlife management these two objectives are also important, but additional reasons include encouraging some form of rotational grazing, reducing the wild-fire hazard and controlling animal diseases and parasites (Trollope et al., in press). In the savannas of the central-northern areas of the Transvaal the current frequency of fire is extremely variable from annually to only every few decades in more protected areas although the more common frequency may be described as between one and five years. Fires in savannas are usually surface fires, and crown fires are very rare although localized burning of canopies can occur. Controlled fires are almost always applied in or near the dry season, usually between May and November in southern African savanna vegetation areas. In these areas it is generally recognized that a fire earlier in this period is less effective in controlling tree growth than a fire later in this period (West, 1965). Effectiveness of late burns as a means of woody plant control is usually ascribed to the fire being hotter and the trees being more susceptible to heat injury at a time when growth has just started. However, Deeming et al., (1972) point out that this stage of plant development (the rapid growth stage) indicates a 'high moisture content throughout the plant' which (as in other living fuels) acts as a heat sink since it takes considerable amounts of energy to dessicate this material. Only after dessication can such material itself act as a heat source. In contrast to burning in the dry season, it has been found (Anon, 1960) in the Transvaal lowveld that very little damage is done to trees and shrubs during late summer (February) veld burning when grass is green.
That fire can fundamentally affect many com ponents of an ecosystem has long been recognized. Interest has been shown in many diverse effects, for example, the cycling of nutrients (Trapnell et al., 1976;Christensen, 1977), changes in the water balance and the possible effects on past carbon dio xide changes in the atmosphere (Reiners & Wright, 1976), while work on woody plants has included classifying savanna woody plant species according to different degrees of fire tolerance with the study of the evolution of pyrophytic habits (Jackson, 1974), derivation of allometric fuel load prediction formu lae (Kaul and Jain, 1967), modelling fuel mass (McNab et al., 1978) and determining fire scar tree ring chronologies (Zackrisson, 1977).
The scientific Committee on Problems of the Environment (set up by the International Council of Scientific Unions) informally co-ordinates as part of its mid term project 2 a short term  interna tional programme for the review of the ecological effects of fire. South African participation in this programme is co-ordinated by the Working Group for Fire Ecology of the Committee for Terrestrial Ecosystems initiated by the National Programme for Environmental Sciences of the Council for Scientific and Industrial Research. The present paper forms part of the contribution by the South African Savanna Ecosystem Project on the ecological effects of fire (Anon, 1978). Results of an experimental burn within the South African Savanna Ecosystem Project are being prepared for publication as a synthesis report by M. V. Gandar (in press) where an overview is given of effects of fire on the abiotic, decomposer, primary producer and consumer components of the ecosystem. The present paper is limited to the more detailed effects of fire on the woody plant compo nent.
The South African Savanna Ecosystem Project is being conducted on a portion of the recentlyestablished Nylsvley Nature Reserve (3 120 ha in ex tent), 10 km south of Naboomspruit in the northern Transvaal. The basic ecological characteristics of the study area are described in Huntley and Morris (1978) while the project's overall objectives and research programme are outlined in Huntley (1978).
The study area lies on the edge of the Springbok flats on a slightly raised plateau at about 1 100 m above sea level. Most of the Waterberg System sand stone bedrock is covered by sandy soils belonging mainly to the Hutton and Clovelly forms (Harmse, 1977). Mean annual rainfall is about 630 mm and occurs mainly in summer. The mean annual air temperature is 18,6°C. The study site's past manage ment has included light summer grazing by cattle with small populations of impala and fluctuating populations of kudu present. The main vegetation type of the study area has been classified as Eragrostis pallens-Burkea africana Tree Savanna (Coetzee et al., 1976) with the most extensive varia tion of this being the Eragrostis pallens-Dombeya rotundifolia variation with dominant trees Burkea africana and Terminalia sericea and dominant shrubs Ochna pulchra and Grewia flavescens. Huntley (1977) has suggested that the broad-leaf savanna of the study site is related to the mesic and moist broadleaf savanna biome of Africa. In the year before the experimental burning, there had been no grazing by cattle in the area to be burned. Since the area had been unburned for some years and had a recent history of very low grazing pressure the graminoid fuel load was probably above average for the given type of vegetation although the nutrient-poor sandy soils may be expected to result in a generally lower heat intensity and slower fire than in several other vegetation types (Anon, 1960). The presence of trees also tends to result in a reduction in graminoid fuel load beneath them so that intensity and frequency of burning may be expected to be reduced by the trees (West, 1965). Given the tree leaf litter input each year (from June to August) and the relatively low decom position rates (especially for leaves of Ochna pulchra), thick layers, sometimes up to 10 cm deep, occurred below certain trees. On the whole, the ground fuel composition of the experimental site was mixed with graminoid, tree leaf litter and some wood pieces of varying dimensions.
The main objectives of the present study were to determine the short term effects of two intensities of spring burn on the individuals of each woody plant species population in a selected area of the Nylsvley study site, with particular attention being given to mortality/survival, degree of canopy reduction, degree of basal regeneration, changes in leaf biomass and to relationships between these aspects.

SEL EC TIV E L IT E R A TU R E REVIEW
Work on the effects of fire on individual woody plants in African savanna appears to have been main ly concentrated on mortality or survival of plants after fire. Less attention has been given to effects of fire on plant canopies and the stimulation of basal shoots. Very little work has apparently been done on establishing and quantifying the interrelationships between different plant dimensions and mass com ponents after fire (pyro-allometry). Such relations may be used to predict, for example, the degree of dependency of leaf mass re-distribution within the plant on the size of the plant following fire. The limited effort in this last mentioned respect possibly relates to the difficulties involved in determining the fire effects on the different organs of the woody plant individual and to the variable degree to which ground fires reach into the canopy stratum of woody vegetation as well as to the often very heterogeneous spatial distribution of woody plants in savannas. Published work on the particular fire effects and plant species considered in the present study has included various, sometimes conflicting, findings.

