The successful analysis of LiDAR data for archaeological research requires an evaluation of effects of different vegetation types and the use of adequate visualization techniques for the identification of archaeological features. The Ceibal-Petexbatun Archaeological Project conducted a LiDAR survey of an area of 20 x 20 km around the Maya site of Ceibal, Guatemala, which comprises diverse vegetation classes, including rainforest, secondary vegetation, agricultural fields, and pastures. We developed a classification of vegetation through object-based image analysis (OBIA), primarily using LiDAR-derived datasets, and evaluated various visualization techniques of LiDAR data. We then compared probable archaeological features identified in the LiDAR data with the archaeological map produced by Harvard University in the 1960s and conducted ground-truthing in sample areas. This study demonstrates the effectiveness of the OBIA approach to vegetation classification in archaeological applications, and suggests that the Red Relief Image Map (RRIM) aids the efficient identification of subtle archaeological features. LiDAR functioned reasonably well for the thick rainforest in this high precipitation region, but the densest parts of foliage appear to create patches with no or few ground points, which make the identification of small structures problematic.

O crescimento acelerado das cidades e os reflexos desse aumento das populações urbanas é preocupação constante na atualidade. Nesse processo, o surgimento de ocupações precárias, especialmente nas regiões metropolitanas, torna-se uma das características mais explicitas, caracterizando a própria lógica de ocupação, uso e direito desigual ao território. O monitoramento dessas áreas, sua formação e expansão, são uma necessidade crescente, em diversos locais no mundo, visto que a inclusão dessas áreas à cidade formal é tido como gatilho para a melhoria das condições de vida de mais de 100 milhões de pessoas que vivem em favelas no mundo todo, como colocam as Metas de Desenvolvimento do Milênio propostas pela Organização das Nações Unidas. Contudo, para que os habitantes das favelas sejam atendidos em seu direito a uma vida digna, faz-se necessário seu conhecimento e principalmente quantas são e onde estão. Um importante instrumento, com relação benéfica entre tempo de aquisição, custo de aplicação e possibilidade de replicabilidade e transferência de conhecimento é o uso de dados de Sensoriamento Remoto. Estes possibilitam o estabelecimento de metodologias através de procedimentos de detecção de feições e classificação do uso do solo, para identificação dessas áreas. Não obstante, os métodos de classificação clássicos quando aplicados a imagens de altíssima resolução espacial não conseguem extrair de forma satisfatória, em determinados casos, informações para uso intraurbano. Nesse ínterim surgem novos paradigmas de classificação de imagens como a Análise Orientada a Objeto, onde o processo de classificação parte do objeto geográfico definido a partir da segmentação da imagem, aproximando o objeto de feições do mundo real. Sobre estes objetos é possível a aplicação de regras de pertinência e de contexto através de linguagens e softwares específicos que permitem a transposição do conhecimento humano de fotointerpretação relação contextual para o meio computacional. Este trabalho objetivou avaliar o uso desta técnica de classificação para a detecção e mapeamento de favelas no município de Taboão da Serra/SP, utilizando dados auxiliares para a caracterização destas áreas e seus graus e tipos de precariedade. Os resultados demonstram a validade da aplicação da técnica. / The accelerated growth of the cities and the reflections of the increase of the urban population has been a constant concern nowadays. In this process, the occurrence of precarious occupancies, mainly in the metropolitan regions, has become one of the most explicit characteristics, describing the logic of occupancy itself, unequal use and right to the territory. The monitoring of these areas, their lineup and expansion, are an increasing need in several places in the world, as the inclusion of these areas in the formal city is considered a trigger for the living conditions improvement of over 100 million people who live in slums all over the world, as the Developments Goals of the Millennium proposed by the United Nations Organization. However, in order to meet the rights to a dignified life of the slums inhabitants, it is necessary to know about them mainly their number and where they are. An important tool related to the beneficial relation among the acquisition time, application cost and possibility of applying again, and transference of knowledge is the use of data from Remote Sensing. These data make it possible to establish the methodologies through the detection of features procedures and classification of the land use for these areas identification. Nevertheless the classical methods of classification cannot obtain, in certain cases, information on the interurban use, in a satisfactory way. In the interim, new paradigms of images classification appear like the Object Based Image Analysis (OBIA) which goes from the defined geographic object to the image segmentation, approaching the object to features of the real world. The application of pertinent rules and context over these objects is possible through specific languages and softwares that allow the transference of human knowledge of photo interpretation and contextual relation to the computing environment. This work aimed at evaluating the use of this classification technique for detection and zoning of slums in Taboão da Serra/SP town using supporting data for the areas characterization, its grades and kinds of precarious conditions. The results show the validity of the technique application.

