There exist numerous approaches to index either spatio-temporal or high-dimensional data. None of them is able to efficiently index hybrid data types, thus spatio-temporal and high-dimensional data. As the best high-dimensional indexing techniques are only able to index point-data and not now-relative data and the best spatio-temporal indexing techniques suffer from the curse of dimensionality, this thesis introduces the Spatio-Temporal Pyramid Adapter (STPA). The STPA maps spatio-temporal data on points, now-values on the median of the data set and indexes them with the pyramid technique. For high-dimensional and spatio-temporal index structures no generally accepted benchmark exists. Most index structures are only evaluated by custom benchmarks and compared to a tiny set of competitors. Benchmarks may be biased as a structure may be created to perform well in a certain benchmark or a benchmark does not cover a certain speciality of the investigated structures. In this thesis, the Interface Based Performance Comparison (IBPC) technique is introduced. It automatically generates test sets with a high code coverage on the system under test (SUT) on the basis of all functions defined by a certain interface which all competitors support. Every test set is performed on every SUT and the performance results are weighted by the achieved coverage and summed up. These weighted performance results are then used to compare the structures. An implementation of the IBPC, the Performance Test Automation Framework (PTAF) is compared to a classic custom benchmark, a workload generator whose parameters are optimized by a genetic algorithm and a specific PTAF alternative which incorporates the specific behavior of the systems under test. This is done for a set of two high-dimensional spatio-temporal indices and twelve variants of the R-tree. The evaluation indicates that PTAF performs at least as good as the other approaches in terms of minimal test cases with a maximized coverage. Several case studies on PTAF demonstrate its widespread abilities.

The monitoring of continuous phenomena like temperature, air pollution, precipitation, soil moisture etc. is of growing importance. Decreasing costs for sensors and associated infrastructure increase the availability of observational data. These data can only rarely be used directly for analysis, but need to be interpolated to cover a region in space and/or time without gaps. So the objective of monitoring in a broader sense is to provide data about the observed phenomenon in such an enhanced form. Notwithstanding the improvements in information and communication technology, monitoring always has to function under limited resources, namely: number of sensors, number of observations, computational capacity, time, data bandwidth, and storage space. To best exploit those limited resources, a monitoring system needs to strive for efficiency concerning sampling, hardware, algorithms, parameters, and storage formats. In that regard, this work proposes and evaluates solutions for several problems associated with the monitoring of continuous phenomena. Synthetic random fields can serve as reference models on which monitoring can be simulated and exactly evaluated. For this purpose, a generator is introduced that can create such fields with arbitrary dynamism and resolution. For efficient sampling, an estimator for the minimum density of observations is derived from the extension and dynamism of the observed field. In order to adapt the interpolation to the given observations, a generic algorithm for the fitting of kriging parameters is set out. A sequential model merging algorithm based on the kriging variance is introduced to mitigate big workloads and also to support subsequent and seamless updates of real-time models by new observations. For efficient storage utilization, a compression method is suggested. It is designed for the specific structure of field observations and supports progressive decompression. The unlimited diversity of possible configurations of the features above calls for an integrated approach for systematic variation and evaluation. A generic tool for organizing and manipulating configurational elements in arbitrary complex hierarchical structures is proposed. Beside the root mean square error (RMSE) as crucial quality indicator, also the computational workload is quantified in a manner that allows an analytical estimation of execution time for different parallel environments. In summary, a powerful framework for the monitoring of continuous phenomena is outlined. With its tools for systematic variation and evaluation it supports continuous efficiency improvement.

