This thesis tests a technique based on objectively identified tropical cloud regimes, in which some cloud characteristics are extrapolated from a single site in the tropics to the entire tropics. Information on cloud top pressure, cloud optical thickness and total cloud cover from 1985-2000 has been derived from the ISCCP D1 data set and has been used to create maps of tropical cloud regimes and maps of total cloud cover over the tropics. The distribution and characteristics of the tropical cloud regimes has been discussed after which total cloud cover values were extrapolated to the cloud regimes over the tropics. After a qualitative and quantitative assessment was used to evaluate the success of the extrapolating method, it was found that the method worked especially well for time averaged extrapolated data sets using the median values of total cloud cover values. / I detta magisterexamensarbete testas en metod som baseras på objektivt framtagna molnregimer, där några molnegenskaper extrapoleras från en plats i tropikerna till resten av tropikerna. Informationen om molntoppstrycket, molnens optiska djup och det totala molntäcket från 1985-2000 har hämtats från ISCCP D1 data set och har använts till att skapa kartor för tropiska molnregimer och för det totala molntäcket över tropikerna. Distributionen och egenskaperna av de tropiska molnregimerna har diskuterats och användes sedan för att extrapolera det totala molntäcket över tropikerna. En kvalitativ och kvantitativ undersökning användes för att utvärdera framgångarna med extrapoleringsmetoden. Det framkom att metoden fungerade särskilt bra för extrapolerade data set med median totala molntäcksvärden över längre tidsperioder.

Improved knowledge of the cryosphere state is paramount for continued model development and for accurate estimates of fresh water supply. This work focuses on evaluation and potential improvements of current snow datasets over the United States. Snow in mountainous terrain is most difficult to quantify due to the slope, aspect, and remote nature of the environment. Due to the difficulty of measuring snow quantities in the mountains, the initial study creates a new method to upscale point measurements to area averages for comparison to initial snow quantities in numerical weather prediction models. The new method is robust and cross validation of the method results in a relatively low mean absolute error of 18% for snow depth (SD). Operational models at the National Centers for Environmental Prediction which use Air Force Weather Agency (AFWA) snow depth data for initialization were found to underestimate snow depth by 77% on average. Larger error is observed in areas that are more mountainous. Additionally, SD data from the Canadian Meteorological Center, which is used for some model evaluations, performed similarly to models initialized with AFWA data. The use of constant snow density for snow water equivalent (SWE) initialization for models which utilize AFWA data exacerbates poor SD performance with dismal SWE estimates. A remedy for the constant snow density utilized in NCEP snow initializations is presented in the next study which creates a new snow density parameterization (SNODEN). SNODEN is evaluated against observations and performance is compared with offline land surface models from the National Land Data Assimilation System (NLDAS) as well as the Snow Data Assimilation System (SNODAS). SNODEN has less error overall and reproduces the temporal evolution of snow density better than all evaluated products. SNODEN is also able to estimate snow density for up to 10 snow layers which may be useful for land surface models as well as conversion of remotely-sensed SD to SWE. Due to the poor performance of previously evaluated snow products, the last study evaluates openly-available remotely-sensed snow datasets to better understand the strengths and weaknesses of current global SWE datasets. A new SWE dataset developed at the University of Arizona is used for evaluation. While the UA SWE data has already been stringently evaluated, confidence is further increased by favorable comparison of UA snow cover, created from UA SWE, with multiple snow cover extent products. Poor performance of remotely-sensed SWE is still evident even in products which combine ground observations with remotely-sensed data. Grid boxes that are predominantly tree covered have a mean absolute difference up to 87% of mean SWE and SWE less than 5 cm is routinely overestimated by 100% or more. Additionally, snow covered area derived from global SWE datasets have mean absolute errors of 20%-154% of mean snow covered area.