a) M ortality/survival o f plants after fire
Relatively low mortality of some woody savanna plants with fire has been indicated by, for example, the classification in Malawi of Burkea afrieana as a 'pyrophytic' species (Jackson, 1974). In the Kavango region of South West Africa, Geldenhuys (1977) tested for mortality in B. africana with fire in an analysis of co-variance and despite the high pro portion of dead trees (35,3% at Rundu), this was not attributable to treatment. Of the four most common woody plant species in a Nigerian savanna after annual (late winter) burns for five years, only B. africana had all individuals (from 4 to 13 m height) survive. B. africana increased relative to the other species with fire and it is suggested that it is through better adaptation to fires that certain B. africana savannas may have arisen (Hopkins, 1965). In con trast to the above findings, van Rensburg (1971) reported that some B. africana trees 'were damaged and killed' in Terminalia sericea Woodland on sand in eastern Botswana after an April burn (toward the end of the normal rainy season). However, mortality of Burkea africana trees has been reported to be not necessarily linked to the effects of the fire in several areas (Rutherford, in press). Tinley (1966) only refers to coppices of B. africana being 'very sensitive to fire' in the northern Okavango Swamps of Botswana.
Effects of long-term application of fire may differ from the expected short-term effects. Thus, for ex ample, in Terminalia sericea-Burkea africana Savanna at Matopos, Zimbabwe, Kennan (1971) reports fewer trees present in long-term fire treated plots than in fire protected plots. There was also a much smaller proportion of those trees below 0,9 m in height in the fire treatments than in the protected plots. In Burkea africana-Terminalia avicennoides -Detarium microcarpum Savanna Woodland in Nigeria, Afolayan (1978) found that annual 'late' (presumably late in the dry season) burning for four years decreased tree density particularly for trees less than 10 cm girth at breast height.
The effect of fire on seedlings has been observed in Matopos savanna, Zimbabwe, where burning at three year intervals resulted in many tree seedlings being killed (West, 1965).
Other woody plant species in the present study have been characterized on other sites in terms of their ability to withstand fire. In the Transvaal lowveld, Van Wyk (1971) reported that Terminalia sericea and Dichrostachys cinerea were to a 'certain extent fire resistant' and despite burning to ground level sprouted vigorously after the burn. In Zambia, Trapnell (1959) found that 'the fire-tolerance of Dom beya rotundifolia has been confirmed by repeated field observation', and that Strychnos pungens was semi-tolerant with Lannea discolor pro bably also so. Lawton (1978) includes Ochna pulchra, Strychnos pungens and Burkea africana in a group that can survive intense dry season fires in parts of north-eastern Zambia. In Wankie National Park, Zimbabwe, Rushworth (1978) found that Ter minalia sericea and Ochna pulchra were 'strikingly frost hardy' and points out the similar coppicing reaction of some other species due to both frost and fire. Geldenhuys (1977), in two regularly burnt areas of the Kavango region of South West Africa/ Namibia, refers to Ochna pulchra as a 'fire-sensitive' species. However, it appears that this finding might be limited to trees with measurable diameter at breast height.
Many other woody plant species in African savanna also appear to be relatively tolerant of fire. Thus in Acacia thornveld (A. karroo) at Matopos, complete killing of woody plants by fire of various fire regimes was extremely rare and nearly all affected plants regrew vigorously from their undamaged bases after the fire (Kennan, 1971). In eastern Cape Acacia kar roo vegetation, Trollope (1974) found in a spring head fire that mortality of A. karroo was 9,9% of which more than three-quarters of the plants were between 1 and 2Vi m tall. In the Molopo area of the northern Cape, Donaldson (1966) reported that even with dry grass artificially packed around the base of Acacia mellifera subsp. detinens plants, the grass burns (at various times of the year) only resulted in a mortality of about 15%. Only when large amounts of wood and twigs were burned at the bases did a 75% mortality obtain here due to much longer duration of heat. In other African savannas, Pterocarpus angolensis has been classed as 'fire-tolerant' (Zam bia: Trapnell, 1959) and in east Africa Balanites aegyptica Savanna no evidence was found of the death of mature trees being affected by a mainly annual fire regime (Harrington & Ross, 1974). Although Trapnell (1959 reported the dominant canopy species of Brachystegia, Julbernardia and Isoberlinia (in Zambia) to be 'fire tender' (but not 'fire intolerant'), West (1971) states that some of these are 'extremely fire tolerant' but states that even these will eventually be eliminated by sufficiently regular, intense, late dry season burns every year. At Matopos, Kennan (1971) found 'that burning had much the same effect on the sandveld trees as in the thornveld . ..' so that mortality in the sandveld woody plants was presumably also very low. The relatively high fire tolerance of woody plants of the present study and other savanna woody plant species is not limited to Africa. For example, in a Texan Pinus taeda-P. echinata forest with a head fire, it was found (Stransky & Halls, 1979) that of the 10 most important understorey woody plant species, three (for example, Cornus florida) had zero mor tality and all others except one had less than 32% mortality.
A possible effective adaptation to fire in savannas is the suffrutex or 'underground tree' form such as that of Dichapetalum cymosum that West (1971) has suggested as a possible evolutionary adaptation that resulted in evading fire. The reaction of trees to fire BURNS IN BURKEA AFRICANA-OCHNA PULCHRA SAVANNA and the possible fire adaptations have resulted in many savannas being regarded as seriously limiting the effectiveness of fire alone as a woody plant con trolling mechanism for management purposes, for example, in eastern Cape Acacia karroo Savanna (Du Toit, 1972a) and Brachystegia spiciform is-Julbernardia globiflora Savanna of the Zimbabwe highveld (Barnes, 1965). In the latter area, burning intervals longer than annually have been found to be ineffec tive in preventing coppice becoming increasingly vigorous (Barnes, 1965). However, Van Wyk (1971 stated that in Dichrostachys cinerea-Terminalia sericea Savanna in the Transvaal lowveld, burning as little as every three years resulted in plants seldom escaping the regular damage to reach maturity. It appears that fire as a tool for the management and control of undesirable woody plants differs in its effectiveness according to area and conditions of application.

b) Effects o f fire on plant canopies
Fire often greatly affects the canopy of woody savanna plants, particularly those of smaller plants.

After a hot November burn in Burkea africana
Savanna in South West Africa, canopy (but not plant) mortality of the plants less than 2 m tall was greater than 75% for each of the 4 most common woody species (Burkea africana, Terminalia sericea, Combretum psidioides and Ochna pulchra) with most (90%) for Ochna pulchra (Rutherford, 1975 plants of up to about 1,2 m were generally burnt back to ground level although some Terminalia sericea trees up to 3,7 m height had also been burned back. In eastern Cape Acacia karroo, 95% of canopy mor tality after fire was limited to plants under 2,5 m tall (Trollope, 1974). In Burkea africana Savanna (regularly burned at the end of the dry season) at Makambu, Kavango in South West Africa, Geldenhuys (1977) found that for shrubs and trees (with stems less than 5 cm DBH) canopy volume dropped by two-thirds relative to that of the control. Mean height of this plant group was 0,9 m compared to 2,2 m of those of the control, that is a decrease of about 60%. That fire generally reduces the canopy heights of lower woody vegetation in many other regions is supported by studies such as that in a Texan forest fire (Stransky & Halls, 1979) where for the ten most important understorey woody species height decreased by 41% from 4,4 m to 2,6 m. After a hot November burn in a Burkea africana Savanna in South West Africa, canopy mortality of Securidaca longipedunculata plants greater than 2 m tall was the lowest of six species, namely zero per cent (Rutherford, 1975).
In Acacia veld at Matopos, Kennan (1971) states that in the case of larger trees, burning invariably caused complete defoliation ('if they were in leaf when burning took place') but seldom did more than to kill branches up to a height of about 1,8 m.
However, in the eastern Cape, it was found that canopy mortality of surviving individuals of Acacia karroo was 79% (Trollope, 1974). Donaldson (1966) found with burning Acacia mellifera subsp. detinens in the northern Cape with fuel (grass, wood, dung or sawdust) at the base of the plants, that generally there were total 'top-growth kills'.
Although it is clear from the evidence that fire may be expected to reduce canopy leaf biomass, it appears that the effect on radial stem growth may be different since in the Kavango region of South West Africa it was found that there was no significant effect of annual fire treatments on stem basal area increment over a nine year period in Burkea africana and other species investigated (Geldenhuys, 1977) while also in Nigerian savanna it was found that controlled burn ing early in the dry season would permit an increase in established trees' basal area (Kemp, 1963).

