Refine
Has Fulltext
- yes (7)
Is part of the Bibliography
- yes (7)
Document Type
- Doctoral Thesis (5)
- Master Thesis (2)
Keywords
- Fernerkundung (5)
- Remote Sensing (2)
- remote sensing (2)
- Arctic (1)
- Australien (1)
- Banks Islands (1)
- Bilharziose (1)
- Blattflächenindex (1)
- Deep learning (1)
- Degradation (1)
Institute
Sonstige beteiligte Institutionen
Burn severity was measured within the Mediterranean sclerophyll forests of south-west Western Australia (WA) using remote sensing data from the Moderate Resolution Imaging Spectroradiometer (MODIS). The region of south-west WA is considered as a high fire prone landscape and is managed by the state government’s Department of Conservation and Land Management (CALM). Prescribed fuel reduction burning is used as a management tool in this region. The measurement of burn severity with remote sensing data focused on monitoring the success and impact of prescribed burning and wildfire in this environment. The high temporal resolution of MODIS with twice daily overpasses in this area was considered highly favourable, as opportunities for prescribed burning are temporally limited by climatic conditions. The Normalised Burn Ratio (NBR) was investigated to measure burn severity in the forested area of south-west WA. This index has its heritage based on data from the Landsat TM/ETM+ sensors (Key and Benson, 1999 [1],[2]) and was transferred from Landsat to MODIS data. The measurement principally addresses the biomass consumption due to fire, whereas the change detected between the pre-fire image and the post-fire image is quantified by the ÄNBR. The NBR and the Normalised Difference Vegetation Index (NDVI) have been applied to MODIS and Landsat TM/ETM+ data. The spectral properties and the index values of the remote sensing data have been analysed within different burnt areas. The influence of atmospheric and BRDF effects on MODIS data has been investigated by comparing uncorrected top of atmosphere reflectance and atmospheric and BRDF corrected reflectance. The definition of burn severity classes has been established in a field trip to the study area. However, heterogeneous fire behaviour and patchy distribution of different vegetation structure made field classification difficult. Ground truth data has been collected in two different types of vegetation structure present in the burnt area. The burn severity measurement of high resolution Landsat data was assessed based on ground truth data. However, field data was not sufficient for rigorous validation of remote sensing data. The NBR index images of both sensors have been calibrated based on training areas in the high resolution Landsat image. The burn severity classifications of both sensors are comparable, which demonstrates the feasibility of a burn severity measurement using moderate spatial resolution 250m MODIS data. The normalisation through index calculation reduced atmospheric and BRDF effects, and thus MODIS top of at-mosphere data has been considered suitable for the burn severity measurement. The NBR could not be uniformly applied, as different structures of vegetation influenced the range of index values. Furthermore, the index was sensitive to variability in moisture content. However, the study concluded that the NBR on MODIS data is a useful measure of burn severity in the forested area of south-west WA.
Worldwide, cold regions are undergoing significant alterations due to climate change. Snow, the most widely distributed cold region component, is highly sensitive to climate change. At the same time, snow itself profoundly impacts the Earth’s energy budget, biodiversity, and natural hazards, as well as hydropower management, freshwater management, and winter tourism/sports. Large parts of the cold regions in Europe are mountain areas, which are densely populated because of the various ecosystem services and socioeconomic well-being in mountains. At present, severe consequences caused by climate change have been observed in European mountains and their surrounding areas. Yet, large knowledge gaps hinder the development of effective regional and local adaptation strategies. Long-term and evidence-based regional studies are urgently needed to enhance the comprehension of regional responses to climate change.
Earth Observation (EO) provides long-term consistent records of the Earth’s surface. It is a great alternative and/or supplement to conventional in-situ measurements which are usually time-consuming, cost-intensive and logistically demanding, particularly for the poor accessibility of cold regions. With the assistance of EO, land surface dynamics in cold regions can be observed in an objective, repeated, synoptic and consistent way. Thanks to free and open data policies, long-term archives such as Landsat Archive and Sentinel Archive can be accessed free-of-charge. The high- to medium-resolution remote sensing imagery from these freely accessible archives gives EO-based time series datasets the capability to depict snow dynamics in European mountains from the 1980s to the present. In order to compile such a dataset, it is necessary to investigate the spatiotemporal availability of EO data, and develop a spatiotemporally transferable framework from which one can investigate snow dynamics.
