Дисертації з теми "Multi-temporal remote sensing"

Multi-temporal image analysis has been widely used in many applications such as urban monitoring, disaster management, and agriculture. With the development of the remote sensing technology, the new generation remote sensing satellite images with High/ Very High spatial resolution (HR/VHR) are now available. Compared to the traditional low/medium spatial resolution images, the detailed information of ground objects can be clearly analyzed in the HR/VHR images. Classical methods of multi-temporal image analysis deal with the images at pixel level and have worked well on low/medium resolution images. However, they provide sub-optimal results on new generation images due to their limited capability of modeling complex spatial and spectral information in the new generation products. Although significant number of object-based methods have been proposed in the last decade, they depend on suitable segmentation scale for diverse kinds of objects present in each temporal image. Thus their capability to express contextual information is limited. Typical spatial properties of last generation images emphasize the need of having more flexible models for object representation. Another drawback of the traditional methods is the difficulty in transferring knowledge learned from one specific problem to another. In the last few years, an interesting development is observed in the machine learning/computer vision field. Deep learning, especially Convolution Neural Networks (CNNs) have shown excellent capability to capture object level information and in transfer learning. By 2015, deep learning achieved state-of-the-art performance in most computer vision tasks. Inspite of its success in computer vision fields, the application of deep learning in multi-temporal image analysis saw slow progress due to the requirement of large labeled datasets to train deep learning models. However, by the start of this PhD activity, few works in the computer vision literature showed that deep learning possesses capability of transfer learning and training without labeled data. Thus, inspired by the success of deep learning, this thesis focuses on developing deep learning based methods for unsupervised/semi-supervised multi-temporal image analysis. This thesis is aimed towards developing methods that combine the benefits of deep learning with the traditional methods of multi-temporal image analysis. Towards this direction, the thesis first explores the research challenges that incorporates deep learning into the popular unsupervised change detection (CD) method - Change Vector Analysis (CVA) and further investigates the possibility of using deep learning for multi-temporal information extraction. The thesis specifically: i) extends the paradigm of unsupervised CVA to novel Deep CVA (DCVA) by using a pre-trained network as deep feature extractor; ii) extends DCVA by exploiting Generative Adversarial Network (GAN) to remove necessity of having a pre-trained deep network; iii) revisits the problem of semi-supervised CD by exploiting Graph Convolutional Network (GCN) for label propagation from the labeled pixels to the unlabeled ones; and iv) extends the problem statement of semantic segmentation to multi-temporal domain via unsupervised deep clustering. The effectiveness of the proposed novel approaches and related techniques is demonstrated on several experiments involving passive VHR (including Pleiades), passive HR (Sentinel-2), and active VHR (COSMO-SkyMed) datasets. A substantial improvement is observed over the state-of-the-art shallow methods.

The study area of this project is located in Makhado Municipality, Limpopo, South Africa. The Limpopo Province is commonly known for being rich in the country?s natural resources. It has a number of villages that are characterized by rich natural resources and a well-known nature reserve, Soutpansberg Mountains. Natural resources such as water, plantations, woodlands and grasslands are commonly found in these villages and are commonly used for alleviating poverty. Rural communities in this municipality are still highly dependent on natural resources. The high dependence on these natural resources subsequently affects negatively the natural environment, e.g. processes such as land degradation. Villages in this region have limited infrastructure development that influence people?s livelihood. Infrastructure developments are commonly known for contributing to growing the economy and it will be no different if such developments are built in these villages. Therefore, it is imperative to find innovative and scientific techniques that provide information which can assist in finding ways of balancing the interaction between the environment and its people. In order to successfully do so, ways of managing and monitoring of natural resources in villages such as Makhado becomes a necessity. Land cover information is required to adequately understand the extent and status of the natural resources of the Makhado region. This information is required for effective monitoring of natural resources. With the aid of remote sensing applications, land cover studies are possible. The applications always aim to provide efficient methods using low cost or freely available data. The main objective of this study was to innovatively and accurately map the land cover classes of Makhado Municipality using Landsat imagery. The study investigated the performance of single and multi-temporal assessment approach. The study found that the results of the multi-temporal approach were more accurate compared to the single-date approach for both periods. The overall accuracy of single-date classifications were 78.1% with Kc of 0.74 and 54.3% with Kc of 0.46 respectively. The classification map results of the multi-temporal approach were 72.9% with Kc of 0.68 and 79.0% and a Kc of 0.76 respectively. The multi-temporal classification maps were used for post-classification change detection. The results of these methods illustrated the major decrease in grasslands from 2006-2009 and 2013-2015 respectively. These results assisted in making further inferences of how the drastic and severe drought that occurred in 2015 till recently had a significant impact on the land cover.
Dissertation (MSc)--University of Pretoria, 2017.
Geography, Geoinformatics and Meteorology
MSc
Unrestricted

Tillage practices have changed dramatically during the past several decades as agricultural specialists have recognized the unfavorable environmental effects of mechanized tillage. Alternatively, conservation tillage management can mitigate adverse environmental impacts of tillage, such as soil and water degradation. Adoption of conservation tillage has continued to increase since its first introduction, which raises questions of when and where it is practiced. Spatial and temporal specifics of tillage practices form important dimensions for development of effective crop management practices and policies.  Because Landsat has been and will continue to image the Earth globally, it provides opportunities for systematic mapping of crop residue cover (CRC) /tillage practices. Thus, the overall objective of this study is to develop methodologies to improve our ability to monitor crop management across different landscapes in a time-efficient and cost-effective manner using Landsat TM and ETM+ imagery, which is addressed in three separate studies. The first study found that previous efforts to estimate CRC along a continuum using Landsat-based tillage indices were unsuccessful because they neglected the key temporal changes in agricultural surfaces caused by tilling, planting, and crop emergence at the start of the growing season. The first study addressed this difficulty by extracting minimum values of multi-temporal NDTI (Normalized Difference Tillage Index) spectral profiles, designated here as the minNDTI method. The minNDTI improves crop residue estimation along a continuum (R2 = 0.87) as well as tillage classification accuracy (overall accuracy > 90%).   A second study evaluated effectiveness of the minNDTI approach for assessing CRC at multiple locations over several years, and compared minNDTI to hyperspectral tillage index (CAI), and the ASTER tillage index (SINDRI). The minNDTI is effective across four different locations (R2 of 0.56 ~ 0.93). The third study, built upon the second study, addressed the Landsat ETM+ missing data issue, and devised methodologies for producing field-level tillage data at broad scales (multiple counties).  In summary, this research demonstrates that the minNDTI technique is currently the best alternative for monitoring CRC and tillage practices from space, and provides a foundation for monitoring crop residue cover at broad spatial and temporal scales.
Ph. D.

In this study we applied the asymmetric Gaussian, double-logistic, and Savitzky-Golay filters to MODIS time-series NDVI data to compare the capability of smoothing algorithms in noise reduction for improving land cover classification in the Great Lakes Basin, and providing groundwork to support cyanobacteria and cyanotoxin monitoring efforts. We used inter-class separability and intra-class variability, at varying levels of pixel homogeneity, to evaluate the effectiveness of three smoothing algorithms. Based on these initial tests, the algorithm which returned the best results was used to analyze how image stratification by ecoregion can affect filter performance. MODIS 16-day 250m NDVI imagery of the Great Lakes Basin from 2001-2013 were used in conjunction with National Land Cover Database (NLCD) 2006 and 2011 data, and Cropland Data Layers (CDL) from 2008 to 2013 to conduct these evaluations. Inter-class separability was measured by Jeffries-Matusita (JM) distances between selected land cover classes (both general classes and specific crops), and intra-class variability was measured by calculating simple Euclidean distance for samples within a land cover class. Within the study area, it was found that the application of a smoothing algorithm can significantly reduce image noise, improving both inter-class separability and intra-class variability when compared to the raw data. Of the three filters examined, the asymmetric Gaussian filter consistently returned the highest values of interclass separability, while all three filters performed very similarly for within-class variability. The ecoregion analysis based on the asymmetric Gaussian dataset indicated that the scale of study area can heavily impact within-class separability. The criteria we established have potential for furthering our understanding of the strengths and weaknesses of different smoothing algorithms, thereby improving pre-processing decisions for land cover classification using time-series data.
Master of Science

Dramatic land-cover changes have occurred in a broad range of spatial and temporal scales over the last decades. Satellite remote sensing, which can observe the earth's surface in a consistent manner, has been playing an important role in monitoring and evaluating land-cover changes. Meanwhile, optical remote sensing, a common approach to acquiring land-cover information, is limited by weather conditions and thus is greatly constrained in areas with frequent cloud cover and rainfall. Recent advances in polarimetric SAR (PolSAR) provide a promising means to extract timely information of land-cover changes regardless of weather conditions. SAR satellite can pass through an area from different orbits, namely ascending orbit and descending orbit. The PolSAR images from the same orbit will have similar backscattering even with different incident angles. But if images are acquired from different orbits, the backscattering will vary greatly, which causes many difficulties to land cover change detection. The proposed algorithms in this study can perform land cover change detection in three situations: 1) repeat-pass images (image from the same orbit and with same incident angle, 2) images from the same orbit but with different incident angle, and 3) images from different orbits. Using images from different orbits will largely reduce the monitoring interval which is important in the surveillance of natural disasters. The present study proposes 1) a sub-pixel automatic registration technique, 2) an automatic change detection technique and 3) an iterative framework to process a time series of PolSAR images that can be applied to the PolSAR images from different orbits. Firstly, automatic registration is crucial to the change detection task because a small positional error will largely degrade the accuracy of change detection. The automatic registration technique is based on the multi-scale Harris corner detector. To improve the efficiency and robustness, the orientation angle differencing method is proposed to reject outliers. This differencing method has been proved effective even in the experiment of using PolSAR images from different orbits when less than 5% of the feature point matches are correct. Secondly, the change detection technique can automatically detect land-cover conversions and classify the newly input image. Hierarchical segmentation has been applied in the change detection which generates objects within the constraint of the previous classification map. Multivariate kernel density estimation is applied to classify newly input PolSAR image. The experiments show that the proposed change detection technique can mitigate the effect of polarimetric orientation shift when the PolSAR images are from different orbits, and it can achieve high accuracy even when complex local deformation is appeared. Lastly, the iterative framework, which integrates the automatic registration and automatic change detection techniques, is proposed to process a time series of PolSAR images. In the iterative process, no obvious decrease of classification accuracy is observed. Therefore, the proposed framework provides a potential treatment to derive land-cover dynamics from a time series of PolSAR images from different orbits.
published_or_final_version
Urban Planning and Design
Doctoral
Doctor of Philosophy

