Este trabalho constou de um estudo das imagens do novo sensor multiespectral do LANDSAT 5, o Thematic Mapper, cujo objetivo foi o de avaliar e selecionar entre as sete bandas espectrais deste sensor, as mais adequadas para mapeamentos geológicos na escala de semidetalhe e para a discriminação de rochas carbonáticas, utilizando-se de técnicas de processamento automático de imagens digitais. A área selecionada para estudo abrangeu uma completa e bem exposta seqüência carbonática do Grupo Bambuí da borda leste da serra do Ramalho, sudoeste do Estado da Bahia, a qual é representativa da extensa cobertura plataformal do domínio da Bacia do São Francisco. Esta região situa-se numa faixa de transição entre os domínios morfoclimáticos da caatinga e do cerrado, onde se observam estreitas relações entre os elementos da paisagem-solo, vegetação, relevo, com a distribuição das unidades litológicas mapeadas na escala 1:50.000. A avaliação das imagens TM foi fundamentada, essencialmente, numa análise referente ao comportamento espectral das litologias presentes na área, tomando-se como base medidas espectrorradiométricas efetuadas no campo, em áreas de afloramentos selecionadas através da interpretação da cada uma das bandas TM originais. Os aspectos referentes às litologias típicas de cada unidade demonstraram ao nível de campo, a possibilidade de se discriminar diferentes faciologias de calcários, dolomitos, rochas pelíticas e seus solos originários, o que em parte pôde ser confirmado com a interpretação das imagens. As diferenças espectrais são não foram mais notáveis devido a presença genérica da matéria orgânica nas rochas carbonáticas, cujo efeito é reduzir os contrastes da reflectância. Porcentagens de matéria orgânica da ordem de 0,05 a 1,5% mostraram exercer forte atenuação nas bandas de absorção ou na reflectância total das rochas, provocando confusões espectrais entre as litologias carbonáticas. As medidas ) radiométricas foram também essenciais para estimar os efeitos das variações do microrrelevo cárstico dos pavimentos rochosos de calcários e dolomitos, nas respostas espectrais registradas nas diferentes bandas TM. Com base nas medidas espectrais de campo das áreas representativas das unidades litológicas e nas medidas estatísticas (distância J-M e entropia gaussiana) dos valores digitais das bandas TM correspondentes a estas mesmas áreas de amostragem, estabeleceu-se uma seleção entre as 7 bandas, de subconjuntos de três e quatro melhores bandas para o estudo das unidades carbonáticas da área, respectivamente, bandas 4, 5 e 7 e 3, 4, 5 e 7. Tais conjuntos de bandas foram então analisados por técnicas de processamento de realces de imagens digitais, com os sistemas automáticos I-100 e SITIM. As técnicas de realce que apresentaram melhores resultados foram as de composições coloridas de sistemas RGB e IHS e de transformações rotacionais tipo componente principal. Esta última foi também testada para o conjunto completo das sete bandas TM. Os resultados da análise fotogeológica das imagens realçadas mostraram aspectos interessantes na separabilidade e discriminação das unidades litológicas, tendo-se concluído que elas foram decorrentes de um comportamento espectral que estava intimamente associado, tanto com as variações composicionais de seus menores constituintes mineralógicos, como com as variações do estado físico das superfícies das rochas, criadas por uma intensa dissolução química cujas formas cársticas resultantes impunham diferentes condições de iluminação das superfícies. A conjunção destas duas condições de efeitos espectrais permitiu a separabilidade de unidades faciológicas, que somente havia sido obtida em mapeamento geológicos na escala 1:2.000, os quais visavam estudos de controle litológico que tais faciologias exercem na ocorrência de depósitos de fluorita e metais básicos, ) da região leste da serra do Ramalho. Com os processamentos de realces de imagens efetuados, também obteve-se uma alta correlação das unidades mapeadas na escala 1:50.000 com as unidades espectrais detectadas nas imagens TM, embora o traçado de alguns limites entre unidades constituídas essencialmente de solos, não mostrasse uma concordância perfeita com o mapa geológico. O estudo pôde comprovar que as novas imagens do sensor TM, com bandas espectrais específicas a aplicações geológicas (bandas 5 e 7), em combinação com a alta resolução espacial de 30m, abrem excelentes perspectivas para mapeamentos geológicos nas escalas 1:50.000 e 1:25.000, bem como para a pesquisa mineral, dada a capacidade de discriminação espectral de litologias que possuem composições mineralógicas às vezes bastante semelhantes entre si, como os calcários e dolomitos da área estudada. / This research consists of a study of the imagery of the new multispectral sensor of the LANDSAT 5 - Thematic Mapper - with the objective to evaluate and make a selection of the most suitable bands for the discrimination of carbonate rocks and geologic units at semidetail scale using computer techniques for digital image treatment. The selected area of study comprises a complete and well exposed carbonate rock sequence of the Bambui Group, east of the \"Serra do Ramalho\". SW of Bahia State. This area is representative of the extensive cover in the San Francisco Basin. The area is situated at the transitional belt between the morphoclimatic doma in of \"Caatinga\" and \"Cerrado\", where there is a close relationship among the elements of landscape - soil, vegetation, and relief -, with the lithologic units. The evaluation of the TM imagery was based mainly on the analysis of the spectral behavior of the lithology, supported by field espectroradiometric measurements taken place in outcrops selected from the interpretation of the original bands. The field spectroradiometry showed the possibility of discriminating different facies of limestones, dolomites, argillites rocks and classes of soils which can also be confirmed with the interpretation of the imagery. Better differentiation was not possible, because of the presence of organic matter in the carbonate rocks which has the effect of decreasing the reflectance. Values of 0.05% to 0.15% of organic matter was determined that would cause strong attenuation in the peaks of absortion of mineral components or in the overall reflectance of the rocks. The spectroradiometric measurement was also essential in the estimate of the effects of the variations of the karstic microrelief of the carbonate rocks in the spectral tonal pattern observed in each TM bands. Based in these spetroradiometric measurements and in digital images analysis (entropy and J-M distance) it was selected group of the best three and four TM bands, respectivelly bands 4, 5, 7 and bands 3, 4, 5 e 7, for the discrimination of the carbonatic units. Then, these groups of bands were analysed by enhancement computer techniques. The best enhancement was obtained by color composition with RGB and IHS systems and principal components. The principal component was also tested in the seven bands. The results of the photogeological analysis of these enhancement images demonstrated that the separability and discrimination of the lithologic units was due to the spectral behavior related to the compositional variations of the secondary mineralogical constituents as well as the variation in the condition of the rock surfaces, with karstic features that create different illumination condition. The union of this two effects (mineralogical and roughness) permitted the separability of some lithologic units, which were until the only recognized in geologic mapping done at the 1:2,000 scale. This detail mapping had as objective the study of the lithologic control in the occurrence of fluorite, zinc and plumb deposits present in that region. The TM imagery showed good correlation among the spectral units and the lithologic units mapped at the scale of 1:50,000, although some units, which were covered by soils were not differentiated as shown in the geologic map. This study confirmed that the new TM imagery, with specific geologic bands (bands 5 and 7) and high spatial resolution (30m), offers excellent perspectives to geologic mapping in the scales 1:50,000 and 1:25,000, as well as to mineral resources, due to the fact that it was possible to separate lithologies with very similar mineralogical composition, like limestones and dolomites within the area of study.