c) Fire and basal shoots
In many studies it has been found that the number of stems are likely to increase with fire. For example, in a Texan forest fire mean stem number of the ten most important understorey woody plant species in creased from 1,5 to 2,1 with a maximum increase for one species (Ilex vomitoria) from 2,2 to 5,1 (Stransky & Halls, 1979). A common phenomenon under total fire protection in savanna, is for woody thickets to tend to develop. However, Harrington (1974) points out that in Uganda despite the densest appearance of Acacia hockii in an unburnt treatment (relative to that in several burning regimes) it had the lowest number of stems per bush (and the lowest number of bushes per hectare).
After a hot November fire in a Burkea africana savanna in South West Africa, it was found that of the three species Terminalia sericea, Burkea africana and Ochna pulchra, the first mentioned had the greatest percentage of plants with basal regeneration shoots present (Rutherford, 1975). In scrub sand veld savanna in Wankie National Park, Zimbabwe, Rushworth (1975) found that whereas the mean number of new coppice stems produced on Ter minalia sericea in an area unburned for at least eight years was zero, those burned in just under three months prior to measurement (an early October burn) was 20,33 (versus 0,05 in an area with approx imately the same burning history as above but without the October burn). Van Wyk (1971) has reported vigorous sprouting of Terminalia sericea after a burn in the Transveld lowveld. Donaldson (1966) has also commented that the multistemmed T. sericea of the Molopo area of the northern Cape have possibly resulted from periodic grassfires in the past.

Interrelationships after fire
After a hot November burn in Burkea africana Savanna in South West Africa, data showed that a higher percentage of plants with canopies killed had basal regeneration shoots present than those with canopies that survived for all species investigated (Rutherford, 1975). These included Burkea africana, Terminalia sericea and Ochna pulchra. In eastern Cape savanna, Trollope (1974) showed that after a spring head fire, of those Acacia karroo trees that survived and had formed basal regeneration shoots after the fire, 86% had canopies killed leaving only 14% with live canopies. James & Smith (1977) state that 'extensive suckering does not usually occur after low-intensity fires' while Farmer's (1962) work on Populus tremuloides demonstrated that suckering was related to the reduction of apical dominance by damage to the above ground parts.

a) The two burns
Three one hectare square blocks of Camp 2 of the Nylsvley study area were burned separately before the remainder of the camp on September 5, 1978. All work on the woody plant species was done in two of these hectare blocks which were about 1 km apart.
Both areas were ignited at one side of the hectare block with flame-throwers directed at the herbaceous layer which allowed the fires to rapidly attain their maximum intensities. Plot 1 was ignited at 19h01 and plot 2 at 18h00. The mean windspeed at 2,0 m above ground from 12h00 on September 5, 1978 to 07h00 September 6, 1978 was 1,8 ms-1 (Harrison, 1978).
Both fires were ignited as head fires (burning in the same direction as the wind) although in plot 1 there was some degree of backburning (burning in the opposite direction to the wind) of some islands left unburned after the main flame front had past. At 19h00 (corresponding to time of burn in plot 1) screen climatic data gave: air temperature 17,2°C; relative humidity 33%; vapour pressure 6,4 mb and saturation vapour pressure deficit 13,1 mb. At 18h00 (corresponding to time of burn in plot 2) screen climatic data gave: air temperature 19,0°C; relative humidity 28%; vapour pressure 8,5 mb and satura tion vapour pressure deficit 13,3 mb (Harrison, 1978). Mean moisture content of plants (mainly grasses) of the herbaceous layer one week prior to the fires was 4,2% (Grunow and Grossman, 1978). Estimated ground fuel loads (see Section 3c) showed more frequent higher levels in plot 2 than in plot 1.
Fuel loads were sometimes very localized, for example, typical individuals of the shrub species Grewia flavescens (type 1, less than 2,5 m height, -Ruther ford, 1979) had 4 800 gm 2 of thin finely divided standing dead wood on the area they covered. Other areas were sometimes virtually bare, that is, less than 10 g of dry herbaceous material for individual square metres. Data on mean herbaceous layer dry mass per unit ground area are not available for the two plots but from many other clipping studies (Huntley & Morris, 1978) on the study site, the mean mass of the standing dead grass lay between 50 and 125 gm -2. Ladder fuel in larger trees was rare, that is, there was usually no continuous fuel path from the herbaceous layer to the tree canopy.
Differences in the behaviour of the burns are given in Table 1. The burn in plot 2 was more than five times faster, fire temperatures were higher, flame heights were greater and burning on an area basis more complete than in plot 1. Although on the basis of these data the fire in plot 2 might be regarded as more intense than in plot 1, because of difficulties in averaging the great differences in heat intensity at different levels above ground, the fire in plot 1 is referred to as the slower burn and that in plot 2 as the faster burn. This designation may be appropriate since speed of fire (whether 'self' generated or wind induced) appears to be important to fire behaviour. That head fires are faster than back fires is com monly observable. Trollope (1978) found greater flame lengths in head fires and a positive correlation between rate of spread and flame height and maxi mum temperatures at grass canopy height in head fires. In laboratory simulated experiments, Gill (1974) indicated an increasing flame height with in creasing wind speed (up to 0,48 m s '1). Wind affects fire behaviour by increasing the flow of oxygen to the fire and (in a head fire) wind bends the flames over the unburned fuel and increases the flow of hot gases from the combustion zone; both processes contribut ing to the pre-heating of the unburned fuels (Deem ing et al., 1972) which is particularly important for realizing the potential of water-conducting woody material as fuel.

b) Experimental layout
The two plots selected, contained woody and her baceous elements that were floristically and structur ally typical of the Nylsvley study area vegetation.
Since the method of recording certain aspects of the plant was different for tree individuals and multi stemmed shrubs (see next section) data were pro cessed separately for these two groups but so as to prevent the same species from occurring in both groups, the groups were defined as the tree species group, that is, individuals that were trees or normally have the potential to grow into tree-sized individuals; and the multistemmed shrub species group whose members seldom form tree-sized individuals on the Nylsvley site. In the slower burn area, tree species were ( 1-6). At the time of the fire many of the woody plants, particularly those of Ochna pulchra, were starting to unfurl new leaves. The most advanced leaves on some individuals of O. pulchra were about 2 weeks old. All the more common species (except Strychnos pungens) had dropped their old leaves before the fire so that almost all of the woody plant leaves formed in the previous growing season were already added to the fuel load on the ground prior to the fire. Also present at the time of the fire were a few woody plant seedlings, mainly those of Burkea africana. The dominant grass in both burn plots was Eragrostis pallens.
In each plot a subplot of 30 x 50 m was placed centrally and demarcated. In the subplots all sizes of standing individuals of woody plant species were tagged with numbered aluminium plates (either on 20 cm high stakes at the base of smaller individuals ( Fig.   1) or on to the trunk of large individuals) in one half of the area. All individuals equal to or larger than 2 m height were tagged in the other half to increase the area. The individuals of this new control set were geographically close to the burn treatments, probably had a very similar treatment history in the past and were also measured in exactly the same way and time as the burned individuals. It transpired that the effects of fire were so profound that differences between treatment and control were usually so great that statistical analyses of most differences, especially those concerning basal regeneration, were super fluous. The control in many respects only served to confirm the obvious. References to unburned control data are thus kept to a minimum with more attention being given to the differences between the two burns.