The ability of human beings to perceive the environment around them with their visual system is called motion perception. This means that the attention of our visual system is primarily focused on those objects that are moving. The property of human motion perception is used in this dissertation to infer human activity from data using artificial neural networks. One of the main aims of this thesis is to discover which modalities, namely RGB images, optical flow and human keypoints, are best suited for HAR in omnidirectional data. Since these modalities are not yet available for omnidirectional cameras, they are synthetically generated and captured with an omnidirectional camera. During data generation, a distinction is made between synthetically generated omnidirectional data and a real omnidirectional dataset that was recorded in a Living Lab at Chemnitz University of Technology and subsequently annotated by hand. The synthetically generated dataset, called OmniFlow, consists of RGB images, optical flow in forward and backward directions, segmentation masks, bounding boxes for the class people, as well as human keypoints. The real-world dataset, OmniLab, contains RGB images from two top-view scenes as well as manually annotated human keypoints and estimated forward optical flow. In this thesis, the generation of the synthetic and real-world datasets is explained. The OmniFlow dataset is generated using the 3D rendering engine Blender, in which a fully configurable 3D indoor environment is created with artificially textured rooms, human activities, objects and different lighting scenarios. A randomly placed virtual camera following the omnidirectional camera model renders the RGB images, all other modalities and 15 predefined activities. The result of modelling the 3D indoor environment is the OmniFlow dataset. Due to the lack of omnidirectional optical flow data, the OmniFlow dataset is validated using Test-Time Augmentation (TTA). Compared to the baseline, which contains Recurrent All-Pairs Field Transforms (RAFT) trained on the FlyingChairs and FlyingThings3D datasets, it was found that only about 1000 images need to be used for fine-tuning to obtain a very low End-point Error (EE). Furthermore, it was shown that the influence of TTA on the test dataset of OmniFlow affects EE by about a factor of three. As a basis for generating artificial keypoints on OmniFlow with action labels, the Carnegie Mellon University motion capture database is used with a large number of sports and household activities as skeletal data defined in the BVH format. From the BVH-skeletal data, the skeletal points of the people performing the activities can be directly derived or extrapolated by projecting these points from the 3D world into an omnidirectional 2D image. The real-world dataset, OmniLab, was recorded in two rooms of the Living Lab with five different people mimicking the 15 actions of OmniFlow. Human keypoint annotations were added manually in two iterations to reduce the error rate of incorrect annotations. The activity-level evaluation was investigated using a TSN and a PoseC3D network. The TSN consists of two CNNs, a spatial component trained on RGB images and a temporal component trained on the dense optical flow fields of OmniFlow. The PoseC3D network, an approach to skeleton-based activity recognition, uses a heatmap stack of keypoints in combination with 3D convolution, making the network more effective at learning spatio-temporal features than methods based on 2D convolution. In the first step, the networks were trained and validated on the synthetically generated dataset OmniFlow. In the second step, the training was performed on OmniFlow and the validation on the real-world dataset OmniLab. For both networks, TSN and PoseC3D, three hyperparameters were varied and the top-1, top-5 and mean accuracy given. First, the learning rate of the stochastic gradient descent (Stochastic Gradient Descent (SGD)) was varied. Secondly, the clip length, which indicates the number of consecutive frames for learning the network, was varied, and thirdly, the spatial resolution of the input data was varied. For the spatial resolution variation, five different image sizes were generated from the original dataset by cropping from the original dataset of OmniFlow and OmniLab. It was found that keypoint-based HAR with PoseC3D performed best compared to human activity classification based on optical flow and RGB images. This means that the top-1 accuracy was 0.3636, the top-5 accuracy was 0.7273 and the mean accuracy was 0.3750, showing that the most appropriate output resolution is 128px × 128px and the clip length is at least 24 consecutive frames. The best results could be achieved with a learning rate of PoseC3D of 10-3. In addition, confusion matrices indicating the class-wise accuracy of the 15 activity classes have been given for the modalities RGB images, optical flow and human keypoints. The confusion matrix for the modality RGB images shows the best classification result of the TSN for the action walk with an accuracy of 1.00, but almost all