Completed doctoral thesis (since 2003)

Tim Wai NG (2018)

Application of object based image analysis and very high resolution satellite imagery for delineation of prosopis invasion in selected areas in Somalia and Kenya

Betreuung: Clement Atzberger



Markus IMMITZER (2017)

Mapping Tree Species, Forest Calamities and Growing Stock using High Resolution Satellite Imagery – Possibilities and Limits

Increasing requirements for monitoring to ensure sustainable forest management result in a higher demand for detailed, reliable and up-to-date forest information. In this work, earth observation (EO) data with high spectral (5 to 10 spectral bands) and high spatial resolutions (0.5 to 30 m) are used for tree species classification, mapping calamities, and growing stock. Classifications of individual tree crowns from ten tree species using WorldView-2 (WV2) data have an overall accuracy of 82%. Classifications of several scenes using automatically generated segments representing tree collectives show significantly higher overall accuracies of 88 to 97%. A direct comparison of three sensors shows that, in addition to spectral resolution, spatial resolution has a great influence on classification accuracy. Seven tree species are distinguished with an overall accuracy of 74% using WV2 data, 68% with Sentinel-2 and 50% with Landsat-8. To map tree species on large areas the spruce and pine classification using WV2 data as a reference, and metrics derived from Landsat time series data as explanatory variables are combined to derive species-specific fractional cover for the entire state of Bavaria. An independent validation shows error values of 12.1% for pine and 14.2% for spruce. The detection of calamities also shows promising results. Three classes – ‘healthy’, ‘bark beetle infested but still green’ and ‘dead’ – can be distinguished at the tree crown level with accuracies of more than 70% using WV2 data. Windthrows are detected using a change detection approach based on RapidEye data recorded before and after the event with accuracies of around 90%. Canopy height models generated from WV2 stereo data are used for growing stock modeling, with ground-based forest inventory results as input data. The best model has an error of 29.5%. The studies presented underline the high potential of earth observation data for the estimation of important forest parameters.

Supervisor: Clement Atzberger

Josef Pennerstorfer  (2017)

Modellierung der Phänologie des Buchdruckers (Ips typographus) unter Verwendung von GIS und Fernerkundung als wesentliche Komponente einer Borkenkäferprädisposition im Rahmen einer umfassenden Risikoanalyse

The European spruce bark beetle (Ips typographus) is considered to be one of the most important pests with increasing significance for forest protection in Central Europe. During the last years, repeated storm events and climate extremes have led to bark beetle outbreaks. The need for risk assessment systems as a base for decision support in forest management becomes even more urgent, considering the predicted increase in temperature. The main goal of this thesis was the development of models for spatio-temporal prediction of phenology and reproductive potential of bark beetles. A central task was the development of global radiation models (SOLMOD), which are essential to estimate bark temperatures required for simulating bark beetle development. Bark beetle phenology was modeled using PHENIPS. Temperature and solar radiation were interpolated using DAYMET and RECLIP:MORE. The integration of the models into geographic information systems was a technical challenge. Local models (50mx50m) using climate data of a local network were designed for the Wilderness Area Dürrenstein and regional models (250mx250m) aiming mainly at online-modeling on coarser scales and using data from the ZAMG (Central Institute for Meteorology and Geodynamics) were developed for the central alpine part of Austria. The models were validated using data from trapping trees and from a monitoring program. Local models showed reliable predictions of all development stages, while total development was overestimated in regional models due to the accumulation of deviations. The validation of SOLMOD regarding suitability of predicted phenology showed only slight deviations. The bark beetle development and predisposition assessment systems provide a long-term retrospective consideration of bark beetle population development and a better understanding of the causal complexes of insect calamities as well as a spatially differentiated view regarding risk patterns.

Supervisor: Werner Schneider

Javier Esteban Portillo (2016)

Assimilation of Leaf Area Index into the crop growth models GRAMI and SAFY to monitor maize crop: a case of study in the Marchfeld region

Simple crop growth models (CGM) integrated with Leaf Area Index (LAI) retrieved from earth observation (EO) data allow monitoring crop growth at regional and plot scales. The global research objective of this study was to develop a method based on the GRAMI and SAFY crop growth models to estimate maize yield in the Marchfeld region, characterized by a restricted availability of input data. Maize yield data were collected for the seasons 2013 and 2014. The LAI was estimated by the CLAIR model from Landsat and DEIMOS data with a NRMSE of 12%. The Look up Table (LUT) and Particle Swarm optimization (PSO) methods were tested at plot and pixel scale. Ensemble Kalman Filter (EnKF) and Particle Filter (PF) were implemented at pixel scale. SAFY calibrated at plot scale by PSO achieved a better performance compared to GRAMI with an estimation error at plot scale (EEP) of 12.5% and at regional scale (EER) of 0.2%. The calibration techniques performed well in forecasting maize yield one to two months before harvesting with an EER of 3% and 7.1% respectively. The yield estimation accuracy was increased by using pixel data. The EEP was reduced to 8.3% using SAFY calibrated at pixel scale for a reduced number of free parameters. The updating at pixel scale showed a good performance. The EEP using EnKF was 9.7% and the EER 2.3%. Moreover, the time of processing was reduced by 17 times compared to the PSO calibration at plot scale. PF showed in average for both crop seasons better accuracy than EnKF. Additionally, the low computational demand of PF makes the updating of SAFY at pixel scale possible even for larger regions.

Supervisor: Clement Atzberger, Francesco Vuolo, Josef Eitzinger

Bruno SCHULTZ (2016)

Land use and land cover classification in Brazil

Supervisor: Clement Atzberger

Sayeh Kassaw AGEGNEHU (2015)

Urban Expansion and Peri-urban Land Tenure Security in Ethiopia Cases of Bahir Dar and Debre Markos.

In this study, analysis of three inter-related issues in peri-urban areas of Amhara National Regional State of Ethiopia has been made, taking Bahir Dar and Debre Markos as case studies. These are the assessment of the rate of urban expansion and peri-urban land tenure security, expropriation and compensation practices, and peri-urban land tenure conflicts and resolution systems. The main purpose of the study is to assess peri-urban land tenure security situation and to recommend strategies for securing peri-urban land tenure. A variety of qualitative and quantitative data generating and analyzing methods are employed in the study. Time series LANDSAT data show that both study cities are expanding at high rates. While cities are expanding rapidly, the formalizing and securing of land rights in the study peri-urban areas is stalled. The land registration practice was launched at the same time with the rural areas but interrupted without proceeding to further processes that lead to certification. Consequently, there are variations in land tenure security feelings with certified farmers exhibiting better sense of tenure security compared to non-certified ones. This lack of formalization peri-urban land rights has also reduced their motivation to undertake land-related investments. The study also showed that the current compensation payment is inadequate. This influences the tenure security feeling of peri-urban dwellers, besides its impact on their livelihood. Adequate compensation payments can also motivate farmers to undertake land-related investments. In addition, making expropriation practices participatory enable farmers to have confidence on their land and also for the expropriating party to understand the feelings and aspirations of the farmers which in turn provide insights for proper implementation of expropriation and compensation practices. Besides, strong support to farmers by preparing alternative livelihood strategies is of paramount importance. The study also assessed the extent of land tenure conflicts and analyzed resolution mechanisms. Results showed that the incidence of land tenure conflicts is high in peri-urban areas compared to rural areas. Mediation played a significant role in resolving land tenure conflicts both in peri-urban and rural areas.

Supervisor: Fuchs Helmut , Mansberger Reinfried , Seher Walter

Menale Wondie MULU (2015)

Spatial and temporal analysis of management impacts on ecosystem services – scenario modeling in a highland watershed of northwest Ethiopia.

The objective of the study was to find ecosystem-service-based development options to improve the living conditions of the people in the highlands of northwest Ethiopia. A set of tools was applied to a demonstration watershed. Participatory resource assessment focusing on ecosystem services (ESs) was conducted by using interviews, focus group discussions, reviews, field surveys and informal meetings with stakeholders. The dynamics of the landscape from 1957 to 2013 was analyzed using aerial photographs and satellite images. The Drivers-Pressures-States-Impacts-Responses (DPSIR) frameworkwas used for the ecosystem-services-scenario-modeling. The pathways of scenarios were evaluated by a combination of qualitative and quantitative analyses of land cover – land use types (LCUTs) and ESs. Subsistence, cultural or religious, and environmental services were the major ESs found in Tara Gedam. The analysis of remote sensing data showed an overall increase of cropland and a decrease of other LCUTs. Land use conversion predominantly occurred from forest or shrubland to cropland. Improved crop varieties and management can increase the yield 2-3 fold. The increased productivity would offer sufficient time and land for fallowing, exclosures and rehabilitation. Business as Usual, Transition Agriculture, Intensive Agriculture and Optimized Ecosystem Services were the scenarios defined as development options suiting the highland watershed using the DPSIR framework. TAG could be chosen as a rapid response of a low economy community. INA emphasizes food security, whereas OPE pays due attention to food security and environment protection simultaneously. The application of either INA or OPE will improve the living condition of subsistence farming. OPE better ensures the provision of diversified and economically valuable ES. Population growth, policy and technology are presumed to be the major ESs stressors. Future ES development requires integrated approaches to improve food security.

Supervisor: Katzensteiner Klaus, Schneider Werner, Mansberger, Reinfried

Gebeyehu Belay SHIBESHI (2014)

Cadastral procedure and spatial framework for the development of an efficient land administration system for the rural lands of anrs (amhara national regional state) of Ethiopia.

The objective of this thesis is to developed suitable methodologies for second level certification program in the Amhara National Regional State (ANRS) of Ethiopia. The main addition in the second level certification program is the inclusion of parcel maps. The development of new way of working requires closer look of the existing situation. The study deals with development of toolbox for institutional set up, cadastral and registration proclamation preparation, creation and densification of ground control points and in the development of suitable survey methods in ANRS. The core legal cadastral model was used to describe both formal and informal settings in ANRS. Modifications were made for proper descriptions of both formal and informal settings. The study identified more similarities than difference between the two systems. Large scale cadastral projects are planned all over the country as part of the country’s five years growth and transformation plan. But cadastral and registration proclamation is not enacted to facilitate and guide the implementation of cadastral projects. The tool that can be used for the development of cadastral and registration proclamation for rural land administration in Ethiopia was developed. Remote sensing and ground surveying techniques were compared to produce cadastral map of different holding types that is good enough to satisfy the needs of users. The study, identified that trust on a system, dependability and repeatability are more important than geometric accuracy. Cost effective options that can fulfill the demands of second level certification program in ANRS by connecting the cadastral maps to a national grid are identified. However, the main part of the study deals with the development of methods suitable for ANRS, the developed methods largely can be used for conditions in the other regional states in the country and even in the developing world.

Supervisor: Fuchs Helmut, Mansberger Reinfried

Abdou Karim OUEDRAOGO (2005)

Sorghum and millet crop production forecasting in the Province of Namentenga (Burkina Faso) by using remote sensing and crop modelling

Quantifying crop production reliably at a local scale is important for the efficient management of food security in African developing countries. In this study, carried out in the province of Namentenga in the north centre of Burkina Faso in West Africa, a method is developed to forecast yield and assess millet and sorghum grain production at provincial level. The method is an integration of crop modelling and optical remote sensing. By inputting a minimum data set, running the CERES-Millet and the CERES-Sorghum models, and analysing outputs, both models are parameterised and evaluated in 2001 and 2003, using field data collected in twenty sample fields during the crop year 2003, and recorded yield by DGPSA in 2001. In order to forecast provincial production, simulated grain yield is extrapolated to provincial crop acreages issued by performing a supervised classification of ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) multi spectral images available for the study area during the crop season 2003. Additionally, the relationship the temporal development of NDVI (Normalized Difference Vegetation Index) from MODIS (Moderate Resolution Imaging Spectroradiometer) 16-day NDVI product and crop parameters (dry grain yield, dry biomass and leaf area index) is investigated with the aim to quantify grain yield remotely. The results of this study suggests that simulated yield is highly correlated to the measured one, or may need some adjustment in case of flooding and weeds infestation not considered in the CERES-Millet and CERES-Sorghum models under DSSAT v3.5. Despite the complexity of the farming system and the field structures, it is pointed out that the ASTER multi spectral images are suitable for a reliable estimation of crop acreages, and can significantly improve the traditional method of crop production assessment based on population census.

Supervisor: Werner Schneider

Abebe Haile SOLOMON. (2005)

Bridging the land rights demarcation gap in Ethiopia - usefulness of high resolution satellite image (HRSI) data

The land sector in Ethiopia is beset by a variety of problems. There are reasons to believe that these problems would continue to stifle agriculture and poverty reduction. The population is increasing at an alarming rate. Conservation measures are not practiced in a sustainable manner largely due to the disincentive that the state ownership of land and the attendant practice of allocating land rights have caused. Also, support systems and tools like land administration (LA) have for long been totally neglected or ineffective. Following the establishment of LA agencies, LA has recently started to witness some changes. These agencies have begun taking actions to improve tenure security. Despite these developments, the country has continued to face the same old challenges mainly because of desultory and irresolute policies. As current LA-related initiatives in the country are also struggling with a choice of suitable modality for building LA spatial framework, the research looks into the viability of using high-resolution satellite imagery (HRSI). This has involved conducting a case study in Dera Woreda (district), Amhara Region. Due to no or feeble physical evidence of boundaries on the ground, HRSI is at the moment not useful for generating land information in the case study area and other parts of the country where similar conditions prevail. The research concludes the inquiry by pointing at the way forward which, together with a host of investment appraisal guidelines, could be considered as means of acquiring land information and reforming land rights.

Betreuung: Werner Schneider, Reinfried Mansberger

Ali FARZANEH (2005)

Land cover mapping employing fusion of remotely sensed high-spatial-resolution pan and medium-spatial-resolution multispectral images in the region of Sari (Iran)

Abstract Land cover mapping on the basis of satellite images is a promising technique both for primary map production and for map updating and monitoring. For mapping at a scale of 1:25.000, high spatial resolution satellite images have to be used in order to obtain the required spatial detail. At the same time, spectral information necessary for identifying certain land cover types can most economically be derived from multi-spectral images of medium spatial resolution. Fusion techniques have to be used to combine information from both sources. Image fusion can be performed at three different levels: at pixel level, at object (image segment) level and at decision (map object) level. In this thesis, a method was developed for object-level fusion of IRS-1C/D pan images (5,8 m pixel size) and LANDSAT TM multispectral images (30 m pixel size) and subsequent classification to produce a land cover map. The method was developed and tested in a study area in the Sari region in Iran (located to the south of Caspian Sea), mainly covered with industrial forests and representing a fragile Mediterranean ecosystem, while suffering from socio-economic problems too. The individual processing steps included segmentation of a multi-band image consisting of both the high-spatial-resolution pan image band and medium-spatial-resolution multispectral bands, with proper weighting of the individual bands in the segmentation procedure in order to obtain both fine detail from the pan image and coarser boundary delineations which show up only in multispectral image. For classification, fuzzy logic membership functions were used. Verification of the classification was carried out and checked with error matrix and kappa calculation on two selected transects. The results showed that employing object-based fusion procedure using medium- and high-resolution data was an appropriate method that improved classification. Pixel-based fusion procedure was demonstrated to be an acceptable method to create a satellite image map (Sat-map) for visual monitoring activities and programs.

Betreuung: Werner Schneider, Helmut Fuchs

Kurt FRIEDL (2004)

Computergestützte Modellierung von Sedimenttransport in alpinen Einzugsgebieten unter Anwendung Geografischer Informationssysteme

Ziel der Untersuchung ist die Prüfung der Frage, ob die derzeitigen physikalischen Modelle zur Ermittlung von Fließgeschwindigkeiten und sohlgebundenem Feststofftransport in Wildbächen in Verbindung mit einem 1D mathematisch-numerischen Modell geeignet sind, Abflusshöhen und Feststofftransport zu ermitteln. Zur Beschreibung eines Wildbaches wird ein Gerinnenmodell hergestellt, welches in seiner Genauigkeit der terrestrischen Vermessung entspricht. Dieses Modell bildet die Basis für die eindimensionale Ausbreitung einer kinematisch beschriebenen Hochwasserwelle, die sohlgebundenen Feststofftransport in einem abstrahierten Rechteck- oder Parabelgerinne induziert. Eine Anwendung des Modells in drei Beispieleinzugsgebieten zeigt einerseits vergleichbare Ergebnisse mit anderen numerischen Modellen. Andererseits wird die starke Variation der Modellantworten bei ähnlichen, d.h. gering variierenden Eingangsgrößen durch die Sensitivität des Gesamtmodells bezüglich dieser Größen aufgezeigt. Dieses Verhalten zeigen auch reale Wildbäche

Supervisor: Helmut Fuchs

Tatjana KOUKAL (2004)

Nonparametric assessment of forest attributes by combination of field data of the Austrian forest inventory and remote sensing data

The Austrian forest inventory provides data of the Austrian forest based on systematic field sampling. A method for mapping forest attributes and computing regional statistics by combination of field data and satellite image data is presented. The reliability of statistics particularly of small sub-areas (e.g. forest districts) is increased due to auxiliary information from satellite images. Both detailed maps and statistical information are important for tasks like planning and monitoring at regional level, e.g. for protection forest management. It is investigated which forest attributes surveyed by the Austrian forest inventory (e.g. forest area, forest cover type, volume) can be derived from satellite images and what accuracies are achieved.n

Supervisor: Werner Schneider

Download: Nonparametric Assessment of Forest Attributes by Combination of Field Data of the Austrian Forest Inventory and Remote Sensing Data

Joachim STEINWENDNER (2003)

Graphen-basierte Subpixelsegmentierung von optischen Satellitenbilddaten

Die automatische Interpretation von Satellitenbilddaten gewinnt in Zeiten von Terabytes an Bilddaten und immer komplexeren Fernerkundungsaufgaben mehr und mehr an Bedeutung. Diese Arbeit bedient sich dreier Wissenschaftsbereiche - Fernerkundung, Bildverstehen und Graphentheorie - zur Lösung von typischen Problemen der automatischen Klassifikation von Satellitenbilddaten. Konventionelle pixelbasierte Klassifikationsverfahren zeigen Schwächen in Fällen von Mischpixeln. Ein neu entwickeltes Optimierungsverfahren für die räumliche Subpixelanalyse verringert signifikant das Mischpixel-Problem. Ein weiteres Problem pixelbasierter Methoden ist, dass für die Klassifikation von Objekten nicht nur spektrale Signaturen, sondern auch die Objektform ein wesentliches Bestimmungsmerkmal ist. Dafür wurde ein Bildsegmentierungsverfahren zur Objektextraktion unter Ausnutzung der Subpixelresultate entwickelt. Eine wichtige Information zur Objektklassifikation sind die Nachbarschaftsbeziehungen, die hervorragend durch Graphen abgebildet werden können. Die entwickelte Bildsegmentierung basiert daher auf graphentheoretischen Methoden zur Bestimmung der Topologie der Objekte.

Supervisor: Werner Schneider