c) Measurements
Recordings of plants were made, (i) shortly before the burns, (ii) just after the burns, (iii) at monthly intervals after the burns and (iv) after completion of one season's growth but before commencement of leaf fall. Most detailed measurements were made in periods (i) and (iv).
One to two weeks before the burns, each woody sample size of the larger individuals. All standing dead individuals equal to or larger than 2 m height in the remainder of each one hectare plot were also tag ged and numbered. In the slower burn plot there were altogether 607 tagged individuals in the subplot and two additional dead individuals in the rest of the hec tare plot. In the faster burn plot 425 individuals were tagged in the subplot and an additional 12 dead individuals in the remainder of the plot. Altogether 1 046 individuals were thus tagged for recording the effects of fire. The day following the burns it was found that in the slower burn area 37% and in the faster burn area 3% of tagged individuals had escaped the fire altogether on large unburned islands with each of these individuals having no vegetation burned on their canopy ground projection area.
Since such a relatively high proportion of individuals was altogether untouched by fire, it was decided to use these unburned individuals (from both plots) as a control rather than the originally envisaged tagged individuals already being monitored as part of a separate programme in Camp 3 of the Nylsvley study plant individual was allocated a numbered alumi nium tag, and the tag position was also recorded for relocating the plant later. The species and live or dead state of the plant was recorded. Also measured were height of plant above ground level, number of live basal shoots (these constituted the whole in dividual in small non-canopied individuals) and number of dead basal shoots. Estimates were made of the proportion of canopy volume that was dead using a five point scale (0-9,9; 10-34,9; 35-64,9; 65-89,9; 90-100%) based on zones of dead twigs. Also estimated was the relative amount and composi tion of ground layer fuel load under the individual on a three point scale and with 3 type classes, namely, woody plant leaf litter, dead standing grass and pieces of wood material (e.g. Fig. 2). This ground layer fuel classification only applied well to in dividuals up to about 2Vi m height. The area under large individuals was relatively large with often great differences in ground fuel load under the same tree.
Photographs of various parts of the vegetation were taken from reference positions. One and two days after the burns, all labels were checked for legibility and position and a few replaced where necessary. Whether the plant was burned or totally untouched by fire was recorded as was an estimate of degree of burn namely: burned, for a broad class including plants that fire had touched at least at some point of where ground fire occured under at least part of the canopy; severely burned for plants where most leaves or other parts of the plant were at least severely scorched; and completely burn ed where the plant's aboveground parts were com pletely burned away or at most only a very short stub (1 or 2 cm high) remained. Photographs were taken from the reference positions.
At monthly intervals for six months following the burns, the photographs of selected components of vegetation were taken from the reference positions. Checks of animal (almost entirely insect) browsing of leaves were made to confirm that this remained at a low level (less than 5%) so that leaf changes could be attributed to fire effects.
Six months after the burns (March), the following were recorded for all tagged individuals: species (a recording check); live or dead state of plant; canopy live or dead; height of top parts of canopies; diameter of stem at 20 cm above ground level (for all stems > 1,0 cm diameter); number of new basal shoots; number of old basal shoots still present; number of basal shoots killed by the fire and still pre sent; number of old dead basal shoots that survived the fire; estimation of proportion of original canopy volume dead (this was difficult to apply in species such as G rem a flavescens, where the original outline of canopies was usually altogether lost through fire - Fig. 3); biomass of basal twig regeneration (clipped at ground level); biomass of basal leaf regeneration; biomass of all leaves in canopy (for individuals <5 m height). All biomass data were obtained oven dry at 85°C and total sampling was used, that is, no sub sampling was employed.
The diameters of stems were taken for use in already established formulae that predict canopy leaf mass (Rutherford, 1979). These regression formulae were first tested by destructive sampling for each of the more common species in the unburnt control populations since the formulae were derived for populations several kilometres distant and three years previous to the time of fire. If, as was found in the Ochna pulchra population, there was very good agreement between predicted canopy leaf mass and actual destructively sampled canopy leaf mass of the unburnt population, the regression formula was applied to the burnt population of the appropriate species (using original plant height if height was reduced by fire) to obtain expected leaf mass had the population not been burned. Therefore, on condition that the regression formulae still proved suitable, this procedure provided a more sensitive measure of the degree of canopy leaf biomass reduction by fire than that provided by using treatments and controls which sometimes had greatly differing distributions of heights within each height class. Canopy leaf biomass reduction per height class was thus calculated in these cases by taking the predicted unburned canopy leaf biomass of the burned individuals and subtracting the actual canopy leaf biomass obtained by direct harvest from the burned individuals.
It should be noted that for practical reasons a small proportion of the individuals included in the sample for numerical counts of stems, mortality and so on were not harvested for biomass. Therefore those numerical non-biomass data concerning differ ences before and after the fires are not necessarily precisely interrelateable with the biomass data set since there are possible differences in plant size distribution in the whole sample set and in the biomass data subset.
Only some data are tabulated since tabulation of all data for all species, all plant size classes and all the types of possible fire effects, results in many large tables with very many empty cells. This difficulty is inherent in studies such as the present, and arises not only from variation in the natural vegetation com position and structure but also from the inapplica bility of some measures of the effects of fire to cer tain plant growth forms. The very unbalanced total data set resulted in no full analysis of variance being attempted. Instead, for categories where sufficient data existed, all data from the category were used in standard statistical tests of significance between means, often using the non-parametric Wilcoxon test where appropriate. Most data are indicated graphic ally where use was made of data grouped into classes in the presentation of relations.

