Abstract After deregulation of the agricultural markets in South Africa in 1997, the estimated maize crop could no longer be verified against the actual crop, due to the lack of control data from the Maize Control Board. This drove the need to explore remotely sensed data as a supplement to the current crop estimation methodology to improve crop estimations. Input data for the development of a Geographic Information System (GIS)-based model consisted of objective yield point data, Moderate Resolution Imaging Spectroradiometer (MODIS) Normalised Difference Vegetation Index (NDVI) images and rainfall grids. Rainfall grids were interpolated from weather station data. NDVI values were obtained from the MODIS sensor aboard the Terra platform. Objective yield point field survey data for the 2001/2002 growing season were utilised since dry-land or irrigated conditions were recorded for that season. MODIS NDVI values corresponded well with the growing stages and age of the maize plants after being adjusted to reflect the crop’s age rather than the Julian date. Rainfall values were extracted from rainfall grids and also aligned with the age of the maize plants. This is a suggested alternative to the traditional method of using the mean NDVI for several districts in a region over a Julian growing period of 11 months according to Julian dates. South African maize production areas extend over seven (7) provinces with eight (8) different temperature and rainfall zones (du Plessis, 2004). Planting-date zones based on the uniform age of the maize plants were developed from objective yield Global Positioning System (GPS) points for the 2001/2002 growing season and compared with the 2004/2005 growing season (Frost and Kneen, 2006). Planting dates were interpolated from these planting zones for objective yield GPS points which were missing planting dates in the survey database. MODIS imagery is affordable (free) and four (4) images cover the whole of South Africa daily, while one (1) image covers the study area daily. Several recommendations, such as establishing yield equations for a normal, dry, and wet season were made. It is also suggested that dry-land and irrigated areas continue to be evaluated separately in future.

Anthropogenic changes to forests, agriculture and hydrology are being driven by a need to provide water, food and shelter to more than six billion people. Unfortunately, these changes have a major impact on hydrology, biodiversity, climate, socio-economic stability and food security. The most pervasive form of land-cover change in South Africa is human settlement expansion. In many cases, new human settlements and settlement expansion are informal and occur in areas that are typically covered by natural vegetation. Settlements are infrequently mapped on an ad-hoc basis in South Africa which makes information on when and where new settlements form very difficult. Determining where and when new informal settlements occur is beneficial from not only an ecological but also a social development standpoint. The objective of this thesis is to make use of coarse resolution satellite data to infer the location of new settlement developments in an automated manner by making use of machine learning methods. The specific sensor that is considered in this thesis is the MODIS sensor on-board the Terra and Aqua satellites. By using samples taken at regular intervals (8 days), a hyper-temporal time-series is constructed and consequently used to detect new human settlement formations in South Africa. Two change detection methods are proposed in this thesis to achieve the goal of automated new settlement development detection using this high-temporal coarse resolution satellite time-series data. / Thesis (PhD(Eng))--University of Pretoria, 2012. / Electrical, Electronic and Computer Engineering / unrestricted

Accurate, fast detection of clouds in satellite imagery has many applications, for example Numerical Weather Prediction (NWP) and climate studies of both the atmosphere and of the Earth’s surface temperature. Most operational techniques for cloud detection rely on the differences between observations of cloud and of clear-sky being more or less constant in space and in time. In reality, this is not the case - different clouds have different spectral properties, and different cloud types are more or less likely in different places and at different times, depending on atmospheric conditions and on the Earth’s surface properties. Observations of clear sky also vary in space and time, depending on atmospheric and surface conditions, and on the presence or absence of aerosol particles. The Bayesian approach adopted in this project allows pixel-specific physical information (for example from NWP) to be used to predict pixel-specific observations of clear sky. A physically-based, spatially- and temporally-specific probability that each pixel contains a cloud observation is then calculated. An advantage of this approach is that identification of ambiguously classed pixels from a probabilistic result is straightforward, in contrast to the binary result generally produced by operational techniques. This project has developed and validated the Bayesian approach to cloud detection, and has extended the range of applications for which it is suitable, achieving skills scores that match or exceed those achieved by operational methods in every case. High temperature gradients can make observations of clear sky around ocean fronts, particularly at thermal wavelengths, appear similar to cloud observations. To address this potential source of ambiguous cloud detection results, a region of imagery acquired by the AATSR sensor which was noted to contain some ocean fronts, was selected. Pixels in the region were clustered according to their spectral properties with the aim of separating pixels that correspond to different thermal regimes of the ocean. The mean spectral properties of pixels in each cluster were then processed using the Bayesian cloud detection technique and the resulting posterior probability of clear then assigned to individual pixels. Several clustering methods were investigated, and the most appropriate, which allowed pixels to be associated with multiple clusters, with a normalized vector of ‘membership strengths’, was used to conduct a case study. The distribution of final calculated probabilities of clear became markedly more bimodal when clustering was included, indicating fewer ambiguous classifications, but at the cost of some single pixel clouds being missed. While further investigations could provide a solution to this, the computational expense of the clustering method made this impractical to include in the work of this project. This new Bayesian approach to cloud detection has been successfully developed by this project to a point where it has been released under public license. Initially designed as a tool to aid retrieval of sea surface temperature from night-time imagery, this project has extended the Bayesian technique to be suitable for imagery acquired over land as well as sea, and for day-time as well as for night-time imagery. This was achieved using the land surface emissivity and surface reflectance parameter products available from the MODIS sensor. This project added a visible Radiative Transfer Model (RTM), developed at University of Edinburgh, and a kernel-based surface reflectance model, adapted here from that used by the MODIS sensor, to the cloud detection algorithm. In addition, the cloud detection algorithm was adapted to be more flexible, making its implementation for data from the SEVIRI sensor straightforward. A database of ‘difficult’ cloud and clear targets, in which a wide range of both spatial and temporal locations was represented, was provided by M´et´eo-France and used in this work to validate the extensions made to the cloud detection scheme and to compare the skill of the Bayesian approach with that of operational approaches. For night land and sea imagery, the Bayesian technique, with the improvements and extensions developed by this project, achieved skills scores 10% and 13% higher than M´et´eo-France respectively. For daytime sea imagery, the skills scores were within 1% of each other for both approaches, while for land imagery the Bayesian method achieved a 2% higher skills score. The main strength of the Bayesian technique is the physical basis of the differentiation between clear and cloud observations. Using NWP information to predict pixel-specific observations for clear-sky is relatively straightforward, but making such predictions for cloud observations is more complicated. The technique therefore relies on an empirical distribution rather than a pixel-specific prediction for cloud observations. To try and address this, this project developed a means of predicting cloudy observations through the fast forward-modelling of pixel-specific NWP information. All cloud fields in the pixel-specific NWP data were set to 0, and clouds were added to the profile at discrete intervals through the atmosphere, with cloud water- and ice- path (cwp, cip) also set to values spaced exponentially at discrete intervals up to saturation, and with cloud pixel fraction set to 25%, 50%, 75% and 100%. Only single-level, single-phase clouds were modelled, with the justification that the resulting distribution of predicted observations, once smoothed through considerations of uncertainties, is likely to include observations that would correspond to multi-phase and multi-level clouds. A fast RTM was run on the profile information for each of these individual clouds and cloud altitude-, cloud pixel fraction- and channel-specific relationships between cwp (and similarly cip) and predicted observations were calculated from the results of the RTM. These relationships were used to infer predicted observations for clouds with cwp/cip values other than those explicitly forward modelled. The parameters used to define the relationships were interpolated to define relationships for predicted observations of cloud at 10m vertical intervals through the atmosphere, with pixel coverage ranging from 25% to 100% in increments of 1%. A distribution of predicted cloud observations is then achieved without explicit forward-modelling of an impractical number of atmospheric states. Weights are applied to the representation of individual clouds within the final Probability Density Function (PDF) in order to make the distribution of predicted observations realistic, according to the pixel-specific NWP data, and to distributions seen in a global reference dataset of NWP profiles from the European Centre for Medium Range Weather Forecasting (ECMWF). The distribution is then convolved with uncertainties in forward-modelling, in the NWP data, and with sensor noise to create the final PDF in observation space, from which the conditional probability that the pixel observation corresponds to a cloud observation can be read. Although the relatively fast computational implementation of the technique was achieved, the results are disappointingly poor for the SEVIRI-acquired dataset, provided by M´et´eo-France, against which validation was carried out. This is thought to be explained by both the uncertainties in the NWP data, and the forward-modelling dependence on those uncertainties, being poorly understood, and treated too optimistically in the algorithm. Including more errors in the convolution introduces the problem of quantifying those errors (a non-trivial task), and would increase the processing time, making implementation impractical. In addition, if the uncertianties considered are too high then a PDF flatter than the empirical distribution currently used would be produced, making the technique less useful.

