The spectral separability of southern pines is a perplexing issue due to limited variance of spectral reflectance in species with similar morphological characteristics. Understory vegetation reflectance may exacerbate the ability to accurately identify various overstory tree species, specifically those of longleaf and loblolly pines in the southeastern US. In this study, identification of target level overstory crowns with varying degrees of understory vegetation cover based on fire return frequency was used to assess the role of understory reflectance on target crown species discernment. Seasonal variations of understory vegetation in late dormant and late growing seasons were compared for disparities in potential reflectance contribution from understory vegetation. Overall, the impact of understory vegetation was considered negligible in the spectral separability of longleaf and loblolly pines based on discriminant analysis results. Classification of WorldView-2 relative spectral profiles resulted in overall accuracies of 92% for dormant season and 96% for growing season imagery.

Primary productivity is a fundamental ecological process and an important measure of ecosystem response to environmental change. Currently, there is a considerable lapse in our understanding of primary productivity in hot and cold deserts, due to the difficulty of measuring production in cryptogam vegetation. However, remote sensing can provide long-term, spatially-extensive estimates of primary production and are particularly well suited to remote environments, such as in the McMurdo Dry Valleys (MDV) of Antarctica, where cyanobacterial communities are the main drivers of primary production. These microbial communities form multi-layered sheets (i.e., microbial mats) on top of desert pavement. The cryptic nature of these communities, their often patchy spatial distribution, and their ability to survive desiccation make assessments of productivity challenging. I used field-based surveys of microbial mat biomass and pigment chemistry in conjunction with analyses of multispectral satellite data to examine the distribution and activity of microbial mats. This is the first satellite-derived estimate of microbial mat biomass for Antarctic microbial mat communities. I show strong correlations between multispectral satellite data (i.e., NDVI) and ground based measurements of microbial mats, including ground cover, biomass, and pigment chemistry. Elemental (C, N) and isotopic composition (15N, 13C) of microbial mats show that they have significant effects on biogeochemical cycling in the soil and sediment of this region where they occur. Using these relationships, I developed a statistical model that estimates biomass (kg of C) in selected wetlands in the Lake Fryxell Basin, Antarctica. Overall, this research demonstrates the importance of terrestrial microbial mats on C and N cycling in the McMurdo Dry Valleys, Antarctica. / Master of Science / Primary productivity is an essential ecological process and a useful measure of how ecosystems respond to climate change. Primary production is more difficult to measure in polar desert ecosystems where there is little to no vascular vegetation. Polar regions are also ecosystems where we expect to see significant responses to a changing climate. Remote sensing and image analysis can provide estimates of primary production and are particularly useful in remote environments. For example, in the McMurdo Dry Valleys (MDV) of Antarctica, cyanobacterial communities are the main primary producers. These microbial communities form multi-layered sheets (i.e., microbial mats) on top of rocks and soil. These communities are cryptic, do not cover large areas of ground continuously, and are able to survive desiccation and freezing. All of these characteristics make assessments of productivity especially challenging. For my master’s research, I collected microbial mat samples in conjunction with the acquisition of a satellite image of my study area in the MDV, and I determined biological parameters (e.g., percent ground cover, organic matter, and chlorophyll-a content) through laboratory analyses using these samples. I used this satellite image to extract spectral data and perform a vegetation analysis using the normalized difference vegetation index (i.e., NDVI), which determines areas in the image that contain vegetation (i.e., microbial mats). By linking the spectral data to the biological parameters, I developed a statistical model that estimates biomass (i.e., carbon content) of my study areas. These are the first microbial mat biomass estimates using satellite imagery for this region of Antarctica. Additionally, I researched the importance of microbial mats on nitrogen cycling in Taylor Valley. Using elemental and isotopic analyses, I determined microbial mats have significant effects on the underlying soil and nutrient cycling. Overall, this research demonstrates the importance of terrestrial microbial mats on C and N fixation in Antarctic soil environments.

Les peuplements forestiers sont une unité de mesure de base pour l'inventaire forestier et la cartographie. Ils sont définis comme de grandes zones forestières (par exemple, de plus de 2 ha) et de composition homogène en terme d'essence d'arbres et d'âge. Leur délimitation précise est généralement effectuée par des opérateurs humains grâce à une analyse visuelle d'images infrarouges à très haute résolution (VHR). Cette tâche est fastidieuse, nécessite beaucoup de temps et doit être automatisée pour un suivi de l'évolution et une mise à jour efficace. Une méthode fondée sur la fusion des données lidar aéroportées et des images multispectrales VHR est proposée pour la délimitation automatique des peuplements forestiers contenant une essence dominante (pureté supérieure à 75%). C'est une principale tâche préliminaire pour la mise à jour de la base de données de la couverture forestière. Les images multispectrales donnent des informations sur les espèces d'arbres alors que les nuages de point Lidar 3D fournissent des informations géométriques sur les arbres et permettent leur extraction individuelle. Les attributs multimodaux sont calculées, à la fois au niveau des pixels et des objets (groupements de pixels ayant une taille similaire aux arbres). Une classification supervisée est ensuite effectuée au niveau de l'objet afin de discriminer grossièrement les espèces d'arbres existantes dans chaque zone d'intérêt. Les résultats de la classification sont ensuite traités pour obtenir des zones homogènes avec des bordures lisses par la minimisation d'une énergie, où des contraintes supplémentaires sont proposées pour former la fonction énergie à minimiser. Les résultats expérimentaux montrent que la méthode proposée fournit des résultats très satisfaisants en termes d'étiquetage et de délimitation, et ce pour des régions géographiquement très éloignées / Forest stands are the basic units for forest inventory and mapping. Stands are defined as large forested areas (e.g., 2 ha) of homogeneous tree species composition and age. Their accurate delineation is usually performed by human operators through visual analysis of very high resolution (VHR) infra-red images. This task is tedious, highly time consuming, and should be automated for scalability and efficient updating purposes. A method based on the fusion of airborne lidar data and VHR multispectral images is proposed for the automatic delineation of forest stands containing one dominant species (purity superior to 75%). This is the key preliminary task for forest land-cover database update. The multispectral images give information about the tree species whereas 3D lidar point clouds provide geometric information on the trees and allow their individual extraction. Multi-modal features are computed, both at pixel and object levels: the objects are groups of pixels having a size similar to trees. A supervised classification is then performed at the object level in order to coarsely discriminate the existing tree species in each area of interest. The classification results are further processed to obtain homogeneous areas with smooth borders by employing an energy minimum framework, where additional constraints are joined to form the energy function. The experimental results show that the proposed method provides very satisfactory results both in terms of stand labeling and delineation, even for spatially distant regions

Master of Science / Department of Biological & Agricultural Engineering / Ajay Sharda / Discovered in 1971, sudden death syndrome (SDS), caused by the fungus Fusarium virguliforme, has spread from the US to South American and European countries. It has potential to infect soybean crops worldwide, causing yield losses of 10% to 15% and even 70% in extreme cases. There is a need for rapid spatial assessment of SDS. Currently, the extent and severity of SDS are scored using visual symptoms as indicators. This method can take hours to collect and is subject to human bias and changing environmental conditions. Color infrared (CIR) and thermal infrared (TIR) imagery detect changes in light reflectance (visible and near-infrared bands) and emittance (canopy temperature), respectively. Stressed crops may show deviations in light reflectiveness, as well as elevated canopy temperatures. The use of CIR and TIR imagery and flexible aerial remote sensing platforms offer an alternative for SDS detection and diagnosis compared to hand scoring methods. Crop stress and diseases have been detected using manned and unmanned aerial systems previously. Yet, to date, SDS has not been remotely assessed using CIR or TIR imagery collected with aerial platforms. The following research utilizes high throughput CIR and TIR imagery collected using a small unmanned aerial system (sUAS) to detect and assess SDS. A comparative evaluation of ground-based and aerial CIR methods for assessing SDS was conducted to understand the effectiveness of novel aerial SDS detection methods. Furthermore, a TIR case study investigating the use of potential thermal canopy changes for SDS detection was conducted to investigate the possibility of using TIR as an SDS indicator. CIR reflectance measured from a ground-based spectrometer and sUAS was collected data over a two-year period. Ground-based spectrometer data were collected weekly, while a sUAS collected aerial imagery late in the growing season each year before plant maturity. Pigment index (PI) values were derived from ground-based and aerial data. Results showed a strong negative correlation between SDS score and PI values. Aerial and ground-based data both showed strong correlations to SDS score, however, aerial data displayed a stronger relationship possibly due to minimal changes in environmental conditions. High SDS scores correlated strongly to aerial derived PI (R2 = 0.8359). Rapidly assessed high SDS allows for accurate screening of SDS critical for soybean breeding. The second year of the study investigated each component of SDS score, severity, and incidence. PI proved to have the best correlation with severity (R2 = 0.6313 and ρ = -0.8016) rather than incidence or SDS score. PI also correlated to SDS scores with R2 = 0.6159 and ρ = -0.7916. A sUAS mounted TIR camera collected imagery four times during the growing season when SDS foliar symptoms were just starting to appear. At the start of the study period, the correlation between canopy temperature and SDS is low (ρ = -0.2907), but increases over the growing season as SDS prevalence increases ending with a strong correlation (ρ = -0.7158). Early identification of SDS leads to the implementation of mitigation practices and changes in irrigation scheduling before the disease reaches severe symptoms. Early mitigation of SDS reduces yield loses for farmers. The use of both CIR and TIR aerial imagery captured using sUAS can provide rapid spatial assessments of SDS, which is required by both producers and plant breeders. PI derived from CIR imagery showing strong correlations to SDS score reinforce the idea of replacing the time-consuming traditional ground-based systems with the more flexible, faster, sUAS methods. TIR imagery was shown to be reliable in assessing SDS in soybeans further establishing another possible aerial method for early detection of SDS.

The goal of this study was to demonstrate the application of aerial imagery as a tool in detecting water quality indicators in a three mile segment of Tibbee Creek in, Clay County, Mississippi. Water samples from 10 transects were collected per sampling date over two periods in 2010 and 2011. Temperature and dissolved oxygen (DO) were measured at each point, and water samples were tested for turbidity and total suspended solids (TSS). Relative reflectance was extracted from high resolution (0.5 meter) multispectral aerial images. A regression model was developed for turbidity and TSS as a function of values for specific sampling dates. The best model was used to predict turbidity and TSS using datasets outside the original model date. The development of an appropriate predictive model for water quality assessment based on the relative reflectance of aerial imagery is affected by the quality of imagery and time of sampling.

Due to the changing climate and inherent atypically occurring meteorological events in the Arctic regions, more accurate snow quality predictions are needed in order to support the Sámi reindeer herding communities in northern Sweden that struggle to adapt to the rapidly changing Arctic climate. Spatial snow depth distribution is a crucial parameter not only to assess snow quality but also for multiple environmental research and social land use purposes. This contrasts with the current availability of affordable and efficient snow monitoring methods to estimate such an extremely variable parameter in both space and time. In this thesis, a novel approach to determine spatial snow depth distribution in challenging alpine terrain is presented and tested during a field campaign performed in Tarfala, Sweden in April 2019. A multispectral camera capturing five spectral bands in wavelengths between 470 and 860 nanometers on board of a small Unmanned Aerial Vehicle is deployed to derive 3D snow surface models via photogrammetric image processing techniques. The main advantage over conventional photogrammetric surveys is the utilization of accurate RTK positioning technology that enables direct georeferencing of the images, and thus eliminates the need for ground control points and dangerous and time-consuming fieldwork. The continuous snow depth distribution is retrieved by differencing two digital surface models corresponding to the snow-free and snow-covered study areas. An extensive error assessment based on ground measurements is performed including an analysis of the impact of multispectral imagery. Uncertainties and non-transparencies due to a black-box environment in the photogrammetric processing are, however, present, but accounted for during the error source analysis. The results of this project demonstrate that the proposed methodology is capable of producing high-resolution 3D snow-covered surface models (< 7 cm/pixel) of alpine areas up to 8 hectares in a fast, reliable and cost-efficient way. The overall RMSE of the snow depth estimates is 7.5 cm for data acquired in ideal survey conditions. The proposed method furthermore assists in closing the scale gap between discrete point measurements and regional-scale remote sensing, and in complementing large-scale remote sensing data by providing an adequate validation source. As part of the Swedish cooperation project ’Snow4all’, the findings of this project are used to support and validate large-scale snow models for improved snow quality prediction in northern Sweden. / På grund av klimatförändringar och naturliga meteorologiska händelser i arktis behövs mer exakta snökvalitetsprognoser för att stödja samernas rensköttsamhällen i norra Sverige som har problem med att anpassa sig till det snabbt föränderliga arktiska klimatet. Rumslig snödjupsfördelning är en avgörande parameter för att inte bara bedöma snökvaliteten utan även för flera miljöforskning och sociala markanvändningsändamål. Detta står i motsats till den nuvarande tillgången till överkomliga och effektiva metoder för snöövervakning för att uppskatta sådan extremt varierande parameter i tid och rum. I detta arbete presenteras och testas en ny metod för att bestämma rumslig snödjupssdistribution i utmanande alpin terräng under en fältstudie som genomfördes i Tarfala i norra Sverige i april 2019. Via fotogrammetrisk bildbehandlingsteknik hämtades snöytemodeller i 3D med hjälp av en multispektral kamera monterad på en liten obemannad drönare. En viktig fördel, i jämförelse med konventionella fotogrammetriska undersökningar, är användningen av exakt RTK-positioneringsteknik som möjliggör direkt georeferencing och eliminerar behovet av markkontrollpunkter. Den kontinuerliga snödjupfördelningen hämtas genom att ytmodellerna delas upp i snöfria respektive snötäckta undersökningsområden. En omfattande felsökning som baseras på markmätningar utförs, inklusive en analys av effekten av multispektrala bilder. Resultaten från denna studie visar att den famtagna metoden kan producera högupplösta snötäckta höjdmodeller i 3D (< 7 cm/pixel) av alpina områden på upp till 8 hektar på ett snabbt, pålitligt och kostnadseffektivt sätt. Den övergripande RMSE för det beräknade snödjupet är 7,5 cm för data som förvärvats under idealiska undersökningsförhållanden. Som ett led i det svenska projektet “Snow4all” används resultaten från projektet för att förbättra och validera storskaliga snömodeller för att bättre förutse snökvaliteten i norra Sverige.

Aedes aegypti and Aedes albopictus are the main vector species for dengue disease and zika, two arboviruses that affect a substantial fraction of the global population. These mosquitoes breed in very slow-moving or standing pools of water, so detecting and managing these potential breeding habitats is a crucial step in preventing the spread of these diseases. Using high-resolution images collected by unmanned aerial vehicles (UAV) and their multispectral mapping data, this paper investigated bathymetry retrieval model in shallow water areas to help improve the habitat detection accuracy. While previous studies have found some success with shallow water bathymetry inversion on satellite imagery, accurate centimeter-level water depth regression from high-resolution, drone multispectral imagery still remains a challenge. Unlike previous retrieval methods generally relying on retrieval factor extraction and linear regression, this thesis introduced CNN methods, considering the nonlinear relationship between image pixel reflectance values and water depth. In order to look into CNN’s potential to retrieve shallow water depths from multispectral images captured by a drone, this thesis conducts a variety of case studies to respectively specify a proper CNN architecture, compare its performance in different datasets, band combinations, depth ranges and with other general bathymetry retrieval algorithms. In summary, the CNN-based model achieves the best regression accuracy of overall root mean square error lower than 0.5, in comparison with another machine learning algorithm, random forest, and 2 other semi-empirical methods, linear and ratio model, suggesting this thesis’s practical significance. / Aedes aegypti och Aedes albopictus är de viktigaste vektorarterna för dengue och zika, två arbovirus som drabbar en stor del av den globala befolkningen. Dessa myggor förökar sig i mycket långsamt rörliga eller stillastående vattensamlingar, så att upptäcka och hantera dessa potentiella förökningsmiljöer är ett avgörande steg för att förhindra spridningen av dessa sjukdomar. Med hjälp av högupplösta bilder som samlats in av obemannade flygfarkoster (UAV) och deras multispektrala kartläggningsdata undersöktes i den här artikeln en modell för att hämta batymetri i grunda vattenområden för att förbättra noggrannheten i upptäckten av livsmiljöer. Även om tidigare studier har haft viss framgång med inversion av bathymetri på grunt vatten med hjälp av satellitbilder, är det fortfarande en utmaning att göra en exakt regression av vattendjupet på centimeternivå från högupplösta, multispektrala bilder från drönare. Till skillnad från tidigare metoder som i allmänhet bygger på extrahering av återvinningsfaktorer och linjär regression, infördes i denna avhandling CNN-metoder som tar hänsyn till det icke-linjära förhållandet mellan bildpixlarnas reflektionsvärden och vattendjupet. För att undersöka CNN:s potential att hämta grunda vattendjup från multispektrala bilder som tagits av en drönare genomförs i denna avhandling en rad fallstudier för att specificera en lämplig CNN-arkitektur, jämföra dess prestanda i olika datamängder, bandkombinationer, djupintervall och med andra allmänna algoritmer för att hämta batymetri. Sammanfattningsvis uppnår den CNN-baserade modellen den bästa regressionsnoggrannheten med ett totalt medelkvadratfel som är lägre än 0,5, i jämförelse med en annan maskininlärningsalgoritm, random forest, och två andra halvempiriska metoder, linjär och kvotmodell, vilket tyder på den praktiska betydelsen av denna avhandling.

Les dispositifs de protection de sites sensibles doivent permettre de détecter des menaces potentielles suffisamment à l’avance pour pouvoir mettre en place une stratégie de défense. Dans cette optique, les méthodes de détection et de reconnaissance d’aéronefs se basant sur des images infrarouge multispectrales doivent être adaptées à des images faiblement résolues et être robustes à la variabilité spectrale et spatiale des cibles. Nous mettons au point dans cette thèse, des méthodes statistiques de détection et de reconnaissance d’aéronefs satisfaisant ces contraintes. Tout d’abord, nous spécifions une méthode de détection d’anomalies pour des images multispectrales, combinant un calcul de vraisemblance spectrale avec une étude sur les ensembles de niveaux de la transformée de Mahalanobis de l’image. Cette méthode ne nécessite aucune information a priori sur les aéronefs et nous permet d’identifier les images contenant des cibles. Ces images sont ensuite considérées comme des réalisations d’un modèle statistique d’observations fluctuant spectralement et spatialement autour de formes caractéristiques inconnues. L’estimation des paramètres de ce modèle est réalisée par une nouvelle méthodologie d’apprentissage séquentiel non supervisé pour des modèles à données manquantes que nous avons développée. La mise au point de ce modèle nous permet in fine de proposer une méthode de reconnaissance de cibles basée sur l’estimateur du maximum de vraisemblance a posteriori. Les résultats encourageants, tant en détection qu’en classification, justifient l’intérêt du développement de dispositifs permettant l’acquisition d’images multispectrales. Ces méthodes nous ont également permis d’identifier les regroupements de bandes spectrales optimales pour la détection et la reconnaissance d’aéronefs faiblement résolus en infrarouge / Surveillance systems should be able to detect potential threats far ahead in order to put forward a defence strategy. In this context, detection and recognition methods making use of multispectral infrared images should cope with low resolution signals and handle both spectral and spatial variability of the targets. We introduce in this PhD thesis a novel statistical methodology to perform aircraft detection and classification which take into account these constraints. We first propose an anomaly detection method designed for multispectral images, which combines a spectral likelihood measure and a level set study of the image Mahalanobis transform. This technique allows to identify images which feature an anomaly without any prior knowledge on the target. In a second time, these images are used as realizations of a statistical model in which the observations are described as random spectral and spatial deformation of prototype shapes. The model inference, and in particular the prototype shape estimation, is achieved through a novel unsupervised sequential learning algorithm designed for missing data models. This model allows to propose a classification algorithm based on maximum a posteriori probability Promising results in detection as well as in classification, justify the growing interest surrounding the development of multispectral imaging devices. These methods have also allowed us to identify the optimal infrared spectral band regroupments regarding the low resolution aircraft IRS detection and classification