a) M ortality/survival o f plants after the burns
Mortality was defined as the proportion of the number of plant individuals that were alive before the fire but dead (above ground) six months after the fire ('Root kill9 of Niering et al., 1970). It is possible that mortality includes individuals that survived or regenerated after fire but died from other causes within the six month period following the fire.
That mortality of unburned individuals would be close to zero was confirmed by the data for the con trol that showed a mortality of 1% of the tree species group's individuals and 0% of the multistemmed shrub species group's individuals. In the slower burn area mortality of the tree species individuals was 5% which was significantly greater (P=0,005) than that of the unburned control. For the shrub species in dividuals mortality was 0%. In the faster burn area mortality of the tree species individuals was 2% which was not significantly different from that of the control. The mortality of the multistemmed shrub species individuals was significantly greater than that of the unburned control. This mortality is, however, subject to further interpretation since one species was involved, namely, Dichapetalum cymosum, where a mortality of 64% for its individuals up to 25 cm tall was indicated. Re-examination in August of most of these individuals recorded as dead in March showed that although most had no aboveground parts visible there were live belowground parts that were in the process of initiating new shoot growth. In these cases, therefore, fire possibly only delayed new growth by one growth season. However, given the ex tensive underground branching and interconnections between 'individuals' of this geoxylic suffruticose species, new regenerative growth possibly occured after the fire but not in the immediate vicinity of the labelled 'individual '. Rushworth (1978) also found that woody suffrutices such as D. cymosum did, con trary to other woody plants, not produce additional stems per plant unit after fire in Wankie National Park, Zimbabwe. That the high mortality of 'in dividuals' of D. cymosum is merely an apparent mor tality, means that, in fact, the mortality values for the multistemmed shrub species individuals was very close to zero.
Despite individuals of Grewia flavescens having the highest fuel loads within the plant and that many plants were completely consumed by the fire (Fig. 3), there was no mortality within these populations in either burn. Mortality of Ochna pulchra plants in the slow burn (5%) was significantly greater (P=0,005) than that in the faster burn (1%). However, when grouped into plant height classes (Table 2) there was no significant difference in mortality for those plants taller than 0,25 m but only for the group below 0,25 m height. In Burkea africana, mortality was limited to small plants under 10 cm tall, that is, 54% for those of both burns. Most of these plants killed were seedlings.

b) Effects o f fire on plant canopies
Plants were defined as canopied where leaves were carried on stems more than one year old. Non canopied individuals were thus made up of only (young) basal shoots whereas canopied individuals had an older main stem bearing the canopy leaves with or without basal shoots present. Because of some difficulties in ageing basal shoots before the burns a few non-canopied individuals possibly had basal shoots slightly older than one year but such basal shoots were morphologically similar to oneyear old shoots. The effects of the burns on canopies are of course limited to the canopied plants. The effects of fire on canopies can be expressed in various ways, depending upon the canopy attribute con sidered. Canopy mortality occurs where the whole canopy dies, but where the plant still survives in the form of basal regeneration shoots ('Stem kill9 of Niering et al., 1970) (Fig. 4). The occurrence of abnormal leaf growth refers to the amount of leaves  that have grown in a convoluted manner and are usually produced not from terminal twigs but from thicker, older wood parts (Fig. 5). Also in terms of shoot extension in Ugandan savanna, Harrington (1974) refers to the tops of taller burned bushes behaving similarly to bushes in an unburnt controlled treatment. Changes in the height and canopy volume of plants refers to measurements as described in the section on methods. Another attribute used is the change in the total amount of leaf biomass in the canopy.

i) Canopy M ortality
Only 0,7% of the unburned control tree species plants' canopies died. In the faster burn, 43,2% canopies were killed of which 92,1% were under 2 m tall with none taller than 5 m. In the faster burn, these plants up to 1,5 m tall had 9,2 canopies killed to each one that survived. For those over 1,5 m tall there were only 0,2 canopies killed for each surviving canopy. In the slower burn, 23,5% canopies were killed of which 90,1% were under 2,5 m tall and also with none taller than 5 m. In the slower burn, those plants up to 1 m tall had 1,6 canopies killed for each one that survived. For plants between 1 and 1,5 m tall the ratio was about 1 : 1 while for those more than 1,5 m tall there were also only 0,2 canopies killed for each surviving canopy. Relative to the control tree species plants, therefore, canopy mortality in both burns was highly significant particularly for in dividuals less than 2 or 2,5 m tall.
In the Ochna pulchra population, canopy mortal ity was 32,0% in the slower burn and 44,1% in the faster burn. Only in one species, namely Vitex rehmannii was it clear that tall individuals had a relatively high canopy mortality, that is, 64,3% for individuals between 2,5 and 5,0 m tall. Despite the sensitivity of canopies of V. rehmannii to fire, no plants of this species died after fire. The hollow main stem of one tree individual, Securidaca longipedun culata, was observed to burn vigorously for several hours after the main flame front had passed, but at the end of the growth season the canopy showed no obvious effects of the fire (Fig. 6).

ii) Occurrence o f abnormal canopy leaf growth
Apart from observing that the incidence of abnor mal canopy leaf growth was generally higher in the faster burn that in the slower burn, the mere presence of abnormal leaf growth was found to be less informative than the actual value of abnormal leaf biomass compared to that of normal leaf biomass. Since a clear qualitative recognition of this abnor mality is required for expressing such ratios, con sideration was limited to the Ochna pulchra popula tion where such distinction was most reliable (Fig. 5). The ratio of normal to abnormal canopy leaf bio mass was found to increase exponentially with plant height (Fig. 7) and is discussed further in Section 4e.

iii) Changes in canopy height and canopy volume
For the unburned tree species control plants, mean change in height was + 4% with the greatest relative reduction in any height class being -2% . Although in unburned plants height appeared to be unimpor tant in affecting tree height changes, in both burns canopied plants of the lowest height classes had the greatest relative reduction in height (about 100%) with tallest plants having a zero reduction in height. The relative decrease in plant height after the faster burn was greater for each height class than that for the slower burn, for example in Ochna pulchra (

-R e la tio n sh ip b etw een m ea n plan t h eig h t an d th e r a tio o f th e m a ss o f n o rm a lly to a b n o rm a lly fo r m e d c a n o p y le a v e s fo r
Ochna pulchra in b oth bu rn s.  8c). In the multistemmed shrub species group, changes in height were more variable and less strong ly correlated with height.

iv) Changes in canopy leaf biomass
Effects o f the burns on the canopy leaf biomass could be determined as described earlier on the basis o f application of allometric biomass prediction fo r mulae once appropriate tests o f the validity of application had been made. It was found that the allometric formulae, applied to unburned plants (stem diam eter >1 cm at 20 cm height) of Ochna pulchra, overestimated the actual harvested canopy leaf biomass value by only 0,1% . In Terminalia sericea it was found that canopy leaf mass was underestim ated but remained within 20% of the actual harvested am ount. In Burkea africana it was found that canopy leaf mass was overestim ated by more than 20% (in the relatively small tested sample population of the control) and the biomass predic tion form ulae were thus not applied to the burned B. africana populations. Since the allometric form ulae cannot be applied to the smallest individuals, canopy leaf biomass data from unburned control plants were utilized for the lowest size classes. In