Sea ice surface roughness is a geophysical property which can be defined and quantified on a variety scales, and consequently affects processes across various scales. The sea ice surface roughness influences various mass, gas, and energy fluxes across the ocean-sea ice-atmosphere interface. Utilizing synthetic aperture radar (SAR) data to understand and map sea ice roughness is an active area of research. This thesis provides new techniques for the estimation of sea ice surface roughness in the Canadian Arctic Archipelago using synthetic aperture radar (SAR). Estimating and isolating sea ice surface properties from SAR imagery is complicated as there are a number of sea ice and sensor properties that influence the backscattered energy. There is increased difficulty in the melting season due to the presence of melt ponds on the surface, which can often inhibit interactions from the sensor to the sea ice surface as shorter microwaves cannot penetrate through the melt water. An object-based image analysis is here used to quantitatively link the winter first-year sea ice surface roughness to C-band RADARSAT-2 and L-band ALOS-2 PALSAR-2 SAR backscatter measured at two periods: winter (pre-melt) and advanced melt. Since the sea ice in our study area, the Canadian Arctic Archipelago, is landfast, the same ice can be imaged using SAR after the surface roughness measurements are established. Strong correlations between winter measured surface roughness, and C- and L-band SAR backscatter acquired during both the winter and advanced melt periods are observed. Results for winter indicate: (1) C-band HH-polarization backscatter is correlated with roughness (r=0.86) at a shallow incidence angle; and (2) L-band HH- and VV-polarization backscatter is correlated with roughness (r=0.82) at a moderate incidence angle. Results for advanced melt indicate: (1) C-band HV/HH polarization ratio is correlated with roughness (r=-0.83) at shallow incidence angle; (2) C-band HH-polarization backscatter is correlated with roughness (r=0.84) at shallow incidence angle for deformed first-year ice only; and (3) L-band HH-polarization backscatter is correlated with roughness (r=0.79) at moderate incidence angle. Retrieval models for surface roughness are developed and applied to the imagery to demonstrate the utility of SAR for mapping roughness, also as a proxy for deformation state, with a best case RMSE of 5 mm in the winter, and 8 mm during the advanced melt. / Graduate

Evapotranspiration (ET; evaporation + transpiration) is central to a wide range of biological, chemical, and physical processes in the Earth system. Accurate remote sensing of ET is challenging due to the interrelated and generally scale dependent nature of the physical factors which contribute to the process. The evaporation of water from porous media like sands and soils is an important subset of the complete ET problem. Chapter 1 presents a laboratory investigation into this question, examining the effects of grain size and composition on the evolution of drying sands. The effects of composition are found to be 2-5x greater than the effects of grain size, indicating that differences in heating caused by differences in reflectance may dominate hydrologic differences caused by grain size variation. In order to relate the results of Chapter 1 to the satellite image archive, however, the question of information loss between hyperspectral (measurements at 100s of wavelength intervals) laboratory measurements and multispectral (≤ 12 wavelength intervals) satellite images must be addressed. Chapter 2 focuses on this question as applied to substrate materials such as sediment, soil, rock, and non-photosynthetic vegetation. The results indicate that the continuum that is resolved by multispectral sensors is sufficient to resolve the gradient between sand-rich and clay-rich soils, and that this gradient is also a dominant feature in hyperspectral mixing spaces where the actual absorptions can be resolved. Multispectral measurements can be converted to biogeophysically relevant quantities using spectral mixture analysis (SMA). However, retrospective multitemporal analysis first requires cross-sensor calibration of the mixture model. Chapter 3 presents this calibration, allowing multispectral image data to be used interchangeably throughout the Landsat 4-8 archive. In addition, a theoretical explanation is advanced for the observed superior scaling properties of SMA-derived fraction images over spectral indices. The physical quantities estimated by the spectral mixture model are then compared to simultaneously imaged surface temperature, as well as to the derived parameters of ET Fraction and Moisture Availability. SMA-derived vegetation abundance is found to produce substantially more informative ET maps, and SMA-derived substrate fraction is found to yield a surprisingly strong linear relationship with surface temperature. These results provide context for agricultural applications. Chapter 5 investigates the question of mapping and monitoring rice agricultural using optical and thermal satellite image time series. Thermal image time series are found to produce more accurate maps of rice presence/absence, but optical image time series are found to produce more accurate maps of rice crop timing. Chapter 6 takes a more global approach, investigating the spatial structure of agricultural networks for a diverse set of landscapes. Surprisingly consistent scaling relations are found. These relations are assessed in the context of a network-based approach to land cover analysis, with potential implications for the scale dependence of ET estimates. In sum, this thesis present a novel approach to improving ET estimation based on a synthesis of complementary laboratory measurements, satellite image analysis, and field observations. Alone, each of these independent sources of information provides novel insights. Viewed together, these insights form the basis of a more accurate and complete geophysical understanding of the ET phenomenon.