A geophysical survey was conducted at David Crockett Birthplace State Park (40GN205, 40GN12) using ground-penetrating radar (GPR) and magnetometry. The data indicated multiple levels of occupation that were investigated by Phase II and Phase III archaeological excavations. New cultural components were discovered, including the remnants of a Protohistoric Native American structure containing European glass trade beads and Middle Woodland artifacts that suggest trade with Hopewell groups from Ohio. A circular Archaic hearth was uncovered at one meter below surface and similar deep anomalies were seen in the GPR data at this level. A semi-automated object-based image analysis (OBIA) was implemented to extract Archaic circular hearths from GPR depth slices using user-defined spatial parameters (depth, area, perimeter, length to width ratio, and circularity index) followed by manual interpretation. This approach successfully identified sixteen probable hearths distributed across the site in a semi-clustered pattern.

Wetlands are vital ecosystems providing essential services to both humans and the environment, yet they face threats from human activities leading to loss and disturbance. This study utilizes remote sensing (RS) methods, including object-based image analysis (OBIA), to map and assess wetland health in Kristianstad’s Vattenrike in the southernmost part of Sweden between 2015 and 2023. Objectives include exploring RS capabilities in detecting wetlands and changes, deriving wetland health indicators, and assessing classification accuracy. The study uses Sentinel-2 imagery, elevation data, and high-resolution aerial images to focus on wetlands along the river Helge å. Detection and classifications were based on Sentinel-2 imagery and elevation data, and the eCognition software was employed. The health assessment was based on the spectral indices Normalized Difference Vegetation Index (NDVI) and Modified Normalized Difference Water Index (mNDWI). Validation was conducted through aerial photo interpretation. The derived classifications demonstrate acceptable accuracy levels and the analysis reveals relatively stable wetland conditions, with an increase in wetland area attributed to the construction of new wetlands. Changes in wetland composition, such as an increase in open meadows and swamp forests, were observed. However, an overall decline in NDVI values across the study area indicates potential degradation, attributed to factors like bare soil exposure and water presence. These findings provide insights into the local changes in wetland extent, composition, and health between the study years.