Changes in watershed vegetative cover from natural and anthropogenic causes including, climatic fluctuations, wildfires and land management practices, can result in increased surface water runoff and erosion. Hydrologic models play an important role in the decision support process for managing these landscape alterations. However, model parameterization requires quantified measures of watershed biophysical condition to generate accurate results. These inputs are often obtained from nationally available land cover data sets that are static in terms of vegetation condition and phenology. Obtaining vegetative data for model input of sufficient spatiotemporal resolution for long-term, watershed-scale change analysis has been a challenge. The purpose of this research was to assess the implications of parameterizing the event-based, Rangeland Hydrology and Erosion Model (RHEM) with dynamic, remotely sensed foliar cover data. The study was conducted on a small, instrumented, grassland watershed within the Walnut Gulch Experimental Watershed surrounding Tombstone, Arizona. A time series of foliar cover rasters was produced by calibrating Landsat-based Soil Adjusted Total Vegetation Index (SATVI) scenes with field measurements. Estimates of basal and litter cover were calculated using allometric relationships derived from ground-based transect data. The model was parameterized using these remotely sensed inputs for all recorded runoff events from 1996-2014. Model performance was improved using the remotely sensed foliar cover compared to using an a priori value based on static national land cover classes. Significant (p<0.05) correlation was shown for the linear relationships between foliar cover and SATVI, foliar cover and basal cover, and foliar cover and litter cover. The integration of Landsat-based vegetative data into RHEM shows potential for modelling on a broadened spatiotemporal scale, allowing for improved landscape characterization and the ability to track watershed response to long-term vegetation changes.

