Towards land change management using ecosystem dynamics and land cover change in rural Eastern Cape

Munch, Zahn (2019-12)

Thesis (PhD)--Stellenbosch University, 2019.

Thesis

ENGLISH ABSTRACT: Land cover change, triggered by natural and anthropogenic land use change, affects ecosystem services provided by grasslands. Woody encroachment into the grasslands is a threat to function and productivity of rangelands, and threaten rural livelihoods, intensified by rising CO2 levels associated with climate change. Processes of change can only effectively be identified after spatial land transition has been revealed and patterns of change quantified. Accurately quantifying the rates and extent of land cover change is the first step in relating underlying land use processes and the environmental effects thereof to land cover change trajectories involving grassland transformation. The study aims to demonstrate how land cover change, in particular woody encroachment influences landscape functions provided by grasslands in the Eastern Cape. The study seeks to determine how accurately land cover transformation can be quantified and modelled using existing datasets that may contain map error and raises the question how the error pattern can affect modelling of future evapotranspiration and carbon storage. A further question is how the drivers of change vary between regions under different land tenure, i.e. dualistic or commercial systems. Systematic land cover change analysis and future land change modelling were used to characterise land cover change trajectories and flows in the landscape. Flows were described using (1) an indicator-based approach, and (2) intensity analysis and change budget. Hypothetical map error was determined for observed and modelled land cover maps. Overall change was partitioned into quantity, exchange and shift disagreement and intensity. The change budget was computed both at catchment and local level. Map error was further investigated using a local geographically weighted method. Local geographically weighted correspondence matrices were constructed to determine spatially explicit probabilities of change and error at catchment level and per land cover class. By consulting the overall allocation difference maps, hotspots of change and probable error were identified for further investigation. Trends in remote sensing-derived biophysical variables were analysed to determine how land cover change would affect the surface energy budget and the carbon cycle, as proxies for water use and rangeland productivity. Primary drivers of landscape modification comprised rangeland degradation, woody encroachment, urbanisation, increased dryland cultivation and commercial afforestation, with the latter concentrated in the commercial catchment. Though grassland persistence still dominated land cover in the landscape, catchments under dualistic land tenure experienced steeper declines in the grassland area. Woody encroachment was also found predominantly in these catchments. Overall accuracy for the input land cover maps were reported as >80%, equating to theoretical land cover change accuracy of 67–72%. Landscape change varied between 18% and 42%, with 19% estimated from direct overlay of land cover maps with 30 m resolution pixels. By applying a multi-resolution aggregation technique, the study showed that lower resolution input data would identify less change in the landscape, mainly because the allocation error diminishes at lower resolutions. For higher change accuracies and reliability, the accuracy of input land cover maps would have to be increased. Hypothetical map error in observed land cover change maps were found to be higher in catchments under dualistic tenure for gaining transitions, whereas losing transitions showed higher error in catchments practicing commercial farming. The hypothetical error accounted for almost 50% of the reported change. The modelled land cover change showed higher allocation disagreement, suggesting that the land change model was not very reliable, particularly for the commercial catchment. Analysis of remotely sensed data products such as albedo, net primary production and evapotranspiration, in combination with land cover change data has led to better understanding of the landscape of the catchments. Though grasslands are predicted to decrease in favour of woody invasive plant species and cultivated land, this study predicted a decrease of 12% and 6% respectively in net carbon storage and water use by vegetation. Information from multiple sources, in both quality and type, were integrated to better understand rangeland productivity degradation and to compare the impact of climate versus land management in the different catchments. Quantifying changes in biophysical parameters can assist scientists and managers in addressing global challenges.