Buildings are not only central to the day-to-day activities but also serve as critical indicators of urban development and transformation. The automatic extraction of building footprints from high-resolution Remote Sensing Imagery (RSI) has emerged as an important and popular tool in urban studies. It helps to enhance the understanding and management of urban sprawl, urban planning, population estimation, resource allocation, and post-disaster damage assessment. In this context, having an automated and robust building detection model is crucial. Deep Learning (DL) model and Object-Based Image Analysis (OBIA) techniques are the main and commonly used for automated building detection. This study investigates the efficacy of a pre-trained DL model and a rule-based model OBIA techniques in building detection across varied resolutions and geographic settings. Employing orthophotos from Luleå, Gävle, and Stockholm, the research assesses the adaptability and robustness of these methods under image properties and urban densities. The DL model was initially trained on 0.25m resolution data of Sweden by Lantmäteriet (Sweden mapping agency). The rule-based model was developed by applying OBIA techniques on behalf of this study. Models were analyzed through six feature agreement statistics including Critical Success Index (CSI), Precision, and Detection Probability (POD). The findings reveal that the DL model consistently outperformed the OBIA approach across all study areas, particularly at the original 25 cm resolution. Gävle showed superior precision with a CSI of 0.8139 for the DL model against a CSI of 0.7493 for OBIA at 25 cm.  The evaluation was improved by considering 50*50 sq. m subsets and building sizes. These evaluations highlight that building size and urban density significantly influence detection accuracy. Larger (> 2500 sq. m) buildings and less dense areas tend to yield higher accuracy across both detection methods. The DL model exhibited high CSI values for very large buildings (>5500 sq. m) in Gävle, surpassing 0.8, while the detection of very small (< 50 sq. m) buildings remained challenging for both methods. Overall, the pre-trained DL model is very sensitive to resolution changes compared to OBIA. Importantly, both give their best performance at the original resolution while DL is superior than OBIA. A rule-based OBIA model is affected by the geographical characteristics more heavily than a DL model. Both models have their best performance in the area with medium building density when medium to very large buildings exist. This study highlights how big the impact of building size, geographic characteristics, and image resolution on the performance of DL and OBIA techniques. However, further investigation is recommended to draw a strong conclusion regarding the impact of resolution on the model performance.

The interface between protected and communal lands presents certain challenges for wildlife conservation and the sustainability of local livelihoods. This is a particular case in South Africa, where foot-and-mouth disease (FMD), mainly carried by African buffalo (Syncerus caffer) is transmitted to cattle despite a fence surrounding the protected areas.The ultimate objective of this thesis was to improve knowledge of FMD transmission risk by analyzing behavioral patterns of African buffalo and cattle near the Kruger National Park, and by modelling at fine spatial scale the seasonal risk of contact between them. Since vegetation is considered as a primary bottom-up regulator of grazers distribution, I developed fine-scale seasonal mapping of vegetation. With that purpose, I explored the utility of WorldView-2 (WV-2) sensor, comparing object- (OBIA) and pixel-based image classification methods, and various traditional and advanced classification algorithms. All tested methods produced relatively high accuracy results (>77%), however OBIA with random forest and support vector machines performed significantly better, particularly for wet season imagery (93%).In order to investigate the buffalo and cattle seasonal home ranges and resource utilization distributions I combined the telemetry data with fine-scale maps on forage (vegetation components, and forage quality and quantity). I found that buffalo behaved more like bulk feeders at the scale of home ranges but were more selective within their home range, preferring quality forage over quantity. In contrast, cattle selected forage with higher quantity and quality during the dry season but behaved like bulk grazers in the wet season.Based on the resource utilization models, I generated seasonal cost (resistance) surfaces of buffalo and cattle movement through the landscape considering various scenarios. These surfaces were used to predict buffalo and cattle dispersal routes by applying a cumulative resistant kernels method. The final seasonal contact risks maps were developed by intersecting the cumulative resistant kernels layers of both species and by averaging all scenarios. The maps revealed important seasonal differences in the contact risk, with higher risk in the dry season and hotspots along a main river and the weakest parts of the fence. Results of this study can guide local decision makers in the allocation of resources for FMD mitigation efforts and provide guidelines to minimize overgrazing. / Doctorat en Sciences / info:eu-repo/semantics/nonPublished