The Midwest US has experienced significant changes in agricultural land use and management practices in recent decades. Cropland expansion, crop rotation change, and crop phenology changes could lead to divergent environmental impacts on linked ecosystems. The overall objective is to examine agricultural land use and management changes and their impacts on water quality in the Midwest US, which is addressed in three separate studies. The first study examined spatial and temporal dimensions of agricultural land use dynamics in east-central Iowa, 2001-2012. Results of this study indicated that increases in corn production in response to US biofuel policies had been achieved mainly by altering crop rotation. This study also examined spatial relationships between cultivated fields and crop rotation practices with respect to underlying soils and terrain. The most intensively cultivated land had shallower slopes and fewer pedologic limitations than others, and the corn was planted on the most suitable soils. The second study characterized key crop phenological parameters (SOS and EOS) for corn and soybean and analyzed their spatial patterns to evaluate their change trends in the Midwest US, 2001-2015. Results showed that MODIS-derived SOS and EOS values are sensitive to input time-series data and threshold values chosen for crop phenology detection. The non-winter MODIS NDVI time-series input data, and a lower threshold value (i.e., 40%) both generated better results for SOS and EOS estimates. Spatial analyses of SOS and EOS values displayed clear south-north gradient for corn and trend analyses of SOS revealed only a small percentage of counties showed statistically significant earlier trends within a user-defined temporal window (2001-2012). The third study integrated remote sensing-derived products from the first two studies with the SWAT model to assess impacts of agricultural management changes on sediment and nutrient yields for three selected watersheds in the Midwest US. With satisfied calibration and validation results for stream flows, sediment and nutrient yields, considered under differing management scenarios, were compared at different spatial scales. Results showed that intensive crop rotation, advancing the planting date with the same length of growing season, and longer growing seasons, dramatically increased, maintained, and slightly reduced sediment, total nitrogen, and total phosphorous yields, respectively. Overall, these studies together illuminate relationships between broad-scale agricultural policies, management decisions, and environmental impacts, and the value of multi-temporal, broad-scale, geospatial analysis of agricultural landscapes.
Ph. D.

Made available in DSpace on 2016-12-23T14:37:35Z (GMT). No. of bitstreams: 1 mileide.pdf: 1266050 bytes, checksum: 2566f7a76319b066b6b3af42e3af0d23 (MD5) Previous issue date: 2008-03-04
The Brazilian Northeast (NEB) region presented different vegetation types that are essential component of its ecosystem. With remote sensing techniques it is possible, for example, to analyzed variations in vegetation community and alterations in vegetation phenological. Analysis the main objective of this work is to evaluate the temporal behavior of the Enhanced Vegetation Index (EVI) from the Moderate Resolution Imaging Spectroradiometer (MODIS), of different vegetation types in the NEB over period between February/2000 and July/2006. The study area was the NEB, where it was used to characterize the vegetations types a vegetation map of Brazil, in the 1:5,000,000 scale from Brazilian Institute of Geography and Statistics (IBGE). A total of 140 cloud-free EVI images with spatial resolution 250 m were acquired from National Aeronautics and Space Administration (NASA). Four CBERS-2/CCD images spatian resolution 20 m were also acquired from National Institute for Espace Research (INPE) to assist EVI data sample collection for each vegetation type. Precipitation data of the cities Petrolina and Pesqueira (Pernambuco), São Luiz and Carolina (Maranhão) located in regions of Caatinga, Atlantic Forest, Amazon and Savannah biome vegetation, respectively, were used to analyze its relationship with EVI from these vegetation. Also, EVI from irrigated area at Petrolina were used in these analysis. Results obtained showed that: i) multi-temporal EVI data from different vegetation types were sensitive to the vegetation phenological cycles, with minor and greater values of EVI in the periods of less and greater precipitation, respectively; ii) amazon biome vegetation presented lesser variation in the multitemporal EVI, however with greater values, justified by vegetation species the are always with green leaf; iii) Caatinga biome vegetation presented greater EVI values variation because the vegetation species on the dry periods occur total defoliation and on wet period the vegetation became green; iv) all EVI data from the vegetations studied presented significant relationship with precipitation (p-value< 0.05).
O Nordeste Brasileiro (NEB) apresenta diferentes tipos de vegetação, sendo importantes para o seu ecossistema. Com a utilização de técnicas de sensoriamento remoto é possível, por exemplo, analisar variações de comunidades de vegetação e suas alterações fenológicas. O objetivo principal deste trabalho é avaliar o comportamento temporal do Índice de Vegetação Melhorado (EVI) do sensor Spectroradiômetro de Resolução Espacial Moderada (MODIS), de diferentes tipos de vegetação do NEB no período entre fevereiro de 2000 a julho de 2006. A área de estudo foi a região do NEB, sendo utilizado para caracterização dos tipos de vegetação um mapa de vegetação na escala de 1:5.000.000 do Instituto Brasileiro de Geografia e Estatística (IBGE). Um total de 140 imagens EVI livres de nuvens com resolução espacial de 250 m foram adquiridas da Agência Nacional Aeroespacial Norteamericana (NASA). Quatro imagens CBERS-2/CCD com resolução espacial de 20 m foram também adquiridas do Instituto Nacional de Pesquisas Espaciais (INPE) para auxiliar na coleta das amostras de dados de EVI dos diferentes tipos de vegetação. Dados de precipitação das cidades de Petrolina e Pesqueira (Pernambuco), Barra do Corda e Carolina (Maranhão) localizadas nas regiões de vegetação do tipo Caatinga, Floresta Atlântica, Amazônia e Cerrado, respectivamente, foram utilizados para avaliar sua relação com os dados de EVI sob estas vegetações. Dados de EVI sobre área irrigada também foram utilizados para esta análise. Os resultados obtidos mostraram que: i) os dados multitemporais EVI de diferentes tipos de vegetação foram sensíveis às respectivas variações fenológicas, com os menores e maiores valores de EVI ocorrendo nos períodos de seca e chuva respectivamente; ii) a vegetação Amazônia apresentou a menor variação multitemporal dos valores de EVI, todavia apresentando os valores mais elevados, podendo-se justificar pela maior quantidade de folhas e por estarem sempre verdes; iii) a vegetação de caatinga analisada apresentou a maior variação dos valores de EVI, pois na época de seca, perde todas as folhas e na época de chuva, se torna verde devido a menor variabilidade da precipitação; iv) todos os dados de EVI das vegetações apresentaram relação significativa (valor-p<0,05) com a precipitação.

Land cover change detection in urban areas provides valuable data on loss of forest and agricultural land to residential and commercial development. Using Landsat 5 Thematic Mapper (1991) and Landsat 7 ETM+ (2000) imagery of Gwinnett County, GA, change images were obtained using image differencing of Normalized Difference Vegetation Index (NDVI), principal components analysis (PCA), and Tasseled Cap-transformed images. Ground truthing and accuracy assessment determined that land cover change detection using the NDVI and Tasseled Cap image transformation methods performed best in the study area, while PCA performed the worst of the three methods assessed. Analyses on vegetative and vegetation changes from 1991- 2000 revealed that these methods perform well for detecting changes in vegetation and/or vegetative characteristics but do not always correspond with changes in land use. Gwinnett County lost an estimated 13,500 hectares of vegetation cover during the study period to urban sprawl, with the majority of the loss coming from forested areas.

Subtropical deforestation in Latin America is thought to be driven by demand for agricultural land, particularly to grow soybeans. However, existing remote sensing methods that can differentiate crop types to verify this hypothesis require high spatial or spectral resolution data, or extensive ground truth information to develop training sites, none of which are freely available for much of the world. I developed a new method of crop classification based on the phenological signatures of crops extracted from multi-temporal MODIS vegetation indices. I tested and refined this method using the USDA Cropland Data Layer from Kansas, USA as a reference. I then applied the method to classify crop types for a study site in Pellegrini, Santiago Del Estero, Argentina. The results show that this method is unable to effectively separate summer crops in Pellegrini, but can differentiate summer crops and non-summer crops. Unmet assumptions about agricultural practices are primarily responsible for the ineffective summer crop classification, underlining the need for researchers to have a complete understanding of ground conditions when designing a remote sensing analysis.

Master of Arts
Department of Geography
Kevin P. Price
Due primarily to changes in land management practices, eastern redcedar (Juniperus virginiana L.), a native Kansas conifer, is rapidly invading onto valuable rangelands. The suppression of fire and increase of intensive grazing, combined with the rapid growth rate, high reproductive output, and dispersal ability of the species have allowed it to dramatically expand beyond its original range. There is a growing interest in harvesting this species for use as a biofuel. For economic planning purposes, density and biomass quantities for the trees are needed. Three methods are explored for mapping eastern redcedar and quantifying its biomass in Riley County, Kansas. First, a land cover classification of redcedar cover is performed using a method that utilizes a support vector machine classifier applied to a multi-temporal stack of Landsat TM satellite images. Second, a Small Unmanned Aircraft System (sUAS) is used to measure individual redcedar trees in an area where they are encroaching into a pasture. Finally, a hybrid approach is used to estimate redcedar biomass using high resolution multispectral and light detection and ranging (LiDAR) imagery. These methods showed promise in the forestry, range management, and bioenergy industries for better understanding of an invasive species that shows great potential for use as a biofuel resource.

