Cassells’ (1995) vegetation map of the park, together with local knowledge and observation were used as the basis of a stratified sampling strategy of miombo, chipya and mateshe vegetation types. A total of 25 belt transects were located within these habitats. Placement of transects was not random as an effort was made to select homogenous areas of vegetation and to avoid ecotones and atypical landscape features such as tracks and roads. Another consideration was accessibility, with most transects located within sight of a road to ensure that they could be easily found again and revisited in future studies. Transect positions were determined using a portable GPS (Garmin 12XL). Standard transects measured 50 x 10 m (= 500 m²), rectangular plots being chosen for ease of sampling and to maximise species diversity (Condit et al., 1996). Transects incorporated a minimum of 50 trees/shrubs or at least 15 specimens of a dominant species. Where these criteria were not met, transects were broadened or lengthened accordingly (Taylor & Walker, 1978).

Data collected in each plot is summarised in Table 1. Within each transect, woody plants (above 1 m in height) were identified, counted and measured for diameter at breast height. Herbs and grasses were identified, and ground cover density estimated on a scale of 1-5. Evidence of damage to vegetation by fire or other causes was also determined and noted. Soil pits were dug to a depth of 0.5 m, and soil colour and texture noted. Soil samples (200g) were taken from all plots at two different depths (5-10 cm and 25-40 cm). Other criteria noted for each plot included topographic features, and number and activity of termite mounds. Photographs were taken of all plots.

Woody vegetation data was ordinated using the multivariate analysis programme TWINSPAN (Hill, 1979), a cluster analysis programme which repeatedly divides a set of samples and the set of species that they contain on the basis of modified first axes of reciprocal averaging ordinations. TWINSPAN was used as an objective test of the chipya-mateshe versus miombo site selection carried out in the stratification process, and the results used as an indication as to whether these vegetation types can easily be separated according to their woody composition. For more details of TWINSPAN and other polythetic divisive classification methods see Kent & Coker (1992). Pseudospecies cut levels were set at 0, 1, 5, 10, 20, 40, 60 and 90. All other criteria were set to default. No omissions of indicators were designated.

Soil samples were oven dried (30° C, 7 days) and then measured for the following: pH; weight loss on ignition (%); organic carbon content (%); total nitrogen (%); total phosphorus (mg/kg). Soil analysis was carried out by Eclipse Voelcker Science, London.

A typical stand of chipya in Kasanka National Park

Miombo woodland, Kasanka National Park

RESULTS

The most striking feature concerning the floristic composition of the chipya plots (Table 2) was the complete absence of the woody miombo genera Brachystegia, Julbernardia, Isoberlinia and Uapaca. These were present in none of the chipya plots and, with the exception of Brachystegia spiciformis, were also absent from the mateshe. In contrast, the miombo plots were dominanted by these three genera. These results agree closely with Lawton’s (1978) ecological groups, comprising fire-tolerant species in the chipya group, a mixture of fire-tolerant and fire-sensitive species in miombo, and fire-sensitive species in the mateshe. Brachystegia spiciformis is a common constituent of dry evergreen forest (Fanshawe, 1961).

TWINSPAN analysis provided an objective test of the differentiation of miombo from chipya-mateshe, clearly separating the chipya-mateshe plots from the miombo plots on the basis of woody plant composition (Figure 1). The sole exception was miombo plot 7, a transect with mature, spaced tree physiognomy, dominated by Isoberlinia angolensis (absent from chipya) and Lannea discolor (common in chipya), but with no Julbernardia, Brachystegia or Uapaca species.

The taxa noted in the herbaceous layer of the three vegetation types is given in Table 3. Floristically, all of the chipya plots were dominated by Aframomum alboviolaceum, Pteridium aquilinum subsp. centrali-africanum and/or Smilax anceps. The herbaceous component in the miombo plots was more diverse than that of the chipya, with no obviously dominant species. In the very shady environment of the mateshe, the herbaceous layer was almost non-existent, with only a few species occuring.

The plots were most easily separated by the density of their herbaceous layers (Figure 2). Chipya plots averaged a density of 4.7 (on a scale of 0-5); miombo, 2.4; and mateshe 0.5. In the chipya,

plots the herbaceous layer was invariably dense and tall. In the miombo plots the herbaceous layer was moderately well-covered in most plots, sparse (0-1) in the Uapaca dominated plots and dense (3-4) in the plots dominated by widely spaced tall trees. In the mateshe plots, the herbaceous layer was invariably sparse. These results should be treated with some caution due to the subjective nature of an ocular estimate of herbaceous density; in addition, these estimates were made in January, in the middle of the rainy season, not at the end, when the herbaceous layer is fully developed.

