Understanding landscape dynamics over thousand years : combining field and model work : with case study in the Drakensberg foothill, KwaZulu-Natal, South Africa

The title of this thesis is “Understanding landscape dynamics over thousands of years : combining field and model work, with a case study in the Drakensberg Foothills, KwaZulu-Natal, South Africa”. As the title clearly states, the overall objective is an increased knowledge of landscape dynamics through the combination of fieldwork and landscape evolution modelling. <br/> Fieldwork is the topic of Chapter 2. The 50 kilo-annum (ka) landscape evolution of the research area in Okhombe valley in the Drakensberg Foothills is studied. Results are presented from extensive fieldwork in Okhombe valley, combined with laboratory work. <br/> Starting around 50 ka and continuing until around 30 ka, with cooler temperatures and more rainfall than at present, the slow processes of solifluction and creep transported material from the steep upper slopes of the research area to the concave areas that were immediately downstream. At least two major mudflow events partly or completely covered the solifluction deposits at the end of this period, around 29 ka. When temperatures and rainfall decreased toward the Last Glacial Maximum, grassland was likely replaced by denser shrubland. Overland flow and water erosion were inhibited. At the onset of warmer and wetter climate around 15 ka, shrubby vegetation retreated to higher altitudes and Okhombe valley was again covered with grassland. This decrease in vegetation cover, together with increased rainfall, resulted in higher rates of fluvial redistribution. Presently, erosion is still widespread in the area. <br/> The knowledge of landscape evolution was put to the test in a landscape evolution model in Chapter 5. Chapters 3 and 4 prepared the LAPSUS model for this task by discussing two important aspects of landscape evolution modelling. <br/> Chapter 3 presents a method to deal with an important conceptual and technical issue in long-term landscape evolution modelling. Conventional models consider depressions in Digital Elevation Models (DEMs) spurious, and remove them before modelling. Long-term multi-process landscape evolution models predict depressions, that therefore must be considered non-spurious. A method is detailed that allows these models to identify and include these depressions in dynamic landscapes. Identification first finds sinks, then adds neighbouring cells to the corresponding depression until a saddle is crossed. Inclusion of depressions in the dynamic landscape led to a procedure to deal with flows of water and sediment into and out of depressions. Depressions can be completely or partly filled with sediment. Partial filling, from each of the neighbouring cells, takes the shape of an above- and below-water delta with user-defined slope. <br/> Chapter 4 discusses ways to more formally list, make and report choices involved in setting-up multi-process landscape evolution models. This discussion is necessary now that models are increasingly combining multiple processes in one study. Choices in model set-up must be made regarding the extent and resolution of time, space and processes. A scheme is presented that can guide workers in making these choices, and tests to determine case-optimal set-ups are discussed using four case studies. <br/> In Chapter 5 , LAPSUS is used with the lessons from Chapters 3 and 4 in mind, to test the landscape reconstruction developed in Chapter 2. Adding to existing process descriptions, the processes of creep, solifluction and biological and frost weathering were developed for LAPSUS. A sensitivity analysis was performed, both for individual processes and for the overall model. Model calibration was trial and error and of qualitative nature. It attempted to simultaneously match model results to fieldwork conclusions for three outputs: zonal process activity over time, relative process activity over time and zonal development of soildepth. After calibration, model results suggested that a very slow wave of sediment moved through the landscape after the onset of the Holocene. Waves of sediment this slow have not been reported before. It is also suggested that erosion following this wave is continuing until today. Chapter 5 also shows that landscape evolution model results allow significant refinements of single-process interpretations of deposits, and can fill in erosional hiatuses in stratigraphical records. <br/> Chapter 6 goes one step further and tests whether the LAPSUS version of Chapter 5 is able to discriminate between landscape responses to stable and changed climate for the next millenium in Okhombe valley. This is an important first step in the use of landscape evolution models in the assessment of the effect of human-induced changing climate. Results of landscape evolution models are, of course, uncertain. This chapter tests the influence of parameter uncertainty, assumes that the influence of uncertainty in process descriptions and model structure is minor, and ignores uncertainty in input values (e.g. climatic records). LAPSUS was run hundreds of times, using random parameter values drawn from their joint probability distributions for three levels of assumed uncertainty and for stable and changed climate. Results indicate that LAPSUS can discriminate between the two climate scenarios in most cases, even at the highest level of parameter uncertainty. An explorative, uncertain and relative conclusion about changes in landscape evolution as a result of climate change can be drawn: erosion will likely be stronger in the concave positions, and deposition will likely be stronger further downstream than under stable climate. <br/> Chapter 7 combines results of the previous chapters. A subdivision of similar deposits in KwaZulu-Natal in four types is proposed using knowledge about the conditions that resulted in the deposits in Okhombe valley. Then, four innovations in landscape evolution modelling that the work in chapter 3-6 has contributed to, are summarized. These innovations are combined into a proposal for iterative model-fieldwork combinations in geomorphology. Eventually the focus is on the role that landscape evolution models can play in studies of land dynamics, given their inherent complex systems’ properties. <br/>