other actions are also classified as walking in real-world data. The classification of human actions based on optical flow works best on the action sit in chair and stand up with an accuracy of 1.00 and walk with 0.50. Furthermore, it is noticeable that almost all actions are classified as sit in chair and stand up, which indicates that the intra-class variance is low, so that the TSN is not able to distinguish between the selected action classes. Validated on real-world data for the modality keypoint the actions rugpull (1.00) and cleaning windows (0.75) performs best. Therefore, the PoseC3D network on a time-series of human keypoints is less sensitive to variations in the image angle between the synthetic and real-world data than for the modalities RGB images and optical flow. The pipeline for the generation of synthetic data with regard to a more uniform distribution of the motion magnitudes needs to be investigated in future work. Random placement of the person and other objects is not sufficient for a complete coverage of all movement magnitudes. An additional improvement of the synthetic data could be the rotation of the person around their own axis, so that the person moves in a different direction while performing the activity and thus the movement magnitudes contain more variance. Furthermore, the domain transition between synthetic and real-world data should be considered further in terms of viewpoint invariance and augmentation methods. It may be necessary to generate a new synthetic dataset with only top-view data and re-train the TSN and PoseC3D. As an augmentation method, for example, the Fourier Domain Adaption (FDA) could reduce the domain gap between the synthetically generated and the real-world dataset.:1 Introduction 2 Theoretical Background 3 Related Work 4 Omnidirectional Synthetic Human Optical Flow 5 Human Keypoints for Pose in Omnidirectional Images 6 Human Activity Recognition in Indoor Scenarios 7 Conclusion and Future Work A Chapter 4: Flow Dataset Statistics B Chapter 5: 3D Rotation Matrices C Chapter 6: Network Training Parameters

During the last decades “Climate Change” has become a vital topic on national and international political agendas. There it is presented as an irrevocable fact of global impact and thus of universal relevance. What has often been neglected are local discourses of marginalized groups and their specific contextualization of “Climate Change” phenomena. The aim of this project, to develop another perspective along these dominant narratives, has resulted in the research question How is social reality reconstructed on the phenomenon of “Climate Change” among the “Emerging Black Farmers” in the Swartland region in Western Cape, South Africa? Taken as an example, “Climate Change Knowledge” is reconstructed through a case study on the information exchange between the NGO Goedgedacht Trust and local small-scale farmers in the post-Apartheid context of on-going political, social, economic and educational transition in South Africa. Applying a constructivist approach, “Climate Change Knowledge” is not understood as an objectively given, but a socially constructed “reality” that is based on the interdependency of socio-economic conditions and individual assets, including language skills and language practice, sets of social norms and values, as well as strategies of knowledge transfer. The data set consists of qualitative data sources, such as application forms and interview material, which are triangulated. The rationale of a multi-layered data analysis includes a discursive perspective as well as linguistic and ethical “side perspectives”. Epistemologically, the thesis is guided by assumptions of complexity theory, framing knowledge around “Climate Change” as a fluid, constantly changing system that is shaped by constant intra- and inter-systemic exchange processes, and characterized by non-linearity, self-organization and representation of its constituents. From this point of departure, a theoretical terminology has been developed, which differentiates between symbols, interrelations, contents and content clusters. These elements are located in a system of spatio-temporal orientation and embedded into a broader (socio-economic) context of “historicity”. Content clusters are remodelled with the help of concept maps. Starting from that, a local perspective on “Climate Change” is developed, adding an experiential notion to the global narratives. The thesis concludes that there is no single reality about “Climate Change” and that the farmers’ “Climate Change Knowledge” highly depends on experiential relativity and spatio-temporal immediacy. Furthermore, analysis has shown that the system’s historicity and social manifestations can be traced in the scope and emphasis of the content clusters discussed. Finally the thesis demonstrates that characteristics of symbols, interconnections and contents range between dichotomies of direct and indirect, predictable versus unpredictable, awareness and negligence or threat and danger, all coexisting and creating a continuum of knowledge production.