Landslides in the landscape exhibit predictable properties of shape, structure and orientation. These properties are reflected to varying degrees in their depiction in a satellite image. Landslides can be isolated along with similar objects in a digital image using differential and template operators. Extraction of the landslide features from these images can proceed using a logic-based model which draws on an appropriate object definition approximating the depiction of the landslides in an edge-operated image and a digital elevation model. An object extraction algorithm based on these concepts is used in repeated trials to ascertain the effectiveness of this automated approach. A low resolution linear object definition (Fischler et al. , 1981) is used to isolate candidate pixel segments in three enhanced images. These segments are classified as landslides or non-landslides according to their image pixel intensity, length, slope, and orientation. Digital elevation data is used to evaluate slope and orientation criteria. Results are compared to an inventory of landslides made using aerial photographs. Study results indicate that 17% to 28% of landslides in the image are identified for trials that produce a commission error rate of less than 50%. Commission errors are dominated by image objects related to roads and waste wood areas in clearcuts. A higher rate of successful identification was noted for landslides which occurred within 15 years of image acquisition (24% to 32%), and was most apparent for the subset of that group which was located in areas that were harvested more than 15 years before acquisition or were unharvested (29% to 38%). Successful identifications in the trials are dominated by events greater than 300 metres long and wider than 20 metres. The results suggest that the approach is more reliable in unharvested areas of the image. The poor quality of the digital elevation data, specifically artifacts produced by the contour-to-grid algorithm, was partly responsible for errors of commission and omission. The simplicity of the object definition used is another factor in error production. The methodology is not operational, but represents a realistic approach to scene segmentation for resource management given further refinement. / Forestry, Faculty of / Graduate

Since the failure of the Scan Line Corrector (SLC) instrument on Landsat 7, observable gaps occur in the acquired Landsat 7 imagery, impacting the spatial continuity of observed imagery. Due to the highly geometric and radiometric accuracy provided by Landsat 7, a number of approaches have been proposed to fill the gaps. However, all proposed approaches have evident constraints for universal application. The main issues in gap-filling are an inability to describe the continuity features such as meandering streams or roads, or maintaining the shape of small objects when filling gaps in heterogeneous areas. The aim of the study is to validate the feasibility of using the Direct Sampling multiple-point geostatistical method, which has been shown to reconstruct complicated geological structures satisfactorily, to fill Landsat 7 gaps. The Direct Sampling method uses a conditional stochastic resampling of known locations within a target image to fill gaps and can generate multiple reconstructions for one simulation case. The Direct Sampling method was examined across a range of land cover types including deserts, sparse rural areas, dense farmlands, urban areas, braided rivers and coastal areas to demonstrate its capacity to recover gaps accurately for various land cover types. The prediction accuracy of the Direct Sampling method was also compared with other gap-filling approaches, which have been previously demonstrated to offer satisfactory results, under both homogeneous area and heterogeneous area situations. Studies have shown that the Direct Sampling method provides sufficiently accurate prediction results for a variety of land cover types from homogeneous areas to heterogeneous land cover types. Likewise, it exhibits superior performances when used to fill gaps in heterogeneous land cover types without input image or with an input image that is temporally far from the target image in comparison with other gap-filling approaches.