The objective of this licentiate thesis is to develop novel algorithms and improve existing methods for urban land cover mapping and urban extent extraction using multi-temporal remote sensing imagery. Past studies have demonstrated that synthetic aperture radar (SAR) have very good properties for the analysis of urban areas, the synergy of SAR and optical data is advantageous for various applications. The specific objectives of this research are: 1. To develop a novel edge-aware region-growing and -merging algorithm, KTH-SEG, for effective segmentation of SAR and optical data for urban land cover mapping; 2. To evaluate the synergistic effects of multi-temporal ENVISAT ASAR and HJ-1B multi-spectral data for urban land cover mapping; 3. To improve the robustness of an existing method for urban extent extraction by adding effective pre- and post-processing. ENVISAT ASAR data and the Chinese HJ-1B multispectral , as well as TerraSAR-X data were used in this research. For objectives 1 and 2 two main study areas were chosen, Beijing and Shanghai, China. For both sites a number of multitemporal ENVISAT ASAR (30m C-band) scenes with varying image characteristics were selected during the vegetated season of 2009. For Shanghai TerraSAR-X strip-map images at 3m resolution X-band) were acquired for a similar period in 2010 to also evaluate high resolution X-band SAR for urban land cover mapping. Ten  major landcover classes were extracted including high density built-up, low density built-up, bare field, low vegetation, forest, golf course, grass, water, airport runway and major road. For Objective 3, eleven globally distributed study areas where chosen, Berlin, Beijing, Jakarta, Lagos, Lombardia (northern Italy), Mexico City, Mumbai, New York City, Rio de Janeiro, Stockholm and Sydney. For all cities ENVISAT ASAR imagery was acquired and for cities in or close to mountains even SRTM digital elevation data. The methodology of this thesis includes two major components, KTH-SEG and KTH-Pavia Urban Extractor. KTH-SEG is an edge aware region-growing and -merging algorithm that utilizes both the benefit of finding local high frequency changes as well as determining robustly homogeneous areas of a low frequency in local change. The post-segmentation classification is performed using support vector machines. KTH-SEG was evaluated using multitemporal, multi-angle, dual-polarization ASAR data and multispectral HJ-1B data as well as TerraSAR-X data. The KTH-Pavia urban extractor is a processing chain. It includes: Geometrical corrections, contrast enhancement, builtup area extraction using spatial stastistics and GLCM texture features, logical operator based fusion and DEM based mountain masking. For urban land cover classification using multitemporal ENVISAT ASAR data, the results showed that KTH-SEG achieved an overall accuracy of almost 80% (0.77 Kappa ) for the 10 urban land cover classes both Beijign and Shanghai, compared to eCognition results of 75% (0.71 Kappa) In particular the detection of small linear features with respect to the image resolution such as roads in 30m resolved data went well with 83% user accuracy from KTH-SEG versus 57% user accuracy using the segments derived from eCognition. The other urban classes which in particular in SAR imagery are characterized by a high degree of heterogeneity were classified superiorly by KTH-SEG. ECognition in general performed better on vegetation classes such as grass, low vegetation and forest which are usually more homogeneous. It is was also found that the combination of ASAR and HJ-1B optical data was beneficial, increasing the final classification accuracy by at least 10% compared to ASAR or HJ-1B data alone. The results also further confirmed that a higher diversity of SAR type images is more important for the urban classification outcome. However, this is not the case when classifying high resolution TerraSAR-X strip-map imagery. Here the different image characteristics of different look angles, and orbit orientation created more confusion mainly due to the different layover and foreshortening effects on larger buildings. The TerraSAR-X results showed also that accurate urban classification can be achieved using high resolution SAR data alone with almost 84% for  eight classes around the Shanghai international Airport (high and low density built-up were not separated as well as roads and runways). For urban extent extraction, the results demonstrated that built-up areas can be effectively extracted using a single ENVISAT ASAR image in 10 global cities reaching overall accuracies around 85%, compared to 75% of MODIS urban class and 73% GlobCover Urban class. Multitemporal ASAR can improve the urban extraction results by 5-10% in Beijing. Mountain masking applied in Mumbai and Rio de Janeiro increased the accuracy by 3-5%.The research performed in  this thesis has contributed to the remote sensing community by providing algorithms and methods for both extracting urban areas and identifying urban land cover in a more detailed fashion. / <p>QC 20140625</p>