These results agree closely with Trapnell’s floristic and physiognomic definition of the characteristic herbaceous layer found in chipya (Trapnell, 1943, Trapnell et al., 1947). In these studies, Aframomum alboviolaceum, Pteridium aquilinum subsp. centrali-africanum and Smilax anceps are given as indicator species of chipya, and major constituents of the unusually dense herbaceous layer characteristic of this vegetation type.

The soils in the chipya plots were invariably sandy loams of a free-draining nature, often with a very dark humic horizon. Chipya subsoils tended to be dark grey or brown in colour. The miombo soils were generally more clayey, but very variable in both texture and colour. Mateshe soils were generally sandy loams, and dark brown or grey in colour. However, soil texture and colour characteristics were not found to be useful for differentiating vegetation types. Chipya and mateshe soils tended to be more sandy than miombo soils, but some miombo soils were equally sandy. Similarly, chipya humic horizons tended to be deeper and darker in colour than those of miombo or mateshe soils, but some chipya soils showed no obvious humic horizon, and some miombo soils showed deep, dark humic horizons. The chemical characteristics of plot topsoils (5-10 cm depth) are summarised in Table 4, below. Subsoils (25-40 cm depth) show identical trends. Values from termite nest soils are given as a comparison.

Chipya soils are characterised by high organic carbon (Figure 3a), high organic Nitrogen (Figure 3b), low Phosphorus (Figure 3c) and low pH (Figure 3d). Mateshe soils are intermediate in all categories. Miombo soils have low organic matter, low organic Nitrogen, high phosphorus and (comparatively) high pH. Soil samples from termite nests show high Carbon, Nitrogen and Phosphorus values, and intermediate pH.

Figure 1: TWINSPAN divisions of chipya, mateshe and miombo vegetation plots in Kasanka National Park.

Figure 2: Estimated density of herbaceous layer (scale 1-5) in chipya, miombo and mateshe plots in Kasanka National Park

Figure 3: Chemical analysis of chipya, miombo, mateshe and termite mound soils in Kasanka National Park : (a) Organic carbon (%) and organic matter (%); (b) Nitrogen (%); (c) Phosphorus (mg/Kg); (d)

Primary productivity is the main determinator for soil organic matter (Webster, 1970) and it is likely that the high organic matter recorded in chipya soils is the product of the characteristically rich herbaceous layer in this vegetation type.

Figure 4 is a scatter diagram of estimated herbaceous layer density versus recorded organic matter (% loss on ignition) in topsoil. Linear regression analysis of this graph produces a statistically significant (P = 0.0246) positive correlation between these two factors (R² = 0.1796), but the indication is that other factors are probably involved as well. It is possible that the small sample size may be important, as well as the subjective nature of the herbaceous density estimate, as suggested above.

It is probable that humus-feeding termites, such as Cubitermes species, are important in degrading soil organic matter in chipya, miombo and mateshe soils (Trapnell et al., 1976), and therefore the presence or absence of these would be expected to impact on soil organic matter content. In this study, the occurrence of Cubitermes mounds in the plots was used as an estimate of their activity in each vegetation type. However, no significant difference was found between numbers of termite mounds in miombo and chipya plots. In addition, when number of mounds were plotted against organic matter in topsoil (Figure 5), linear regression analysis suggests that no significant correlation exists between these factors (R² = 0.000113; P = 0.96). Again, small sampling size may be a limitation with this analysis. In addition, definite identification of Cubitermes was not always possible, and some mounds may have contained non-humus feeding termites.

Table 2: Woody composition of chipya, miombo and mateshe vegetation types in Kasanka plots.

Table 3: Density and floristic composition of the herbaceous layer in chipya, miombo and mateshe in Kasanka plots.

Table 4: Characteristics of chipya, mateshe and miombo topsoil (1-10 cm) in the Kasanka plots. Mean values shown. One-tailed t-test, chipya and miombo, * P<0.05, **P<0.005, †P<0.01, ‡P<0.0005.

Figure 4: Scatter diagram of estimated herbaceous layer density versus recorded organic matter (% loss on ignition) in topsoil in chipya, miombo and mateshe plots in Kasanka National Park. Linear regression analysis R² = 0.1796; P = 0.0246

Figure 5: Scatter diagram of number of termite mounds plotted against organic matter in topsoil in chipya, miombo and mateshe plots in Kasanka National Park. Linear regression analysis R² = 0.000113; P = 0.96.

