This study was designed to provide the means for implementing formal scientific vegetation management 1n the succulent valley bushveld of the eastern Cape, South Africa. Nowhere in the world has a detailed, effective and practical veld management system being developed entirely from research, and even the most successful management systems rely heavily on the intuition of people. A process, formally called 'adaptive management', combines this intuition with scientific testing and the overall objective of this study was to provide a framework for formalized adaptive management in succulent valley bushveld. On analyzing the process of adaptive management, the following knowledge 'tools' were identified: (i) a management system for immediate implementation; (ii) a technique for vegetation assessment; (iii) a technique for monitoring vegetation change; (iv) a technique for monitoring forage use and recovery; (v) a list of key forage species; (vi) a model to set initial stocking rates; (vii) a method of recording essential information; and (viii) a database of ecological principles. Providing these 'tools' became the goals of this study. These topics covered almost all facets of rangeland science, and the approach was to address these in a 'top down' manner, rather than sub-optimize by specializing on anyone component. Most of the 'tools' were achieved to a greater or lesser extent and are presented as a series of publications. However, a central tool, that for monitoring vegetation change, remains outstanding despite comprehensive testing of a range of traditional botanical methods. Indeed, critical review revealed that this 'missing tool' is a problem which is common in all vegetation communities in South Africa - despite the impression created by vegetation researchers that adequate techniques are indeed available. This is serious because land managers are not able to evaluate the impact of their efforts and the government is unable to monitor the effectiveness of their research and extension services, costing millions of public monies annually. The implication also, is that vegetation cannot be managed scientifically (management implies monitoring). Either formal adaptive management is not practicable, or researchers are operating from an inappropriate paradigm; specifically that of providing techniques for their research projects and claiming that these (or derivatives of these) are adequate for farm or regional scale monitoring. More generally, research has often become an end in it's self, with research quality being judged by criteria which are of little significance to the real world and which damage efficiency. Perhaps, the real value of vegetation research lies in the experiential learning which the researcher gains not the inevitably parochial results. / Thesis (Ph.D.)-University of Natal, Pietermaritzburg, 1993.

The work reported in this thesis was designed to develop a better understanding of the fate of fertiliser nitrogen applied to a tropical pasture under field conditions, with the eventual objective of improving the economy of livestock production off such pastures. This involved an examination of the concentrations of soil total nitrogen, ammonium nitrogen and nitrate nitrogen at different depths within the soil profile following the application of different levels of fertiliser nitrogen to a grazed kikuyu (Pennisetum clandestinum) pasture, as well as the influence of such applications on pasture yield and some elements of pasture quality. The trial was conducted over a two year period at Broadacres in the Natal Mistbelt. A labelled [15]NH[4]N0[3] fertiliser experiment was also conducted to ascertain how the labelled ammonium ion moved through the soil, roots and herbage after being applied in spring onto a kikuyu pasture. In the absence of fertiliser N, a total of 15.45 t/ha of soil N was recorded at an average concentration of 0.15%. More than 30% of the soil total N was, however, situated within the top 10cm of soil. organic matter (OM) content in the top 0-10cm of the profile was high (4.75%), reflecting an accumulation of organic matter in this zone. However, as organic C (and thus c: N ratios) declined with depth, so too did soil total N concentration. Not surprisingly, fertiliser measurably increase soil total N, N applications did not but indirectly may have affected soil N dynamics by increasing net mineralisation (due to its "priming" effect) thereby stimulating plant growth and thus increasing the size of the organic N pool through greater plant decay. Total soil N concentration did not change significantly from the first to the second season. This could be attributed to the fact that N gains and losses on the pastures, being over 15 years old, were probably in equilibrium. Generally similar trends in soil total N down the profile over both seasons was further confirmation of this. Before the application of any fertiliser, 331.9 kg NH[4]-N was measured in the soil to a depth of 1m, on average, over both seasons. This amount represented only 2.1% of the soil total N in the profile. The concentration of NH[4]-N followed a quadratic trend down the soil profile, irrespective of the amount of fertiliser N applied, with the largest concentrations accumulating, on average, in the 0-10cm and 75-100cm depth classes and lowest concentrations in the 50-75cm depth class. Laboratory wetting/drying experiments on soil samples