This thesis analyses the highly structured and densely populated farmland surrounding Kakamega Forest (western Kenya) in a spatially-explicit manner. The interdisciplinary approach combines methodologies and technologies from different scientific disciplines: remote sensing with OBIA, GIS and spatially explicit modelling (geomatics and geographic science) with socio-economic as well as agro-economic considerations (human and social sciences) as well as cartographic science. Furthermore, the research is related to conservation biology (biological sciences). Based on an in-situ ground truthing and visual image interpretation, very high spatial resolution QuickBird satellite imagery covering 466 km² of farmland was analysed using the concept of object-based image analysis (OBIA). In an integrative workflow, statistical analysis and expert knowledge were combined to develop a sophisticated rule set. The classification result distinguishing 15 LULC classes was used alongside with temporally extrapolated and spatially re-distributed population data as well as socio-/agro-economic factors in order to create a spatially-explicit typology of the farmland and to model scenarios of rural livelihoods. The farmland typology distinguishes ten types of farmland: 3 sugarcane types (covering 48% of the area), 3 tea types (30%), 2 transitional types (15%), 1 steep terrain type (2%), and 1 central type (5%). The scenarios consider different developments of possible future yields and prices for the main agricultural products sugarcane, tea, and maize. Out of all farmland types, the ‘marginal sugarcane type’ is best prepared to cope with future problems. Besides a comparably low population density, a high share of land under cultivation of food crops coupled with a moderate cultivation of cash crops is characteristic for this type. As part of the research conducted, several novel methodologies were introduced. These include a new conceptual framework for categorizing parameter optimization studies, the area fitness rate (AFR) as a novel discrepancy measure, the technique of ‘classification-based nearest neighbour classification’ for classes which are difficult to separate from others, and a novel approach for accessing the accuracy of OBIA classifications. Finally, this thesis makes a number of recommendations and elaborates promising starting points for further scientific research. / Die vorliegende Arbeit untersucht räumlich-expliziten das stark strukturierte und dicht besiedelte Agrarland um den Kakamega Wald (Westkenia). Dabei kombiniert der interdisziplinäre Ansatz Methoden und Technologien verschiedener Wissenschaftsbereiche: die Fernerkundung mit der objekt-basierten Bildanalyse (OBIA), GIS und die räumlich-explizite Modellierung (Geoinformatik und Geographie) mit sozio- und agro-ökonomische Aspekten (Human- und Sozialwissenschaft) sowie der Kartographie. Zudem steht die Arbeit in Bezug zum Schutz der biologischen Vielfalt (Biologie). Ausgehend von einer Referenzdatenerfassung vor Ort und einer visuellen Bildinterpretation wurden räumlich sehr hochauflösende QuickBird-Satellitenbilddaten, die 466 km² des Agrarlandes abdecken, mit Hilfe von OBIA ausgewertet. In einem integrativen Ansatz wurden dabei statistische Verfahren und Expertenwissen kombiniert, um einen ausgefeilten Regelsatz zur Klassifizierung zu erzeugen. Das Klassifizierungsergebnis unterscheidet 15 Klassen der Landnutzung bzw. -bedeckung; zusammen mit zeitlich extrapolierten und räumlich neu verteilten Bevölkerungsdaten sowie sozio- und agro-ökonomischen Faktoren ermöglichte es, eine räumlich-explizite Typologie des Agrarlandes zu erstellen und Szenarien zum ländlichen Auskommen zu modellieren. Die Agrarlandtypologie unterscheidet zehn Landtypen: 3 Zuckerrohr-dominierte Typen (48% des Gebietes), 3 Tee-dominierte Typen (30%), 2 Übergangstypen (15%), 1 Typ steilen Geländes (2%) und 1 zentralen Typ (5%). Die Szenarien betrachten mögliche zukünftige Entwicklungen der Erträge und Preise der Hauptanbauarten Zuckerrohr, Tee und Mais. Von allen Agrarlandtypen ist der „marginal Zuckerrohr-dominierte Typ“ am besten gerüstet, um zukünftigen Problemen zu begegnen. Bezeichnend für diesen Typ sind – neben einer vergleichsweise geringen Bevölkerungsdichte – ein hoher Anteil an Nahrungsmittelanbau zusammen mit einem gemäßigten Anbau von exportorientierten Agrarprodukten. Als Teil der Forschungsarbeit werden verschiedene neuartige Methoden vorgestellt, u.a. ein neuer konzeptioneller Rahmen für das Kategorisieren von Studien zur Parameteroptimierung, die „area fitness rate“ (AFR) als neue Messgröße für Flächendiskrepanzen, die klassifikations-basierte Nächster-Nachbar Klassifizierung sowie ein Ansatz zum Bestimmen der Güte von OBIA-Klassifizierungen. Schließlich gibt die Arbeit eine Reihe von Empfehlungen und bietet vielversprechende Ausgangspunkte für weiterführende wissenschaftliche Forschungen.