 

DISCUSSION

The physiognomy and floristic composition of chipya in Kasanka National Park is entirely consistent with the descriptions of lake basin chipya provided by Trapnell (1943) when he introduced the term to the scientific literature. In KNP chipya and miombo appear to occur in a mosaic, not a continuum, thus the miombo woody dominants Brachystegia, Julbernardia, Isoberlinia and Uapaca, were not present in any of the chipya plots. This trend is reflected in the TWINSPAN woody species analysis, which clearly separates chipya from miombo. The corollary of this is that the characteristically luxuriant herbaceous layer of chipya, dominated by Pteridium aquilinum subsp. centrali-africanum, Aframomum alboviolaceum, Smilax anceps and tall grasses of the genera Hyparrhenia and Andropogon was not found in any of the miombo plots.

In contrast, chipya and mateshe occur in three plots as a continuum, clearly associated, and with readily recognisable ecotones, hence the TWINSPAN grouping of these plots with chipya at the second division (Figure 1). These findings have implications for chipya dynamics, which are discussed below (Section 2.4.3) .

The KNP chipya soils are typically sandy loams with a deep, dark humic horizon (Table 4). These results are consistent with the findings of Trapnell (1943); Trapnell et al. (1947); Lawton (1964); and Fanshawe (1969), all of whom report freely drained soils of a permeable and friable nature in chipya. However, Astle (1968) and Lawton (1978) caution against differentiating between miombo and chipya soils according to soil texture, because miombo may also occur on freely draining sandy loams. This was found to be the case in KNP, where miombo plots 8-10 were on sandy loams, and miombo plots 6-9 exhibited soils with a deep, dark humic horizon.

Chemical analysis of the chipya soils in the KNP (Table 4) showed that, compared to KNP miombo soils, they have significantly higher organic C and organic N, and lower P and pH. These results, apparently the first published full chemical analyses of chipya soils, agree closely with Trapnell (1943) who characterises chipya soils as being ‘typically of a richly humic nature’ and ‘markedly acid’. The miombo analyses compare closely with those published by Trapnell et al. (1976) for miombo woodlands in Ndola (e.g. topsoil % organic C, 0.86; % N, 0.068; pH 5.2). Again, these characteristics cannot be used in isolation to define the vegetation types. In KNP the chemical characteristics measured were found to be variable in all vegetation types, thus some miombo soils showed high organic C (e.g. miombo plot 7: 1.45) and some chipya soils showed low organic C (e.g. chipya plot 9: 0.5).

Primary productivity is the main determinator for soil organic matter (Webster, 1970) and it is most likely that the high organic matter in chipya soils is either:

the product of the characteristically rich herbaceous layer of chipya, or;

a legacy of the evergreen forest from which chipya is derived

Our results demonstrate that the mateshe from which it is believed the chipya is derived in KNP has a lower soil organic matter content than chipya, but higher than that of miombo. This finding suggests that the high soil organic matter in chipya soils is at least partially a product of the rich herbaceous layer. The question remains, though, how did this luxuriant layer arise? A possible explanation (Trapnell, pers. comm.) is that the burning of mateshe may produce a fertilising effect, such as that utilised in slash and burn agriculture (Stent, 1933). Once established, the dense herbaceous component becomes a source of organic matter itself. Fires are not necessarily an annual event in chipya, and in fire-free years, the dense herbaceous layer will break down into a mulch or green manure, which will boost organic matter concentrations in the topsoil. During years in which the chipya does burn, charcoal and soot is produced as a result of the fire. Chipya fires would be expected to produce a great deal of fine soot, due to the amount of green growth in the herbaceous layer, and soot and charcoal will further contribute to the un-decomposed organic reserves in the soil. In the present study, when % organic matter in the topsoil was plotted against estimated density of the herbaceous layer (all plots), a statistically significant positive correlation was shown (Figure 4), but it is probable that other factors (e.g. vegetation history, fire regime etc.) are involved as well.

Trapnell et al. (1976) have suggested that termites are of great importance in the breakdown of organic matter in miombo soils. These authors found that the number of Cubitermes mounds present in miombo woodland increased significantly in late burnt plots (160 mounds per 0.4 ha) compared to early burnt (120) and completely protected plots (80). However, Ferrar (1982) showed that humus-feeding termites foraging in a recently burnt broadleaf savanna was greatly reduced for two months after the fire. In the present study, no clear fire-related trends were seen in the presence or absence of termites in chipya and miombo, although Cubitermes were apparently absent from the fire-protected mateshe (Figure 5). In addition, it was not possible to demonstrate a statistically significant relationship between termites and organic matter. It is perhaps interesting to note that soil organic matter content is higher in termite mounds than in any of the vegetation types (Table 4). It has been suggested (Scholes and Walker, 1993) that the higher concentrations of organic matter found in the walls of termite mounds compared to that in the surrounding soils implies that most of the energy requirements of these termites are met by small plant fragments in the soil rather than by humified soil organic matter. If this is so, the role of humus–feeding microbes may well be more significant.