Submitted by Neusa Fagundes (neusa.fagundes@unioeste.br) on 2017-09-15T17:50:23Z No. of bitstreams: 1 Rafaela_Nicolau2017.pdf: 2526644 bytes, checksum: b34f04c83f091610940b76b8a104e99a (MD5) / Made available in DSpace on 2017-09-15T17:50:23Z (GMT). No. of bitstreams: 1 Rafaela_Nicolau2017.pdf: 2526644 bytes, checksum: b34f04c83f091610940b76b8a104e99a (MD5) Previous issue date: 2017-02-15 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / The acquisition of effective technologies for prediction and monitoring of agricultural crops highlights the search for methodologies that make this information available before harvesting. Currently, the monitoring of agricultural production is still partially carried out through subjective and onerous techniques by Brazilian official bodies. The study of the agricultural monitoring and/or estimation of winter crops yield, using vegetation indexes extracted from multitemporal images of the MODIS sensor, is a reality that has been tested by several authors, in the search for greater objectivity to the generated figures. In this context, this research aims to map and estimate areas with winter and maize crops, using temporal series of the EVI vegetation index from the MODIS sensor of the Terra and Aqua satellites, for the 2012, 2013, and 2014 harvests for the state of Paraná. As a way of adjusting the mapping through the MODIS sensor (250 meters), the visual analysis was performed in which images of medium spatial resolution (30 meters) were used to identify the chosen cultures. In article 1, color compositions were generated using images from the pre-planting period until the initial development and images representing the vegetative peak of the crops. Subsequently, the extraction of cultivated areas with the crops of interest was performed, so that these could be compared with official data through statistics and correlations, as well as accuracy analyzes. In Article 2, colored compositions were generated using only the vegetative peak images of the cultures to be classified using the Spectral Angle Mapper (SAM) algorithm. Subsequently, the masks were compared with official data through statistics and correlations, as well as accuracy analyzes. In Article 1, an underestimation of the safflower and winter crops areas was found for the 2012 and 2013 crops, and an overestimation for the 2014 safflower, and for the winter crops, overestimation. By the accuracy analyzes, the masks were classified with excellence. In Article 2, it was verified that the data of areas were overestimated for the safflower corn and underestimated for the winter crops. The accuracy analyzes were classified as excellent, in relation to the medium resolution image. / A obtenção de tecnologias eficazes para a previsão e o acompanhamento de safras agrícolas ressalta a busca de metodologias que disponibilizem essas informações antes da colheita. Atualmente, o acompanhamento da produção agrícola é ainda em parte realizado por meio de técnicas subjetivas e onerosas por órgãos oficiais brasileiros. O estudo do monitoramento agrícola e/ou estimativa de safras das culturas de inverno, utilizando índices de vegetação extraídos de imagens multitemporais do sensor MODIS, é uma realidade que tem sido testada por diversos autores na busca de maior objetividade para os valores gerados. Nesse contexto, esta pesquisa tem por objetivo mapear e estimar áreas com as lavouras de inverno e de milho safrinha, utilizando séries temporais do índice de vegetação EVI, provenientes do sensor MODIS dos satélites Terra e Aqua, nas safras 2012, 2013 e 2014 para o estado do Paraná. Como forma de ajustar o mapeamento por meio do sensor MODIS (250 metros), foi realizada uma análise visual em que foram utilizadas imagens de média resolução espacial (30 metros) para identificação das culturas desejadas. No Artigo 1 foram geradas composições coloridas utilizando imagens do período de pré-plantio até o desenvolvimento inicial e imagens que representam o pico vegetativos das lavouras. Posteriormente, foi realizada a extração de áreas cultivadas com as lavouras de interesse para que pudessem ser comparadas com dados oficiais por meio de estatísticas e correlações, como também análises de acurácia. No Artigo 2 foram geradas composições coloridas utilizando somente as imagens que representam os picos vegetativos das lavouras para serem classificadas, utilizando o algoritmo SAM (Spectral angle mapper). Posteriormente, as máscaras foram comparadas com dados oficiais por meio de estatísticas e correlações, como também análises de acurácia. No Artigo 1 foi verificada subestimação para o milho 2ª safra nas safras 2012 e 2013 e superestimação em 2014 e, para lavouras de inverno, superestimação. Pelas análises de acurácia, as máscaras foram classificadas com excelência. No Artigo 2 foi verificado que os dados de áreas foram superestimados para o milho 2ª safra e subestimados para as lavouras de inverno. As análises de acurácia foram classificadas como excelentes em relação à imagem de média resolução.

Submitted by Mario BC (mario@bc.ufrpe.br) on 2016-08-15T13:14:14Z No. of bitstreams: 1 Anderson Santos da Silva.pdf: 1059889 bytes, checksum: ff989424df01788dbda8e075b1d48a91 (MD5) / Made available in DSpace on 2016-08-15T13:14:14Z (GMT). No. of bitstreams: 1 Anderson Santos da Silva.pdf: 1059889 bytes, checksum: ff989424df01788dbda8e075b1d48a91 (MD5) Previous issue date: 2016-02-26 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - CAPES / Conselho Nacional de Pesquisa e Desenvolvimento Científico e Tecnológico - CNPq / This research is based on estimated and observed agricultural productivity in an area of commercial sugarcane production located at São Francisco’s Agroindustry – AGROVALE S.A., Juazeiro – BA, Brazilian northeast. The new yield estimation models were obtained by multiple linear regression, in which the inputs variables were: irrigation, precipitation, average air temperature, vapor saturation deficit of the air, photoperiod, normalized difference vegetation index (NDVI), leaf area index (LAI) and fractional soil cover (FC). To obtain these models, it was used the statistics program Statística version 10. Futhermore, the meteorological data were obtained from an automatic weather station located at the Farm Brasil Uvas, Juazeiro – BA such as: precipitation (mm), temperature (°C), relative humidity (%), evapotranspiration (mm), current vapor pressure (hPa) and saturation vapor pressure (hPa). The crop yield data and parameters related to crop development were obtained from AGROVALE Agriculture Department. The spectral data, NDVI, IAF and FC, were extracted from MODIS sensor images (Spectroradiometer Imager Moderate Resolution). The data used to models validation were obtained from the same sources previously mentioned. The data were analyzed by mean absolute error (DMA) and mean relative error (DMR). The comparison of yield observed and estimated values showed that the spectral agrometeorological model (SAM) presented the lower and better mean relative error (DMR) with a mean variation of 0.34 %, followed by agrometeorological model with a mean variation of 1.37 % and, finally, the spectral model presented larger mean relatives errors in comparison with other two models, showing a mean variation of 6.58%, approaching AGROVALE’s technicians estimation that presented a mean variation of 6.75%. At the validation’s model for the 2004/2005 crop year, the spectral surpassed the agrometeorological and agrometeorological spectral with average relative errors of 5.05%, while for other models the difference were 15.11% and 16.19%, reflecting a productivity of 93.05 t ha-1 versus 83.19 t ha-1 and 82.13 t ha-1 of agrometeorological and agrometeorologicalspectral models, respectively, for an observed yield of 98 t ha-1. Soon after the 2011/2012 years crop there was a planting renovation with a new variety, with different physiology and consequently a distinct productive power and, from 2013/2014 crop year, the models underestimated the productivity compared to the real. The estimate made by the technicians, based on the crop development since planting until next harvest, showed satisfactory results as well as the tested models. / Esta pesquisa baseou-se na avaliação de produtividade agrícola estimada e observada em uma área de cultivo comercial de cana-de-açúcar localizada na Agroindústria do Vale do São Francisco – AGROVALE S.A., Juazeiro – BA, sertão nordestino. Novos modelos de estimativas de produtividades foram obtidos por regressão linear múltipla utilizando-se, como variáveis de entrada: a irrigação, a precipitação, a temperatura média do ar, o déficit de saturação de vapor do ar, o fotoperíodo, o índice de vegetação por diferença normalizada (NDVI), o índice de área foliar (IAF) e a fração de cobertura do solo (FC). Para obtenção desses modelos utilizou-se o programa estatístico Statística versão 10. Além disso, os meteorológicos foram obtidos na estação meteorológica automática instalada na Fazenda Brasil Uvas, em Juazeiro – BA sendo elas: precipitação, temperatura, umidade relativa, evapotranspiração, pressão atual de vapor e pressão de saturação de vapor. Os dados de rendimento agrícola e parâmetros inerentes ao desenvolvimento da cultura foram disponibilizados pelo Departamento Agrícola da usina AGROVALE. Os dados espectrais: NDVI, IAF e FC foram extraídos de produtos derivados de imagens orbitais do sensor MODIS (Espectrorradiômetro Imageador de Resolução Moderada). Os dados para validação dos modelos também foram obtidos nas mesmas fontes citadas anteriormente. Os dados foram avaliados por meio do cálculo do erro médio absoluto e do erro médio relativo ou percentual. A comparação dos valores observados e estimados de produtividades mostra que o modelo agrometeorológico-espectral (MAE) apresentou as menores e melhores diferenças médias relativas com uma variação média de 0,34%, seguido do modelo agrometeorológico (MA) com uma variação média de 1,37% e por último o modelo espectral (ME) apresentou as maiores diferenças médias relativas, quando comparado com os outros dois modelos obtendo uma variação média de 6,58%, aproximando-se mais da estimativa feita pelos técnicos da usina que apresentou variação média de 6,75%. Na validação dos modelos para o ano-safra de 2004/2005 o espectral superou os agrometeorológico e o agrometeorológico-espectral com diferenças médias relativas na ordem de 5,05% enquanto nos demais modelos as diferenças foram de 15,11% e 16,19%, refletindo numa produtividade de 93,05 t ha-1 contra 83,19 t ha-1 e 82,13 t ha-1 dos modelos agrometeorológicos e agrometeorológico-espectral, respectivamente, para uma produtividade observada de 98 t ha-1. Logo após a safra de 2011/2012 ocorreu uma renovação de plantio com nova variedade, fisiologia diferenciada e, consequentemente, um poder produtivo distinto e a partir da safra de 2013/2014 os modelos subestimaram a produtividade quando comparadas com o real. A estimativa feita pelos técnicos da usina baseada no desenvolvimento da cultura desde o plantio até próximo da colheita, apresentou resultados satisfatórios assim como os modelos testados.