collected from a depth of 75-100cm showed that NH[4]-N concentrations declined only marginally from their original concentrations. A high organic C content of 1.44% at this depth was also probable evidence of nitrification inhibition. Analysis of a similar Inanda soil form under a maize crop did not exhibit the properties eluded to above, suggesting that annual turn-over of the soil was causing mineralisation-immobilisation reactions to proceed normally. Addition of fertiliser N to the pasture significantly increased the amount of NH[4]-N over that of the control camps. Furthermore, the higher the application rate, the greater the increase in NH[4]-N accumulation within the soil profile. As N application rates increased, so the NH[4]-N:N0[3]-N ratio narrowed in the soil complex. This was probably due to NH[4]-N being applied ln excess of plant requirements at the high N application rates. On average, 66.7 kg more NH[4]-N was present in the soil in the first season than in the second after fertilisation. Although this amount did not differ significantly from spring through to autumn, during early spring and late summer/autumn concentrations were higher than in mid-summer. Observed soil NH4-N trends were also very similar to the soil total N trends within both seasons, suggesting that soil total N concentrations might well play an important role in determining soil NH4-N concentrations. Before fertilisation, only 45.6 kg N0[3]-N, representing 0.29% of the soil total N, was on average, found in the profile to a depth of 1m. The highest concentration of N0[3]-N was lodged in the top 10cm of the soil. Nitrate-N declined, on average, with depth down the profile. However, during the second season, even though the concentration of N03-N declined down the profile, it increased with depth during relative to that of the first season, suggesting the movement of N0[3]-N down the profile during this period. Fertilisation significantly increased the concentration of N0[3]-N above that of the control camps. Concentrations increased as fertiliser application rates increased, as did N0[3]-N concentrations with depth. This has important implications regarding potential leaching of N03-N into the groundwater, suggesting that once applications reach levels of 300 kg N/ha/season or more, applications should become smaller and more frequent over the season in order to remove this pollution potential. On average, 94.3 kg N0[3]-N/ha was present down to a depth of 1m over both seasons. However, significantly more N0[3]-N was present in the second season than in the first. This result is in contrast to that of the NH[4]-N, wherein lower concentrations were found in the second season than in the first. No specific trends in N0[3]-N concentration were observed within each season. Rather, N0[3]-N concentrations tended to vary inconsistently at each sampling period. Nitrate N and ammonium N concentrations within each month followed a near mirror image. A DM yield of 12.7 t/ha, averaged over all treatments, was measured over the two seasons. A progressive increase in DM yield was obtained with successive increments of N fertiliser. The response of the kikuyu to the N applied did, however, decline as N applications increased. A higher yield of 1.8 t DM/ha in the first season over that of the second was difficult to explain since rainfall amount and distribution was similar over both seasons. On average, 2.84% protein N was measured in the herbage over both seasons. In general, protein N concentrations increased as N application rates increased. On average, higher concentrations of protein-N were measured within the upper (>5cm) than in the lower <5cm) herbage stratum, irrespective of the amount of N applied. Similar bi-modal trends over time in protein-N concentration were measured for all N treatments and within both herbage strata over both seasons, with concentrations tending to be highest during early summer (Dec), and in early autumn (Feb), and lowest during spring (Oct), mid-summer (Jan) and autumn (March). spring and autumn peaks seemed to correspond with periods of slower growth, whilst low mid-summer concentrations coincided with periods of high DM yields and TNC concentrations. The range of N0[3]-N observed in the DM on the Broadacres trial was 0.12% to 0.43%. As applications of fertiliser N to the pasture increased, N0[3]-N concentrations within the herbage increased in a near-linear fashion. On average, higher concentrations of N0[3]-N, irrespective of the amount of fertiliser N applied, were measured wi thin the upper (>5cm) than the lower <5cm) herbage stratum. A similar bi-modal trend to that measured with protein-N concentrations was observed in both seasons for N0[3]-N in the herbage. High concentrations of N0[3]-N were measured during spring (Nov) and autumn (Feb), and lower concentrations in midsummer (Dec & Jan), very early spring (Oct) and early autumn (March). During summer, declining N0[3]-N concentrations were associated with a corresponding increase in herbage DM yields. A lack of any distinctive trend emerged on these trials in the response of TNC to increased fertilisation with N suggests that, in kikuyu, applied N alone would not materially alter TNC concentrations. Higher concentrations of TNC were determined in the lower <5cm) height stratum, on average, than in the corresponding upper (>5cm) stratum. This may be ascribed to the fact that TNCs tend