Student Number : 0310612G - MSc research report - School of Geography, Archaeology and Environmental Studies - Faculty of Science / Wildfire prediction and management is an issue of safety and security for many rural communities in South Africa. Wildfire prediction and early warning systems can assist in saving lives, infrastructure and valuable resources in these communities. Timely and accurate data are required for accurate wildfire prediction on both weather conditions and the availability of fuels (vegetation) for wildfires. Wildfires take place in large remote areas in which land use practices and alterations to land cover cannot easily be modelled. Remote sensing offers the opportunity to monitor the extent and changes of land use practices and land cover in these areas. In order for effective fire prediction and management, data on the quantity and state of fuels is required. Traditional methods for detecting vegetation rely on the chlorophyll content and moisture of vegetation for vegetation mapping techniques. Fuels that burn in wildfires are however predominantly dry, and by implication are low in chlorophyll and moisture contents. As a result, these fuels cannot be detected using traditional indices. Other model based methods for determining above ground vegetation biomass using satellite data have been devised. These however require ancillary data, which are unavailable in many rural areas in South Africa. A method is therefore required for the detection and quantification of dry fuels that pose a fire risk. ASTER and MAS (MODIS Airborne Simulator) imagery were obtained for a study area within the Lowveld region of the Limpopo Province, South Africa. Two of the ASTER and two of the MAS images were dated towards the end of the dry season (winter) when the quantity of fuel (dry vegetation) is at its highest. The remaining ASTER image was obtained during the middle of the wet season (summer), against which the results could be tested. In situ measurements of above ground biomass were obtained from a large number of collection points within the image footprints. Normalised Difference Vegetation Index and Transformed Vegetation Index vegetation indices were calculated and tested against the above ground biomass for the dry and wet season images. Spectral response signatures of dry vegetation were evaluated to select wavelengths, which may be effective at detecting dry vegetation as opposed to green vegetation. Ratios were calculated using the respective bandwidths of the ASTER and MAS sensors and tested against above ground biomass to detect dry vegetation. The findings of this study are that it is not feasible, using ASTER and MAS remote sensing data, to estimate brown and green vegetation biomass for wildfire prediction purposes using the datasets and research methodology applied in this study. Correlations between traditional vegetation indices and above ground biomass were weak. Visual trends were noted, however no conclusive evidence could be established from this relationship. The dry vegetation ratios indicated a weak correlation between the values. The removal of background noise, in particular soil reflectance, may result in more effective detection of dry vegetation. Time series analysis of the green vegetation indices might prove a more effective predictor of biomass fuel loads. The issues preventing the frequent and quick transmission of the large data sets required are being solved with the improvements in internet connectivity to many remote areas and will probably be a more viable path to solving this problem in the near future.

Thesis advisor: Noah P. Snyder / Thesis advisor: Gabrielle C. David / The Reservoir Sedimentation Database (ResSed), a catalogue of reservoirs and depositional data that has recently become publicly available, allows for rapid calculation of sedimentation and capacity-loss rates over short (annual to decadal) timescales. This study is a statistical investigation of factors controlling average sediment yield (Y) in eastern United States watersheds. I develop an ArcGIS-based model that delineates watersheds upstream of ResSed dams and calculate drainage areas to determine Y for 191 eastern US watersheds. Geomorphic, geologic, regional, climatic, and land use variables are quantified within study watersheds using GIS. Sediment yield exhibits a large amount of scatter, ranging from 4.7 to 3336 tonnes1km 2year-1. A weak inverse power law relationship between drainage area (A) and Y (R2 = 0.09) is evident, similar to other studies (e.g., Koppes and Montgomery, 2009). Linear regressions reveal no relationship between mean watershed slope (S) and Y, possibly due to the relatively low relief of the region (mean S for all watersheds is 6°). Analysis of variance shows that watersheds in formerly glaciated regions exhibit a statistically significant lower mean Y (159 tonnes1km-2year-1) than watersheds in unglaciated regions (318 tonnes1km-2year-1), while watersheds with different dam purposes show no significant differences in mean Y. Linear regressions reveal no relationships between land use parameters like percent agricultural, and percent impervious surfaces (I) and Y, but classification and regression trees indicate a threshold in highly developed regions (I > 34%) above which the mean Y (965 tonnes1km-2year-1) is four times higher than watersheds in less developed (I < 34%) regions (237 tonnes1km 2year-1). Further, interactions between land use variables emerge in formerly glaciated regions, where increased agricultural land results in higher rates of annual capacity loss in reservoirs (R2 = 0.56). Plots of Y versus timescale of measurement (e.g., Sadler and Jerolmack, 2014) show that nearly the full range of observed Y, including the highest values, are seen over short survey intervals (< 20 years), suggesting that whether or not large sedimentation events (such as floods) occur between two surveys may explain the high degree of variability in measured rates. / Thesis (MS) — Boston College, 2014. / Submitted to: Boston College. Graduate School of Arts and Sciences. / Discipline: Earth and Environmental Sciences.

