The subtropical forests located along South Africa’s Wild Coast region, declared as one of the biodiversity hotspots, provide benefits to the local and national economy. However, there is evidence of increased pressure exerted on the forests by growing population and reduced income from activities not related to forest products. The ability of remote sensing to quantify subtropical forest changes over time, perform species discrimination (using field spectroscopy) and integrating field spectral and multispectral data were all assessed in this study. Investigations were conducted at pixel, leaf and sub-pixel levels. Both per-pixel and sub-pixel classification methods were used for improved forest characterisation. Using SPOT 6 imagery for 2013, the study determined the best classification algorithm for mapping sub-tropical forest and other land cover types to be the maximum likelihood classifier. Maximum likelihood outperformed minimum distance, spectral angle mapper and spectral information divergence algorithms, based on overall accuracy and Kappa coefficient values. Forest change analysis was made based on spectral measurements made at top of the atmosphere (TOC) level. When applied to the 2005 and 2009 SPOT 5 images, subtropical forest changes between 2005-2009 and 2009-2013 were quantified. A temporal analysis of forest cover trends in the periods 2005-2009 and 2009-2013 identified a decreasing trend of -3648.42 and -946.98 ha respectively, which translated to 7.81 percent and 2.20 percent decrease. Although there is evidence of a trend towards decreased rates of forest loss, more conservation efforts are required to protect the Wild Coast ecosystem. Using field spectral measurements data, the hierarchical method (comprising One-way ANOVA with Bonferroni correction, Classification and Regression Trees (CART) and Jeffries Matusita method) successfully selected optimal wavelengths for species discrimination at leaf level. Only 17 out of 2150 wavelengths were identified, thereby reducing the complexities related to data dimensionality. The optimal 17 wavelength bands were noted in the visible (438, 442, 512 and 695 nm), near infrared (724, 729, 750, 758, 856, 936, 1179, 1507 and 1673 nm) and mid-infrared (2220, 2465, 2469 and 2482 nm) portions of the electromagnetic spectrum. The Jeffries-Matusita (JM) distance method confirmed the separability of the selected wavelength bands. Using these 17 wavelengths, linear discriminant analysis (LDA) classified subtropical species at leaf level more accurately than partial least squares discriminant analysis (PLSDA) and random forest (RF). In addition, the study integrated field-collected canopy spectral and multispectral data to discriminate proportions of semi-deciduous and evergreen subtropical forests at sub-pixel level. By using the 2013 land cover (using MLC) to mask non-forested portions before sub-pixel classification (using MTMF), the proportional maps were a product of two classifiers. The proportional maps show higher proportions of evergreen forests along the coast while semi-deciduous subtropical forest species were mainly on inland parts of the Wild Coast. These maps had high accuracy, thereby proving the ability of an integration of field spectral and multispectral data in mapping semi-deciduous and evergreen forest species. Overall, the study has demonstrated the importance of the MLC and LDA and served to integrate field spectral and multispectral data in subtropical forest characterisation at both leaf and top-of-atmosphere levels. The success of both the MLC and LDA further highlighted how essential parametric classifiers are in remote sensing forestry applications. Main subtropical characteristics highlighted in this study were species discrimination at leaf level, quantifying forest change at pixel level and discriminating semi-deciduous and evergreen forests at sub-pixel level.

No description available.

No description available.

Semi-natural grasslands are amongst the most biodiverse ecosystems in Europe, though their importance they are experiencing a declining trend. To monitor and assess the health of these ecosystems is generally costly, personnel demanding and time-consuming. With satellite imagery and machine learning becoming more accessible, this can offer a cheap and effective way to gain ecological information about semi-natural grasslands.This thesis explores the possibilities to predict plant species richness in semi-natural grasslands with high resolution satellite imagery through machine learning. Five different machine learning models were employed with various subsets of spectral- and geographical features to see how they performed and why. The study area was in southern Sweden with satellite and survey data from the summer of 2019.Geographical features were the features that influenced the machine learning models most. This can be explained by the geographical spread of the semi-natural grasslands, as well as difficulties in finding correlations in the relatively noisy satellite data. The most important spectral features were found in the red edge- and the short-wave infrared spectrums. These spectrums represent leaf chlorophyll content and water content in vegetation, respectively. The most accurate machine learning model was Random Forest when it was trained using with all the spectral- and geographical features. The other models; Logistic Regression, Support Vector Machine, Voting Classifier and Neural Network, showed general inabilities to interpret feature subsets containing the spectral data.This thesis shows that with deeper knowledge about the satellite-biodiversity relationship and how to apply it with machine learning have the possibilities of cheaper, more efficient and standardized monitoring of ecologically valuable areas such as semi-natural grasslands.

No description available.

abstract: There are more than 20 active missions exploring planets and small bodies beyond Earth in our solar system today. Many more have completed their journeys or will soon begin. Each spacecraft has a suite of instruments and sensors that provide a treasure trove of data that scientists use to advance our understanding of the past, present, and future of the solar system and universe. As more missions come online and the volume of data increases, it becomes more difficult for scientists to analyze these complex data at the desired pace. There is a need for systems that can rapidly and intelligently extract information from planetary instrument datasets and prioritize the most promising, novel, or relevant observations for scientific analysis. Machine learning methods can serve this need in a variety of ways: by uncovering patterns or features of interest in large, complex datasets that are difficult for humans to analyze; by inspiring new hypotheses based on structure and patterns revealed in data; or by automating tedious or time-consuming tasks. In this dissertation, I present machine learning solutions to enhance the tactical planning process for the Mars Science Laboratory Curiosity rover and future tactically-planned missions, as well as the science analysis process for archived and ongoing orbital imaging investigations such as the High Resolution Imaging Science Experiment (HiRISE) at Mars. These include detecting novel geology in multispectral images and active nuclear spectroscopy data, analyzing the intrinsic variability in active nuclear spectroscopy data with respect to elemental geochemistry, automating tedious image review processes, and monitoring changes in surface features such as impact craters in orbital remote sensing images. Collectively, this dissertation shows how machine learning can be a powerful tool for facilitating scientific discovery during active exploration missions and in retrospective analysis of archived data. / Dissertation/Thesis / Doctoral Dissertation Exploration Systems Design 2019

abstract: In this thesis, I investigate possible formation processes in the northern Claritas Fossae and the large Thaumasia graben on Mars. In particular, I assess three proposed formation hypotheses for the region: a mega-landslide across the Thaumasia plateau, originating in Tharsis and moving to the south-west; a rift system pulling apart Claritas Fossae and opening the large Thaumasia graben generally propagating in a north-south direction: and extension caused by uplifting from underlying dike swarms. Using digital terrain models (DTMs) from the High Resolution Stereo Camera (HRSC) aboard Mars Express and visual images from the Context Camera (CTX) aboard the Mars Reconnaissance Orbiter (MRO), I analyzed the geomorphic and structural context of the region. Specifically, I produced geomorphologic and structural feature maps, conducted sector diagram analyses of fault propagation direction, calculated and compared extension and strain in local and regional samples, analyzed along strike throw-profiles of faults, and conducted surface age estimates through crater counting. I found that no single formation mechanism fully explains the surface features seen in Northern Claritas Fossae today. Instead I, propose the following sequence of events led to the surface characteristics we now observe. The region most likely underwent two episodes of uplift and extension due to sub-surface magmatic intrusions, then experienced an extensional event which produced the large Thaumasia graben. This was followed by the emplacement of a layer of lava burying the bottom of the Thaumasia graben and the eastern edge of the region. Additional extension followed across the eastern portion of the study area, and finally of a young lava flow was emplaced abutting and overprinting the southwestern edge. / Dissertation/Thesis / Masters Thesis Geological Sciences 2019

No description available.

No description available.

No description available.