to be found in higher concentrations where plant protein-N and N0[3]-N concentrations are low. A P concentration of 0.248% before N fertilisation, is such that it should preclude any necessity for P supplementation, at least to beef animals. Herbage P concentrations did, however, drop as N fertiliser application rates were increased on the pasture, but were still high enough to preclude supplementation. Even though no significant difference in P concentration was measured between the two herbage strata, a higher P content prevailed within the lower <5cm) herbage stratum. On average, 2.96% K was present within the herbage material in this trial. The norm for pastures ranges between 0.7 and 4.0%. On these trials, applications of fertiliser N to the camps did not significantly affect K concentrations within the herbage. The lower <5cm) herbage stratum, comprising most of the older herbage fraction, was found to contain the highest K concentration in the pasture. The presence of significantly (although probably biologically non-significantly) less K within the herbage in the second season than in the first may be linked to depletion of reserves of · this element in the soil by the plant and/ or elemental interactions between K and other macro-nutrients. An average Ca content of 0.35% within the herbage falls within the range of 0.14 to 1.5% specified by the NRC (1976) as being adequate for all except high-producing dairy animals. Increasing N application rates to the pasture increased the Ca content within the herbage . No significant differences in Ca concentration were found between the upper (>5cm) and lower <5cm) herbage strata over both seasons, even though the lower stratum had a slightly higher Ca concentration, on average, than the upper stratum. Calcium concentrations did not vary between seasons, probably because concentrations tend rather to vary according to stage of plant maturity, season or soil condition. However, higher concentrations of the element were measured in the second season than in the first. The reason for this is unknown. On average, 0.377% Mg was present within the herbage over both seasons. This compares favourably with published data wherein Mg concentrations varied from 0 . 04 to 0.9% in the DM, with a mean of 0.36%. All camps with N applied to them contained significantly more Mg in their herbage than did the material of the control camps. On these trials, the Ca :Mg ratio is 0.92: 1, which 1S considered to be near the optimum for livestock and thus the potential for tetany to arise is minimal. Magnesium concentrations remained essentially similar within both herbage strata, regardless of the rate of fertiliser N applied. As in the case of Ca, Mg concentrations within the herbage were significantly higher in the second season than in the first. Calcium:phosphate ratios increased, on average in the herbage, as N application rates increased. This ratio was high in spring, dropped off in summer and increased again into autumn, suggesting that the two ions were following the growth pattern of the kikuyu over the season. The K/Mg+Ca ratios were nearly double that of the norm, suggesting that the pasture was experiencing luxury K uptake which may be conducive to tetany in animals grazing the pasture. This ratio narrowed as N application rates were increased, probably as a result of ion dilution as the herbage yields increased in response to these N applications. The ratio was low in spring (October), but increased to a peak in December, before declining again to a low in March. / Thesis (Ph.D.)-University of Natal, Pietermaritzburg, 1994.

The seasonal production of 11 pastures was evaluated in dryland and irrigated trials at eight sites in the Winter Rainfall Region. These data were related to climatic conditions using the Growth Index concept to produce a model for pasture growth. Under dryland at Tygerhoek, the animal production potential of lucerne and medic was compared in grazing trials. Lucerne was found to be the higher producing of the two. At this site also, the influence of chemical control of volunteer grasses in dryland pastures on animal production potential was tested. Weed control had a positive influence on animal production at low, but not at high stocking rates. Under irrigation at complex grass/legume Tygerhoek, the grazing capacity of a mixture was established under continuous and rotational grazing. While rotationally grazed pastures produced the highest yields, the clover component of these pastures proved to be most productive under continuous grazing. As a result, rotationally grazed pastures, could carry more animals, but animal production was generally highest under continuous grazing. Under irrigation at Outeniqua, seven grass and grass/legume mixtures were compared in grazing trials. Pastures based on fescue generally had the highest grazing capacity, but those based on white clover the highest animal production potential. These data were used to produce a climate:pasture:animal which was validated using independent This model was used to predict animal production model, grazing trial data. performance of two-species mixtures at a number of sites. These results suggested that while grass pastures allowed more animals to be carried than did mixtures, both animal performance and gross returns were highest in grass/legume mixtures. / Thesis (Ph.D.)-University of Natal, Pietermaritzburg, 1986.

Two management blocks were set aside in Umfolozi Game Reserve (UGR) to compare different management approaches. In one, the non-cull block, a noninterventionist policy was followed and no large mammals were removed, while the other, the cull block was subjected to the same game removal treatment as the remainder of the reserve. The main objectives of this study were to determine the relationships between vegetation, soil surface variables and both soil loss and rainfall run-off, to derive predictive models for run-off and soil loss based on vegetation and soil surface variables, to determine the relationship between different levels of soil erosion and the production potential of soils, and to determine the relationship between different levels of soil erosion and herbaceous species diversity. Rainfall simulator trials and natural run-off plots were used to collect quantitative data on soil loss and run-off. Bivariate scattergrams showed that the relationship between soil surface and vegetation variables plotted against soil loss was curvilinear. "Susceptibility to erosion" showed the highest positive correlation, and "surface cover" the highest negative correlation with soil loss. The relationship between annual run-off and both the soil surface and vegetation variables was also curvilinear, with "soil capping" showing the highest positive and "litter cover" the highest negative correlation with run-off. Using multiple regression analysis it was found that "susceptibility to erosion" and predictors of annual soil loss. "surface cover" were the best "Soil capping" and "percentage contribution of run-off. forbs" were the best predictors of annual No clear relationship between either soil loss and stocking rate, or run-off and stocking rate was apparent in the two experimental blocks, and the differences in soil loss and run-off could not be explained by differences in stocking rate alone. There were however defects in the experimental design which invalidated the assumption that the stocking ~ate diffe~entia1 management blocks would inc~ease with time. between the two Because of the above deficiency, an a1te~native study a~ea on the weste~n bounda~y fence, which allowed fo~ pai~ed sampling sites on eithe~ side of the fence, was chosen. Ge~lach t~oughs we~e used to measu~e soil loss. The g~eatest va~iabi1ity in soil loss was explained by the position of the plots on the slope ~athe~ than whethe~ the plots we~e in UGR o~ in adjacent KwaZu1u. Simi1a~ly, diffe~ences in topog~aphy, ~athe~ than diffe~ences in 1anduse, exe~ted an ove~~iding effect on A-ho~izon depth, he~bage accumulation and g~ass species ~ichness. Conside~ing the ~esu1ts obtained, the opinion that a noninte~ventionist policy would lead to a decline in vegetation p~oductivity and to a 10ng-te~m ~eduction in species dive~sity appea~s to be unfounded. Finally, based on the data collected and on a ~eview of cu~~ent scientific 1ite~atu~e, changes to the Natal Pa~ks Boa~d soils policy and objectives a~e suggested, and the objectives a~e t~ans1ated into ope~ationa1 management goals. / Thesis (Ph.D.)-University of Natal, Pietermaritzburg, 1988.

The overall objective was to develop stocking rate models for three veld types. namely the Lowveld. the Southern Tall Grassveld and the Natal Sour Sandveld. in Natal. Sub-objectives were to determine the 1) residual herbage mass at the end of the summer, 2) residual herbage mass at the end of winter and 3) individual animal performance under grazing conditions, and the effect of stocking rate on these three variables. Multiple linear regression component models were successfully developed to meet all three of the sub-objectives for each of the three veld types. Results show that veld condition is an extremely important factor determining animal production from veld, and that stocking rate on veld in good condition could possibly be double that on veld in poor condition. Stocking rate did not have the expected impact on individual animal performance in the summer. although it did have an important moderating influence under any particular set of environmental conditions. Stocking rate did. however. have a marked effect on herbage production and therefore on herbage availability in winter and so also on the ability to overwinter cattle without having to supply additional supplementary feed. Stocking rate in summer therefore had a major indirect effect on animal production in the winter. Carryover of residual herbage from one year to the next is probably not as important in these veld types as in some other parts of the country. / Thesis (Ph.D.)-University of Natal, Pietermaritzburg, 1990.

Lolium perenne L. (perennial ryegrass) generally fails to persist under the sub-tropical cpnditions of South Africa. Furthermore, little research data are available on how to manage this species locally. This study was designed to identify the management options, particularly with r espect to grazing defoliation, which would help enhance the longevity of perennial ryegrass pastures. This was addressed by: 1) reviewing on-farm management practices of perennial ryegrass in KwaZulu-Natal; 2) conducting a detailed two-year field study of the effects of grazing frequency (HF, MF and LF = high, medium and low frequency, respectively) and intensity (HI, MI and LI = high, medium and low intensity, respectively), rotationally applied with the addition of a continuous grazing treatment (CG), on parameters linked to persistency. tiller population dynamics, dry matter These included: (DM) yield and quality, perennial ryegrass vigour, weed invasion and root development; and 3) examining effects of different levels of applied nitrogen (N) during the establishment year on various parameters linked to persistency. These included: tiller population densities, DM yield and quality, perennial ryegrass vigour, weed invasion and root development. The review of on-farm management practices of perennial ryegrass growers in KwaZulu-Natal revealed that reasonably high rates of N application (e.g. 350 and 250 kg N ha¯¹ a¯¹ to perennial ryegrass as pure and clover-based stands, respectively) are important for pasture survival. However, a consistent distribution of the applied N is even more important (i. e. at least seven split applications of N onto pure stands of perennial ryegrass and five onto perennial ryegrass-clover). In terms of grazing management, the period of absence of animals from the pasture during summer was identified as the most important grazing variable affecting pasture survival (i.e. ≥ 21 days). Also, the length of the period of occupation by animals should be as short as possible, particularly during summer (i. e. ≤ 3 days). Paying careful attention to summer irrigation is also an important variable contributing to pasture survival. Grazing intensity was not highlighted as an important contributor to pasture survival. In terms of tillering potential, DM yield and quality (cellulose dry matter disappearance and herbage N) and perennial ryegrass vigour, perennial ryegrass followed definite seasonal trends. These were highest during autumn and spring and were lowest during the mid to late summer period. Perennial ryegrass was most susceptible to general sward degradation through poor management during the mid to late summer period when the danger from weed invasion is greatest and its growth potential, vigour and tillering abilities are lowest within these seasonal periods, grazing defoliation produced marked effects. In terms of tiller survival, DM yield, plant vigour, reduced weed invasion and root production, treatments incorporating low frequency grazing (e.g. LFLI and LFHI) generally out-performed (P≤0.05) those incorporating high frequency grazing, irrespective of the intensity (e.g. HFHI, HFLI, and continuous grazing (CG)). The defoliation treatment incorporating medium frequency and intensity (MFMI) (currently the recommended defoliation strategy for perennial ryegrass) was also out-performed in many instances (P~0.05) by the low frequency treatments (e.g. LFHI and LFLI) . During the establishment year, increasing levels of applied N increased (P≤O. 05) perennial ryegrass DM yields and herbage quality. Models predicting the response of DM yield and quality to applied N suggest linear responses up to 720 kg N ha¯¹ a¯¹. Further refinement of such models and the inclusion of animal production parameters is recommended. Maximum (P≤0.05) tiller population densities occurred at applied N levels of 480 kg ha¯¹ a¯¹. Perennial ryegrass vigour increased (P≤O. 05) with increasing levels of applied N up to 480 kg h¯¹ a¯¹, but individual tiller vigour decreased. Increasing levels of applied N up to 360 kg ha¯¹ a¯¹ suppressed (P≤O. 05) weed tiller densities. Increasing levels of applied N (up to 600 kg h¯¹) increased (P≤O. 05) the root organic matter (OM) per unit volume of soil in the top 5 cm of the soil and decreased root OM per unit volume in the 10 - 20 cm soil depth category. / Thesis (Ph.D.)-University of Natal, Pietermaritzburg, 1994.