Mit dieser Arbeit ist es erstmalig gelungen, funktionelle MRT und Dipol-Quellenlokalisation in einer Weise zu kombinieren, die es erlaubt, ein und dieselbe kortikale Aktivität simultan mit beiden Verfahren zu erfassen. Insbesondere wurde dies durch (a) Korrektur eines vom Tomographen induzierten Artefaks in den EKPs und (b) durch eine deutliche Verbesserung des experimentellen Designs, und damit einer wesentlich effektiveren Nutzung von EEG und fMRT-Messzeit erreicht. So wurde es dadurch möglich, mit beiden Methoden die kortikale Aktivität einer Einzelpulsstimulation noch aufzulösen. Eine wesentliche Voraussetzung für die simultane Kombination beider Verfahren: Aufgrund der sehr verschiedenen Latenzen von elektrophysiologischer (< 1ms) und vaskulärer (SII->ant. Inseln und medialeWand) in Übereinstimmung mit der Literatur (Forss et al., 1996; Mauguiere et al., 1997b) blieb. So ergab die Quellenlokalisation für die Wahlreaktionsaufgabe fünf Dipole innerhalb des Gehirns, welche mittels Koregistrierung den Aktivierungen des primären somatosensorschen Kortex (20 - 140ms), des sekundären somatosensorischen Kortex (50 - 150ms), der beiden anterioren Inseln (80 - 140ms) und des supplementär-motorischen Region (90 - 140ms, 220 - 270ms) aus der funktionellen MRT zugeordnet wurden. Durch einen Vergleich der Aktivierungsmuster von Wahl- und Einfachreaktionsaufgabe jeweils in der Dipol-Quellenanalyse und in der funktionellen MRT konnten weitere Belege dafür gefunden werden, daß, wie in der Literatur postuliert (Romo and Salinas, 2001), der kontralaterale sekundäre somatosensorische Kortex an der Kategorisierung somatosensorischer Stimulusattribute beteiligt ist. So ergab ein Vergleich der Dipolzeitverläufe für Wahl- und Einfachreaktionsaufgabe lediglich für den Dipol im kontralateral somatosensorischen Kortex im Intervall 57-62 ms nach Stimulusapplikation einen signifikant unterschiedlichen Aktivierungsverlauf (p < 0,001). Übereinstimmend zeigte die funktionelle MRT für die Wahlreaktionsaufgabe neben einer stärkeren Aktivierung der SMA eine hochsignifikant stärkere Aktivierung im Areal des kontralateralen sekundären somatosensorischen Kortex (p-cluster < 0,001). / In this study, we have shown that it is feasible to perform dipole source analysis and fMRI based on the same neuronal activity associated with somatosensory categorization. This was possible by reduction of scanner-induced baseline artifact interfering with the ERPs as well as an optimized experimental protocol for interleaved EEG and fMRI acquisition. We consider this study to be a further step toward imaging brain activity simultaneously at high spatial and temporal resolution. Since an event-related protocol with a single brief pulse stimulation paradigm was successfully employed, this approach seems to be suitable for the investigation of cognitive tasks. By further technical improvements also the exploration of brain activity in single subjects might become possible, opening the field of clinical applications. In particular for the characterization of irregular and nonreproducible events, a substantial contribution of combined EEG–fMRI studies toward a more detailed understanding of physiological processes underlying cerebral activations is expected.