All of the authors who have written about the mateshe-chipya sequence are agreed that fire is an essential element driving these vegetation changes (Trapnell, 1943; Schmitz, 1950; Lawton, 1963; Fanshawe, 1969; White, 1983). Although all concur that fire is crucial to the creation of chipya, the main area of argument surrounding the chipya-evergreen sequence is the role of miombo (Brachystegia, Julbernardia) species in the presence or absence of fire. The results of the present study suggest the following:

Chipya-miombo succession

The apparent absence of a chipya-miombo continuum in KNP argues against a chipya-miombo succession. However, there may be encroachment of miombo species into areas of chipya through the mechanism suggested by Lawton (1978, 1996), i.e. through colonisation by root sucker Uapaca species, which create a fire-free environment allowing the fire-sensitive species of Brachystegia, Julbernardia etc. to establish themselves. In addition, we did see some evidence of Brachystegia spiciformis colonisation of chipya, originating from areas of mateshe in which parent trees were present.

Miombo-chipya regression

The Ndola burning experiments (Trapnell, 1959) showed quite clearly that the burning of miombo does not produce chipya. This finding is borne out by the results of the present study, where our observations suggested a miombo-chipya mosaic rather than the continuum observed by Lawton (1978). In KNP no Brachystegia, Julbernardia, Isoberlinia or Uapaca species were found in the chipya plots, and no luxuriant herbaceous layer with chipya indicators was found in the miombo plots.

Chipya-mateshe succession

The clear association of chipya and mateshe in many of the plots in KNP, and the recorded ecotones suggest that mateshe and chipya are part of the same sequence. The idea of a chipya-mateshe succession (e.g. Lawton, 1963) is partly a question of climate, i.e. whether a forest climax is possible in today’s climate. Evidence from the Ndola plots (Trapnell, 1959; Lawton, 1996) shows that a succession to evergreen forest is possible in fire-protected miombo, in today’s climate. However, even here the protected plots have latterly been accidentally burnt, and this raises the equally pertinent question of whether any area in Zambia is likely, in practice, to be spared from fire. It might be argued that in the unlikely event that a pocket of chipya was protected from fire for a considerable period of time, the build up of litter would ensure that when the fire did come, it would be all the more fierce and would surely kill fire-sensitive seedlings, saplings or young trees. Lawton (1978) suggests a ‘chipya’-Uapaca-Brachystegia-mateshe succession in the presence of fire. However, his definition of chipya encompasses fire-degraded miombo, which is very different from the chipya of Trapnell in which the luxuriant herbaceous layer ensures fierce, hot fires. Thus, although Lawton’s proposed succession sequence may well occur in miombo, where the herbaceous layer is relatively sparse and fires consequently less severe, it is hard to envisage any succession sequence in chipya proper.

Mateshe-chipya regression

Trapnell (1943); Schmitz (1950); Lawton (1964); Fanshawe (1969) and White (1983) have all proposed that chipya is a regression product of mateshe, driven by fire and/or cultivation. The results obtained in KNP appear to confirm this. The clear association of chipya and mateshe in KNP suggests that the two vegetation types are related, and given today’s climate and the prevalence of fire in Zambia, a mateshe-chipya regression sequence is very likely. Further evidence from the current study was found in the floristic composition of the mateshe plots, which were characterised by different dominant species at different sites. This pattern suggests that these patches of mateshe are relics, rather than groups of pioneer species establishing themselves.

The results from Kasanka National Park suggest that chipya can be differentiated from miombo by the absence of the fire-sensitive genera Brachystegia, Julbernardia, Isoberlinia and Uapaca, and by its luxuriant herbaceous layer comprising the indicator species Pteridium aquilinum, Smilax anceps and Aframomum alboviolaceum. In addition, chipya soils have been shown to typically possess high organic matter, low phosphorus and low pH compared to miombo soils. Finally, the evidence in KNP suggests that chipya arises from a mateshe-chipya regression driven by fire. There is no evidence that the fire-sensitive miombo genera Brachystegia, Julbernardia, Isoberlinia and Uapaca are part of the sequence.