In South Africa, the loss of wetlands and their associated benefits has been considerable. A need was identified for a system that, using available information, would assist in achieving a balance between local, mainly short-term benefits to individuals and spatially wider and longer term benefits to society. Such a system, termed WETLAND-USE, was developed with the philosophy that:(l) wetlands have been well demonstrated to supply several indirect benefits to society (e.g. water quality enhancement); (2) the impact on these benefits can be described on a qualitative basis using field indicators that characterize the wetland and the disturbance associated with a particular land-use; (3) this information can be communicated to wetland users, which will contribute to achieving a desired balance, provided there is an enabling organizational environment and due consideration is taken of the socio-economic and organizational factors affecting wetland management. The primary conceptual framework underlying WETLAND-USE was the pressure-state-perceptions-policy framework, which depicts: the mode of use (i.e. the pressure); how this affects the state of the system (including its underlying processes and the goods and services it delivers); which in turn shape the perceptions that ultimately determine the policy pertaining to further use. This cycle is repeated at a range of organizational levels from local to national and takes place within a particular socioeconomic context. WETLAND-USE, which was designed for use by fieldworkers, and built using a rule-based, expert system approach, has two main parts, dealing largely with biophysical and social aspects respectively. Part 1, which guides the collection of data relating to the state of the wetland, assists in: (1) predicting the likely impacts of disturbances associated with a proposed land-use (the pressure) on the wetland state, and (2) providing ongoing management guidelines for particular land-uses. Part 2 assists in: (1) describing the social, land tenure and policy contexts of the wetland; and (2) establishing and maintaining organizational arrangements, local policy and management objectives and goals. Several discrete investigations were required for the development and refinement of WETLAND-USE, which was done in an iterative fashion. Initial discrete investigations fed into the development of a prototype system which was refined through evaluation using a questionnaire survey and further discrete investigations. The revised system was re-evaluated using a fieldworkshop approach and, based on the performance of the system in the field, it was revised further to produce the final system. In the two initial discrete studies, protocols were developed for characterizing key physical determinants of wetland functioning, notably: (1) degree of wetness, one of the primary functional determinants, described in the field using readily identifiable soil morphological indicators (e.g. matrix chroma and mottles) and (2) landform setting, which strongly influences local flow patterns and lateral exchange of water and water-borne materials. Graminoid plant species composition and functional groups (defined in terms of photosynthetic pathway) were then described in relation to the above physical determinants, together with rainfall, temperature and soil texture, within wetlands spanning a wide altitudinal range. This revealed that degree of wetness and altitude had the strongest influence over the vegetation parameters examined. An investigation into incorporating cumulative impacts into wetland decision making revealed that consideration should be given to: wetland loss in relation to ecoregions and catchments, and the relation of change in wetland extent, spatial configuration and context respectively to wetland function. Current conservation initiatives in KwaZulu-Natal were shown to account poorly for cumulative impacts on wetlands. Rules of thumb for making such considerations, given severe data limitations, were developed with reference to the high turn-over of species along the altitudinal gradient observed in the vegetation study. The "rules" were than applied to a case-study, the upper Mgeni catchment, as part of an initiative to engage a diversity of stakeholders in wetland information gathering and use. This resulted in the selection of priority wetlands in the catchment and an examination of the extent to which integration had been achieved vertically (across hierarchical levels) and horizontally (across organizations within particular hierarchical levels). In order to broaden the range of land-uses accounted for by the WETLAND-USE prototype, it was applied to a communally used wetland, Mbongolwane, and found to account poorly for the traditional cultivation and vegetation harvesting practices encountered. WETLAND-USE was modified to include a greater diversity of land-use types as well as enhancing its capacity to allow assessments to be conducted using the system's general criteria, thereby making WETLAND-USE more robust. In enhancing the capacity of WETLAND-USE to account for the social and organizational dimension of wetland management, the involvement of local and outside organizations in influencing wetland resource use in five sites was examined in relation to predefined frameworks. The sites, Mandlazini wetland, Mbongolwane wetland, Blood River vlei, Ntabamhlope vlei and Wakkerstroom vlei were chosen to represent a diversity of social contexts and management authorities. This revealed that in communally used areas in particular, a wide range of organizations are involved to varying degrees in influencing the use of different wetland resources. The level to which the local organizational environment contributed to sustainable use varied greatly among wetlands, but in all cases had important deficiencies: (1) self-governing resource-management organizations were largely lacking and in communal areas were weakening under contemporary conditions; and (2) although a formal management system was in place in two of the five wetlands, it was largely absent in the remaining three. There has been little involvement from extension services in facilitating local policy development and in promoting alternative land-uses which have less pressure on the state of the wetland. Local wetland management policy and collaboration among land-owners in wetlands under multiple separate ownership such as Blood River vlei was identified as being particularly poor. The evaluations of WETLAND-USE revealed that, in relation to the underlying philosophy of the thesis, WETLAND-USE had been improved through field application and incorporation of the findings of the discrete investigations. Nevertheless, important limitations of the study were highlighted, including: its high level of reliance on expert opinion in the face of a paucity of empirical data relating to the functioning of local wetlands and their attendant benefits (and how these are affected by anthropogenic disturbances), and a particularly shallow representation of socio-economic factors. The identification of these limitations was useful in highlighting key areas for further research. / Thesis (Ph.D.)-University of Natal, Pietermaritzburg, 1999.