Breitstreifenlaser haben eine breite Emissionsapertur, die es ermöglicht eine hohe Ausgangsleistung zu erreichen. Gleichzeitig führt sie jedoch zu einer Verringerung der lateralen Strahlqualität und zu ihrem nicht-stationären Verhalten. Forschung in diesem Gebiet ist anwendungsgetrieben und somit ist das Hauptziel eine Erhöhung der Brillanz, die sowohl die Ausgangsleistung als auch die laterale Strahlqualität beinhaltet. Um die zugrunde liegenden raumzeitlichen Phänomene zu verstehen und dieses Wissen zu nutzen, um die Kosten der Brillanz-Optimierung zu minimieren, ist ein selbst-konsistentes Simulationstool notwendig, welches die wichtigsten Prozesse beinhaltet. Zunächst wird in dieser Arbeit ein quasi-dreidimensionales elektro-optisch-thermisches Model präsentiert, welches wesentliche qualitative Eigenschaften von realen Bauteilen gut beschreibt. Zeitabhängige Wanderwellen-Gleichungen werden genutzt, um die inhärent nicht-stationären optischen Felder zu beschreiben, welche an eine Ratengleichung für die Überschussladungsträger in der aktiven Zone gekoppelt sind. Das Model wird in dieser Arbeit um eine Injektionsstromdichte erweitert, die laterale Stromspreizung und räumliches Lochbrennen korrekt beschreibt. Des Weiteren wird ein Temperaturmodel präsentiert, das kurzzeitige lokale Aufheizungen in der Nähe der aktiven Zone und die Formierung einer stationären Temperaturverteilung beinhalten. Im zweiten Teil wird das beschriebene Modell genutzt, um die Gründe von Brillanz-Degradierung, das heißt sowohl die Ursprünge der Leistungssättigung als auch des nicht diffraktionslimitierten Fernfeldes zu untersuchen. Abschließend werden im letzten Teil Laserentwürfe besprochen, welche die laterale Brillanz verbessern. Hierzu gehört ein neuartiges “Schachbrettlaser” Design, bei dem longitudinal-laterale Gewinn-Verlust-Modulation mit zusätzlicher Phasenanpassung ausgenutzt wird, um eine sehr geringe Fernfeld-Divergenz zu erhalten. / Broad-area lasers are edge-emitting semiconductor lasers with a wide lateral emission aperture that enables high output powers, but also diminishes the lateral beam quality and results in their inherently non-stationary behavior. Research in the area is driven by application and the main objective is to increase the brightness which includes both the output power and lateral beam quality. To understand the underlying spatio-temporal phenomena and to apply this knowledge in order to reduce costs for brightness optimization, a self-consistent simulation tool taking into account all essential processes is vital. Firstly, in this work a quasi-three-dimensional opto-electronic and thermal model is presented, that describes well essential qualitative characteristics of real devices. Time-dependent traveling-wave equations are utilized to describe the inherently non-stationary optical fields, which are coupled to dynamic rate equations for the excess carriers in the active region. This model is extended by an injection current density model to accurately include lateral current spreading and spatial hole burning. Furthermore a temperature model is presented that includes short-time local heating near the active region as well as the formation of a stationary temperature profile. Secondly, the reasons of brightness degradation, i.e. the origins of power saturation and the spatially modulated field profile are investigated and lastly, designs that mitigate those effects that limit the lateral brightness under pulsed and continuous-wave operation are discussed. Amongst those designs a novel “chessboard laser” is presented that utilizes longitudinal-lateral gain-loss modulation and an additional phase tailoring to obtain a very low far-field divergence.