This thesis analyses the highly structured and densely populated farmland surrounding Kakamega Forest (western Kenya) in a spatially-explicit manner. The interdisciplinary approach combines methodologies and technologies from different scientific disciplines: remote sensing with OBIA, GIS and spatially explicit modelling (geomatics and geographic science) with socio-economic as well as agro-economic considerations (human and social sciences) as well as cartographic science. Furthermore, the research is related to conservation biology (biological sciences). Based on an in-situ ground truthing and visual image interpretation, very high spatial resolution QuickBird satellite imagery covering 466 km² of farmland was analysed using the concept of object-based image analysis (OBIA). In an integrative workflow, statistical analysis and expert knowledge were combined to develop a sophisticated rule set. The classification result distinguishing 15 LULC classes was used alongside with temporally extrapolated and spatially re-distributed population data as well as socio-/agro-economic factors in order to create a spatially-explicit typology of the farmland and to model scenarios of rural livelihoods. The farmland typology distinguishes ten types of farmland: 3 sugarcane types (covering 48% of the area), 3 tea types (30%), 2 transitional types (15%), 1 steep terrain type (2%), and 1 central type (5%). The scenarios consider different developments of possible future yields and prices for the main agricultural products sugarcane, tea, and maize. Out of all farmland types, the ‘marginal sugarcane type’ is best prepared to cope with future problems. Besides a comparably low population density, a high share of land under cultivation of food crops coupled with a moderate cultivation of cash crops is characteristic for this type. As part of the research conducted, several novel methodologies were introduced. These include a new conceptual framework for categorizing parameter optimization studies, the area fitness rate (AFR) as a novel discrepancy measure, the technique of ‘classification-based nearest neighbour classification’ for classes which are difficult to separate from others, and a novel approach for accessing the accuracy of OBIA classifications. Finally, this thesis makes a number of recommendations and elaborates promising starting points for further scientific research.:1. Introduction 2. Geodata and reference data 3. Object-based image analysis (OBIA) 4. Optimization of segmentation parameters 5. Feature selection and threshold determination 6. OBIA classification: rule set development and realisation 7. Classification results 8. Spatial farmland typology 9. Spatially explicit planning scenarios of rural livelihoods 10. Discussion / Die vorliegende Arbeit untersucht räumlich-expliziten das stark strukturierte und dicht besiedelte Agrarland um den Kakamega Wald (Westkenia). Dabei kombiniert der interdisziplinäre Ansatz Methoden und Technologien verschiedener Wissenschaftsbereiche: die Fernerkundung mit der objekt-basierten Bildanalyse (OBIA), GIS und die räumlich-explizite Modellierung (Geoinformatik und Geographie) mit sozio- und agro-ökonomische Aspekten (Human- und Sozialwissenschaft) sowie der Kartographie. Zudem steht die Arbeit in Bezug zum Schutz der biologischen Vielfalt (Biologie). Ausgehend von einer Referenzdatenerfassung vor Ort und einer visuellen Bildinterpretation wurden räumlich sehr hochauflösende QuickBird-Satellitenbilddaten, die 466 km² des Agrarlandes abdecken, mit Hilfe von OBIA ausgewertet. In einem integrativen Ansatz wurden dabei statistische Verfahren und Expertenwissen kombiniert, um einen ausgefeilten Regelsatz zur Klassifizierung zu erzeugen. Das Klassifizierungsergebnis unterscheidet 