Using landscape classification within complex, heterogeneous systems Carola Cullum and Kevin Rogers Centre for Water in the Environment University of the Witwatersrand Traditional Landscape Classification • Assumes that landscape units and classes are homogenous – Average value applies over whole area – Units are replaceable within classes – A reference site represents a whole class – Can extrapolate knowledge between units – Can upscale/ downscale in a nested hierarchy – Samples can represent the whole system BUT ... landscapes are heterogeneous, complex and dynamic • Every location is unique – History, neighbours, context – Heterogeneity matters! • Not noise or complication • May be as /more important than general trends in determining outcomes • Complexity • Locations change through time Heterogeneity, complexity and dynamics mean that… • Average values do NOT apply equally over a whole unit / class • Reference sites and samples do NOT represent the whole class equally well • It is difficult to – Extrapolate between units – Upscale/ downscale • Units change –definition issues AND... Many different classifications/maps are possible • Many ways of carving up the world – Soil? Veg? Geology? Combination? • We need to decide – Why are these particular units useful? – What factors define the units? – At what spatial and temporal scales? – Why are the similarities that define classes of units important? Does heterogeneity, complexity, dynamics, lack of a priori classification structure…. Mean we have to • Measure everything at every location? • Have multiple models and maps for every purpose? • Have different management plans for every piece of land? We need some practical solutions! Our framework for landscape classification can help ... • For water-controlled ecosystems – Many ecological patterns coupled to water distribution in time and space • Captures repeating landscape patterns – Catenas, drainage networks … • Wide range of applications – Integrates diverse disciplines, terrestrial and aquatic, science and management Landscape hierarchy informs definition of landscape units • Based around catchments • Guides choice of – Scales – Attributes to define units and classes of units • Factors associated with water distribution • Vegetation, Soils, Topography.... PHYSIOGRAPHIC ZONE CATCHMENT CATENAL ELEMENT Heterogeneity demands fuzzy classes • Landscapes do not fit neatly into predefined categories – Transitions, overlaps – Not all units fit neatly into a class • Archetypes not rules – Archetype is an idealised style – Degree of similarity to an archetype determines degree of membership of a class – All classifications not equally right or wrong – Can assess heterogeneity within/ between classes by mapping membership values • Different archetypes in different settings Landscape elements in N’waswitshaka: Southern Granites Physiographic zone Geology and landscape dissection Catchments Hillslopes and Channels Catenal elements in CASSs of different stream orders 1st order CASS 2nd order CASS Catenal elements Koppies Woodier midslopes Grassier toeslopes Woodier crests Grassier midslopes Possible sodic sites Grassier crests Woodier toeslopes Channels and banks 3rd order CASS 4th order CASS 5th order CASS Heterogeneity in assemblages of catenal elements in 1st order catchments in N’waswitshaka Stream Poss sodic site Grassy midslope, woody crest Woodier crest Grassier midslope Koppie Grassier crest Woodier midslope Grassier toeslope Woodier toeslope Channel and bank Woody midslope, woody crest Where does heterogeneity matter? • Low order catchment in N’waswitshaka • Conceptual model for archetype → Hypothesised directions of water flow Woody, Sandy Crest Seepline Grassy, Clayey Midslope Clayey banks, gravel bed channels Where does heterogeneity matter? • What matters to direction and intensity of water flows? – Vegetation, slope, area ++++ – Arrangement of catenal elements – boundaries permeable? – Geology: Dykes, koppies ... – Humans: Roads, dams, gravel pits... – Different arrangement of catenal elements: eg grassy crest, woody midslope • Outlier or new class? How the heterogeneity approach is different Traditional classification Classification under heterogeneity paradigm Hard boundaries Fuzzy boundaries Yes/No class membership Degrees of class membership Class is represented by…. A reference unit / sample Archetype and range of variation Focus on means/ similarity Focus on variance/ dissimilarity Inflexible classes – one size fits all Flexible – adapt to incorporate new learning Thank you!
© Copyright 2026 Paperzz