-R e la tio n sh ip s b e tw e e n m e a n p lan t h e ig h t and : a , relative ch an ge in plan t h eig h t; b , p r o p o r tio n o f in d iv id u a ls that d ecrease in h eigh t (grap h in v e r te d fo r c o m p a r is o n ); c , relative ch a n g e in c a n o p y v o lu m e fo r Ochna pulchra in (1) th e slow er burn and (2) th e fa ster b u rn .
canopy leaf biomass by 92% , and this also became less marked with taller plants but with a 50% decrease for plants between 2,5 and 5 m tall.
In Terminalia sericea of both burns, trees > 2,7 m tall had a canopy leaf biomass o f 79% o f that predicted, but for the smaller trees only 6% of the predicted value. This contrasts with the 83% of that predicted in the unburned control plants. Therefore, allowing for the shift in the prediction equation (given by the control), a reduction of at least 90% in canopy leaf mass after fire occured for the T. sericea plants less than or equal to 2,6 m tall. In Burkea africana of both burns, stem basal area was used as a measure of plant size to relate to the harvested canopy leaf biomass. For trees up to a cross-sectional stem basal area (at 20 cm height) approaching 50 cm2 canopy leaf biomass was less in burned plants than unburned plants, but for plants with a basal area of about 50 cm 2 or more this difference no longer held ( Fig. 9). For taller plants (approxim ately greater than 2,5 m) Ochna pulchra had a greater canopy leaf biom ass reduction than Terminalia sericea and Burkea africana. For smaller plants the differences were less pronounced between these three main tree species.

c) Effect o f fire on basal shoot numbers
Most basal shoots that were live before the burns were killed by the fire. Although 14% o f the basal shoots of control tree species plants died without fire, 91% were killed in the slower burn and 100% in the faster burn. The killing of live basal shoots by fire was independent of plant height and species. In the m ultistemmed shrub species plants killing of live basal shoots was virtually 100% in both burns. The form ation of new post-fire basal shoots was a common phenom enon in burned individuals (Figs. 10 & 11) and these were virtually completely grown by December, that is, about three m onths after the burn (Fig. 3). Whereas a mean of 0,09 new basal shoots were formed per unburned control plant, the corres ponding values were 2,42 for the slower burn and 4,55 for the faster burn. In the unburned control plants, however, new basal shoots were not formed from individuals more than 1 m tall, whereas in burned individuals these were formed irrespective of plant height, but with a tendency for fewer basal shoots to be produced per tall individual. In the m ultistemmed shrub species plants approxim ately 5% of the unburned individuals produced new basal shoots, whereas corresponding values were 93% in the slower burn area and 100% in the faster burn area.
The possible relationship between the num ber of live basal shoots before the fire and those formed after the fire was investigated to determine to what  In Ochna pulchra, although the num ber of new basal shoots tended to decrease with increasing plant height class, it was found that replacement o f basal shoots increased with increasing plant height. No such trend was discernible in Burkea africana. The basal shoot replacement relation is given in Fig. 12 for the mean of both burns. The degree o f ability of these plants to produce new basal shoots after fire appears to depend on not only the size of the plant, but also on the number of basal shoots prior to the fire.

F ig . 12 .-R e la tio n sh ip b etw een m ean p la n t h eig h t a nd th e ratio o f nu m b er o f p ost burn basal r eg en era tio n s h o o ts to n u m b er o f p re-b u rn live ba sa l s h o o ts (b a sa l s h o o t rep la cem en t r atio)
in Ochna pulchra.

d) Effect o f fire on production o f new basal regene ration shoot biomass
Basal shoot production (over the 6 m onth period following the burns) averaged per individual from both burns was 9,59 g for Ochna pulchra, 14,52 g for Burkea africana and 1,09 g for Terminalia sericea (Table 3). In each of these populations, basal regeneration shoot mass of unburned plants was usually considerably less than 10% o f basal mass in burned plants. This was also valid for the rarer species, but in some there were very small sized samples for the control, for example, there was no Dichapetalum cymosum in the control area. From Table 3 it can be seen that species with low values (< 1 0 g) for basal regeneration shoot mass per in dividual were Dichapetalum cymosum, Terminalia sericea, Strychnos pungens and Ochna pulchra. The species with highest values (> 5 0 g) were Grewia flavescens, Dombeya rotundifolia and Vitex rehman nii with Euclea natalensis having the highest value of all (285,62 g). Those species with greatest basal regeneration shoot mass tended to be non-suffrutex shrub forms whereas tree growth form s usually had relatively low basal regeneration shoot mass.
In terms of mean biomass o f basal shoots per in dividual shoot (where these were clearly distinguish able), Lannea discolor and Burkea africana had highest values while lowest values were found in Dichapetalum cymosum and Terminalia sericea.
Species may be divided into three groups according to their ratio o f basal leaf mass to basal twig mass. Those with a ratio o f less than 1, that is, there was more twig mass than leaf mass, included Securidaca longipedunculata and Terminalia sericea. Those with up to twice the am ount o f leaf mass to twig mass included the largest num ber of species. Those with more than twice as much leaf mass as twig mass included the two most im portant geoxylic suffrutex species Dichapetalum cymosum and Fadogia mon ticola.
Differences in mean basal shoot regeneration mass between different species were tested (for all species with more than 3 individuals) often using the nonparam etric Wilcoxon test where appropriate (Table  4). At a level of significance o f P = 0,05 more than half the com binations were significantly different.
Since the reaction o f Grewia flavescens was so differ ent in two burns, these were treated separately.
In Ochna pulchra in the slower burn, basal shoot regeneration per individual (4,86 g) was significantly lower than that (12,28 g) in the faster burn. In Grewia flavescens the corresponding values were 65,00 g and 160,02 g. This greater production of basal regeneration shoot mass in the faster burn area was also confirm ed in Burkea africana for each height class o f individuals. In Grewia flavescens, the live basal shoot mass per unburned individual was significantly greater (P = 0,005) than the post fire regeneration basal shoot mass in the slower burn area, but there was no significant difference to that produced in the faster burn.
The possible relationship between degree of burn recorded in both burns and basal regeneration mass of the woody plants was shown by 63,73 g for those recorded as burned, 115,12 g for those recorded as severely burned and 107,68 g for those completely burned.
The species with the greatest range in basal regeneration shoot mass was Euclea natalensis (1,74 to 1143,37 g). When the population was divided into three equal size classes that is up to 0,5 m tall, 0,5 -1 ,0 m and 1,0-1,5 m tall, the respective mean masses were 40,44 g, 248,38 g and 768,72 g and each value was significantly different from the other. There was thus a clearly greater basal shoot regenera tion mass with increased plant size for this shrub species.
It was noted during observations that the individual leaf size, especially o f Ochna pulchra, was markedly larger in basal regeneration shoots than in the canopy.