15 Klassen der Landnutzung bzw. -bedeckung; zusammen mit zeitlich extrapolierten und räumlich neu verteilten Bevölkerungsdaten sowie sozio- und agro-ökonomischen Faktoren ermöglichte es, eine räumlich-explizite Typologie des Agrarlandes zu erstellen und Szenarien zum ländlichen Auskommen zu modellieren. Die Agrarlandtypologie unterscheidet zehn Landtypen: 3 Zuckerrohr-dominierte Typen (48% des Gebietes), 3 Tee-dominierte Typen (30%), 2 Übergangstypen (15%), 1 Typ steilen Geländes (2%) und 1 zentralen Typ (5%). Die Szenarien betrachten mögliche zukünftige Entwicklungen der Erträge und Preise der Hauptanbauarten Zuckerrohr, Tee und Mais. Von allen Agrarlandtypen ist der „marginal Zuckerrohr-dominierte Typ“ am besten gerüstet, um zukünftigen Problemen zu begegnen. Bezeichnend für diesen Typ sind – neben einer vergleichsweise geringen Bevölkerungsdichte – ein hoher Anteil an Nahrungsmittelanbau zusammen mit einem gemäßigten Anbau von exportorientierten Agrarprodukten. Als Teil der Forschungsarbeit werden verschiedene neuartige Methoden vorgestellt, u.a. ein neuer konzeptioneller Rahmen für das Kategorisieren von Studien zur Parameteroptimierung, die „area fitness rate“ (AFR) als neue Messgröße für Flächendiskrepanzen, die klassifikations-basierte Nächster-Nachbar Klassifizierung sowie ein Ansatz zum Bestimmen der Güte von OBIA-Klassifizierungen. Schließlich gibt die Arbeit eine Reihe von Empfehlungen und bietet vielversprechende Ausgangspunkte für weiterführende wissenschaftliche Forschungen.:1. Introduction 2. Geodata and reference data 3. Object-based image analysis (OBIA) 4. Optimization of segmentation parameters 5. Feature selection and threshold determination 6. OBIA classification: rule set development and realisation 7. Classification results 8. Spatial farmland typology 9. Spatially explicit planning scenarios of rural livelihoods 10. Discussion

Thesis (MSc)--Stellenbosch University, 2015. / ENGLISH ABSTRACT: Anthropological influences are threatening the state of many savanna ecosystems in most rural landscapes around the world. Effective monitoring and management of these landscapes requires up to date maps and data on the state of the environment. Degradation data over a range of scales is often not readily available due to a lack of financial resources, time and technical capabilities. The aim of this research was to use a medium resolution multispectral SPOT 5 image from 2010 and Landsat 8 images from 2014 to map habitat intactness in the Bushbuckridge and Kruger National Park (KNP) region. The images were pre-processed and segmented into meaningful image objects using an object based image analysis (OBIA) approach. Five image derivatives namely: brightness, compactness, NIR standard deviation, area and the normalised difference vegetation index (NDVI) were evaluated for their capability to model habitat intactness. A habitat intactness index was generated by combining the five derivatives and rescaling them to a data range of 0 to 10, with 0 representing completely transformed areas, 10 being undisturbed natural vegetation. Field data were collected in October 2014 using a field assessment form consisting of 10 questions related to ecosystem state, in order to facilitate comparisons with the remote sensing habitat intactness index. Both satellite data sets yielded low overall accuracies below 30%. The results were improved by applying a correction factor to the reference data. The results significantly improved with SPOT 5 producing the highest overall accuracy of 62.6%. The Landsat 8 image for May 2014 achieved an improved accuracy of 60.2%. The SPOT 5 results showed to be a better predictor of habitat intactness as it assigned natural vegetation with better accuracy, while Landsat 8 correctly