Among the available EO image archives, the Landsat Archive has the longest uninterrupted records of the Earth’s land surface. Furthermore, its 30 m spatial resolution fulfils the requirements for snow monitoring in complex terrains. Landsat data can yield a time series of snow dynamics in mountainous areas from 1984 to the present. However, severe Landsat data gaps have occurred across certain regions of Europe. Moreover, the Landsat Level 1 Precision and Terrain (L1TP) data is scarcer (up to 50% less) in high-latitude mountainous areas than in low-latitude mountainous areas. Given the abovementioned facts, the Regional Snowline Elevation (RSE) is selected to characterize the snow dynamics in mountainous areas, as it can handle cloud obstructions in the optical images. In this thesis, I present a five-step framework to derive and densify RSE time series in European mountains, i.e. (1) pre-processing, (2) snow detection, (3) RSE retrieval, (4) time series densification, and (5) Regional Snowline Retreat Curve (RSRC) production.
The results of the intra-annual RSE variations show a uniquely high variation in the beginning of the ablation seasons in the Alpine catchment Tagliamento, mainly toward higher elevation. As for inter-annual variations of RSE, median RSE increases in all selected catchments, with an average speed of around 4.66 m ∙ a−1 (median) and 5.87 m ∙ a−1 (at the beginning of the ablation season). The fastest significant retreat is observed in the catchment Drac (10.66 m ∙ a−1, at the beginning of the ablation season), and the slowest significant retreat is observed in the catchment Uzh (1.74 m ∙ a−1, at the beginning of the ablation season). The increase of RSEs at the beginning of the ablation season is faster than the median RSEs, whose average difference is nearly 1.21 m ∙ a−1, particularly in the catchment Drac (3.72 m ∙ a−1). The results of the RSRCs show a significant rise in RSEs at the beginning of the ablation season, except for the Alpine catchment Alpenrhein and Var, and the Pyrenean catchment Ariege. It indicates that 11.8 and 3.97 degrees Celsius less per year are needed for the regional snowlines to reach the middle point of the RSRC in the Tagliamento and Tysa, respectively. The variation of air temperature is regarded as an example of a potential climate driver in this thesis. The retrieved monthly mean RSEs are highly correlated (mean correlation coefficient "R" ̅ = 0.7) with the monthly temperature anomalies, which are more significant in months with extremely low/high temperature. Another case study that investigates the correlation between river discharges and RSEs is carried out to demonstrate the potential consequences of the derived snowline dynamics. The correlation analysis shows a good correlation between river discharges and RSEs (correlation coefficient, R=0.52).
In this thesis, the developed framework signifies a better understanding of the snow dynamics in mountain areas, as well as their potential triggers and consequences. Nonetheless, an urgent need persists for: (1) validation data to assess long-term snow-related observations based on high-resolution EO data; (2) further studies to reveal interactions between snow and its ambient environment; and (3) regional and local adaptation-strategies coping with climate change. Further studies exploring the above-mentioned research gaps are urgently needed in the future.
The ecosystem of the high northern latitudes is affected by the recently changing environmental conditions. The Arctic has undergone a significant climatic change over the last decades. The land coverage is changing and a phenological response to the warming is apparent. Remotely sensed data can assist the monitoring and quantification of these changes. The remote sensing of the Arctic was predominantly carried out by the usage of optical sensors but these encounter problems in the Arctic environment, e.g. the frequent cloud cover or the solar geometry. In contrast, the imaging of Synthetic Aperture Radar is not affected by the cloud cover and the acquisition of radar imagery is independent of the solar illumination. The objective of this work was to explore how polarimetric Synthetic Aperture Radar (PolSAR) data of TerraSAR-X, TanDEM-X, Radarsat-2 and ALOS PALSAR and interferometric-derived digital elevation model data of the TanDEM-X Mission can contribute to collect meaningful information on the actual state of the Arctic Environment. The study was conducted for Canadian sites of the Mackenzie Delta Region and Banks Island and in situ reference data were available for the assessment. The up-to-date analysis of the PolSAR data made the application of the Non-Local Means filtering and of the decomposition of co-polarized data necessary.
The Non-Local Means filter showed a high capability to preserve the image values, to keep the edges and to reduce the speckle. This supported not only the suitability for the interpretation but also for the classification. The classification accuracies of Non-Local Means filtered data were in average +10% higher compared to unfiltered images. The correlation of the co- and quad-polarized decomposition features was high for classes with distinct surface or double bounce scattering and a usage of the co-polarized data is beneficial for regions of natural land coverage and for low vegetation formations with little volume scattering. The evaluation further revealed that the X- and C-Band were most sensitive to the generalized land cover classes. It was found that the X-Band data were sensitive to low vegetation formations with low shrub density, the C-Band data were sensitive to the shrub density and the shrub dominated tundra. In contrast, the L-Band data were less sensitive to the land cover. Among the different dual-polarized data the HH/VV-polarized data were identified to be most meaningful for the characterization and classification, followed by the HH/HV-polarized and the VV/VH-polarized data. The quad-polarized data showed highest sensitivity to the land cover but differences to the co-polarized data were small. The accuracy assessment showed that spectral information was required for accurate land cover classification. The best results were obtained when spectral and radar information was combined. The benefit of including radar data in the classification was up to +15% accuracy and most significant for the classes wetland and sparse vegetated tundra. The best classifications were realized with quad-polarized C-Band and multispectral data and with co-polarized X-Band and multispectral data. The overall accuracy was up to 80% for unsupervised and up to 90% for supervised classifications. The results indicated that the shortwave co-polarized data show promise for the classification of tundra land cover since the polarimetric information is sensitive to low vegetation and the wetlands. Furthermore, co-polarized data provide a higher spatial resolution than the quad-polarized data.
The analysis of the intermediate digital elevation model data of the TanDEM-X showed a high potential for the characterization of the surface morphology. The basic and relative topographic features were shown to be of high relevance for the quantification of the surface morphology and an area-wide application is feasible. In addition, these data were of value for the classification and delineation of landforms. Such classifications will assist the delineation of geomorphological units and have potential to identify locations of actual and future morphologic activity.
Remote sensing for disease risk profiling: a spatial analysis of schistosomiasis in West Africa
(2014)
Global environmental change leads to the emergence of new human health risks. As a consequence, transmission opportunities of environment-related diseases are transformed and human infection with new emerging pathogens increase. The main motivation for this study is the considerable demand for disease surveillance and monitoring in relation to dynamic environmental drivers. Remote sensing (RS) data belong to the key data sources for environmental modelling due to their capabilities to deliver spatially continuous information repeatedly for large areas with an ecologically adequate spatial resolution.
A major research gap as identified by this study is the disregard of the spatial mismatch inherent in current modelling approaches of profiling disease risk using remote sensing data. Typically, epidemiological data are aggregated at school or village level. However, these point data do neither represent the spatial distribution of habitats, where disease-related species find their suitable environmental conditions, nor the place, where infection has occurred. As a consequence, the prevalence data and remotely sensed environmental variables, which aim to characterise the habitat of disease-related species, are spatially disjunct.
The main objective of this study is to improve RS-based disease risk models by incorporating the ecological and spatial context of disease transmission. Exemplified by the analysis of the human schistosomiasis disease in West Africa, this objective includes the quantification of the impact of scales and ecological regions on model performance.
In this study, the conditions that modify the transmission of schistosomiasis are reviewed in detail. A conceptual underpinning of the linkages between geographical RS measures, disease transmission ecology, and epidemiological survey data is developed. During a field-based analysis, environmental suitability for schistosomiasis transmission was assessed on the ground, which is then quantified by a habitat suitability index (HSI) and applied to RS data. This conceptual model of environmental suitability is refined by the development of a hierarchical model approach that statistically links school-based disease prevalence with the ecologically relevant measurements of RS data. The statistical models of schistosomiasis risk are derived from two different algorithms; the Random Forest and the partial least squares regression (PLSR). Scale impact is analysed based on different spatial resolutions of RS data. Furthermore, varying buffer extents are analysed around school-based measurements. Three distinctive sites of Burkina Faso and Côte d’Ivoire are specifically modelled to represent a gradient of ecozones from dry savannah to tropical rainforest including flat and mountainous regions.
The model results reveal the applicability of RS data to spatially delineate and quantitatively evaluate environmental suitability for the transmission of schistosomiasis. In specific, the multi-temporal derivation of water bodies and the assessment of their riparian vegetation coverage based on high-resolution RapidEye and Landsat data proofed relevant. In contrast, elevation data and water surface temperature are constraint in their ability to characterise habitat conditions for disease-related parasites and freshwater snail species. With increasing buffer extent observed around the school location, the performance of statistical models increases, improving the prediction of transmission risk. The most important RS variables identified to model schistosomiasis risk are the measure of distance to water bodies, topographic variables, and land surface temperature (LST). However, each ecological region requires a different set of RS variables to optimise the modelling of schistosomiasis risk. A key result of the hierarchical model approach is its superior performance to explain the spatial risk of schistosomiasis.
Overall, this study stresses the key importance of considering the ecological and spatial context for disease risk profiling and demonstrates the potential of RS data. The methodological approach of this study contributes substantially to provide more accurate and relevant geoinformation, which supports an efficient planning and decision-making within the public health sector.
Information on the state of the terrestrial vegetation cover is important for several ecological, economical, and planning issues. In this regard, vegetation properties such as the type, vitality, or density can be described by means of continuous biophysical parameters. One of these parameters is the leaf area index (LAI), which is defined as half the total leaf area per unit ground surface area. As leaves constitute the interface between the biosphere and the atmosphere, the LAI is used to model exchange processes between plants and their environment. However, to account for the variability of ecosystems, spatially and temporally explicit information on LAI is needed both for monitoring and modeling applications.
Remote sensing aims at providing such information. LAI is commonly derived from remote sensing data by empirical-statistical or physical models. In the first approach, an empirical relationship between LAI measured in situ and the corresponding canopy spectral signature is established. Although this method achieves accurate LAI estimates, these relationships are only valid for the place and time at which the field data were sampled, which hampers automated LAI derivation. The physical approach uses a radiation transfer model to simulate canopy reflectance as a function of the scene’s geometry and of leaf and canopy parameters, from which LAI is derived through model inversion based on remote sensing data. However, this model inversion is not stable, as it is an under-determined and ill-posed problem.
Until now, LAI research focused either on the use of coarse resolution remote sensing data for global applications, or on LAI modeling over a confined area, mostly in forest and crop ecosystems, using medium to high spatial resolution data. This is why to date no study is available in which high spatial resolution data are used for LAI mapping in a heterogeneous, natural landscape such as alpine grasslands, although a growing amount of high spatial and temporal resolution remote sensing data would allow for an improved environmental monitoring. Therefore, issues related to model parameterization and inversion regularization techniques improving its stability have not yet been investigated for this ecosystem.
This research gap was taken up by this thesis, in which the potential of high spatial resolution remote sensing data for grassland LAI estimation based on statistical and radiation transfer modeling is analyzed, and the achieved accuracy and robustness of the two approaches is compared. The objectives were an ecosystem-adapted radiation transfer model set-up and an optimized LAI derivation in mountainous grassland areas. Multi-temporal LAI in situ measurements as well as time series of RapidEye data from 2011 and 2012 over the catchment of the River Ammer in the Bavarian alpine upland were used. In order to obtain accurate in situ data, a comparison of the LAI derivation algorithms implemented in the LAI-2000 PCA instrument with destructively measured LAI was performed first. For optimizing the empirical-statistical approach, it was then analyzed how the selection of vegetation indices and regression models impacts LAI modeling, and how well these models can be transferred to other dates. It was shown that LAI can be derived
with a mean accuracy of 80 % using contemporaneous field data, but that the accuracy decreases to on average 51 % when using these models on remote sensing data from other dates. The combined use of several data sets to create a regression which is used for LAI derivation at different points in time increased the LAI estimation accuracy to on average 65 %. Thus, reduced field measurement labor comes at the cost of LAI error rates being increased by 10 - 30 % as long as at least two campaigns are conducted. Further, it was shown that the use of RapidEye’s red edge channel improves the LAI derivation by on average 5.4 %.