L’Arganier [Argania spinosa (L.) Skeels] est un arbre de la famille des Sapotacées, endémique du sud-ouest marocain. C’est un arbre multi-usages, qui constitue une ressource primordiale pour les populations de cet espace semi-aride et aride du Maroc. Il constitue la clef-de-voûte de l’agro-écosystème traditionnel de l’arganeraie reposant sur un équilibre entre ressources et exploitation humaine, et joue également un rôle important dans la lutte contre la désertification et l’érosion. Actuellement, la menace de dégradation de l’arganeraie est une préoccupation majeure aussi bien pour la population que pour les scientifiques. On assiste en effet depuis plusieurs décennies à une diminution du couvert arboré, à la fois en surface occupée et en densité d’arbres. Face à cette préoccupation, nous avons étudié l’espace multidimensionnel de l’arganeraie en cherchant à identifier les principales caractéristiques de cet espace, ainsi que les facteurs responsables de sa dégradation. Ensuite, nous avons dévoilé l’originalité de cet espace à partir de son organisation sociale et spatiale, ainsi que le mode de fonctionnement et de gestion de ce territoire. Dans la deuxième partie nous avons montré l’apport de la télédétection spatiale et des systèmes d’information géographique pour la caractérisation de l’occupation du sol et l’identification des changements à partir d’un suivi diachronique, en utilisant une série d’images SPOT, Landsat, Google Earth, Ikonos. Nous avons également testé la possibilité d'évaluer la densité des arganiers à partir des images à haute résolution spatiale Ikonos et Google Earth. Nous concluons à la nécessité d’un suivi de ce territoire afin de pouvoir évaluer les changements et prendre les mesures d’aménagement et de protection nécessaires
The Argan [Argania spinosa (L.) Skeels] is a species of tree endemic to the calcareous semi-desert Sous valley of southwestern Morocco. It is the sole species in the genus Argania (family of Sapotaceae). It is a multi-purpose tree, and the main resource provider for the population of this semi-arid and arid area (source of forage, oil, timber and fuel). Argan is the keystone species of the traditional agro-ecosystem of the Berber society, ensuring a meta-stable equilibrium between resource availability and anthropic use; it plays a major role in preventing erosion and desertification damages.Currently, in spite of the Biosphere Reserve label attributed by UNESCO in 1998, the threat of degradation of the sparse Argan forest is a main concern for both local population and scientists. Since several decades, a decrease of extension area of the species and of tree density has been observed. According to this preoccupation, we have studied the multidimensional space of the Argan forest, in view of identifying its main features and the potential drivers of degradation processes. Then the originality of this area has been demonstrated through the assessment of its social and spatial organization, and of land-use and management practices.In the second part, we have shown the possible use of remotely sensed data and of Geographic Information Systems for surveying land-use/land-cover and for monitoring changes through a multi-temporal analysis of satellite images: SPOT, Landsat, Ikonos and Google Earth imagery. The evaluation of tree density has been performed through object-oriented classification of high spatial resolution satellite imagery (Ikonos, Google Earth). In conclusion, we recommend the effective use of a monitoring system to follow environmental changes in the Argan tree area, and to produce the detailed information needed for implementation of management and conservation strategies ensuring a sustainable development of the area

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior
This study had as objective to do a multi-temporal analysis from Arroio Grande watershed area, located in Santa Maria, RS, to detect the forest covering changes, its localization and quantification, as well as, to monitor the deforestation and regeneration processes and its main determinant factors. Four satellite images were used: LANDSAT 5 (1987), LANDSAT 5 (1995), LANDSAT 7 (2002) and CBERS 2 (2005). The SPRING applicative was used to elaborate the cartography dada basis and to digitally process the images. The images were segmented with threshold 10 to similarity and 20 to area and classified with Bhattacharya algorithm in the following land uses: forest, field, exposed soil, agriculture, irrigated agriculture and water lamina. After the images classification, a thematic maps cross was done with LEGAL programming. As a result, maps with the following land uses were obtained: forest maintenance, regeneration and deforestation. For a period of 18 years, the forest covering increased 25,59% or 10,24% in the area, mainly in the hillside and in the plateaus, changing from 14.135,42 ha (40,01%) in 1987 to 17.752,20 ha (50,25%) in 2005. However, there is still a great deficit of riparian forest in plain (depression), mainly due to rice cultivation. The obtained results show the potential of Remote Sensing and Geoprocessing techniques in mapping the land use. They also can be used to support researches, territorial planning, economical development and environmental preservation in this region. With the data bank obtained, it will be possible to create models able to simulate the forest covering dynamic in the studied area.
Este trabalho teve como objetivo realizar uma análise multitemporal da área da microbacia do Arroio Grande, localizada em Santa Maria, RS, a fim de detectar mudanças na cobertura florestal, sua localização e quantificação, além de monitorar os processos de desmatamento e regeneração e seus principais fatores determinantes. Foram utilizadas quatro imagens de satélite: LANDSAT 5 (1987), LANDSAT 5 (1995), LANDSAT 7 (2002) e CBERS 2 (2005). Utilizou-se o aplicativo SPRING para a elaboração da base de dados cartográficos e do processamento digital das imagens. As imagens foram segmentadas com limiar de 10 para similaridade e 20 para área e classificadas com auxílio do algoritmo Bhattacharya nos seguintes usos da terra: floresta, campo, solo exposto, agricultura, agricultura irrigada e lâmina d água. Após a classificação das imagens, foi realizado o cruzamento dos mapas temáticos com ajuda da programação LEGAL. Como resultado, obtiveram-se mapas com os seguintes usos da terra: manutenção florestal, regeneração e desmatamento, ou seja, o que permaneceu inalterado de uma época para outra, o que regenerou e o que foi desmatado. Para um período de 18 anos, a cobertura florestal aumentou 25,59% ou 10,24% da área da microbacia, principalmente na encosta (rebordo) e no planalto, passando de 14.135,42 ha (40,01%) em 1987 para 17.752,20 ha (50,25%) em 2005. Porém, ainda há um déficit muito grande de mata ciliar na planície (depressão), principalmente devido ao cultivo de arroz. Os resultados obtidos mostram o potencial das técnicas de Sensoriamento Remoto e Geoprocessamento no mapeamento do uso da terra. Também servem para apoiar as mais diversas iniciativas de pesquisa, planejamento territorial, desenvolvimento econômico e preservação ambiental nesta região. Com o banco de dados gerado, será possível confeccionar modelos capazes de simular a dinâmica da cobertura florestal na área pesquisada.

Conselho Nacional de Desenvolvimento Científico e Tecnológico
The monitoring of the use and coverage is very important when studying determined regions, just because it helps knowing the environmental reality and contributes to solve problems that can probably appear. This research was done from the images of ALOS and LANDSAT satellites. Its main objective was to have a multi-temporal analysis of Arroio Grande micro watershed, central region of Rio Grande do Sul. The specific purposes were to identify and to quantify the different classes of land use found in this micro watershed along the 1987, 1998, 2002, 2005, 2007 and 2009 periods, as well as cross the land use information to show the forest coverage changes during the 22 years of analysis. The software SPRING 5.1.7 was employed to classify the supervised images through Bhattacharya, a sorter algorithm, and the map spatial analysis was done through the Spatial Language of Algebraic Geoprocessing program with the same computational application. The classes of land use as forest, field, agriculture, irrigated agriculture, exposed soil and water layer were observed in the images of each year in this analysis. These classes were utilized in the spatial analysis of the forest coverage in which forest monitoring parameters have been defined (forest maintenance and regeneration, deforestation). In this research the principal results that have been noticed were the increase of 17,98% on the distributed forest coverage, mainly in the areas of bigger declination, and the reduction of 16,32% on the field area. The analysis of the spatial forest coverage has presented stability with the landscape, in a gradual progression, because the area of forest maintenance, found in these 22 years, was 12.252,60ha, the forest regeneration was 4.389,12ha and only 1.853,82ha of deforested area.
O monitoramento do uso e cobertura da terra faz-se importante no estudo de determinadas regiões, pois auxilia no conhecimento da realidade ambiental e contribui na busca por soluções de problemas que possam se apresentar. A partir do uso de imagens dos satélites, ALOS e LANDSAT, foi realizada esta pesquisa com o objetivo principal de fazer uma análise multitemporal na microbacia do Arroio Grande, região central do Rio Grande do Sul, cujos objetivos específicos foram: Identificar e quantificar as diferentes classes de uso da terra encontradas na microbacia nos períodos de 1987, 1998, 2002, 2005, 2007 e 2009; além de cruzar as informações de uso da terra, evidenciando a cobertura florestal que sofreu alterações no decorrer dos 22 anos de análise. Para tanto, foram utilizados o software SPRING 5.1.7 para a classificação supervisionada das imagens, com a adoção do algoritmo classificador Bhattacharya, e a análise espacial dos mapas com a programação LEGAL do mesmo aplicativo computacional. Para esta análise, foram observadas as classes de uso do solo: floresta, campo, agricultura, agricultura irrigada, solo exposto e lâmina d água, nas imagens de cada ano. Estas classes foram empregadas na análise espacial da cobertura florestal em que foram definidos parâmetros para o monitoramento florestal (manutenção florestal, regeneração florestal e desmatamentos). Os principais resultados notados, nesta pesquisa, foram o aumento de 17,98% na cobertura florestal distribuída, principalmente, nas áreas de maiores declividade, e a redução de 16,32% sobre a área de campo. Quanto à análise espacial da cobertura florestal, esta mostrou-se em estabilidade com a paisagem, e em gradual progressão, pois a área de manutenção florestal encontrada, nestes 22 anos, foi de 12.252,60 ha, a regeneração florestal foi de 4.389,12 ha e apenas 1.853,82 ha de área desmatada.

Remotely sensed data has potential value for vegetation change detection and mapping in arid/semi-arid environments, which can be used and analysed to extract information relevant to the understanding of useful information for the study of desert environments and desertification monitoring, assessment and mapping. In Libya, the vast agricultural development which occurred during the last decades was accompanied by desertification ofdifferent areas. Desertification varied in type and degree according to the geographical site, irregularity of rainfall and the prevalence of strong wind which significantly affects the stability of the fragile ecosystem. The potential of this study was to offer answers to the understanding of desertification indicators and has identified criteria for desertification assessment and the creation of land degradation maps using remote sensing data and a geographic information system (GIS). The indicators which mainly impact the study area are wind erosion, vegetation degradation, salinization, and deterioration of water resources etc. Landsat TM imagery has been used as a source of data to monitor land cover and its change over large areas.In this study, multi-temporal Landsat TM imagery has been used in order to map land cover and their changes during five-year intervals from 1988 to 2000. This was achieved by using a soil adjusted vegetation index formula to detect vegetation Thealgorithm classification technique has been used to map vegetation cover, Eolian Mapping (EM) vegetation of various densities, by used the Soil Adjusted Vegetation Index (SA VI) images: TM 1988, TM 1992, TM 1996 and TM 2000. The results of this technique show areas that have vulnerability to wind erosion susceptibility. and change detection algorithm has been used to calculate the vegetation changes in theperiod from 1988 to 2000. This is therefore one land degradation factor that can be created from remotely sensed data. The analysis clearly demonstrates a net decrease in vegetation cover. This situation exemplifies the deterioration of the naturalvegetation cover. The information derived from remotely sensed data has beenintegrated in a GIS to identify relevant factors for developing a spatial model for desertification assessment and mapping. A Geographic Information System was used to combine and interpret a range of parameters (land cover, soil type, topography,climate, etc.).This This study presents an efficient methodology to delineate the land degradation factors in study area, in a GIS environment. In this study have used one of the multi-criteria decision-making techniques, Analytical Hierarchical Process (AHP) which provides a systematic approach for assessing and integrating the impact of various factors,involving several levels. The methodology has been present for computing a composite index of land degradation factors derived from topographical, land cover, soil type and climate data. All data are finally integrated in a GIS environment to prepare a final desertification map. This land degradation factors computed from AHP method not only considers susceptibility of each area to emphasize thevulnerability of land to erosion but also takes into account the factors that are related to desertification.