assigned mostly degraded areas. These findings suggest that the method was not easily transferable between the different satellite sensors in this savanna landscape, with a high occurrence of forest plantations and rural settlements too. These areas caused high omission errors in the reference data, resulting in the moderate overall accuracies obtained. It is recommended that these sites be clipped out of the analysis in order to obtain acceptable accuracies for non-transformed areas. The study nevertheless demonstrated that the habitat intactness index maps derived can be a useful data source for mapping general patterns of degradation especially on a regional scale. Therefore, the methods tested in this study can be integrated in habitat mapping projects for effective conservation planning. / AFRIKAANSE OPSOMMING: Antropologiese invloede bedreig die toestand van savanna-ekostelsels in die meeste landelike landskappe regoor die wêreld. Doeltreffende monitering en bestuur van hierdie landskappe vereis op datum kaarte en inligting oor die toestand van die omgewing. Agteruitgangsdata van verskillende skale is dikwels nie geredelik beskikbaar nie weens 'n gebrek aan finansiële hulpbronne, tyd en tegniese vermoëns. Die doel van hierdie navorsing was om ‘n hoë resolusie multispektrale SPOT 5 beeld van 2010 en Landsat 8 beelde van 2014 te gebruik om die habitatongeskondenheid in die Bushbuckridge en Kruger Nasionale Park (KNP) streek te karteer. Die beelde is voorverwerk en gesegmenteer om sinvolle beeldvoorwerpe te skep deur die gebruik van ‘n voorwerp gebaseerde beeldanalise (OBIA) benadering. Vyf beeldafgeleides naamlik: helderheid, kompaktheid, NIR standaardafwyking, area en die genormaliseerde verskil plantegroei-indeks (NDVI) is geëvalueer vir hul vermoë om habitat ongeskondenheid te modelleer. ‘n Habitatongeskondenheidsindeks is gegenereer deur die kombinasie van die vyf afgeleides wat herskaal is na 'n datareeks van 0 tot 10, met 0 om totaal getransformeerde gebiede te verteenwoordig en 10 om ongestoorde natuurlike plantegroei voor te stel. Velddata is versamel in Oktober 2014 met gebruik van 'n veldassesseringsvorm, bestaande uit 10 vrae wat verband hou met die toestand van die ekostelsel, om vergelykings met die afstandswaarneming habitatongeskondenheidsindeks te fasiliteer. Beide satellietdatastelle het lae algehele akkuraatheid onder 30% opgelewer. Die resultate is deur die toepassing van 'n regstellingsfaktor tot die verwysing data verbeter. Die resultate het aansienlik verbeter met SPOT 5 wat die hoogste algehele akkuraatheid van 62.6% gelewer het. Die Landsat 8 beeld vir Mei 2014 bereik 'n verbeterde akkuraatheid van 60.2%. Die SPOT 5 resultate het geblyk om ‘n beter voorspeller van habitatongeskondenheid te wees as gevolg van ‘n beter akkuraatheid vir natuurlike plantegroei, terwyl Landsat meestal gedegradeerde gebiede kon voorspel. Hierdie bevindinge dui daarop dat die metode nie maklik oordraagbaar was tussen die verskillende satelliet sensors in hierdie savanna landskap nie, veral as gevolg van ‘n hoë voorkoms van bosbouplantasies en landelike nedersettings. Hierdie gebiede veroorsaak hoë weglatingsfoute in die verwysing data, wat lei tot gematigde algehele akkuraatheid. Dit word aanbeveel dat hierdie areas gemasker word tydens die ontleding om aanvaarbare akkuraatheid te verkry vir nie-getransformeerde gebiede. Nogtans het die studie getoon dat die afgeleide habitatongeskondenheidsindekskaarte ‘n nuttige bron van data kan wees vir die kartering van algemene patrone van agteruitgang, veral op 'n plaaslike skaal. Daarom kan die getoetsde metodes in die studie in habitatkarteringsprojekte vir doeltreffende bewaring beplanning geïntegreer word. Stellenbosch University https://scholar.sun.ac.za