e) Relationships between plant height and woody plant biomass components after fire
Earlier reference has been made to effects of the burns on plant height and canopy volume in Ochna pulchra and to effects on canopy leaf mass in Burkea T A B L E 3 .-Basal regeneration shoot mass data for the m ore com m on w ood y plant species When basal leaf mass is com pared to canopy leaf mass in Ochna pulchra it is found that the ratio of canopy leaf mass to basal leaf mass increases nonlinearly with increasing tree height (Fig. 13). Plants under 1 m tall have m ore than half their leaf mass in the form o f basal leaves; plants around 1 m tall have basal and canopy leaf mass about equal, while in plants taller than 1 m canopy leaf mass becomes rapidly much greater than basal leaf mass.
In Ochna pulchra it was found for both burns together that the ratio of normally formed to ab n o r mally formed canopy leaf mass increased exponen tially with plant height (Fig. 7). For plants about 1 m tall abnormally formed leaf mass more or less equalled norm ally formed leaves. For plants above 1 m tall, norm ally produced leaf mass quickly became many times that o f the abnorm ally produced leaf mass. In the faster burn plants, abnormally formed leaves contributed relatively more than in the slower burn.
In the faster burn, extreme values for the ratios were also obtained, that is zero for the lowest height class and oo for the highest height class.
Differences in leaf mass between burned and un burned Ochna pulchra for both burns for each height class are indicated in Fig. 14. It can be seen that an increasing proportion of canopy leaf mass was lost after fire with decreasing plant height to a point where plants were too small to be canopied. The rela tionship approxim ates an exponential decay curve where the percentage change in canopy leaf mass decreases exponentially with lower plant height. However, in contrast to this m onotonic relationship, an increasing proportion of total leaf mass is lost with decreasing height only until a height o f 1-1,5 m is reached, thereupon a maximum reduction in total leaf mass having been attained, less leaf mass is lost until for smallest plants the mean total leaf mass becomes a net increase. It is also clear that it is those plants roughly 0,75-2,5 m tall, that on average have more than 50% of total leaf mass reduced by fire whereas for the smallest and largest individuals this was not so. It was found tht in Ochna pulchra, there was an increase in basal regeneration shoot mass with height until a maximum was reached for heights about 1 to 1,5 m after which there was a decline in basal regeneration shoot mass with plant height (Fig. 15). A relationship between plant height and the ratio of leaf mass to twig mass o f the basal regeneration shoots was found for both Ochna pulchra and Burkea africana (Fig. 16). This inverse relationship M e a n p la n t h e ig h t

( g r o u p e d d a t a ) ( c m )
assumed the form o f a logarithmically decreasing ratio with increasing tree height. There was very little distinction in the relationship between the two burns in Ochna pulchra. The ratio in Burkea africana was generally much higher than that for Ochna pulchra for each height class. This inverse relationship is very different to that in canopies of unburned Ochna pulchra where there is no such inverse relationship with plant height. a) Com bining the above inverse relationship equation for Ochna pulchra (Fig. 16) with the parabolic dependence o f basal shoot mass on plant height, the result is given in Fig. 15. This may be contrasted with corresponding d ata in norm al canopies (Fig. 17),

f) Other pyro-allometric biomass relations
In previous sections, much evidence has been given to show that the more damage the canopy of, for ex- am ple, Ochna pulchra, is subjected to, the greater the am ount o f basal regeneration. Now using the relative reduction in leaf biomass o f canopies as a measure of canopy damage this is related to basal shoot regen eration mass to provide a more precise relationship.
In deriving these relationships, individuals that were completely killed by the fire were om itted as were the very few' individuals that actually increased slightly in canopy leaf mass after fire. Grouping all plants (of all heights) o f O. pulchra into five equal classes of relative reduction in canopy leaf mass, generally increased basal shoot mass with classes of increasing reduction in canopy leaf mass (Fig. 18) was apparent in both burns. Expressed as a relation (Fig. 19) it was found that there was an exponential increase in basal shoot production with an increasing proportion of canopy leaf mass lost through fire (until a few in dividuals pass beyond a certain threshold and die). Particularly for the uppermost canopy leaf mass reduction class, the plants of the faster burn area produced a greater mass o f basal regeneration shoots than those of the slower burn. Although basal shoot regeneration mass was found to increase with increased damage to canopies, the concept should possibly not be extended to very small ( < 0,25m height) uncanopied plants o f O. pulchra. Using the recorded degree of burn as a measure of damage to these plants, it was found that plants recorded as burned had a mean shoot mass o f 10,35 g, whereas those recorded as severely or completely burned had a mass of only 5,37 g, which is signifi cantly (P = 0,005) lower. Another feature that emerged for O. pulchra of both burns was not only an inverse but an approx imately rectangular hyperbolic relation between number of basal regeneration shoots and the mean mass per basal shoot. In the slow'er burn area the data set was more tightly grouped ( Fig. 20a) in that there was no occurrence o f individuals with both more than 6 basal shoots and a mean shoot mass of m ore than 2 g. In the faster burn area (Fig. 20b), although the distribution was still hyperbolic, the data were more variable. In both these distributions, plant height appeared not to be im portant.

g) Effects o f fire on standing dead individuals
O f the 45 dead standing woody plants labelled before the fire (including controls), most belonged to Ochna pulchra, Burkea africana and Terminalia sericea. No significant differences in the fire effects between the two burns were found for these dead plants.
The effects o f fire on the height o f dead standing individuals was found to tend strongly to one o f two extrem es, namely no reduction in height (69%) or m axim um reduction to ground level (24%). All those individuals felled by the fire were under 1,5 m tall, that is, no individuals over 1,5 m tall were fell ed by the fire despite the observed active burning at the base o f two o f these a day after the fire. Eightysix percent o f those felled by fire were totally con sum ed. A lthough 24% o f burned individuals were felled, this was not much higher than the 19% o f in dividuals th at fell over in the absence o f fire in the six m onth period. There was no apparent selectivity of felling of individuals by fire or other agents accor ding to plant species. In the main live sample in dividuals, basal shoots that were dead before the burn were alm ost always totally consum ed by fire. W oody plant leaf litter on the ground, although in flam m able, som etim es did not ignite fully, for exam ple, even in thick layers in Ochna pulchra patches, burning was often lim ited to very superficial surface layers o f leaves (Fig. 21).

D IS C U S S IO N A N D C O N C L U S IO N S
In keeping with com m on veld managem ent prac tice, the experim ental burns were carried out in early spring when the various deciduous plant populations were at different stages o f bud break development.
A lthough stage o f bud break development at which the plant is burned may be considered possibly critical to the subsequent fire effects m easured, data o f H opkins (1963) for work on selected tree species in south-western N igerian savanna, appear to indicate that burning in the period o f a few weeks before and after norm al bud break, does not m arkedly change the period needed for recovery to start.

a) M ortality/su rvival o f plants after the burns
That m ortality was significant only in the slower burn and was also largely limited to very low tree species plants, indicates that the effects o f a slower head fire may differ to those of a faster head fire in a way similar to the different effects between a back burn and a head burn at ground level in grassland (Trollope, 1978). The apparently anom alous lower m ortality with greater damage to the canopy in the faster burn, relates to the com pensatory basal regeneration effect described. Therefore, even in a faster fire, with a probably more intense overall heat than in the slower burn, the m ortality effects may relate to more intense heat nearer the ground and not to heat loads at canopy level. Even where fire was reported with wind twice the speed o f that in the p re sent study, m ortality remained low (Stransky & Halls, 1979).
That most species in the present study had m ortal ity after fire varying from 0 to 5% shows a fire tolerance that appears com m on in many other woody plants in Africa. The limitation of m ortality in Burkea africana to very low plants, mainly seedlings, agrees well with Jackson's (1974) classification o f it as a pyrophytic species. The quoted fire 'tolerance' or 'resistance' (Trapnell, 1959;Van Wyk, 1971;Law ton, 1978) for Dichrostachys cinerea, Dom beya rotundifolia, Lannea discolor, Strychnos pungens and Terminalia sericea is confirmed by the less than 5% m ortality for each in the present study. Ochna pulchra was also shown in this study to be generally fire tolerant in agreement with indications o f Lawton (1978) and Rushworth (1978) but not with a 'firesensitive' description (Geldenhuys, 1977).
The very low m ortality rate after fire of m ulti stem m ed shrubs (om itting Dichapetalum cym osum ), which includes the 0% m ortality of Grewia fla ves cens despite its relatively high fuel loads, is paralleled in eastern Cape Acacia karroo Savanna where only the m ultistem m ed plants o f Rhus lucida recovered fully after a fire (Trollope, 1974).