Spatial, temporal and population dynamics have influenced learner enrolments in Vhembe District primary schools in Limpopo, South Africa. Vhembe District primary schools have in recent years experienced closure of some of its primary schools due to declining learner enrolments. The dynamics of demographic factors such as migration, fertility and mortality cause fluctuations in the school-age population over time and across space. Poor economic development, the location of the district and the spatial distribution of primary schools make the situation in this rural-based district even more complex. The communities of Vhembe District are discontent about the state of affairs in the area regarding the provision of education and the closure of schools. The closure of schools destabilises the social cohesion amongst members of the community and disempowers them, while inadequate provision of primary school education makes them feel neglected and robbed of their constitutional right. This study investigated the causes of declining learner enrolment and the effect of the closure of schools on the communities. To achieve the objectives data came from questionnaires at household level and from interviews conducted with educators, circuit managers, officials in the Limpopo Education Department and traditional leaders. The study found that declining fertility and out-migration from the area are responsible for a declining school-age population. This is the reality and the communities of Vhembe District will have to live with it because overall learner enrolments continue to decline. Lack of a planning model in the former Venda territory led to an over-supply of primary schools thus schools in close proximity had to compete for learners. Poorly equipped schools and general development of the area exacerbate the problem and some members of the community perceive education in this district to be inferior. Younger economically active people are increasingly moving to places with better opportunities. This study offers some recommendations to alleviate the problems identified in Vhembe District. Application of a geographical approach to an adaptive strategy considers the natural environment in political, social and economic context. It suggests that education authorities could apply such a strategy to make the schools in rural areas more sustainable / Geography / D. Phil. (Geography)

Spatial, temporal and population dynamics have influenced learner enrolments in Vhembe District primary schools in Limpopo, South Africa. Vhembe District primary schools have in recent years experienced closure of some of its primary schools due to declining learner enrolments. The dynamics of demographic factors such as migration, fertility and mortality cause fluctuations in the school-age population over time and across space. Poor economic development, the location of the district and the spatial distribution of primary schools make the situation in this rural-based district even more complex. The communities of Vhembe District are discontent about the state of affairs in the area regarding the provision of education and the closure of schools. The closure of schools destabilises the social cohesion amongst members of the community and disempowers them, while inadequate provision of primary school education makes them feel neglected and robbed of their constitutional right. This study investigated the causes of declining learner enrolment and the effect of the closure of schools on the communities. To achieve the objectives data came from questionnaires at household level and from interviews conducted with educators, circuit managers, officials in the Limpopo Education Department and traditional leaders. The study found that declining fertility and out-migration from the area are responsible for a declining school-age population. This is the reality and the communities of Vhembe District will have to live with it because overall learner enrolments continue to decline. Lack of a planning model in the former Venda territory led to an over-supply of primary schools thus schools in close proximity had to compete for learners. Poorly equipped schools and general development of the area exacerbate the problem and some members of the community perceive education in this district to be inferior. Younger economically active people are increasingly moving to places with better opportunities. This study offers some recommendations to alleviate the problems identified in Vhembe District. Application of a geographical approach to an adaptive strategy considers the natural environment in political, social and economic context. It suggests that education authorities could apply such a strategy to make the schools in rural areas more sustainable / Geography / D. Phil. (Geography)

One of the major applications of wireless sensor networks is to sense accurate and reliable data from the physical environment with or without a priori knowledge of data statistics. To extract accurate data from the physical environment, we investigate spatial data correlation among sensor nodes to develop data accuracy models. We propose three data accuracy models namely Estimated Data Accuracy (EDA) model, Cluster based Data Accuracy (CDA) model and Distributed Cluster based Data Accuracy (DCDA) model with a priori knowledge of data statistics. Due to the deployment of high density of sensor nodes, observed data are highly correlated among sensor nodes which form distributed clusters in space. We describe two clustering algorithms called Deterministic Distributed Clustering (DDC) algorithm and Spatial Data Correlation based Distributed Clustering (SDCDC) algorithm implemented under CDA model and DCDA model respectively. Moreover, due to data correlation in the network, it has redundancy in data collected by sensor nodes. Hence, it is not necessary for all sensor nodes to transmit their highly correlated data to the central node (sink node or cluster head node). Even an optimal set of sensor nodes are capable of measuring accurate data and transmitting the accurate, precise data to the central node. This reduces data redundancy, energy consumption and data transmission cost to increase the lifetime of sensor networks. Finally, we propose a fourth accuracy model called Adaptive Data Accuracy (ADA) model that doesn't require any a priori knowledge of data statistics. ADA model can sense continuous data stream at regular time intervals to estimate accurate data from the environment and select an optimal set of sensor nodes for data transmission to the network. Data transmission can be further reduced for these optimal sensor nodes by transmitting a subset of sensor data using a methodology called Spatio-Temporal Data Prediction (STDP) model under data reduction strategies. Furthermore, we implement data accuracy model when the network is under a threat of malicious attack.