To determine whether spaceborne C-band SAR data could be used alone, or in conjunction with optical data, for accurately mapping mangrove forests of the Mexican Pacific, four scenes of dual-polarized ENVISAT ASAR data, at two incidence angles, were collected for the Teacapán-Agua Brava-Las Haciendas estuarine-mangrove complex. Several combinations of these ASAR data were classified to determine the most optimal arrangement for mangrove mapping. In addition, corresponding Landsat TM data were classified using the same training sites. The overall accuracy in mapping these mangroves did improve when more than one polarization mode was employed. In general, the higher incidence angle data (∼41° vs ∼23°) provided better results. In all circumstances, the optical data alone provided higher classification accuracies. When contained as one mangrove class, the highest overall accuracy achieved using the ASAR data was 54% as compared to 76% for the optical data. When considering four separate mangrove classes, representing the four conditions typical of this system (dead, poor condition, healthy, tall healthy), overall accuracies dropped to 45% and 63%, respectively. With the limited penetration of C-band into canopies, it was difficult to separate healthy and tall healthy mangrove from palm and other terrestrial forests using the ASAR data. In addition to confusion amongst the four mangrove classes, the dead mangrove stands created considerable misclassification as they were readily misidentified with water and saltpan areas in the optical data and with agricultural lands in the ASAR data procedure. Given the advantage of ASAR for identifying dead stands from open water and saltpan, these data were then used in conjunction with the optical data to reduce the misclassification of these areas.

Hurricane Isaac’s landfall on the coast of Louisiana spawned a hydrological research project between Mississippi State University (MSU), the Northern Gulf Institute (NGI), and the National Oceanic and Atmospheric Administration (NOAA) in the Lower Pearl River Basin (LPRB). Unmanned aerial systems data collection missions were scheduled every two months in the LPRB. This research provides a comparison between Landsat-8 imagery and corresponding UAS imagery with regards to the four remote sensing resolutions: spatial, spectral, radiometric, and temporal. Near-infrared (NIR) imagery from each platform was compared by land-water masks and statistical comparisons. A classification method known as natural breaks with Jenks Optimization determined threshold values between land and water for each image. Land-water masks revealed substantial differences between areas of land and water in comparing imagery. The overall difference in average land and water percentages between the two platforms was 1.77%; however, a larger percentage was 20.41% in a single comparison.

Multiple linear and ordinal logistic regression methods were used to develop cubic foot volume (outside bark to a pulpwood diameter top) estimation models for the central Mississippi Institute for Forest Inventory (MIFI) inventory region of Mississippi, USA based on multi-scene Landsat derived variables. These models were used to stratify the region into volume classes to estimate the statistical gains made from a stratified random sample versus a complete random sample. Ordinal logistic regression produced higher accuracy statistics for all forest cover classes except the mixed forest cover class and the method is recommended to be used to estimate cubic foot volume (outside bark to a pulpwood diameter top) for the study area. Statistical gains from ordinal logistic regression averaged 30.34% and relative precision averaged 1.53 for the study area. For each forest cover type volume model that was produced, it was found that the interaction variable between Landsat TM band 5 and the GIS age variable was statistically significant.

No description available.

Simulated Landsat reflectance spectra of soil samples were compared to actual Landsat radiance values of soils in two fields (1 and 3) near Vance, Mississippi. The simulated reflectance spectra were calculated by combining Landsat spectral sensitivity with laboratory-based spectrophotometer reflectance values. The actual radiance data were obtained by extracting pixel values from Landsat images. Simple linear regression (SLR) yielded significant linear relationships for 1997 field-1 and 2001 field-3 data. Multiple linear regression (MLR) and weighted linear regression (WLR), which indirectly accounted for moisture content and spatial resolution, respectively, yielded improvement in R2 for most of the studied bands. The analyses generally satisfied the normality and constant variance assumptions, and removal of outliers improved the validity of the assumptions and R2. It was concluded that indirect measures of soil moisture content and spatial uncertainty can substantially improve the relationship between remotely sensed bare-soil spectra and laboratory spectra.

No description available.