La paramétrisation du transfert de chaleur par télédétection, basée sur le schéma SEBS, s'est déjà avérée très adaptée pour l'estimation de l'évapotranspiration (ET) sur des surfaces naturelles homogènes. Cependant, l'utilisation d'une telle méthode pour des paysages hétérogènes (e.g. régions semi-arides ou surfaces agricoles) est plus délicate, puisque le principe de la théorie de la similarité est compromis par la présence de différentes sources de chaleur et de hauteurs variées. Dans un premier temps, cette thèse a pour objectif de proposer et d'évaluer différents modèles basés sur la géométrie de la végétation qui permettent d'estimer la longueur de rugosité pour le transfert de quantité de mouvement à la surface (z0m), cette dernière étant un paramètre clé dans la caractérisation du transfert de chaleur. En revanche, une telle investigation ne peut être menée qu'à une petite échelle et à l'aide de données de télédétection très haute résolution permettant ainsi une description très détaillée de la surface. Ensuite, le second aspect de ce travail est de caractériser le transfert de chaleur dans le cas d'études régionales. Puis, la capacité de SEBS à estimer les flux de chaleur turbulents à de grandes échelles spatiales et temporelles sera évaluée. Pour ce faire, l’approche multi-échelle de SEBS (MSSEBS) a été implémentée afin de traiter une zone de 2,4 millions km2, incluant le Plateau du Tibet et l’amont des principaux fleuves d’Asie du sud-est. La combinaison de données horaires de température de surface FY-2 avec un rayonnement net journalier et des paramètres de surface avancés, permet de produire une série temporelle d’ET sur le Plateau du Tibet pour la période 2008-2010, et à une fréquence journalière
The parameterization of heat transfer by remote sensing, and based on SEBS scheme for turbulent heat fluxes retrieval, already proved to be very convenient for estimating evapotranspiration (ET) over homogeneous land surfaces. However, the use of such a method over heterogeneous landscapes (e.g. semi-arid regions or agricultural land) becomes more difficult, since the principle of similarity theory is compromised by the presence of different heat sources with various heights. This thesis aims at first to propose and evaluate some models based on vegetation geometry for retrieving the surface roughness length for momentum transfer (z0m), which is a key parameter in the characterization of heat transfer. Such an investigation can only be led at a small scale with very-high resolution remote sensing data, for a precise description of the land surface. Therefore, the second aspect of this work is to determine how to address the characterization of heat transfer for regional studies. Then, the reliability of SEBS for estimating turbulent heat fluxes at large spatial and temporal scales has been evaluated. To do so, the Multi-Scale SEBS approach (MSSEBS) has been implemented for a 2.4 million km2 area including the Tibetan Plateau and the headwaters of the major rivers of East and South Asia. The addition of gap-filled hourly FY-2 LST data to advanced daily averaged net radiation and land surface parameters, allows to compute time-series of land surface ET over the Tibetan Plateau during the period 2008-2010, and on a daily basis

The goal of this study is to develop a smoke detecting algorithm using digital image processing techniques on multi-spectral (visible & infrared) video. By utilizing principal component analysis (PCA) followed by spatial filtering of principal component images the location of smoke can be accurately identified over a period of exposure time with a given frame capture rate. This result can be further analyzed with consideration of wind factor and fire detection range to determine if a fire is present within a scene. Infrared spectral data is shown to contribute little information concerning the smoke signature. Moreover, finalized processing techniques are focused on the blue spectral band as it is furthest away from the infrared spectral bands and because it experimentally yields the largest footprint in the processed principal component images in comparison to other spectral bands. A frame rate of .5 images/sec (1 image every 2 seconds) is determined to be the maximum such that temporal variance of smoke can be captured. The study also shows eigenvectors corresponding to the principal components that best represent smoke and are valuable indications of smoke temporal signature. Raw video data is taken through rigorous pre-processing schemes to align frames from respective spectral band both spatially and temporally. A multi-paradigm numerical computing program, MATLAB, is used to match the field of view across five spectral bands: Red, Green, Blue, Long-Wave Infrared, and Mid-Wave Infrared. Extracted frames are aligned temporally from key frames throughout the data capture. This alignment allows for more accurate digital processing for smoke signature. v Clustering analysis on RGB and HSV value systems reveal that color alone is not helpful to segment smoke. The feature values of trees and other false positives are shown to be too closely related to features of smoke for in solely one instance in time. A temporal principal component transform on the blue spectral band eliminates static false positives and emphasizes the temporal variance of moving smoke in images with higher order. A threshold adjustment is applied to a blurred blue principal component of non-unity principal component order and smoke results can be finalized using median filtering. These same processing techniques are applied to difference images as a more simple and traditional technique for identifying temporal variance and results are compared.

碩士
國立成功大學
測量工程學系
85
Rice-field inventory is the principal mission of the Food Bureau of Taiwan. In the past years, the inventory is based on aerial photos interpreted by human operators. Due to the large number of rice fields, this task requires a large amount of photos and processing time. In this research, we suggest an approach that incorporate existing cultivating-field maps, multi -temporal SPOTXS images, and the spectral knowledge of rice growth to improve the automationof interpretation of rice fields. Spectral data are the key factors for satellite image interpretation. Because the land cover of rice fields changes as the rice grows, their spectral reflectance characteristics will change respectively. Selecting images at appropriate rice growing stages will help theinterpretation of rice-field. To achieve this goal, a thorough understanding about the knowledge of rice is the first requirement. In this research, the domain knowledge about rice interpretation includes knowledge of rice growth, the land-cover types of varying periods, the spectral characteristic of features, and their locations. We adopted region-based classification in which the regions are provided by using cultivating-field maps. Our experiment consists of three steps. First, the rice transplanting date is determined based on its location. Such information is served as the basis for deciding which rice growing stages of the images are related to. Second, we analyze spectral statistics of rice fields to determine the appropriate combinations of bands for interpreting images in different periods of time. Images in different rice growing stages are further processed to distinguish rice fields and non-rice fields. Finally, ISODATA algorithm is introduced for automatic interpretation experiment based on the region provided by cultivating-field maps. SPOT images in 1993 and 1995 were tested in this research. Three approaches, the NDVI method, the NDVI differences method, the disparate band differences method, were suggested and tested respectively. The results are compared against with the result of region-based maximum likelyhood classification. Except NDVI method, the suggested approaches provide competitive accuracy as the maximum likelyhood classification. We can therefore conclude that the "automatic" rice-field interpretation method suggested in this research is feasible and can provide competitive accuracy as the current methods do.

High-throughput phenotyping using high spatial, spectral, and temporal resolution remote sensing (RS) data has become a critical part of the plant breeding chain focused on reducing the time and cost of the selection process for the “best” genotypes with respect to the trait(s) of interest. In this study, the potential of accurate and reliable sorghum biomass prediction using hyperspectral and LiDAR data acquired by sensors mounted on UAV platforms is investigated. Experiments comprised multiple varieties of grain and forage sorghum, including some photoperiod sensitive varieties, providing an opportunity to evaluate a wide range of genotypes and phenotypes.

Feature extraction is investigated, where various novel features, as well as traditional features, are extracted directly from the hyperspectral imagery and LiDAR point cloud data and input to classical machine learning (ML) regression based models. Predictive models are developed for multiple experiments conducted during the 2017, 2018, and 2019 growing seasons at the Agronomy Center for Research and Education (ACRE) at Purdue University. The impact of the regression method, data source, timing of RS and field-based biomass reference data acquisition, and number of samples on the prediction results are investigated. R2 values for end-of-season biomass ranged from 0.64 to 0.89 for different experiments when features from all the data sources were included. Using geometric based features derived from the LiDAR point cloud and the chemistry-based features extracted from hyperspectral data provided the most accurate predictions. The analysis of variance (ANOVA) of the accuracies of the predictive models showed that both the data source and regression method are important factors for a reliable prediction; however, the data source was more important with 69% significance, versus 28% significance for the regression method. The characteristics of the experiments, including the number of samples and the type of sorghum genotypes in the experiment also impacted prediction accuracy.

Including the genomic information and weather data in the “multi-year” predictive models is also investigated for prediction of the end of season biomass. Models based on one and two years of data are used to predict the biomass yield for the future years. The results show the high potential of the models for biomass and biomass rank predictions. While models developed using one year of data are able to predict biomass rank, using two years of data resulted in more accurate models, especially when RS data, which encode the environmental variation, are included. Also, the possibility of developing predictive models using the RS data collected until mid-season, rather than the full season, is investigated. The results show that using the RS data until 60 days after sowing (DAS) in the models can predict the rank of biomass with R2 values of around 0.65-0.70. This not only reduces the time required for phenotyping by avoiding the manual sampling process, but also decreases the time and the cost of the RS data collections and the associated challenges of time-consuming processing and analysis of large data sets, and particularly for hyperspectral imaging data.

In addition to extracting features from the hyperspectral and LiDAR data and developing classical ML based predictive models, supervised and unsupervised feature learning based on fully connected, convolutional, and recurrent neural networks is also investigated. For hyperspectral data, supervised feature extraction provides more accurate predictions, while the features extracted from LiDAR data in an unsupervised training yield more accurate prediction.

Predictive models based on Recurrent Neural Networks (RNNs) are designed and implemented to accommodate high dimensional, multi-modal, multi-temporal data. RS data and weather data are incorporated in the RNN models. Results from multiple experiments focused on high throughput phenotyping of sorghum for biomass predictions are provided and evaluated. Using proposed RNNs for training on one experiment and predicting biomass for other experiments with different types of sorghum varieties illustrates the potential of the network for biomass prediction, and the challenges relative to small sample sizes, including weather and sensitivity to the associated ground reference information.