Bush encroachment is a major problem for the savannas supporting livestock in the Eastern Cape. Farmers employ chemical poisoning and mechanical clearing of woody vegetation to improve grass production. This thesis addressed the following questions. 1. Does soil fertility and hence, herbaceous production and/or quality increase beneath or between former bushclumps following woody clearing or poisoning? 2. Are chemical or mechanical control methods effective for all woody species? 3. Does bush density and height decline with increasing competition from the herbaceous layer? 4. Can fire and goats retard or revert woody re-establishment, thereby keeping these savannas open? A trial for assessing different methods of controlling bush was conducted near Kei Road. Initial treatments included chemical poisoning, mechanical clearing and a control, each replicated eight times in 0.36 hectare plots. Follow-up treatments were control, fire, goats, and the combination of fire and goats each replicated twice per primary treatment. Two additional mechanically cleared plots were oversown with Chloris gayana seeds. Herbaceous production, species composition, foliage quality and soil fertility, and mortality, recruitment, height increase and density of woody individuals were monitored for five seasons. Mechanical clearing of the woody vegetation increased soil fertility, except total nitrogen, and that explained the dramatic response in grass production that endured for the first four seasons since clearing. The periphery and ex-bushclump zones were characterised by increased colonization of Panicum maximum while there were no changes in frequencies of other key herbaceous species in all vegetation zones. Acacia karroo and woody "weeds" (Solanum incanum, Berkheya bipinnatifida) established from seed while all other woody species recruited mainly from coppicing. Seedling recruitment and resprouting resulted in high densities of woody stems and individuals by the second season after clearing when compared with the pre-clearing levels. Oversowing cleared areas with pasture grasses did not only increase grass production but also reduced the density of coppicing woody plants and "weeds". Chemical treatments mirrored the controls in terms of grass production, except during a very wet season, and species composition. Although encroaching woody species (e.g. Scutia myrtina, Maytenus heterophylla and Trimeria trinervus) were susceptible to poisoning, woody plant density was not reduced. Multi-stemmed woody individuals were resistant to poisoning. Fire and goats kept most coppicing woody plants short, less than half a metre, after three seasons of browsing and also improved grass production in the dense bushclumps suggesting that these clumps were being opened up. However, high browsing pressure forced goats to graze more hence this effect was masked. This study indicates that chemical and mechanical control of bush are economically unsustainable for beef and mutton production, at least in the medium term. Fire and goats are appropriate as a follow-up strategy for retarding woody regrowth, keeping bushclumps open, improving grass production and economic viability of mechanical clearing. / Thesis (M.Sc.)-University of Natal, Pietermaritzburg, 2000.

Temperate grasslands and the biodiversity they sustain are becoming increasingly imperilled. Habitat loss and degradation are considered primary causes of grassland species declines. Since livestock production is a dominant use of remaining temperate grassland, livestock producers and the grazing management decisions they make play a vital role in the recovery of grassland species. In this thesis, I examined social and ecological factors that drive habitat management and abundance of prairie wildlife species so as to contribute to conservation planning for prairie species at risk and their habitat. I focused on grassland songbirds because, of all prairie wildlife species, grassland birds have undergone some of the most dramatic declines in recent years. I employed an interdisciplinary approach, using theory and methodology from ornithology, rangeland management and the human dimension of conservation biology to achieve three objectives: i) to determine the extent to which indices of rangeland health explained variation in grassland songbird abundance for ten grassland bird species, including three species currently listed under Canada’s Species at Risk Act: Sprague’s pipit, McCown’s longspur and Chestnut-collared longspur; ii) to describe livestock producer characteristics, summarize producer awareness of and attitudes towards species at risk and the Canadian Species at Risk Act and evaluate how characteristics, awareness and attitudes affect producer willingness to engage in voluntary stewardship actions that support species at risk conservation; and iii) to explore relationships between select social and ecological factors and bird abundance for the three aforementioned grassland bird species at risk to elucidate novel pathways for achieving their conservation. I address the first objective in Chapter 2, where I identify the rangeland health index as a poor predictor of bird abundance and vegetation structure variables, such as litter and vegetation volume, as strong predictors of bird abundance. These findings make a case for further refining the rangeland health index as a tool for biodiversity assessment. In Chapter 3 I achieve the second objective and summarize producer characteristics, awareness and attitudes towards species. I also identify awareness, attitudes and rangeland management learning approach as important to producer willingness to support species at risk recovery. I accomplish the third objective in Chapter 4, where I present results of a structural equation model that upholds bird-habitat relationships identified in Chapter 2 and distinguishes management jurisdiction, size of land holdings and attitudes as important social factors to consider in conservation planning. Chapters 2 and 3 contribute to theory and methodology related to the ecological and social dimensions of grassland bird conservation, respectively. Chapter 4 demonstrates how structural equation models can be used to integrate social and ecological factors, and thereby inform habitat conservation and management. Both social and ecological data presented in this thesis make valuable contributions to producer engagement and habitat management aspects of conservation planning efforts for species at risk in the Milk River watershed of southwestern Saskatchewan. Overall, my findings point to the importance of a joint effort by regional private and public managers to use livestock grazing to create a mosaic of vegetation structure and habitat conditions suitable for the grassland bird community as a whole. This thesis provides a methodological approach that draws on and integrates social and ecological data, methods and concepts, thereby demonstrating how to conduct interdisciplinary research for biological conservation.