Das Amazonasgebiet hat in den letzten Jahrzehnten eine Intensivierung der menschlichen Aktivitäten erfahren, die in Verbindung mit häufigen schweren Dürren die Umwelt anfälliger für Brände gemacht hat. In dieser Dissertation wurden Fernerkundungsdaten analysiert, um die räumlich-zeitliche Verteilung der Feuer in den letzten 20 Jahren im brasilianischen Amazonasgebiet umfassend zu untersuchen und die verschiedenen Brandursachen zu entschlüsseln. (I) Die erste Forschungsarbeit wertete die Verteilung der verbrannten Fläche aus und zeigte, dass die meisten Brände auf bewirtschafteten Weiden und in den immergrünen Tropenwäldern auftraten, was die Behauptung stützt, dass ihr Auftreten stark auf anthropogene Landnutzungsänderungen reagiert. Die Ergebnisse zeigten auch, dass weder Entwaldung noch Walddegradierung mit Waldbränden korrelierte, wohl aber Feuer, die auf Weiden oder Ackerflächen gelegt wurden und in den angrenzenden Wald übergesprungen sind. (II) Die zweite Forschungsarbeit analysierte einzelne Brände, die durch den auf komplexen Netzwerken basierenden FireTracks-Algorithmus identifiziert wurden. Der Algorithmus wurde verwendet, um Feuerregime für sechs verschiedene Landnutzungsklassen zu ermitteln. Die integrierte Größe, Dauer, Intensität und Ausbreitungsrate dieser räumlich-zeitlichen Brandcluster in den verschiedenen Landnutzungstypen zeigte auf, wie seltene Waldbrände, die natürlicherweise nicht in immergrünen tropischen Wäldern vorkommen, sich zu einem Feuerregime entwickelten, das für Savannenbrände typisch ist. (III) Die dritte Forschungsarbeit analysierte extreme, d. h. die intensivsten Einzelfeuer in immergrünen tropischen Wäldern, und zeigte deren großen Anteil an der insgesamt verbrannten Waldfläche. Während der globale Klimawandel das Potenzial hat, die Trockenheit zu verstärken, sind die anthropogenen Ursachen der Waldzerstörung die Zündquellen, die die Verteilung extremer Brände in den empfindlichen tropischen Wäldern bestimmen. / The Amazon region has experienced an intensification of human activities in the last decades, which combined with frequent severe droughts has led to an environment more susceptible to fire. Remotely sensed data is employed to comprehensively analyse the spatio-temporal fire distribution in the Brazilian Legal Amazon over the past 20 years to disentangle the diverse fire drivers in the region. Special focus is given to burned tropical evergreen forests.  (I) The evaluation of the burned area distribution revealed that most of it occurred in pastures and tropical evergreen forests, supporting the claim that fire incidence responds strongly to anthropogenic land-use changes. The results also showed that neither deforestation nor degradation correlated with forest fires, but escaping fires from pastures and agriculture do. (II) The analysis of individual fires identified by the complex networks based FireTracks algorithm led to the characterization of six different land cover-dependent fire regimes (fire size, duration, intensity, and rate of spread), which uncovered how evergreen forest fires have escalated from being naturally rare to showing characteristics more typical of savanna fires. (III) The analysis of extreme (most intense) fires in evergreen forests showed their large contribution to the total forest burned. While global climate change has the potential to increase drought conditions, anthropogenic drivers of forest degradation provide the ignition sources that determine extreme fire distribution in the tropical forests. The findings call for the development of control and monitoring plans to prevent fires from escaping from managed lands into forests, better management techniques to support effective land use and ecosystem management, targeting forest degradation in addition to deforestation, and considering the human factor in fire ignition and spread in Dynamic Global Vegetation Models in order to reduce uncertainty in fire regime projections.