A dissertation submitted to the School of Geography, Archaeology and Environmental Studies, Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfilment of requirements for the degree of Master of Science in Environmental Sciences. Johannesburg, March 2016. / Mapping Prosopis glandulosa (mesquite) invasion in the arid environment of South Africa using remote sensing techniques Mureriwa, Nyasha Abstract Decades after the first introduction of the Prosopis spp. (mesquite) to South Africa in the late 1800s for its benefits, the invasive nature of the species became apparent as its spread in regions of South Africa resulting in devastating effects to biodiversity, ecosystems and the socio-economic wellbeing of affected regions. Various control and management practices that include biological, physical, chemical and integrated methods have been tested with minimal success as compared to the rapid spread of the species. From previous studies, it has been noted that one of the reasons for the low success rates in mesquite control and management is a lack of sufficient information on the species invasion dynamic in relation to its very similar co-existing species. In order to bridge this gap in knowledge, vegetation species mapping techniques that use remote sensing methods need to be tested for the monitoring, detection and mapping of the species spread. Unlike traditional field survey methods, remote sensing techniques are better at monitoring vegetation as they can cover very large areas and are time-effective and cost-effective. Thus, the aim of this research was to examine the possibility of mapping and spectrally discriminating Prosopis glandulosa from its native co-existing species in semi-arid parts of South Africa using remote sensing methods. The specific objectives of the study were to investigate the spectral separability between Prosopis glandulosa and its co-existing species using field spectral data as well as to upscale the results to different satellites resolutions. Two machine learning algorithms (Random Forest (RF) and Support Vector Machines (SVM)) were also tested in the mapping processes. The first chapter of the study evaluated the spectral discrimination of Prosopis glandulosa from three other species (Acacia karoo, Acacia mellifera and Ziziphus mucronata) in the study area using in-situ spectroscopy in conjunction with the newly developed guided regularized random forest (GRRF) algorithm in identifying key wavelengths for multiclass classification. The GRRF algorithm was used as a method of reducing the problem of high dimensionality associated with hyperspectral data. Results showed that there was an increase in the accuracy of discrimination between the four species when the full set of 1825 wavelengths was used in classification (79.19%) as compared to the classification used by the 11 key wavelengths identified by GRRF (88.59%). Results obtained from the second chapter showed that it is possible to spatially discriminate mesquite from its co-existing acacia species and other general land-cover types at a 2 m resolution with overall accuracies of 86.59% for RF classification and 85.98% for SVM classification. The last part of the study tested the use of the more cost effective SPOT-6 imagery and the RF and SVM algorithms in mapping Prosopis glandulosa invasion and its co-existing indigenous species. The 6 m resolution analysis obtained accuracies of 78.46% for RF and 77.62% for SVM. Overall it was concluded that spatial and spectral discrimination of Prosopis glandulosa from its native co-existing species in semi-arid South Africa was possible with high accuracies through the use of (i) two high resolution, new generation sensors namely, WorldView-2 and SPOT-6; (ii) two robust classification algorithms specifically, RF and SVM and (iii) the newly developed GRRF algorithm for variable selection and reducing the high dimensionality problem associated with hyperspectral data. Some recommendations for future studies include the replication of this study on a larger scale in different invaded areas across the country as well as testing the robustness of the RF and SVM classifiers by making use of other machine learning algorithms and classification methods in species discrimination. Keywords: Prosopis glandulosa, field spectroscopy, cost effectiveness, Guided Regularised Random Forest, Support Vector Machines, Worldview-2, Spot-6

A research report submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in partial fulfilment of the requirements for the degree of Master of Science (Geographical Information Systems and Remote Sensing) at the School of Geography, Archaeology & Environmental Studies) Johannesburg. 2017 / Vegetation mapping using remote sensing is a key concern in environmental application using remote sensing. The new high resolution generation has made possible, the mapping of spatial distribution of vegetation communities. The aim of this research is to test the use of new generation multispectral data for vegetation classification in Ezemvelo Game Reserve, Bronkhorspruit. Sentinel-2 and RapidEye images were used covering the study area with nine vegetation classes: eight from grassland (Mixed grassland, Wetland grass, Aristida congesta, Cynadon dactylon, Eragrostis gummiflua, Eragrostis Chloromelas, Hyparrhenia hirta, Serephium plumosum) and one from woodland (Woody vegetation). The images were pre-processed, geo-referenced and classified in order to map detailed vegetation classes of the study area. Random Forest and Support Vector Machines supervised classification methods were applied to both images to identify nine vegetation classes. The softwares used for this study were ENVI, EnMAP, ArcGIS and R statistical packages (R Development Core, 2012) .These were used for Support Vector Machines and Random Forest parameters optimization. Error matrix was created using the same reference points for Sentinel-2 and RapidEye classification. After classification, results were compared to find the best approach to create a current map for vegetation communities. Sentinel-2 achieved higher accuracies using RF with overall accuracy of 86% and Kappa value of 0.84. Sentinel-2 also achieved overall accuracy of 85% with a Kappa value of 0.83 using SVM. RapidEye achieved lower accuracies using RF with an overall accuracy of 82% and Kappa value of 0.79. RapidEye using SVM produced overall accuracy of 81% and a Kappa value of 0.79. The study concludes that Sentinel-2 multispectral data and RF have the potential to map vegetation communities. The higher accuracies achieved in the study can assist management and decision makers on assessing the current vegetation status and for future references on Ezemvelo Game Reserve. Keywords Random forest, Support Vector Machines, Sentinel-2, RapidEye, remote sensing, multispectral, hyperspectral and vegetation communities / LG2018

A dissertation submitted to the Faculty of Science, University of the Witwatersrand, Johannesburg, in fulfilment of requirements for the degree of Master of Science. September 2016. / The Iron Age is a very critical aspect of South Africa’s history. It represents a technology that laid a solid foundation for the development of South Africa in terms of its economy, politics and society. It is therefore imperative to study Iron Age, or rather its remnants such as stone-walled structures and ash middens because these give insight into this critical time period’s technology and those responsible for it. Remote sensing spatial technology provides the opportunity not only to study these Iron Age remnants but to save time and resources while doing so through satellite imagery. This study employs remote sensing by comparing different multispectral satellite images ̶ GeoEye 1 and SPOT 5 ̶ to find the optimum platform to detect key archaeological remnants  ash middens  from the Iron Age period in the Suikerbosrand Nature Reserve located in Southern Gauteng, South Africa. The performance of GeoEye 1 and SPOT 5 in detecting ash middens was compared through supervised classification techniques, Support Vector Machine and Maximum Likelihood Classification, on different band combinations of the two images. Overall, the band combination of Green, Red and NIR is the best performing on both SPOT 5 and GeoEye 1 compared to Green, Red, and Mid IR on SPOT 5 and Green, Red, and Blue on GeoEye 1. However, higher accuracy of results for the detection of ash middens were obtained on the GeoEye 1 platform. The GeoEye platform performed better than the SPOT platform in the detection and analysis of ash middens. Key Words: Ash Middens, GeoEye, Remote Sensing, Satellite Imagery, SPOT / LG2017