AFRIKAANSE OPSOMMING: Verandering in grondbedekking wat deur natuurlike en antropogeniese verandering in grondgebruik veroorsaak word, beïnvloed die ekostelseldienste wat deur grasvelde gelewer word. Houtagtige indringing van die grasvelde is ʼn bedreiging vir die funksionering en produktiwiteit van weivelde en bedreig landelike lewensbestaan. Hierdie proses word aangehelp deur die toenemende CO2-vlakke wat met klimaatsverandering verband hou. Prosesse van verandering kan eers effektief geïdentifiseer word nadat ruimtelike landoorgang geopenbaar is en patrone van verandering gekwantifiseer is. Die akkurate kwantifisering van die trajek en omvang van grondbedekkingsverandering is die eerste stap om die onderliggende prosesse vir grondgebruik en die omgewingseffekte daarvan aan grondbedekkingstrajeksies te koppel. Laasgenoemde hou direkte verband met die transformasie van grasveld. Hierdie studie het ten doel om te demonstreer hoe grondbedekkingsverandering, veral deur houtagtige indringerplante, die landskapfunksies wat grasvelde in die Oos-Kaap verrig beïnvloed. Die studie poog om te bepaal hoe akkuraat die transformasie van grondbedekking met bestaande datastelle, wat kaartfoute bevat, gekwantifiseer en gemodeller kan word. Daar is onsekerheid oor hoe die foutpatrone in die datastelle die modellering van toekomstige evapotranspirasie en koolstofopberging kan beïnvloed. 'n Verdere navorsingsvraag is hoe die drywers van verandering tussen streke onder verskillende grondbesit, te wete dualistiese of kommersiële stelsels, wissel. Sistematiese ontleding van grondbedekking en toekomstige modellering is gebruik om die trajek en vloei van grondverandering in die landskap te beskryf. Vloei is beskryf met behulp van (1) 'n aanwyser-gebaseerde benadering, en (2) intensiteitsanalise en veranderingsbegroting. Hipotetiese kaartfoute is vir waargenome en gemodelleerde grondbedekkingskaarte bepaal. Algehele verandering is in kwantiteit, wissel en verskuiwingsverskille en intensiteit opgedeel. Die veranderingsbegroting is per opvanggebied sowel as op plaaslike vlak bereken. Kaartfoute is verder met behulp van 'n plaaslike geografies-geweegde metode ondersoek. Plaaslike geografies-geweegde korrespondensie-matrikse is opgestel om ruimtelik-sensitiewe waarskynlikhede vir veranderinge en foute per opvanggebied en grondbedekkingklas te bepaal. Die totale toekenningsverskilkaarte is geraadpleeg om brandpunte van verandering en waarskynlike foute vir verdere ondersoek te identifiseer. Die tendense in biofisiese veranderlikes wat vanaf afstandswaarneming afgely is, is ontleed om te bepaal hoe verandering in grondbedekking die oppervlakte-energiebegroting en die koolstofsiklus, wat watergebruik en graslandproduktiwiteit verteenwoordig, sou beïnvloed. Die resultate het getoon dat weiveldagteruitgang, houtagtige indringing, verstedeliking, verhoogde droëlandverbouing en kommersiële bosbou, met laasgenoemde gekonsentreer in die kommersiële opvanggebied, die primêre drywers van grondbedekkingsverandering was. Alhoewel grasland grondbedekking die landskap steeds oorheers, het dit in opvanggebiede onder dualistiese grondbesit afgeneem. Houtagtige indringing is hoofsaaklik in hierdie opvanggebiede opgemerk. Die algehele akkuraatheid van die inset grondbedekkingskaarte is as >80% gerapporteer, wat teoreties in die konteks van grondbedekkingverandering aan 67-72% gelykstaande is. Landskapverandering het tussen 18% en 42% gewissel, met 19% wat op direkte oorleg van 30 m resolusie grondbedekkingskaarte geraam is. 'n Multi-resolusie-samevoegingstegniek het getoon dat laer resolusie-insetdata minder verandering in die landskap identifiseer, hoofsaaklik omdat die toewysingsfout by laer resolusies verminder. Die akkuraatheid van grondbedekkingskaarte sal verhoog moet word om die akkuraatheid en betroubaarheid van veranderinge te verbeter. Daar is bevind dat hipotetiese kaartfoute in waargenome grondebedekkingsveranderingskaarte hoër was in opvanggebiede onder dualistiese bestuur vir grondbedekkingsklasse wat toegeneem het, terwyl in die kommersiële opgevanggebied groter foute in grondbedekkingsklasse met afnemende oorgange opgetel is. Die hipotetiese fout was verantwoordelik vir byna 50% van die gemelde verandering. Die gemodelleerde gronddekkingverandering het 'n groter toewysigingsverskil getoon, wat daarop dui dat die grondbedekkingsveranderingsmodel nie baie betroubaar was nie, veral nie vir die kommersiële opvanggebied nie. Analise van afstandswaarnemingsdataprodukte soos albedo, netto primêre produksie en evapotranspirasie, in kombinasie met grondbedekkingsveranderingsdata, het gelei tot 'n beter begrip van die landskap in die opvanggebiede. Alhoewel die voorspelling is dat grasvelde ten gunste van houtagtige indringerplantspesies en bewerkte lande sal afneem, het hierdie studie slegs 'n afname van onderskeidelik 12% en 6% in die netto koolstofopberging en watergebruik deur plantegroei voorspel. Inligting uit verskeie bronne, beide in kwaliteit en tipe, is geïntegreer om die agteruitgang van die landskapproduktiwiteit beter te verstaan en om die impak van klimaat op grondbestuur in die verskillende opvanggebiede te vergelyk. Die kwantifisering van veranderinge in biofisiese parameters kan wetenskaplikes en bestuurders help om wêreldwye uitdagings die hoof te bied.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/107081
This item appears in the following collections: