The opening of large archives of satellite data such as LANDSAT, MODIS and the SENTINELs has given researchers unprecedented access to data, allowing them to better quantify and understand local and global land change. The need to analyze such large data sets has led to the development of automated and semi-automated methods for satellite image time series analysis. However, few of the proposed methods for remote sensing time series analysis are available as open source software. In this paper we present the R package dtwSat. This package provides an implementation of the time-weighted dynamic time warping method for land cover mapping using sequence of multi-band satellite images. Methods based on dynamic time warping are flexible to handle irregular sampling and out-of-phase time series, and they have achieved significant results in time series analysis. Package dtwSat is available from the Comprehensive R Archive Network (CRAN) and contributes to making methods for satellite time series analysis available to a larger audience. The package supports the full cycle of land cover classification using image time series, ranging from selecting temporal patterns to visualizing and assessing the results.

Pour prévoir la production, les systèmes de surveillance de la sécurité alimentaire doivent être renseignés par des données sur les surfaces cultivées et sur le rendement. Ces données peuvent être estimées par les systèmes d'observations satellitaires à moyenne résolution spatiale, qui, par leur vision synoptique, constituent une source d'information particulièrement adéquate. En Afrique de l'Ouest, l'estimation des surfaces cultivées par télédétection reste cependant problématique en raison d'un domaine cultivé fragmenté, d'une grande hétérogénéité spatiale due aux conditions environnementales et aux pratiques culturales, et de la synchronisation des phénologies des agrosystèmes et des écosystèmes liée au régime des précipitations. Dans ce contexte, cette thèse présente, en trois volets, des développements méthodologiques originaux pour la caractérisation des systèmes agricoles d'Afrique de l'Ouest par télédétection. Les méthodes ont été développées à partir de séries temporelles MODIS (250 m à 500 m de résolutionspatiale) acquises sur le Mali. (i) La cartographie des surfaces cultivées a été réalisée à partir d'indices spectraux, spatiaux, texturaux et temporels dérivés des images. Deux approches ont été appliquées : une approche de type ISODATA consécutive à une segmentation du territoire basée sur les images MODIS et une approche de fouille de données basée sur des « motifs séquentiels ». Les produits cartographiques obtenus présentent une meilleure précision que les produits globaux « occupation du sol » existants (70% vs 50% en moyenne). Cependant, une part importante des erreurs d'omission et de commission (de 20% à 40%) reste incompressible en raison de la fragmentation du domaine cultivé. (ii) La cartographie des types de systèmes agricoles a nécessité un premier travail de typologie effectué à partir d'une BD d'enquêtes de terrain de l'Institut d' Economie Rurale de Bamako sur 100 villages. Trois types de systèmes agricoles ont été déterminés à l'échelle du village : céréales dominantes (mil, sorgho), cultures intensives dominantes (maïs, coton) et mélange de sorgho et de coton. La classification des systèmes agricoles à partir des indicateurs de télédétection précédemment cités a été produite par un algorithme de type Random Forest avec une précision globale de 60%. Les résultats mettent en évidence une combinaison optimale d'indicateurs comprenant le NDVI ainsi que la texture pour la caractérisation des systèmes agricoles. (iii) Enfin, pour le suivi des cultures, le produit phénologique MODIS a été testé et évalué à partir de variables phénologiques obtenues par simulations agro-météorologiques du modèle de plante SARRA-H. Les résultats montrent que ce produit comporte des incohérences dues au fort ennuagement de début de saison des pluies. Après suppression des données aberrantes, on montre que les dates de transition phénologique des surfaces cultivées issues de MODIS sont plus précoces de 20 jours comparées aux sorties du modèle de culture, en raison notamment de la nature mixte « agro-écosystème » des surfaces à l'échelle du pixel MODIS. Les résultats de cette thèse permettent de dégager de nouvelles pistes de couplage entre télédétection, données de terrain et modélisation agro-météorologique en apportant une information continue dans le temps et dans l'espace sur la caractérisation du domaine cultivé au « Sahel ». / For food security systems, data on cultivated surfaces and yields are a prerequisite for agricultural production forecast. Moderate resolution satellite remote-sensing systems offer a synoptic vision that makes them a particularly appropriate information source for the estimation of such data. However, the estimation of cultivated surfaces is still challenging in West Africa, because of highly fragmented farmland, specific weather conditions resulting in high regional variability in terms of agricultural systems and practices, and synchronized phenology of crops and natural vegetation due to the rainfall regime. In this context, this thesis presents three original methodological approaches for the characterization of agricultural systems in West Africa by remote sensing. These methods were developed using MODIS time series (from 250 to 500 m spatial resolution) acquired for Mali. (i) The mapping of cultivated areas was carried out with spectral, spatial and textural indices derived from the images. Two approaches were chosen: one of ISODATA type following a segmentation of the territory based on MODIS imagery, and the other of data mining type based on ‘sequential patterns'. The crop map obtained showed a better precision than that of the existing land cover global products (70% vs 50% in average). Furthermore, it was shown that a significant part of user and producer errors (20 to 40%) could not be compressed due to farmland fragmentation. (ii) The mapping of agricultural system types first required the definition of a typology derived from an IER (Institute of Rural Economy in Mali) field survey data base on 100 villages. Three types of agricultural systems were determined at the village scale: mainly cereals (millet, sorghum), mainly intensive crops (maize, cotton) and a mixture of sorghum and cotton. The classification of agricultural systems using the aforementioned remote sensing indicators was carried out by a Random Forest type algorithm with an overall accuracy of 62%. Results bring to light the important part played by temporal NDVI and texture in agricultural system characterization. (iii) Finally, for crop monitoring, the MODIS phenological product was tested and assessed using phenological variables obtained from agro-meteorological simulations made by the SARRA-H plant model. Results show that this product contains inconsistencies due to the significant cloud cover linked with the start of the raining season. After the suppression of incongruous data, the phenological transition dates for crop land derived from MODIS were shown to be earlier by 20 days than the SARRA-H-simulated transition dates, due mainly to the ‘agro-ecosystem' mixed nature of surfaces at MODIS pixels scale. The results of this thesis highlight new possibilities for the combinination of remote sensing, field data and agro-meteorological modelling, delivering nonstop information in time and space on the characterization of “Sahel” farmland.

Submitted by Luciana Ferreira (lucgeral@gmail.com) on 2017-12-27T10:27:35Z No. of bitstreams: 2 Tese - Arielle Elias Arantes - 2017.pdf: 12603465 bytes, checksum: d64f9bccc8dcafceb1f000629bf7f031 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Approved for entry into archive by Luciana Ferreira (lucgeral@gmail.com) on 2017-12-27T10:28:07Z (GMT) No. of bitstreams: 2 Tese - Arielle Elias Arantes - 2017.pdf: 12603465 bytes, checksum: d64f9bccc8dcafceb1f000629bf7f031 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) / Made available in DSpace on 2017-12-27T10:28:07Z (GMT). No. of bitstreams: 2 Tese - Arielle Elias Arantes - 2017.pdf: 12603465 bytes, checksum: d64f9bccc8dcafceb1f000629bf7f031 (MD5) license_rdf: 0 bytes, checksum: d41d8cd98f00b204e9800998ecf8427e (MD5) Previous issue date: 2017-12-18 / Cattle ranching is the main land use activity in Brazil, with about 175 million hectares of cultivated pasture, with at least 50% of these being with some degree of degradation. Degraded pastures present low biomass production of little nutritional value, which leads to low animal weight gain in the rainy season and loss of weight in the dry season. Due to its low productive efficiency, if these areas were identified and recovered, they could be intensified, freeing pasture areas for other uses. In this context, the objective of this work was to evaluate the pasture vigor and the potential livestock intensification for Brazil. In order to obtain the vigor and productivity of Brazilian pastures, a Pasture Strength Index (PVI) was generated by integrating the α (intercept) and β (slope) coefficients, derived from the linear regression of the vegetation index (NDVI) over time (2000 to 2017). Pastures with low PVI values were located throughout the Caatinga biome, in the MATOPIBA region of the Cerrado biome, in the eastern portion of Mato Grosso do Sul, southeastern of Mato Grosso and northwestern of Goiás. These areas are associated to regions of higher water deficit, as shown by the relationship between the PVI and the total annual precipitation (R² = 0.40) and evapotranspiration. For the Cerrado biome, the PVI showed high spatial correspondence with the green biomass and percent green cover. Green biomass and percent green cover were generated from extrapolation of field data to the spatial resolution of MODIS images. The areas with lower PVI values in the Cerrado biome also had lower green biomass (< 6000 kg ha-1 ) and percent green cover (< 47%) during the growing season. Considering the accumulated green biomass in the growing season, it was observed that the Cerrado’s cattle stocking rate could increase from 1.11 AU ha-1 (real cattle stocking rate) to 2.56 AU ha-1 (potential cattle stocking rate). The real cattle stocking rate in 2015 was generated through the integration of the 2006 Livestock Census data with the Livestock Production data for the year 2015. The potential cattle stocking rate was obtained from the relationship between the forage production (green biomass and gross primary productivity - GPP) and the forage demand of one animal unit (1 AU = 450 kg). The potential of intensification was determined from the difference between the actual and the potential cattle stocking rates. For all of Brazil, the cattle stocking rate in 2015 was 0.97 AU ha-1 , reaching a potential of 3.60 AU ha-1 , that is, the potential for intensification was 2.63 AU ha-1 . The greatest potential of intensification occurred in the South region (3.62 AU ha-1 ), and the lowest in the North (2.13 AU ha-1 ) and Northeast (2.22 AU ha-1 ) regions of Brazil. / A pecuária é a principal atividade de uso da terra no Brasil, com cerca de 175 milhões de hectares de pastagens cultivadas, sendo que pelo menos 50% destas estão com algum nível de degradação. Pastagens degradadas apresentam baixa produção de biomassa de pouco valor nutritivo, o que leva a um menor ganho de peso animal na estação chuvosa e a perda de peso na estação seca. Pela sua baixa eficiência produtiva, caso haja a identificação e a recuperação destas áreas, estas poderiam ser intensificadas, liberando áreas para outros usos. Neste contexto, este trabalho teve por objetivo avaliar o vigor e o potencial de intensificação das pastagens brasileiras. Para obter o vigor das pastagens, gerou-se um Índice de Vigor do Pasto (PVI), por meio da integração dos coeficientes α (intercepto) e β (slope), obtidos da regressão linear do índice de vegetação (NDVI) ao longo do tempo (2000 a 2017). As pastagens com os menores valores de PVI localizaram-se em todo o bioma Caatinga, na região do MATOPIBA no bioma Cerrado, no leste do Mato Grosso do Sul, sudeste do Mato Grosso e noroeste de Goiás. Estas áreas estão associadas a regiões de maior déficit hídrico, como mostrado pela relação do PVI com a precipitação (R² = 0,40) e a evapotranspiração acumuladas durante o ano (R² = 0,30). Para o bioma Cerrado, o PVI apresentou alta correspondência espacial com a biomassa verde e com o percentual de cobertura verde. A biomassa verde e o percentual de cobertura verde foram geradas a partir da extrapolação de dados de campo para a resolução espacial das imagens MODIS. As áreas com menores valores de PVI no bioma Cerrado também tiveram pouca biomassa verde (< 6.000 kg ha-1 ) e porcentagem cobertura verde (< 47%) durante a estação de crescimento. Considerando o acúmulo de biomassa verde na estação de crescimento, percebeu-se que a lotação bovina do Cerrado poderia aumentar de 1,11 UA ha-1 (lotação real) para 2,56 UA ha-1 (lotação potencial). A lotação bovina real em 2015 foi estimada por meio da integração de dados do Censo Agropecuário de 2006 com dados da Produção Pecuária Municipal de 2015. Já a lotação potencial, foi obtida a partir da relação entre a produção de forragem (biomassa verde ou produtividade primária bruta – GPP) e a demanda de forragem de uma unidade animal (1 UA = 450 kg). A partir da diferença entre a lotação bovina real e potencial determinou-se o potencial de intensificação. Para todo o Brasil, a lotação bovina em 2015 foi de 0,97 UA ha-1 , podendo chegar a um potencial de 3,60 UA ha-1 , ou seja, o potencial de intensificação foi de 2,63 UA ha-1 . O maior potencial de intensificação se deu na região Sul (3,62 UA ha-1 ) e os menores nas regiões Norte (2,13 UA ha-1 ) e Nordeste (2,22 UA ha-1 ) do Brasil.

Land use / land cover (LULC) change assessment is getting more consideration by global environmental change studies as land use change is exposing dryland environments for transitions and higher rates of resource depletion. The semiarid regions of northwestern Ethiopia are not different as land use transition is the major problem of the region. However, there is no satisfactory study to quantify the change process of the region up to now. Hence, spatiotemporal change analysis is vital for understanding and identification of major threats and solicit solutions for sustainable management of the ecosystem. LULC change studies focus on understanding the patterns, processes and dynamics of land use transitions and driving forces of change. The change processes in dryland ecosystems can be either seasonal, gradual or abrupt changes of random or systematic change processes that result in a pattern or permanent transition in land use. Identification of these processes of change and their type supports adoption of monitoring options and indicate possible measures to be taken to safeguard this dynamic ecosystem. This study examines the spatiotemporal patterns of LULC change, temporal trends in climate variables and the insights of the communities on change patterns of ecosystems. Landsat imagery, MODIS NDVI, CRU temperature, TAMSAT rainfall and socio-ecological field data were used in order to identify change processes. LULC transformation was monitored using support vector machine (SVM) algorithm. A cross-tabulation matrix assessment was implemented in order to assess the total change of land use categories based on net change and swap change. In addition, the pattern of change was identified based on expected gain and loss under a random process of gain and loss, respectively. Breaks For Additive Seasonal and Trend (BFAST) analysis was employed for determining the time, direction and magnitude of seasonal, abrupt and trend changes within the time series datasets. In addition, Man Kendall test statistic and Sen’s slope estimator were used for assessing long term trends on detrended time series data components. Distributed lag (DL) model was also adopted in order to determine the time lag response of vegetation to the current and past rainfall distribution. Over the study period of 1972- 2014, there is a significant change in LULC as evidenced by a significant increase in size of cropland of about 53% and a net loss of over 61% of woodland area. The period 2000-2014 has shown a sharp increase of cropland and a sharp decline of woodland areas. Proximate causes include agricultural expansion and excessive wood harvesting; and underlying causes of demographic factor, economic factors and policy contributed the most to an overuse of existing natural resources. In both the observed and expected proportion of random process of change and of systematic changes, woodland has shown the highest loss compared to other land use types. The observed transition and expected transition under random process of gain of woodland to cropland is 1.7%, implies that cropland systematically gains to replace woodland. The comparison of the difference between observed and expected loss under random process of loss also showed that when woodland loses cropland systematically replaces it. The assessment of magnitude and time of breakpoints on climate data and NDVI showed different results. Accordingly, NDVI analysis demonstrated the existence of breakpoints that are statistically significant on the seasonal and long term trends. There is a positive trend, but no breakpoints on the long term precipitation data during the study period. The maximum temperature also showed a positive trend with two breakpoints which are not statistically significant. On the other hand, there is no seasonal and trend breakpoints in minimum temperature, though there is an overall positive trend along the study period. The Man-Kendall test statistic for long term average Tmin and Tmax showed significant variation where as there is no significant trend within the long term rainfall distribution. The lag regression between NDVI and precipitation indicated a lag of up to forty days. This proves that the vegetation growth in this area is not primarily determined by the current precipitation rather with the previous forty days rainfall. The combined analysis showed declining vegetation productivity and a loss of vegetation cover that contributed for an easy movement of dust clouds during the dry period of the year. This affects the land condition of the region, resulting in long term degradation of the environment

Made available in DSpace on 2017-07-10T19:24:19Z (GMT). No. of bitstreams: 1 Clovis_Cechim_MC.pdf: 6987482 bytes, checksum: c33db297dd7ec8aaf8bfde9e1e56c2cc (MD5) Previous issue date: 2016-02-04 / Sugarcane has been cropped and produced in Brazil for a long time, so, it deserves mention because it makes the country as the largest producer, with also representativeness in sugar and ethanol production. The knowledge of reliable estimates concerning their cropped areas is essential for Brazilian agribusiness, as they help in determining prices to producers by power plants as well as allow establishing logistics flow of production. The cropped areas estimates are made by official agencies. Therefore, in order to reduce this subjectivity, geotechnology use comes as an alternative since it has been widely used in mappings agricultural crops. Thus, this study aimed at developing a methodology for mapping sugarcane crop in Paraná State with satellite images as LANDSAT, IRS and spectrum-temporal series of vegetation indexes from MODIS sensor, for 2010/2011 to 2014/2015 harvesting season. The carried out mappings indicated a strong positive correlation concerning Canasat and official IBGE. The developed method was based on Fuzzy ARTMAP classification and was efficient to map and estimate the sugarcane cropped area using vegetation index in Paraná State. / A cana-de-açúcar como cultura cultivada e produzida no Brasil merece destaque, pois torna o País o maior produtor mundial, com representatividade também na produção de açúcar e etanol. O conhecimento de estimativas confiáveis de suas áreas cultivadas é imprescindível para o agronegócio brasileiro, por auxiliar na determinação dos preços aos produtores pelas usinas e permitir estabelecer a logística de escoamento da produção. As estimativas de área cultivada são realizadas de forma subjetiva pelos órgãos oficiais. Com a finalidade de diminuir tal subjetividade, surge como alternativa o uso de geotecnologias, as quais têm sido muito utilizadas em mapeamentos de culturas agrícolas. Diante disto, o objetivo deste trabalho foi o desenvolvimento de uma metodologia para o mapeamento da cultura de cana-de-açúcar para o Estado do Paraná usando imagens dos satélites LANDSAT, IRS e de séries espectro-temporais de índices de vegetação, provenientes do sensor MODIS, para as safras de 2010/2011 a 2014/2015. O mapeamento da cultura foi realizado a partir do modelo de classificação supervisionada Fuzzy ARTMAP, tendo como variáveis de entrada, termos harmônicos de amplitude e fase e as métricas fenológicas da cultura. Os mapeamentos realizados indicaram forte correlação positiva com relação aos dados do Canasat e oficiais IBGE. O método desenvolvido com base na classificação Fuzzy ARTMAP demonstrou ser eficiente para mapear e estimar a área cultivada da cultura de cana-de-açúcar utilizando índices de vegetação no Estado do Paraná.

Made available in DSpace on 2017-05-12T14:47:35Z (GMT). No. of bitstreams: 1 Clovis_Cechim_MC.pdf: 6987482 bytes, checksum: c33db297dd7ec8aaf8bfde9e1e56c2cc (MD5) Previous issue date: 2016-02-04 / Sugarcane has been cropped and produced in Brazil for a long time, so, it deserves mention because it makes the country as the largest producer, with also representativeness in sugar and ethanol production. The knowledge of reliable estimates concerning their cropped areas is essential for Brazilian agribusiness, as they help in determining prices to producers by power plants as well as allow establishing logistics flow of production. The cropped areas estimates are made by official agencies. Therefore, in order to reduce this subjectivity, geotechnology use comes as an alternative since it has been widely used in mappings agricultural crops. Thus, this study aimed at developing a methodology for mapping sugarcane crop in Paraná State with satellite images as LANDSAT, IRS and spectrum-temporal series of vegetation indexes from MODIS sensor, for 2010/2011 to 2014/2015 harvesting season. The carried out mappings indicated a strong positive correlation concerning Canasat and official IBGE. The developed method was based on Fuzzy ARTMAP classification and was efficient to map and estimate the sugarcane cropped area using vegetation index in Paraná State. / A cana-de-açúcar como cultura cultivada e produzida no Brasil merece destaque, pois torna o País o maior produtor mundial, com representatividade também na produção de açúcar e etanol. O conhecimento de estimativas confiáveis de suas áreas cultivadas é imprescindível para o agronegócio brasileiro, por auxiliar na determinação dos preços aos produtores pelas usinas e permitir estabelecer a logística de escoamento da produção. As estimativas de área cultivada são realizadas de forma subjetiva pelos órgãos oficiais. Com a finalidade de diminuir tal subjetividade, surge como alternativa o uso de geotecnologias, as quais têm sido muito utilizadas em mapeamentos de culturas agrícolas. Diante disto, o objetivo deste trabalho foi o desenvolvimento de uma metodologia para o mapeamento da cultura de cana-de-açúcar para o Estado do Paraná usando imagens dos satélites LANDSAT, IRS e de séries espectro-temporais de índices de vegetação, provenientes do sensor MODIS, para as safras de 2010/2011 a 2014/2015. O mapeamento da cultura foi realizado a partir do modelo de classificação supervisionada Fuzzy ARTMAP, tendo como variáveis de entrada, termos harmônicos de amplitude e fase e as métricas fenológicas da cultura. Os mapeamentos realizados indicaram forte correlação positiva com relação aos dados do Canasat e oficiais IBGE. O método desenvolvido com base na classificação Fuzzy ARTMAP demonstrou ser eficiente para mapear e estimar a área cultivada da cultura de cana-de-açúcar utilizando índices de vegetação no Estado do Paraná.

Die weltweite Bevölkerungszunahme, sich ändernde Ernährungsgewohnheiten, und die Nachfrage nach Bioenergie erfordern eine Erhöhung der landwirtschaftlichen Produktion. Die Intensivierung bestehender landwirtschaftlicher Flächen ist hierbei eine mögliche Option. Allerdings verstehen wir nur wenig von den räumlichen Mustern der landwirtschaftlichen Nutzungsintensität, da adäquate Datensätze fehlen. Europa ist eine beispielhafte Region, in der eine Intensivierung als auch ein Rückgang der Landnutzung stattfindet. Ziel dieser Dissertation war es Methoden zu entwickeln, die MODIS NDVI Zeitreihen und statistische Daten kombinieren und eine europaweite Kartierung der landwirtschaftlichen Nutzungsintensität ermöglichen. Für eine Einschätzung der landwirtschaftlichen Nutzungsintensität wurden eine Reihe von Intensitätsindikatoren entwickelt und kartiert. Die resultierenden Karten zeigen eine hohe Landnutzungsintensität in West- und Zentraleuropa und dem Mittelmeerraum, gekennzeichnet durch Mehrfachernten und langen Anbauzeiten. Gebiete mit niedriger Intensität lagen in Osteuropa, in Gebirgsregionen sowie in der Extremadura in Spanien, wo Brachland und die Aufgabe von Agrarflächen häufig sind. Die Aufgabe von Agrarflächen ist ein aktueller Prozess der Landnutzungsveränderung in Osteuropa, während die gleichzeitige Rekultivierung ehemaliger Agrarflächen ebenfalls umfassend ist. Diese räumlichen Muster lassen sich mit unterschiedlichen Agrarumweltbedingungen begründen aber auch mit sozioökonomischen Veränderungen wie die Restrukturierung des osteuropäischen Agrarsektors nach 1989 oder die Marginalisierung landwirtschaftlicher Flächen insbesondere in Gebirgsregionen. Die entstandenen Karten belegen das Potential von MODIS NDVI Zeitreihen, komplexe Phänomene landwirtschaftlicher Nutzungsintensität zu erfassen. Diese könnten genutzt werden um Umweltfolgen der landwirtschaftlichen Produktion zu bewerten oder Zielregionen für eine nachhaltige Intensivierung zu identifizieren. / Global population growth, changing diets, and the demand of bioenergy require an increase in agricultural production. Intensifying agricultural production is one pathway to meet increasing demands. However, our understanding of spatial patterns of agricultural land use remains weak since adequate data sets are lacking. Europe is as a prime example for a region that is undergoing both, intensification as well as decreasing agricultural land use. The goal of this doctoral thesis was to develop methodologies that combine MODIS NDVI time series and agricultural statistics to map spatial patterns of land-use intensity across Europe. To assess land-use intensity, a wide range of intensity indicators were mapped. The resulting maps revealed high-intensity areas in Western and Central Europe and the Mediterranean region, characterized by multi-harvests and long crop duration. Low-intensity areas are mostly located in Eastern Europe, in mountain regions and the Extremadura in Spain, where fallow and abandonment land are widespread. Agricultural abandonment is an ongoing land-use change process in Eastern Europe. At the same time, recultivation of formerly abandoned land is widespread as well. These spatial patterns are the result of agro-environmental conditions but also of changes in socio-economic conditions such as the restructuring of the agricultural sector in eastern European countries after 1989, or the marginalization of farmland especially in mountain regions. The resulting maps show the potential of MODIS time series to assess the complex phenomenon of land-use intensity. They may form a basis to assess the environmental outcomes of agricultural production and to identify target regions for sustainable intensification.