碩士
國立中央大學
國際永續發展碩士在職專班
102
The main objective of this research is aimed at producing multi-temporal land cover maps of the Western Gambia in order to monitor land-cover changes (LCC) using Remote Sensing techniques with a focus on urban growth during the period 1985 – 2013. The study area is the Western Gambia covering West Coast Region and part of North Bank Region lying between Latitudes 13o 36’ N and 13o 03’ N and Longitude 16o 54’ W and 16o 06’ W. About 63.49% of the country’s population resides here with a population density of 478 people per km2. For this purpose, multi-temporal Landsat images for 1985, 1999, and 2013 were acquired. The existing land cover reference maps collected and modified for accuracy assessment. These multi-temporal data were processed using spatial analysis tools of geo-referencing, Support Vector Machine (SVM) classification, and post- classification processes, to map the patterns and extent of land cover in the study area as well as determine the magnitude of changes between the years of interest. The result of the study showed that the built-up areas have been on a constant positive and mostly uncontrolled expansion from 46.42 km2 of the study area in 1985 to 97.35 km2 in 1999 and to 193.02 km2 in 2013. On the other hand, terrestrial vegetation has been on a steady decline, from 1,184.22 km2 in 1985 to 929.69 km2 in 2013, while the cultivation land experienced a slight increase in area. The mangrove forest is fairly stable in the past three decades. Population pressure is the major driving forces of LCC in the Western Gambia; therefore the government and other stakeholders should develop policies and strategies to achieve a balanced, coordinated and sustainable natural resources management.

碩士
國立成功大學
資訊工程學系碩博士班
92
Change detection in remote sensing has become an interesting topic of image processing. Conventionally, change detection in remote sensing depends on expert operators to visually analyze a series of multitemporal images. It usually takes a lot of human labor and time in order to achieve good performance. With development of computer technology, the work by human labor has gradually been accomplished by image processing. We still find it helpful to incorporate some operators’ domain knowledge into the early learning processing for successful change detection. For example, it depends on human operators for the selection of representative training samples in a supervised approach and the identification of change types resulting from an unsupervised approach. Nevertheless, an automatic change identification procedure is still in developing. In this study, a framework is proposed in order to fulfill the requirement of automation for change identification of remotely sensed images. First of all, a change detection process is performed to roughly discriminate between change and low-change areas based on textural and spectral features. Given a condition that class information in the first image is sufficient, those low-change samples are used for radiometric calibration and further provide class statistics for the second and subsequent images. By a post classification means, a series of multi-temporal images thus produce the change types with time. Our experimental results have demonstrated the feasibility of the proposed method.

Soil maps are fundamental for agricultural management. However, mapping soils is a difficult task because of their high spatial variability and the challenge of choosing representative field sites for soil analysis. Globally, soil information is becoming a prioritized agenda, due to the increasing demand for soil information for quantitative ecological, environmental and agronomic modelling. Hence, improved digital soil mapping techniques are required to fulfill this demand. The Plant Available Water-holding Capacity (PAWC) is a key soil property in most agricultural management activities as it determines the maximum water that can be readily extracted by plants. Globally, there is an increasing demand for spatially explicit soil PAWC data for understanding the potential consequences of climate change and development of adaptation strategies. The coarse resolution of current PAWC information limits the spatial detail of future predictions and decision support. Plant growth in water-limited Mediterranean climates is predominantly controlled by soil water availability. In rain-fed cropping systems, differences in PAWC can explain a large proportion of the spatial and temporal crop yield variability. The overall aim of this research was to develop a methodology to estimate spatial pattern of PAWC at a high spatial resolution using satellite-based remote sensing techniques. The underlying hypothesis is that the spatio-temporal plant growth patterns contain integrated information about soil properties and plant-soil-water interaction in the profile. The objective was to evaluate if phenological metrics derived from MODIS-NDVI (Moderate Resolution Imaging Spectroradiometer Normalized Difference Vegetation Index) can be used to infer about PAWC. The study was conducted in the South Australian agricultural region, which is one of the major grain producing regions of the country. Central to facilitating the research was the design and development of a flexible software package (CropPhenology) to extract phenological metrics that are indicators of crop condition at different growth stages. The CropPhenology package was developed in R to be used for analyzing data for all later stages of the project. It is available in the public domain repository GitHub. Initially, the sensitivity of remote sensing phenological metrics for differences in soil PAWC was assessed in a controlled situation. Phenological metrics for crop grown in soils of contrasting PAWC values under identical agricultural management were compared. The results identified potential phenological metrics to be used as indicators for soil PAWC. The findings support that the soil signal can be extracted from time-series vegetation growth dynamics. The research further evaluated the efficacy of the phenological metrics for assessment of spatio-temporal crop growth variability for management practices. The association between phenological metrics and management zones were analyzed in a South Australian cropping field. The result shows that phenological metrics have potential to inform about both spatial variability and temporal variability, highlighting a pathway towards alternative approaches for assessing the spatio-temporal variability in cropping fields. Finally, an approach was developed for spatial PAWC estimation. Multiple linear regression models were developed that analytically associate of the measured soil PAWC values with MODIS-NDVI phenological metrics. The PAWC map shows significant agreement with the landscape-scale soil map of the region with realistic detail of PAWC variability within the soil map units across management units. The evidence from this result indicates the potential of phenological metrics from satellite remote sensing for soil PAWC mapping at unprecedented detail over a broad regional extent. Advances in PAWC mapping as demonstrated in this thesis will improve models assessing future climate change development of adaptation strategies and will narrow the gap in spatial detail between regional decision making and farm-based precision agriculture.
Thesis (Ph.D.) (Research by Publication) -- University of Adelaide, School of Biological Sciences, 2017.

A research report submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in partial fulfilment of the requirement for the degree of Masters of Science (GIS and Remote Sensing) at the School of Geography, Archaeology & Environmental Studies
Tamarix spp, commonly known as tamarisk or salt cedar, belong to the family of Tamaricaceae. It is a phreaphytic halophyte with 55 species in the genus Tamarix. South Africa has one indigenous (Tamarix usneoides) and two exotic (T. ramosissima and T.chinensis). Not only are the exotic Tamarix species becoming infamous invaders, but their hybridisation with the indigenous T. usneoides is also complicating morphological discrimination between the different species, and the prospect of potential use of bio-control agents to curb invasion. Thus, lack of spatial information about the current and the past distribution of tamarisk have hampered the effort to control its invasion. This study aimed at investigating the use of multi-temporal remotely sensed data to map the exotic Tamarix invasion in the riparian ecosystem of the Western Cape Province of South Africa, where it predominantly occurs. Random Forest (RF) and Support Vector Machine (SVM) were tested to classify Tamarix and other land-cover types. Sentinel 2 data and Landsat OLI earth observation data were used to map the current and the temporal exotic Tamarix distribution between 2007 and 2018, respectively. This included mapping the current and the multi-temporal Tamarix extent of invasion using the multi-spectral sensors Sentinel 2 and Landsat 5 and 8, respectively. Sentinel 2 was able to detect and discriminate the exotic Tamarix spp invasion using RF and SVM algorithms. The Random Forest classification achieved an overall accuracy of 87.83% and kappa of 0.85, while SVM achieved an overall accuracy of 86.31% and kappa of 0.83. Multi-temporal Landsat data was able to map the current and previous extent of exotic Tamarix invasion for the period between 2007 and 2018. Six land-cover types were classified using SVM. The overall accuracies achieved for 2007, 2014 and 2018 were 87.66%, 91.10%, and 90.62% respectively, and the kappa were 0.85, 0.89, and 0.88, respectively. It was found that the exotic Tamarix invasion increased from 284.67 ha to 647.10 ha in De Rust area, 74.70 ha to 97.29 ha in Leeu Gamka and 215.01 ha to 544.41 ha in Prince Albert region in a period of 11 years. Sentinel 2 and Landsat data have shown the potential to be used in Tamarix mapping. The results obtained in this study would help in implementation of conservation and rehabilitation plans.
GR 2020

Current changes of biodiversity result almost exclusively from human activities. This anthropogenic conversion of natural ecosystems during the last decades has led to the so-called ‘biodiversity crisis’, which comprises the loss of species as well as changes in the global distribution patterns of organisms. Species richness is unevenly distributed worldwide. Altogether, 17 so-called ‘megadiverse’ nations cover less than 10% of the earth’s land surface but support nearly 70% of global species richness. Mexico, the study area of this thesis, is one of those countries. However, due to Mexico’s large extent and geographical complexity, it is impossible to conduct reliable and spatially explicit assessments of species distribution ranges based on these collection data and field work alone. In the last two decades, Species distribution models (SDMs) have been established as important tools for extrapolating such in situ observations. SDMs analyze empirical correlations between geo-referenced species occurrence data and environmental variables to obtain spatially explicit surfaces indicating the probability of species occurrence. Remote sensing can provide such variables which describe biophysical land surface characteristics with high effective spatial resolutions. Especially during the last three to five years, the number of studies making use of remote sensing data for modeling species distributions has therefore multiplied. Due to the novelty of this field of research, the published literature consists mostly of selective case studies. A systematic framework for modeling species distributions by means of remote sensing is still missing. This research gap was taken up by this thesis and specific studies were designed which addressed the combination of climate and remote sensing data in SDMs, the suitability of continuous remote sensing variables in comparison with categorical land cover classification data, the criteria for selecting appropriate remote sensing data depending on species characteristics, and the effects of inter-annual variability in remotely sensed time series on the performance of species distribution models. The corresponding novel analyses were conducted with the Maximum Entropy algorithm developed by Phillips et al. (2004). In this thesis, a more comprehensive set of remote sensing predictors than in the existing literature was utilized for species distribution modeling. The products were selected based on their ecological relevance for characterizing species distributions. Two 1 km Terra-MODIS Land 16-day composite standard products including the Enhanced Vegetation Index (EVI), Reflectance Data, and Land Surface Temperature (LST) were assembled into enhanced time series for the time period of 2001 to 2009. These high-dimensional time series data were then transformed into 18 phenological and 35 statistical metrics that were selected based on an extensive literature review. Spatial distributions of twelve tree species were modeled in a hierarchical framework which integrated climate (WorldClim) and MODIS remote sensing data. The species are representative of the major Mexican forest types and cover a variety of ecological traits, such as range size and biotope specificity. Trees were selected because they have a high probability of detection in the field and since mapping vegetation has a long tradition in remote sensing. The result of this thesis showed that the integration of remote sensing data into species distribution models has a significant potential for improving and both spatial detail and accuracy of the model predictions
Sämtliche aktuell zu beobachtenden Veränderungen in der Biodiversität lassen sich fast ausschließlich auf menschliche Aktivitäten zurückführen. In den letzten Jahrzehnten hat insbesondere die anthropogene Umwandlung bisher unberührter, natürlicher Ökosysteme zur sogenannten ‚Biodiversitätskrise‘ geführt. Diese umfasst nicht nur das Aussterben von Arten, sondern auch räumliche Verschiebungen in deren Verbreitungsgebieten. Global gesehen ist der Artenreichtum ungleich verteilt. Nur insgesamt 17 sogenannte ‚megadiverse‘ Länder, welche 10% der globalen Landoberfläche umfassen, beherbergen fast 70% der weltweiten Artenvielfalt. Mexiko, das Studiengebiet dieser Arbeit, ist eine dieser außerordentlich artenreichen Nationen. Aufgrund seiner großen Ausdehnung und geographischen Komplexität kann eine verlässliche und detaillierte räumliche Erfassung von Artverbreitungsgebieten in Mexiko jedoch nicht nur auf Basis dieser Datenbanken sowie von Feldarbeiten erfolgen. In den letzten beiden Jahrzehnten haben sich Artverbreitungsmodelle (Species distribution models, SDMs) als wichtige Werkzeuge für die räumliche Interpolation solcher in situ Beobachtungen in der Ökologie etabliert. Artverbreitungsmodelle umfassen die Analyse empirischer Zusammenhänge zwischen georeferenzierten Fundpunkten einer Art und Umweltvariablen mit dem Ziel, räumlich kontinuierliche Vorhersagen zur Wahrscheinlichkeit des Vorkommens der jeweiligen Art zu treffen. Mittels Fernerkundung können Umweltvariablen mit Bezug zu den biophysikalischen Eigenschaften der Landoberfläche in hohen effektiven räumlichen Auflösungen bereitgestellt werden. Insbesondere in den letzten drei bis fünf Jahren ist daher die Verwendung von Fernerkundungsdaten in der Artverbreitungsmodellierung sprunghaft angestiegen. Da es sich hierbei jedoch immer noch um ein sehr neues Forschungsfeld handelt, stellen diese meist nur Einzelstudien mit Beispielcharakter dar. Eine systematische Untersuchung zur Modellierung von Artverbreitungsgebieten mit Hilfe von Fernerkundungsdaten fehlt bisher. Diese Forschungslücke wurde in der vorliegenden Arbeit aufgegriffen. Hierzu wurden spezifische Untersuchungen durchgeführt, welche insbesondere folgende Aspekte betrachteten: die sinnvolle Verknüpfung von Klima- und Fernerkundungsdaten im Rahmen von Artverbreitungsmodellen, den quantitativen Vergleich von kontinuierlichen Fernerkundungsdaten und einer bestehenden kategorialen Landbedeckungsklassifikation, die Identifizierung von Kriterien zur Auswahl geeigneter Fernerkundungsprodukte, welche die Eigenschaften der Studienarten berücksichtigen, sowie der Einfluss inter-annueller Variabilität in fernerkundlichen Zeitreihen auf die Ergebnisse und Leistungsfähigkeit von Artverbreitungsmodellen. Die entsprechenden neuen Analysen wurden mit Hilfe des von Phillips et al. (2004) entwickelten Maximum Entropy Algorithmus zur Artverbreitungsmodellierung durchgeführt. Im Rahmen dieser Arbeit wurde ein umfangreicherer Datensatz an Fernerkundungsvariablen als in der bisherigen Literatur verwendet. Die entsprechenden Fernerkundungsprodukte wurden spezifisch aufgrund ihrer Eignung für die Beschreibung ökologisch relevanter Parameter, die sich auf die Verbreitungsgebiete von Arten auswirken, ausgewählt. Für den Zeitraum von 2001 bis 2009 wurden zwei Terra-MODIS Standardprodukte mit 1 km räumlicher und 16-tägiger zeitlicher Auflösung zu geglätteten, kontinuierlichen Zeitreihen zusammengefügt. Diese Produkte beinhalten den verbesserten Vegetationsindex (Enhanced Vegetation Index, EVI), Reflexionsgrade und die Landoberflächentemperatur (Land Surface Temperature, LST). Diese hochdimensionalen Zeitreihendaten wurden in insgesamt 18 phänologische sowie 35 statistische Maßzahlen überführt, welche auf der Basis einer umfassenden Sichtung der vorhandenen Literatur zusammengestellt wurden. Die Verbreitungsgebiete von zwölf Baumarten wurden mit Hilfe eines hierarchisch aufgebauten Ansatzes, welcher sowohl Klimadaten (WorldClim) als auch Fernerkundungsdaten des MODIS-Sensors berücksichtigt, modelliert. Die Studienarten sind repräsentativ für die in Mexiko vorkommenden Waldtypen und decken eine breite Spannweite ökologischer Eigenschaften wie Größe des Verbreitungsgebietes und Breite der ökologischen Nische ab. Als Studienobjekte wurden Bäume ausgewählt, weil sie im Feld mit hoher Wahrscheinlichkeit richtig erfasst werden und außerdem die fernerkundungsbasierte Kartierung von Vegetation bereits auf eine Vielzahl an Studien zurückgreifen kann. Durch die im Rahmen dieser Dissertation durchgeführten Untersuchungen konnte gezeigt werden, dass die Integration von Fernerkundungsdaten in Artverbreitungsmodelle ein signifikantes Potential zur Verbesserung der räumlichen Detailgenauigkeit und der Güte der Modellvorhersagen bietet

There are considerable expenditures by government and private forest industry to enhance the growth of forests and reduce time required for crop rotation. The effectiveness of some of these treatments is dependent on site productivity. In addition, as responsible stewards of the forest resource and habitat, it is important that the state of forests are actively monitored, especially in the face of a changing climate and increased rates of disturbance. This dissertation reports on the development of a method for estimating and mapping forest productivity. The Shawnigan Lake thinning and fertilization forest installation, established in 1971 by CFS, was selected as the study site largely for its rich mensuration history. Square treatment plots were 0.04ha in area and included two thinning levels (1/3 & 2/3 of the basal area), two fertilization treatments (224kg & 448kg N/ha) with repeated fertilizations and macronutrient experiments (S, P) and control plots. A sample of plots was selected for high precision ground based lidar reference surveys. In September of 2012 a multi-sensor airborne survey of SLP was conducted that collected high-density lidar (up to ~70pnts/m2) and VNIR imaging spectroscopy. A thorough empirical radiometric calibration was conducted in addition to a spatial calibration at the Victoria International Airport. A combination of area based height percentile, point density ratios and statistical moments with individual lidar tree metrics including height distribution and proximity metrics were generated. Topographic metrics were also generated from the lidar ground classified point cloud. A library of spectral indices was computed from the imaging spectrometer data, with an emphasis on those indices known to be associated with vegetation health. These metrics were summarized to the plot level for a coarse scale regression analysis. A control survey and ground based lidar was used to facilitate an individual tree based fine scale of analysis, where reference data could unambiguously be matched to airborne collected data through the projected positions. Regression analysis was conducted applying the best subset regression with exhaustive feature selection search criteria and included a critical evaluation of the resulting selected features. Models were investigated considering the data source and in combination, that is, lidar metrics were considered independent of spectroscopy as well as the converse, and lidar metrics in combination with spectral metrics. The contribution of this study is the revelation that existing area based point cloud metrics are highly correlated, potentially noisy and sensitive to variations in point density, resulting in unstable feature selection and coefficients in model building. The approach offered as an alternative is the gridded lidar treetops method, which is evidently lacking within the literature and which this study overwhelmingly advocates. Additionally, the breadth and diversity of metrics assessed, the size and quality of the reference data applied, and the fine spatial scale of analysis are unique within the research area. This study also contributes to the knowledge base, in that, productivity can be estimated by remote sensing technologies. The use of gridded generalizations of the individual tree approach reduced estimation errors for both structural and productivity attributes. At the plot-level, crown structure and crown health features best estimated productivity. This study emphasizes the dangers of empirical modeling; at the even-aged SLP installation, growth is strongly tied to structure and the extrapolation to other sites is expected to provide biased values. It is my perspective that physical lidar structural models of the dominant and co-dominant crown classes be used to augment spatially explicit tree and stand growth models. In addition, direct measures should be obtained by multi-temporal lidar surveys or as an alternative photogrammetric point clouds after an initial lidar survey to quantify growth and aid in calibrating growth models.
Graduate

A dissertation submitted to the Faculty of Science, University of the Witwatersrand, in fulfilment of the requirements for the Degree of Master of Science. Johannesburg, 20 January 2017.
Sand dune movements can be effectively monitored through the comparison of multitemporal satellite images. However, not all remote sensing platforms are suitable to study sand dunes. This study compares coarse (Landsat 7 and 8) and fine (Worldview 2) resolution platforms, specifically focussing on sand dunes within the Ubārī Sand Sea (Libya), and identified the average migration rate and direction for the linear dunes within a section of the Ubārī sand sea for the time period from 2002-2015 with the use of Landsat imagery. Two band combinations were compared with the use of two supervised classifications. The best combination was found to be red, green, blue and near-infrared band combination and the maximum likelihood classifier. The dune features, namely the crest, slope and interdunal areas were successfully classified based on both the coarse and fine resolution imagery, but the accuracy with which it can be classified are different between the two resolutions. The classifications based on the Worldview 2 imagery had overall accuracies ranging from 55.43 - 60.83% with kappa values of 0.3486 – 0.4225 compared to the overall accuracies and kappa values of the classifications based on the Landsat 8 imagery ranging from 52.11 – 64.67% and 0.3878 – 0.4927 respectively. An average migration rate of 8.64 (± 4.65) m/yr in a generally north western direction was calculated based on the analysis of remote sensing data with some variations in this rate and the size and shape of the dunes. It was found that although Worldview 2 imagery provides more accurate and precise mensuration data, and smaller dunes identified from Worldview data were not delineated clearly on the Landsat imagery. Landsat imagery is sufficient for the studying of dunes at a regional scale. This means that for studies concerned with the dune patterns and movements within sand seas, Landsat is sufficient. In studies where the specific dynamics of specific dunes are to be selected, a finer resolution is required; platforms such as Worldview are needed in order to gain more detailed insight and to link the past and present day climate and environmental change.
MT2017

Irrigated agriculture in the Khorezm region in the arid inner Aral Sea Basin faces enormous challenges due to a legacy of cotton monoculture and non-sustainable water use. Regional crop growth monitoring and yield estimation continuously gain in importance, especially with regard to climate change and food security issues. Remote sensing is the ideal tool for regional-scale analysis, especially in regions where ground-truth data collection is difficult and data availability is scarce. New satellite systems promise higher spatial and temporal resolutions. So-called light use efficiency (LUE) models are based on the fraction of photosynthetic active radiation absorbed by vegetation (FPAR), a biophysical parameter that can be derived from satellite measurements. The general objective of this thesis was to use satellite data, in conjunction with an adapted LUE model, for inferring crop yield of cotton and rice at field (6.5 m) and regional (250 m) scale for multiple years (2003-2009), in order to assess crop yield variations in the study area. Intensive field measurements of FPAR were conducted in the Khorezm region during the growing season 2009. RapidEye imagery was acquired approximately bi-weekly during this time. The normalized difference vegetation index (NDVI) was calculated for all images. Linear regression between image-based NDVI and field-based FPAR was conducted. The analyses resulted in high correlations, and the resulting regression equations were used to generate time series of FPAR at the RapidEye level. RapidEye-based FPAR was subsequently aggregated to the MODIS scale and used to validate the existing MODIS FPAR product. This step was carried out to evaluate the applicability of MODIS FPAR for regional vegetation monitoring. The validation revealed that the MODIS product generally overestimates RapidEye FPAR by about 6 to 15 %. Mixture of crop types was found to be a problem at the 1 km scale, but less severe at the 250 m scale. Consequently, high resolution FPAR was used to calibrate 8-day, 250 m MODIS NDVI data, this time by linear regression of RapidEye-based FPAR against MODIS-based NDVI. The established FPAR datasets, for both RapidEye and MODIS, were subsequently assimilated into a LUE model as the driving variable. This model operated at both satellite scales, and both required an estimation of further parameters like the photosynthetic active radiation (PAR) or the actual light use efficiency (LUEact). The latter is influenced by crop stress factors like temperature or water stress, which were taken account of in the model. Water stress was especially important, and calculated via the ratio of the actual (ETact) to the potential, crop-specific evapotranspiration (ETc). Results showed that water stress typically occurred between the beginning of May and mid-September and beginning of May and end of July for cotton and rice crops, respectively. The mean water stress showed only minor differences between years. Exceptions occurred in 2008 and 2009, where the mean water stress was higher and lower, respectively. In 2008, this was likely caused by generally reduced water availability in the whole region. Model estimations were evaluated using field-based harvest information (RapidEye) and statistical information at district level (MODIS). The results showed that the model at both the RapidEye and the MODIS scale can estimate regional crop yield with acceptable accuracy. The RMSE for the RapidEye scale amounted to 29.1 % for cotton and 30.4 % for rice, respectively. At the MODIS scale, depending on the year and evaluated at Oblast level, the RMSE ranged from 10.5 % to 23.8 % for cotton and from -0.4 % to -19.4 % for rice. Altogether, the RapidEye scale model slightly underestimated cotton (bias = 0.22) and rice yield (bias = 0.11). The MODIS-scale model, on the other hand, also underestimated official rice yield (bias from 0.01 to 0.87), but overestimated official cotton yield (bias from -0.28 to -0.6). Evaluation of the MODIS scale revealed that predictions were very accurate for some districts, but less for others. The produced crop yield maps indicated that crop yield generally decreases with distance to the river. The lowest yields can be found in the southern districts, close to the desert. From a temporal point of view, there were areas characterized by low crop yields over the span of the seven years investigated. The study at hand showed that light use efficiency-based modeling, based on remote sensing data, is a viable way for regional crop yield prediction. The found accuracies were good within the boundaries of related research. From a methodological viewpoint, the work carried out made several improvements to the existing LUE models reported in the literature, e.g. the calibration of FPAR for the study region using in situ and high resolution RapidEye imagery and the incorporation of crop-specific water stress in the calculation
Die vorliegende Arbeit beschäftigt sich mit der Modellierung regionaler Erntemengen von Baumwolle und Reis in der usbekischen Region Khorezm, einem Bewässerungsgebiet das geprägt ist von langjähriger Baumwoll-Monokultur und nicht-nachhaltiger Land- und Wassernutzung. Basis für die Methodik waren Satellitendaten, die durch ihre großflächige Abdeckung und Objektivität einen enormen Vorteil in solch datenarmen und schwer zugänglichen Regionen darstellen. Bei dem verwendeten Modell handelt es sich um ein sog. Lichtnutzungseffizienz-Modell (im Englischen Light Use Efficiency [LUE] Model), das auf dem Anteil der photosynthetisch aktiven Strahlung basiert, welcher von Pflanzen für das Wachstum aufgenommen wird (Fraction of Photosynthetic Active Radiation, FPAR). Dieser Parameter kann aus Satellitendaten abgeleitet werden. Das allgemeine Ziel der vorliegenden Arbeit war die Nutzung von Satellitendaten für die Ableitung der Erntemengen von Baumwolle und Reis. Dazu wurde ein Modell entwickelt, das sowohl auf der Feldebene (Auflösung von 6,5 m) als auch auf der regionalen Ebene (Auflösung von 250 m) operieren kann. Während die Ableitung der Erntemengen auf der Feldebene nur für ein Jahr erfolgte (2009), wurden sie auf der regionalen Ebene für den Zeitraum 2003 bis 2009 modelliert. Intensive Feldmessungen von FPAR wurden im Studiengebiet während der Wachstumssaison 2009 durchgeführt. Parallel dazu wurden RapidEye-Daten in ca. zweiwöchentlichem Abstand aufgezeichnet. Aus den RapidEye-Daten wurde der Normalized Difference Vegetation Index (NDVI) berechnet, der anschließend mit den im Feld gemessenen FPAR-Werten korreliert wurde. Die entstandenen Regressionsgleichungen wurden benutzt um Zeitserien von FPAR auf RapidEye-Niveau zu erstellen. Anschließend wurden diese Zeitserien auf die MODIS-Skala aggregiert um damit das MODIS FPAR-Produkt zu validieren (1 km), bzw. eine Kalibrierung des 8-tägigen 250 m NDVI-Datensatzes vorzunehmen. Der erste Schritt zeigte dass das MODIS-Produkt im Allgemeinen die RapidEye-basierten FPAR-Werte um 6 bis 15 % überschätzt. Aufgrund der besseren Auflösung wurde das kalibrierte 250 m FPAR-Produkt für die weitere Modellierung verwendet. Für die eigentliche Modellierung wurden neben den FPAR-Eingangsdaten noch weitere Daten und Parameter benötigt. Dazu gehörte z.B. die tatsächliche Lichtnutzungseffizienz (LUEact), welche von Temperatur- und Wasserstress beeinflusst wird. Wasserstress wurde berechnet aus dem Verhältnis von tatsächlicher (ETact) zu potentieller, feldfruchtspezifischer Evapotranspiration (ETc), die beide aus einer Kombination von Satelliten- und Wetterdaten abgeleitet wurden. Der durchschnittliche Wasserstress schwankte nur geringfügig von Jahr zu Jahr, mit Ausnahmen in den Jahren 2008 und 2009. Die Modellschätzungen wurden durch feldbasierte Ernteinformationen (RapidEye-Ebene) sowie regionale statistische Daten (MODIS-Ebene) evaluiert. Die Ergebnisse zeigten, dass beide Modellskalen regionale Ernteerträge mit guter Genauigkeit nachbilden können. Der Fehler für das RapidEye-basierte Modell betrug 29,1 % für Baumwolle und 30,4 % für Reis. Die Genauigkeiten für das MODIS-basierte Modell variierten, in Abhängigkeit des betrachteten Jahres, zwischen 10,5 % und 23,8 % für Baumwolle und zwischen -0,4 % und -19,4 % für Reis. Insgesamt gab es eine leichte Unterschätzung der Baumwoll- (Bias = 0,22) und Reisernte (Bias = 0,11) seitens des RapidEye-Modells. Das MODIS-Modell hingegen unterschätzte zwar auch die (offizielle) Reisernte (mit einem Bias zwischen 0,01 und 0,87), überschätzte jedoch die offiziellen Erntemengen für die Baumwolle (Bias zwischen -0,28 und -0,6). Die Evaluierung der MODIS-Skala zeigte dass die Genauigkeiten extrem zwischen den verschiedenen Distrikten schwankten. Die erstellten Erntekarten zeigten dass Erntemengen grundsätzlich mit der Distanz zum Fluss abnehmen. Die niedrigsten Erntemengen traten in den südlichsten Distrikten auf, in der Nähe der Wüste. Betrachtet man die Ergebnisse schließlich über die Zeit hinweg, gab es Gebiete die über den gesamten Zeitraum von sieben Jahren stets von niedrigen Erntemengen gekennzeichnet waren. Die vorliegende Studie zeigt, dass satellitenbasierte Lichtnutzungseffizienzmodelle ein geeignetes Werkzeug für die Ableitung und die Analyse regionaler Erntemengen in zentralasiatischen Bewässerungsregionen darstellen. Verglichen mit verwandten Studien stellten sich die ermittelten Genauigkeiten sowohl auf der RapidEye- als auch auf der MODIS-Skala als gut dar. Vom methodischen Standpunkt aus gesehen ergänzte diese Arbeit vorhanden LUE-Modelle um einige Neuerungen und Verbesserungen, wie z.B. die Validierung und Kalibrierung von FPAR für die Studienregion mittels Feld- und hochaufgelösten RapidEye-Daten und dem Einbezug von feldfrucht-spezifischem Wasserstress in die Modellierung

Over the past decades, improvements in remote sensing technology have led to mass proliferation of aerial imagery. This, in turn, opened vast new possibilities relating to land cover classification, cartography, and so forth. As applications in these fields became increasingly more complex, the amount of data required also rose accordingly and so, to satisfy these new needs, automated systems had to be developed. Geometric distortions in raw imagery must be rectified, otherwise the high accuracy requirements of the newest applications will not be attained. This dissertation proposes an automated solution for the pre-stages of multi-spectral satellite imagery classification, focusing on Fast Fourier Shift theorem based geo-rectification and multi-temporal cloud-cover correction. By automatizing the first stages of image processing, automatic classifiers can take advantage of a larger supply of image data, eventually allowing for the creation of semi-real-time mapping applications.

This study determines and predicts multi-temporal Land-Use-Land-Cover Change (LULC) in a peripheral urban landscape over a 22 year period in relation to the study area‘s greenery. A change detection analysis using post classification Maximum Likelihood algorithm on three multispectral SPOT-4 images was used to determine land-cover transformation. To predict future land coverage, a Land-Cover Change Modeler (LCM) and a Markov Chain were used. Results show that between the year 2000-2006, 2006-2011 and 2000-2011 the study area experienced varied changes in the different LULCs. Built-up areas increased by 10.08%, 3.15% and 13.23% in 2000-2006, 2006-2011, and 2000-2011 respectively. Areas covered by thicket decreased by 0.59% in 2000-2006 but increased by 0.56%, 0.07% in 2006-2011 and 2000-2011 respectively. Forest land-cover increased by 2.59% in 2000-2006, 2.82% in 2006-2011, and 5.41% in 2000-2011. Grassland declined by 8.46% and 2.64% in 2000-2006 and 2000-2011 respectively while degraded grassland declined by 3.62%, 12.45% and 16.07% in 2000-2006, 2006-2011, and 2000-2011 respectively. Projection results indicate a consistent pattern of growth or decline to those experienced between 2000-2011. This study provides insight into LULC patterns within the eThekwini metro area and offers invaluable understanding of the transformation of the urban green spaces. Key words: Land-Use-Land-Cover Change, Change detection, Land-Cover Change Modeler, Markov Chain Process, Land-Cover Change Prediction.
Thesis (M.Sc.)-University of KwaZulu-Natal, Pietermaritzburg, 2013.

Dissertation presented as the partial requirement for obtaining a Master's degree in Geographic Information Systems and Science
Realizou-se uma investigação sobre a aplicação do Modelo de Mistura Linear (MML) numa série multi-temporal. Depois de aplicado a uma sequência anual completa de imagens MODIS com 500 metros de resolução especial, o MML gerou um conjunto de imagens fracção com uma variação intra-anual da abundância de componentes puros vegetação, solo e sombra na mistura do pixel. No pré-processamento das imagens de MOD09A1 utilizou-se uma abordagem do Compósito do Valor Máximo, para obter os dados de base para input do modelo de MML. Baseado em 36 imagens fracção, caracterizaram-se 304 amostras, correspondendo cada amostra a uma de onze classes de ocupação do solo definidas neste estudo. Estabeleceu-se uma relação entre estas classes e o CLC2000. Para a classificação das unidades de ocupação do solo, baseadas na variação do perfil temporal dos componentes puros no pixel das amostras, recorreu-se à utilização de uma rede neuronal de mapas autoorganizados (SOM). A análise da qualidade dos resultados resultou da construção da matriz de convergência, sendo calculados os índices de exatidão global, do produtor e do utilizador para avaliação de resultados. O trabalho reporta uma análise descritiva dos resultados obtidos segundo a metodologia proposta e apresenta-se como um estudo preliminar para análise de mistura espectrais, numa vertente multi-temporal, para caracterização da ocupação do solo.

Given current rates of anthropogenic environmental change, combined with the increasing lethal and non-lethal mortality threat that human activities pose, there is a vital need to understand wildlife movement and behaviour in human-dominated landscapes to help inform conservation efforts and wildlife management. As long-term monitoring of wildlife populations using Global Positioning System (GPS) telemetry increases, there are new opportunities to quantify change in wildlife movement and behaviour. The objective of this PhD research is to develop novel methodological approaches for quantifying change in spatial-temporal patterns of wildlife movement and habitat by leveraging long time series of GPS telemetry and remotely sensed data. Analyses were focused on the habitat and movement of individuals in the threatened grizzly bear (Ursus arctos) population of Alberta, Canada, which occupies a human-dominated and heterogeneous landscape. Using methods in functional data analysis, a multivariate regionalization approach was developed that effectively summarizes complex spatial-temporal patterns associated with landscape disturbance, as well as recovery, which is often left unaccounted in studies quantifying patterns associated with disturbance. Next, the quasi-experimental framework afforded by a hunting moratorium was used to compare the influence of lethal (i.e., hunting) and non-lethal (i.e., anthropogenic disturbance) human-induced risk on antipredator behaviour of an apex predator, the grizzly bear. In support of the predation risk allocation hypothesis, male bears significantly decrease risky daytime behaviours by 122% during periods of high lethal human-induced risk. Rapid behavioural restoration occurred following the end of the hunt, characterized by diel bimodal movement patterns which may promote coexistence of large predators in human-dominated landscapes. A multi-scale approach using hierarchical Bayesian models, combined with post hoc trend tests and change point detection, was developed to test the influence of landscape disturbance and conditions on grizzly bear home range and movement selection over time. The results, representing the first longitudinal empirical analysis of grizzly bear habitat selection, revealed selection for habitat security at broad scales and for resource availability and habitat permeability at finer spatial scales, which has influenced potential landscape connectivity over time. Finally, combining approaches in movement ecology and conservation physiology, a body condition index was used to characterize how the physiological condition (i.e., internal state) of grizzly bears influences behavioral patterns due to costs and benefits associated with risk avoidance and resource acquisition. The results demonstrated individuals in poorer condition were more likely to engage in risky behaviour associated with anthropogenic disturbance, which highlights complex challenges for carnivore conservation and management of human-carnivore conflict. In summary, this dissertation contributes 1) a multivariate regionalization approach for quantifying spatial-temporal patterns of landscape disturbance and recovery applicable across diverse natural systems, 2) support for the growing theory that apex predators modify behavioural patterns to account for temporal overlap with lethal and non-lethal human-induced risk associated with humans, 3) an integrated approach for considering multi-scale spatial-temporal change in patterns of wildlife habitat selection and landscape connectivity associated with landscape change, 4) a cross-disciplinary framework for considering the impacts of the internal state on behavioural patterns and risk tolerance.
Graduate

With the advancement of technology and the development of more sophisticated remote sensing sensor systems, the use of satellite imagery has opened up various fields of exploration and application. There has been an increased interest in analysis of multi-temporal satellite image in the past few years because of the wide variety of possible applications of in both short-term and long-term image analysis. The type of changes that might be of interest can range from short-term phenomena such as flood assessment and crop growth stage, to long-term phenomena such as urban fringe development. This thesis studies flood assessment and land cover mapping of satellite images, and proposes nature inspired algorithms that can be easily implemented in realistic scenarios. Disaster monitoring using space technology is one of the key areas of research with vast potential; particularly flood based disasters are more challenging. Every year floods occur in many regions of the world and cause great losses. In order to monitor and assess such situations, decision-makers need accurate near real-time knowledge of the field situation. How to provide actual information to decision-makers for effective flood monitoring and mitigation is an important task, from the point of view of public welfare. Over-estimation of the flooded area leads to over-compensation to people, while under-estimation results in production loss and negative impacts on the population. Hence it is essential to assess the flood damage accurately, both in qualitative and quantitative terms. In such situations, land cover maps play a very critical role. Updating land cover maps is a time consuming and costlier operation when it is performed using traditional or manual methods. Hence, there is a need to find solutions for such problem through automation. Design of automatic systems dedicated to satellite image processing which involves change detection to discriminate areas of land cover change between imaging dates. The system integrates the spectral and spatial information with the techniques of image registration and pattern classification using nature inspired techniques. In the literature, various works have been carried out for solving the problem of image registration and pattern classification using conventional methods. Many researchers have proved, for different situations, that nature inspired techniques are promising in comparison with that of conventional methods. The main advantage of nature inspired technique over any other conventional methods is its stochastic nature, which converges to optimal solution for any dynamic variation in a given satellite image. Results are given in such terms as to delineate change in multi-date imagery using change-versus-no-change information to guide multi-date data analysis. The main objective of this study is to analyze spatio-temporal satellite data to bring out significant changes in the land cover map through automated image processing methods. In this study, for satellite image analysis of flood assessment and land cover mapping, the study areas and images considered are: Multi-temporal MODerate-resolution Imaging Spectroradiometer (MODIS) image around Krishna river basin in Andhra Pradesh India; Linear Imaging Self Scanning Sensor III (LISS III)and Synthetic Aperture Radar(SAR)image around Kosi river basin in Bihar, India; Landsat7thematicmapperimage from the southern part of India; Quick-Bird image of the central Bangalore, India; Hyperion image around Meerut city, Uttar Pradesh, India; and Indian pines hyperspectral image. In order to develop a flood assessment framework for this study, a database was created from remotely sensed images (optical and/or Synthetic Aperture Radar data), covering a period of time. The nature inspired techniques are used to find solutions to problems of image registration and pattern classification of a multi-sensor and multi-temporal satellite image. Results obtained are used to localize and estimate accurately the flood extent and also to identify the type of the inundated area based on land cover mapping. The nature inspired techniques used for satellite image processing are Artificial Neural Network (ANN), Genetic Algorithm (GA),Particle Swarm Optimization (PSO), Firefly Algorithm(FA),Glowworm Swarm Optimization(GSO)and Artificial Immune System (AIS). From the obtained results, we evaluate the performance of the methods used for image registration and pattern classification to compare the accuracy of satellite image processing using nature inspired techniques. In summary, the main contributions of this thesis include (a) analysis of flood assessment and land cover mapping using satellite images and (b) efficient image registration and pattern classification using nature inspired algorithms, which are more popular than conventional optimization methods because of their simplicity, parallelism and convergence of the population towards the optimal solution in a given search space.

Dissertation submitted in partial fulfillment of the requirements for the Degree of Master of Science in Geospatial Technologies.
Dhaka, the capital of Bangladesh, has undergone radical changes in its physical form, not only in its vast territorial expansion, but also through internal physical transformations over the last decades. In the process of urbanization, the physical characteristic of Dhaka is gradually changing as open spaces have been transformed into building areas, low land and water bodies into reclaimed builtup lands etc. This new urban fabric should be analyzed to understand the changes that have led to its creation. The primary objective of this research is to predict and analyze the future urban growth of Dhaka City. Another objective is to quantify and investigate the characteristics of urban land cover changes (1989-2009) using the Landsat satellite images of 1989, 1999 and 2009. Dhaka City Corporation (DCC) and its surrounding impact areas have been selected as the study area. A fisher supervised classification method has been applied to prepare the base maps with five land cover classes. To observe the change detection, different spatial metrics have been used for quantitative analysis. Moreover, some postclassification change detection techniques have also been implemented. Then it is found that the ‘builtup area’ land cover type is increasing in high rate over the years. The major contributors to this change are ‘fallow land’ and ‘water body’ land cover types. In the next stage, three different models have been implemented to simulate the land cover map of Dhaka city of 2009. These are named as ‘Stochastic Markov (St_Markov)’ Model, ‘Cellular Automata Markov (CA_Markov)’ Model and ‘Multi Layer Perceptron Markov (MLP_Markov)’ Model. Then the best-fitted model has been selected based on various Kappa statistics values and also by implementing other model validation techniques. This is how the ‘Multi Layer Perceptron Markov (MLP_Markov)’ Model has been qualified as the most suitable model for this research. Later, using the MLP_Markov model, the land cover map of 2019 has been predicted. The MLP_Markov model shows that 58% of the total study area will be converted into builtup area cover type in 2019. The interpretation of depicting the future scenario in quantitative accounts, as demonstrated in this research, will be of great value to the urban planners and decision makers, for the future planning of modern Dhaka City.