A poluição atmosférica é hoje uma questão que afeta as megacidades por todo o mundo. As partículas de aerossol participam do balanço radiativo, da formação de nuvens, da química atmosférica, e são prejudiciais à saúde da população exposta. A extensão e o rápido crescimento das megacidades têm levado à necessidade do desenvolvimento de ferramentas para o monitoramento da poluição do ar, urbana e regional, por sensoriamento remoto via satélites. Foram analisadas as propriedades ópticas dos aerossóis da região metropolitana de São Paulo com medidas obtidas por fotômetros da rede mundial AERONET (Aerosol Robotic Network) operada pela NASA. Foi desenvolvida uma nova metodologia para a determinação da espessura óptica das partículas de aerossol com alta resolução espacial de 1x1 km, sobre a região metropolitana de São Paulo, por satélite. Cinco modelos ópticos de aerossol, representativos da região, foram definidos como função do albedo simples. No comprimento de onda de 550 nm, os modelos possuem valores de albedo simples que variam de 0,83 a 0,93. Foram utilizadas medidas de radiâncias obtidas com o sensor MODIS (Moderate-Resolution Imaging Spectroradiometer), a bordo dos satélites Terra e Aqua da NASA. Extensos testes de sensibilidade foram realizados, analisando o papel do albedo simples, parâmetro de assimetria, refletância de superfície, vapor de água e outras propriedades na espessura óptica derivada. O algoritmo desenvolvido utiliza a propriedade de refletância crítica, que pode ser obtida pelo próprio sensor, para determinar o modelo de aerossol a ser empregado. Este procedimento permitiu a identificação mais precisa do modelo de aerossol, de forma dinâmica e interativa, reduzindo a incerteza na determinação da espessura óptica em alta resolução com o sensor MODIS. Os resultados de validação mostraram uma melhora significativa na comparação entre os produtos de espessura óptica obtidos com o sensor MODIS, quando comparado com as medidas de referência obtidas com radiômetros em superfície. Foram obtidos com esta metodologia mapas com a distribuição espacial de aerossóis com resolução de 1x1 km. Os estudos de casos também apresentaram a potencialidade do método em identificar o modelo de aerossol mais adequado, seja em eventos de poluição local, seja de transporte de poluentes de longa distância. Os modelos e procedimentos desenvolvidos podem ser aplicados a outras regiões urbanas, após as devidas validações. Como ferramenta complementar ao monitoramento ambiental de estações de superfície o produto apresentado pode-se tornar operacional e ser utilizado em rotina por órgãos de controle ambiental em megacidades, como, por exemplo, pela CETESB em São Paulo / Urban air pollution is a public concern in all megacities around the world. Aerosol particles are active participants in the atmospheric energy budget, cloud properties, atmospheric chemistry and have adverse effects on human health. The spatial extension and the high growth rate of the megacities show the need of the use of remote sensing technologies on urban air pollution monitoring. Optical properties of São Paulo aerosol particles were analyzed using global sun photometer measurements from the AERONET (Aerosol Robotic Network) operated by NASA. A new methodology was developed to retrieve aerosol optical thickness in 1x1 km resolution over São Paulo metropolitan area from satelites measurements. Five aerosol optical models representative of the region were defined as a function of the single scattering albedo. The single scattering albedo in 550 nm varied from 0,83 to 0,93 in the models. Radiances were used from MODIS (Moderate-Resolution Imaging Spectroradiometer) sensor on Terra and Aqua NASA platforms. Sensitivities studies have shown the importance of the single scattering albedo, assymmetry parameter, surface reflectance, water vapor and other properties in the aerosol optical thickness retrieval from space. The developed algorithm uses the critical reflectance aerosol property, that is obtainable from the sensor measurements, to identify the aerosol model to be used. This procedure allow a more precise and dynamic definition of the aerosol model, reducing the uncertainties in the aerosol optical thickness retrieved from the MODIS sensor. Validation results have shown a significant improvement in a comparison between aerosol optical thickness obtained from MODIS and from surface radiometers measurements. Aerosol optical thickness images with 1x1 km resolution were obtained with this methodology and shows that the increase in the resolution of the aerosol optical thickness provides a more effective monitoring of the aerosol distribution in São Paulo. The case studies have shown the potentiality of this methodology to identify an adequate aerosol model, for both local aerosol pollution and in the long distance transport of pollutants. The models and procedures developed in this work can be applied in other urban regions with the appropriate validation. The presented product can be operational and used as routine measurement by environmental agencies in megacities, as an example, for CETESB in São Paulo, as a complementary tool to the regular ground based particulate matter monitoring.