Stoats are introduced mammalian carnivores implicated in the decline of several of New Zealand�s endemic species. Most research into stoats in New Zealand has focused on beech forest habitat, especially in years of peak stoat abundance following heavy beech seedfall and peak cohorts of mice. In New Zealand, alpine grasslands occur above the altitudinal limit of beech forest (900-1000 m a.s.l.). Although previous research has shown stoats to be present there, little is known about the ecology of stoats in alpine grasslands. This research aimed to test whether alpine grasslands were a marginal habitat occupied by surplus stoats that had spilled over from beech forest populations, i.e. a sink habitat. The alternative is that alpine grasslands are a desirable habitat deliberately exploited by stoats. This question was answered using mark-recapture, radio-tracking, diet analysis and a food addition experiment. Another objective was to determine whether nest survival is higher in alpine grassland compared to beech forest and whether stoats are likely to be a frequent predator of ground nests in alpine grasslands relative to other introduced mammals that inhabit them. If nest survival is higher in alpine grassland then alpine grasslands may be a refuge from predation. However, if it is not then it is important for management to know the relative risk posed by stoats compared with other predators. An artificial nest experiment was used to answer these questions. This research was undertaken during two years of low to intermediate beech seedfall and therefore provided an opportunity to look at the ecology of stoats in a New Zealand National Park outside years of peak abundance. The principal study site for this research was the Borland Valley, Fiordland National Park. Compositional analysis showed that stoats in alpine grassland selected for it over adjacent beech forest. The range cores of these stoats were high up in alpine grassland away from the ecotone with beech forest. Stoats occurred at similar densities in alpine grasslands as they did in beech forest and observed survival was similar between the two habitats (with the exception of 2004 when it may have been higher in alpine grassland). The most frequent prey of stoats inhabiting beech forest were birds and mice. Although stoats in alpine grasslands also ate birds and mice their most frequent prey were ground weta and hare. Food addition appeared to cause diet switching but did not reduce the distances moved by stoats, suggesting that other factors may be more important in regulating their summer home range size in alpine grasslands. All of these factors lead to the conclusion that alpine grasslands in the Borland are not a marginal habitat for stoats, but may instead be a desirable one. Artificial nests had a higher probability of survival in alpine grassland compared to adjacent beech forest, but survival was too low to support the idea that alpine grasslands are a refuge. Stoats were the most frequent predator of artificial nests in both habitats, but 95 % confidence intervals overlapped the predation rate by possums, which was also high. These findings illustrate the need for a comprehensive landscape approach to stoat control in montane National Parks, for two reasons: 1) endemic biodiversity in alpine grasslands may be under threat from stoat predation, 2) alpine grasslands may act as a source for dispersing stoats that reinvade lowland stoat control areas. In the absence of heavy beech seedfall and peak mouse abundance, stoats occurred at densities of around 1 km⁻� in both habitats and there was recruitment into these populations. This raises the important question: What regulates the distribution and abundance of stoats in years of low beech seedfall and low mouse abundance? In these years birds, ground weta and hares may be as important as mice are in years of peak abundance following heavy beech seedfall.