With regard to physical LAI modeling, special interest lay in assessing the accuracy improvements that can be achieved through model set-up and inversion regularization techniques. First, a global sensitivity analysis was applied to the radiation transfer model in order to identify the most important model parameters and most sensitive spectral features. After model parameterization, several inversion regularizations, namely the use of a multiple sample solution, the additional use of vegetation indices, and the addition of noise, were analyzed. Further, an approach to include the local scene’s geometry in the retrieval process was introduced to account for the mountainous topography. LAI modeling accuracies of in average 70 % were achieved using the best combination of regularization techniques, which is in the upper range of accuracies that were achieved in the few existing other grassland studies based on in situ or air-borne measured hyperspectral data. Finally, further physically derived vegetation parameters and inversion uncertainty measures were evaluated in detail to identify challenging modeling conditions, which was mostly neglected in other studies. An increased modeling uncertainty for extremely high and low LAI values was observed. This indicates an insufficiently wide model parameterization and a canopy deviation from model assumptions on some fields. Further, the LAI modeling accuracies varied strongly between the different scenes. From this observation it can be deduced that the radiometric quality of the remote sensing data, which might be reduced by atmospheric effects or unexpected surface reflectances, exerts a high influence on the LAI modeling accuracy.
The major findings of the comparison between the empirical-statistical and physical LAI modeling approaches are the higher accuracies achieved by the empirical-statistical approach as long as contemporaneous field data are available, and the computationally efficiency of the statistical approach. However, when no or temporally unfitting in situ measurements are available, the physical approach achieves comparable or even higher accuracies. Furthermore, radiation transfer modeling enables the derivation of other leaf and canopy variables useful for ecological monitoring and modeling applications, as well as of pixel-wise uncertainty measures indicating the robustness and reliability of the model inversion and LAI derivation procedure. The established look-up tables can be used for further LAI derivation in Central European grassland also in other years.
The use of high spatial resolution remote sensing data for LAI derivation enables a reliable land cover classification and thus a reduced LAI mapping error due to misclassifications. Furthermore, the RapidEye pixels being smaller than individual fields allow for a radiation transfer model inversion over homogeneous canopies in most cases, as canopy gaps or field parcels can be clearly distinguished. However, in case of unexpected local surface conditions such as blooming, litter, or canopy gaps, high spatial resolution data show corresponding strong deviations in reflectance values and hence LAI estimation, which would be reduced using coarser resolution data through the balancing effect of the surrounding surface reflectances. An optimal pixel size with regard to modeling accuracy hence depends on the canopy and landscape structure. Furthermore, a reduced spatial resolution would enable a considerable acceleration of the LAI map derivation.
This illustration of the potential of RapidEye data and of the challenges associated to LAI derivation in heterogeneous grassland areas contributes to the development of robust LAI estimation procedures based on new and upcoming, spatially and temporally high resolution remote sensing imagery such as Landsat 8 and Sentinel-2.
Increasing urbanisation is one of the biggest pressures to vegetation in the City of Cape Town. The growth of the city dramatically reduced the area under indigenous Fynbos vegetation, which remains in isolated fragments. These are subject to a number of threats including atmospheric deposition, atypical fire cycles and invasion by exotic plant and animal species. Especially the Port Jackson willow (Acacia saligna) extensively suppresses the indigenous Fynbos vegetation with its rapid growth.
The main objective of this study was to investigate indicators for a quick and early prediction of the health of the remaining Fynbos fragments in the City of Cape Town with help of remote sensing.
First, the productivity of the vegetation in response to rainfall was determined. For this purpose, the Enhanced Vegetation Index (EVI), derived from Terra MODIS data with a spatial resolution of 250m, and precipitation data of 19 rainfall stations for the period from 2000 till 2008 were used. Within the scope of a flexible regression between the EVI data and the precipitation data, different lags of the vegetation response to rainfall were analysed. Furthermore, residual trends (RESTREND) were calculated, which result from the difference between observed EVI and the one predicted by precipitation. Negative trends may suggest a degradation of the habitats. In addition, the so-called Rain-use Efficiency (RUE) was tested in this context. It is defined as the ratio between net primary production (NPP) – represented by the annual sum of EVI – and the annual rainfall sum. These indicators were analysed for their suitability to determine the health of the indigenous Fynbos vegetation.
Furthermore, the degree of dispersal of invasive species especially the Acacia saligna was investigated. With the specific characteristics of the tested indicators and the spectral signature of Acacia saligna, i.e. its unique reflectance over the course of the year, the dispersal was estimated. Since the growth of invasive species dramatically reduces the biodiversity of the fragments, their presence is an important factor for the condition of ecosystem health.
This work focused on 11 test sites with an average size of 200ha, distributed over the whole area of the City of Cape Town. Five of these fragments are under conservation and the others shall be protected in the near future, too, which makes them of special interest. In January 2010, fieldwork was undertaken in order to investigate the state and composition of the local vegetation.
The results show promising indicators for the assessment of ecosystem health. The coefficients of determination of the EVI-rainfall regression for Fynbos are minor, because the reaction of this vegetation type to rainfall is considerably lower than the one of the invasive species. Thus, a good distinction between indigenous and alien vegetation is possible on the basis of this regression. On the other hand, the RESTREND method, for which the regression forms the basis, is only of limited use, since the significance of these trends is not given for Fynbos vegetation. Furthermore, the RUE has considerable potential for the assessment of ecosystem health in the study area. The Port Jackson willow has an explicitly higher EVI than the Fynbos vegetation and thus its RUE is more efficient for a similar amount of rainfall. However, it has to be used with caution, because local and temporal variability cannot be extinguished in the study area over the rather short MODIS time series.
These results display that the interpretation of the indicators has to be conducted differently from the literature, because the element of invasive species was not considered in most of the previous papers. An increase in productivity is not necessarily equivalent with an improvement in health of the fragment, but can indicate a dispersal of Acacia saligna. This shows the general problem of the term ‘degradation’ which in most publications so far is only measured by productivity and other factors like invasive species are disregarded.
On the basis of the EVI-rainfall regression and statistical measures of the EVI, the distribution of invasive species could be delineated. Generally, a strong invasion of the Port Jackson willow was discovered on the test sites. The results display that a reasoned and sustainable management of the fragments is essential in order to prevent the suppression of the indigenous Fynbos vegetation by Acacia saligna. For this purpose, remote sensing can give an indication which areas changed so that specific field surveys can be undertaken and subsequent management measures can be determined.
Grünflächen stellen einen der wichtigsten Umwelteinflüsse in der Wohnumwelt der Menschen dar. Einerseits wirken sie sich positiv auf die physische und mentale Gesundheit der Menschen aus, andererseits können Grünflächen auch negative Wirkungen anderer Faktoren abmildern, wie beispielsweise die im Laufe des Klimawandels zunehmenden Hitzeereignisse. Dennoch sind Grünflächen nicht für die gesamte Bevölkerung gleichermaßen zugänglich. Bestehende Forschung im Kontext der Umweltgerechtigkeit (UG) konnte bereits aufzeigen, dass unterschiedliche sozio-ökonomische und demographische Gruppen der deutschen Bevölkerung unterschiedlichen Zugriff auf Grünflächen haben. An bestehenden Analysen von Umwelteinflüssen im Kontext der UG wird kritisiert, dass die Auswertung geographischer Daten häufig auf zu stark aggregiertem Level geschieht, wodurch lokal spezifische Expositionen nicht mehr genau abgebildet werden. Dies trifft insbesondere für großflächig angelegte Studien zu. So werden wichtige räumliche Informationen verloren. Doch moderne Erdbeobachtungs- und Geodaten sind so detailliert wie nie und Methoden des maschinellen Lernens ermöglichen die effiziente Verarbeitung zur Ableitung höherwertiger Informationen.
Das übergeordnete Ziel dieser Arbeit besteht darin, am Beispiel von Grünflächen in Deutschland methodische Schritte der systematischen Umwandlung umfassender Geodaten in relevante Geoinformationen für die großflächige und hochaufgelöste Analyse von Umwelteigenschaften aufzuzeigen und durchzuführen. An der Schnittstelle der Disziplinen Fernerkundung, Geoinformatik, Sozialgeographie und Umweltgerechtigkeitsforschung sollen Potenziale moderner Methoden für die Verbesserung der räumlichen und semantischen Auflösung von Geoinformationen erforscht werden. Hierfür werden Methoden des maschinellen Lernens eingesetzt, um Landbedeckung und -nutzung auf nationaler Ebene zu erfassen. Diese Entwicklungen sollen dazu beitragen bestehende Datenlücken zu schließen und Aufschluss über die Verteilungsgerechtigkeit von Grünflächen zu bieten.
Diese Dissertation gliedert sich in drei konzeptionelle Teilschritte. Im ersten Studienteil werden Erdbeobachtungsdaten der Sentinel-2 Satelliten zur deutschlandweiten Klassifikation von Landbedeckungsinformationen verwendet. In Kombination mit punktuellen Referenzdaten der europaweiten Erfassung für Landbedeckungs- und Landnutzungsinformationen des Land Use and Coverage Area Frame Survey (LUCAS) wird ein maschinelles Lernverfahren trainiert. In diesem Kontext werden verschiedene Vorverarbeitungsschritte der LUCAS-Daten und deren Einfluss auf die Klassifikationsgenauigkeit beleuchtet. Das Klassifikationsverfahren ist in der Lage Landbedeckungsinformationen auch in komplexen urbanen Gebieten mit hoher Genauigkeit abzuleiten. Ein Ergebnis des Studienteils ist eine deutschlandweite Landbedeckungsklassifikation mit einer Gesamtgenauigkeit von 93,07 %, welche im weiteren Verlauf der Arbeit genutzt wird, um grüne Landbedeckung (GLC) räumlich zu quantifizieren.
Im zweiten konzeptionellen Teil der Arbeit steht die differenzierte Betrachtung von Grünflächen anhand des Beispiels öffentlicher Grünflächen (PGS), die häufig Gegenstand der UG-Forschung ist, im Vordergrund. Doch eine häufig verwendete Quelle für räumliche Daten zu öffentlichen Grünflächen, der European Urban Atlas (EUA), wird bisher nicht flächendeckend für Deutschland erhoben. Dieser Studienteil verfolgt einen datengetriebenen Ansatz, die Verfügbarkeit von öffentlichem Grün auf der räumlichen Ebene von Nachbarschaften für ganz Deutschland zu ermitteln. Hierfür dienen bereits vom EUA erfasste Gebiete als Referenz. Mithilfe einer Kombination von Erdbeobachtungsdaten und Informationen aus dem OpenStreetMap-Projekt wird ein Deep Learning -basiertes Fusionsnetzwerk erstellt, welche die verfügbare Fläche von öffentlichem Grün quantifiziert. Das Ergebnis dieses Schrittes ist ein Modell, welches genutzt wird, um die Menge öffentlicher Grünflächen in der Nachbarschaft zu schätzen (𝑅 2 = 0.952).
Der dritte Studienteil greift die Ergebnisse der ersten beiden Studienteile auf und betrachtet die Verteilung von Grünflächen in Deutschland unter Hinzunahme von georeferenzierten Bevölkerungsdaten. Diese exemplarische Analyse unterscheidet dabei Grünflächen nach zwei Typen: GLC und PGS. Zunächst wird mithilfe deskriptiver Statistiken die generelle Grünflächenverteilung in der Bevölkerung Deutschlands beleuchtet. Daraufhin wird die Verteilungsgerechtigkeit anhand gängiger Gerechtigkeitsmetriken bestimmt. Abschließend werden die Zusammenhänge zwischen der demographischen Komposition der Nachbarschaft und der verfügbaren Menge von Grünflächen anhand dreier exemplarischer soziodemographischer Gesellschaftsgruppen untersucht. Die Analyse zeigt starke Unterschiede der Verfügbarkeit von PGS zwischen städtischen und ländlichen Gebieten. Ein höherer Prozentsatz der Stadtbevölkerung hat Zugriff das Mindestmaß von PGS gemessen an der Vorgabe der Weltgesundheitsorganisation. Die Ergebnisse zeigen auch einen deutlichen Unterschied bezüglich der Verteilungsgerechtigkeit zwischen GLC und PGS und verdeutlichen die Relevanz der Unterscheidung von Grünflächentypen für derartige
Untersuchungen. Die abschließende Betrachtung verschiedener Bevölkerungsgruppen arbeitet Unterschiede auf soziodemographischer Ebene auf.
In der Zusammenschau demonstriert diese Arbeit wie moderne Geodaten und Methoden des maschinellen Lernens genutzt werden können bisherige Limitierungen räumlicher Datensätze zu überwinden. Am Beispiel von Grünflächen in der Wohnumgebung der Bevölkerung Deutschlands wird gezeigt, dass landesweite Analysen zur Umweltgerechtigkeit durch hochaufgelöste und lokal feingliedrige geographische Informationen bereichert werden können. Diese Arbeit verdeutlicht, wie die Methoden der Erdbeobachtung und Geoinformatik einen wichtigen Beitrag leisten können, die Ungleichheit der Wohnumwelt der Menschen zu identifizieren und schlussendlich den nachhaltigen Siedlungsbau in Form von objektiven Informationen zu unterstützen und überwachen.