b) Effects o f fire on plant canopies
D ata from the present study clearly show the con siderable effect of fire on woody plant canopies, p a r Whereas a two-thirds decrease in canopy volume corresponded to trees and shrubs under 5 cm DBH in Burkea africana Savanna in a fire treatm ent at M akam bu, South West Africa (Geldenhuys, 1977), this decrease corresponded to 1 m tall plants of Ochna pulchra in the present study. The accom pany ing 60% reduction in plant height at M akam bu was, however, not as great as the percentage reduction for the 1 m tall Ochna pulchra.
Although total plant m ortalities may be low in both Acacia and Burkea-Ochna Savannas, it appears that in terms of canopy damage and canopy mortality the Nylsviey study species such as Ochna pulchra and even Vitex rehmannii may be less susceptible to fire than some southern African /lcac/«-dom inated vege tation.
The observed high resistance of the canopy o f a large individual of Securidaca longipedunculata to prolonged burning in the present study is in keeping with the results from the hot November burn in Burkea africana Savanna in South West Africa.

c) Fire and basal shoots
The present study amply confirm ed an expected in crease in number of basal stems per plant individual after fire. However, in contrast to some other cited findings, regeneration in terms o f numbers and mass o f shoots in Terminalia sericea after fire was far less marked and the only tree species that reacted almost as vigorously to fire as the Terminalia sericea in the W ankie study (Rushworth, 1975) was Dombeya rotundifolia. The present study's relatively great basal shoot mass in Lannea discolor agrees with the report for the Transvaal lowveld where after an October burn, Lannea discolor was one o f the two species that 'sprouted well' on certain plots (Anon, 1960).
That more than half the species com binations referred to (Table 4) were significantly different with respect to mean basal shoot regeneration mass after fire appears to indicate a wide and fairly even range of values with sometimes limited variance. In the savanna plant community type studied therefore, there was no m ajor clustering o f woody plant species in terms o f their basal regeneration response to fire.
Although multistemmed shrubs were (in the slower burn) somewhat more susceptible to killing o f live basal shoots than were tree species, there was a clear tendency for most multistemmed shrub species, p ar ticularly non-suffrutices, to have higher mean basal regeneration shoot mass per plant than for most tree species.
The present study data point to an almost paradox ical situation regarding fire induced basal shoot regeneration o f those Ochna pulchra plants with 100% canopy leaf mass reduction through fire. Here a plant tends either to produce the maximum mass of basal shoots or none at all (if it dies).
The ratio of number of new basal shoots to the num ber o f old basal shoots killed by the fire, namely the basal shoot replacement ratio, is possibly a useful attribute of vegetation that is subject to recurring fire, that is where the basal shoots that are produced after one fire are affected in a subsequent fire. It is clear that successive fires may not result in a merely additive process of increasing basal shoot num bers. If the basal shoot replacement ratio is assumed to be constant with successive fires, it may be postulated that basal shoots o f Ochna pulchra will increase more rapidly than those o f Burkea africana on the basis of the existing data. A finding analogous to the present study result is the determ ination of a positive correla tion indicating that species, for example Burkea africana, are dependent upon the num ber o f parent trees for regeneration in regularly burned savanna vegetation at M akam bu, South West Africa (Gelden huys, 1977).

d) Biomass and other relations after fire
That there is a positive relation between the amount of reduction in canopy leaf mass and the am ount o f basal regeneration is an underlying reason for the existence (in Fig. 14) o f a point corresponding to an interm ediate height at which plants exhibit a maximum change in total (basal and canopy) leaf mass and below which smaller plants have a reduced loss in total leaf biomass. Although the data only support a positive change in total leaf mass after fire for smallest plants, it is likely that very large trees exhibit no significant change in total leaf biomass after fire (there is no basis for upw ard extrapolation o f curve b in Fig. 14).
One of the main findings o f the present study may be described as the difference between basal shoot mass after fire and unburnt canopy shoot mass as they depend on plant height (Figs. 15 & 17). Apart from the different forms o f the relation with height and the relative proportion o f the biomass com po nents, there are also the differences in this proportion with plant height (Fig. 16). Du T o it's (1972b) relation o f stem diameter (mm) and mean coppice regrowth (1 x 1 0 g ) of decapitated trees of Acacia karroo in the eastern Cape Province indicates steadily increasingly coppice mass with in creasing tree size. This pattern was not attained in the present fire study since decreased canopy damage effects in the larger trees o f the burn changed this relationship in the upper size range. Only where a severe crown fire is conceivable would a Du Toit type relation be expected to apply to a burned tree popula tion.
The present study has clearly dem onstrated rela tionships between reduction in canopy leaf mass and basal shoot regeneration mass. But it is im portant to note that, although in accordance with expected effects o f reduction in apical dom inance, increasing damage to the canopy increased basal shoot produc tion (up to a threshold value), there was an opposite effect for very small uncanopied O. pulchra plants where increased damage to the plant reduced new basal shoot production. That basal shoot produc tion, in plants with much canopy leaf mass removed by fire, was greater in the faster burn area than in the slower burn area, suggests that the fire heat intensity also had a more direct effect on stim ulation of basal shoot production possibly through more effective killing of buds in the canopy.
An indication o f the possible limited resources of plants in their regeneration reaction to fire is reflected by the inverse hyperbolic distribution between number of basal shoots per plant and the mean individual basal shoot mass: a com pensatory effect between num ber and size o f basal regeneration shoots in O. pulchra.
It was found that in several respects, the effects of the faster fire were m ore variable than in the slower fire, for example, the relationship between number of basal regeneration shoots (in O. pulchra) and mean mass per basal shoot was tighter in the slower burn than in the faster burn. This possibly parallels the findings o f Trollope (1978) in eastern Cape grassland where the rate o f spread o f head fires was far more variable than that of back fires. In the rela tionships between plant height and changes in plant size (Fig. 8), the faster fire consistently had lower correlation coefficients than those for the slower fire. Although most different types o f effects o f fire on the woody plants were found to depend on plant height, a few' effects were found to be independent of plant height namely: the killing o f live basal shoots and the burning away o f old dead basal shoots. Plant height also appeared not to affect the relationship in O. pulchra between num ber of basal regeneration shoots and the mean mass per basal shoot. The