There have been reports that rainfall in East Africa is changing or becoming more variable. This can have significant implications for conservation initiatives and the food security of this populace region that is heavily reliant on the rain fed agricultural system. The perceptions of farmers regarding rainfall along with 30 years of satellite data and 16 years of ground level observations were analyzed in order to characterize rainfall in and around Kibale National Park, a protected area in the Ugandan portion of the Albertine Rift. Two homogenous rainfall regions exist in the area, and the onset, cessation, and amount of rainfall during seasons is highly variable. The perceptions of farmers align with the analysis of rainfall data, indicating that the season beginning in March shows the highest degree of variability. Decreases in the amount of rainfall are found for both rainy seasons.

Made available in DSpace on 2015-07-15T13:31:42Z (GMT). No. of bitstreams: 1 dissertacaoCharles.pdf: 2350650 bytes, checksum: 3a5a6710de4fe0f9f72e936ffc6d9a22 (MD5) Previous issue date: 2007-11-22 / Brazil Northeast s climate is usually described as semi-arid, characterized by hard dry seasons intermingled by hard rainfall seasons. Some areas of Northeast present annual pluviometric measure about 400 mm in mean, while in others the annual pluviometric measure is about 2000 mm. Rain events result from interplay of several physical phenomena, most of which can be individually described on the basic laws of mechanics and thermodynamics in a rather adequate way. Because of this, a huge progress has been achieved in recent years in relation to weather forecast with the use of very precise algorithms in large scale computing resources. They take into account the variables that are relevant for the atmospheric and ocean circulation and input of large amount of physical data obtained from a dense set of stations scattered around the world. In order to improve the interpretation of the accurate data resulting from the description of atmospheric phenomena and rain events, it is necessary to proceed with sophisticated analyses of recorded and simulated data, as spatial and temporal statistical correlations, scale properties, topological properties of spatial event distribution, ad so on. They indicate the extent of statistical relevance of the data, local and global effects, typical patterns, and other topological features related to the phenomena.In this work, we explore the usefulness of complex network framework for the analysis and nderstanding of rain events, based solely on recorded data from a set of stations in Northeast Brazil. The method is inspired on a proposal to characterize actual sequences of earthquake events where, like precipitation phenomena, the available data stems from complex systems with a very large number of physical variables. The potential network nodes are the meteorological stations where the rain events have been recorded, while the network edges are placed according to rules that take into account temporal and spatial correlation criteria between events occurring at different stations, for a time span as large as one month. We evaluate usual network properties based on diameter, node degrees, clustering coefficient, minimal inter-node distance along network edges. This allows for a characterization of networks based on seasonality and on spatial span of the region where the stations are distributed. The obtained results are discussed, taking into account the known precipitation patterns of the investigated region. rainfall variability, complex networks, fractals. / Climaticamente, a regi ao Nordeste do Brasil ?e marcada pela predomin?ncia de clima semi-?arido, caracterizado por per?ıodos de secas severas intercalados por per?odos de chuvas intensas. Eventos clim?aticos como a chuva resultam da intera??o de v?rios fen?menos f?sicos que, em sua maioria, pode ser descrita individualmente pelas leis b?sicas da mec?nica e termodin?mica de forma satisfat?ria. Por esse motivo, um imenso progresso tem sido observado, nos ?ltimos anos, com rela??o ? previs?o de tempo e clima utilizando algoritmos mais precisos em recursos computacionais de larga escala. Estes algoritmos levam em considera??o as vari?veis que s?o relevantes para a circula??o atmosf?rica e oce?nica al?m de uma grande quantidade de dados f?sicos obtidos de um conjunto denso de esta??es distribu?das ao redor do mundo. Com objetivo de prover a interpreta??o dos dados destes algoritmos, ? necess?rio proceder com an?lises sofisticadas dos dados armazenados e simulados, como correla?c oes estat?ısticas temporais e espaciais, propriedades de escalas, propriedades topologicas da distribui??o espacial de eventos, etc. Os resultados falam sobre a relev?ncia estat?stica dos dados, efeitos locais e globais, padr?es t?picos e outros recursos relacionados ao fen?meno. Neste trabalho, n?s exploramos o uso da Teoria de Redes Complexas para a an?lise e interpreta??o de eventos de chuva, baseandonos somente em registros de dados de um conjunto de esta??es pluviom?tricas da regi?o Nordeste do Brasil. Este m?etodo ? inspirado em uma proposta para caracterizar sequ?ncias de eventos s?smicos, eventos em que, assim como no fen?meno das chuvas, a grande quantidade de vari?aveis f?sicas envolvidas motiva a an?lise usando m?todos da Teoria de Sistemas Complexos. Os n?s das redes geradas s?o as esta??es meteorol?gicas onde h? dados de chuva no per?odo analisado, enquanto as arestas s?o criadas de acordo com crit?rios de correla??o temporal e espacial entre eventos de chuva ocorridos em diferentes esta??es pluviom?tricas. Calculamos os ?ndices mais comuns de caracteriza??o de redes complexas, tais como: di?metro, caminho m?nimo m?dio, coeficiente de aglomera??o m?dio. As redes conectam esta??es a diferentes dist?ncias, e a fim de estudar a causalidade n?o-local desse fen?meno foram calculados ?ndices fractais de caracteriza??o. Os valores de di?metro e de caminho m?nimo m?dio s?o menores para os meses de inverno e primavera, t?picos de chuva mais localizada no litoral; enquanto que para os meses de ver?o e outono, t?picos de chuva mais distribu?ıda em toda a regi?o, os valores s?o maiores. A dimens?o fractal calculada para dados do Sul do Nordeste (Bahia) ? semelhante ? calculada para dados do Norte do Nordeste (demais estados da Regi?o), mas ambas s?o diferentes das dimens?es fractais de redes completas e regulares hipot?ticas, o que demonstra que a distribui??o das esta??es pluviom?tricas n?o ? homog?nea. Estes resultados sugerem o estudo mais aprofundado deste m?todo de an?lise de dados pluviom?tricos, que, atrav?s da modelagem em Sistemas Complexos.

Made available in DSpace on 2014-12-17T15:03:32Z (GMT). No. of bitstreams: 1 JuremaMSA_DISSERT.pdf: 2330527 bytes, checksum: f58a53e0a077e8160ead9878b1c5ac07 (MD5) Previous issue date: 2013-08-30 / O estudo da variabilidade da precipita??o ? importante para o planejamento das atividades econ?micas, possibilitando o uso mais eficiente e racional dos recursos h?dricos. Dessa forma, o objetivo desta pesquisa ? caracterizar o estado do Rio Grande do Norte com rela??o ? variabilidade temporal da precipita??o, agrup?-lo em regi?es homog?neas e comparar diferentes t?cnicas de agrupamento. Para o estudo da variabilidade pluvial foram utilizados os ?ndices: Grau de Concentra??o de Precipita??o (PCD), que representa o grau em que a precipita??o ? distribu?da ao longo do ano; e o Per?odo de Concentra??o de Precipita??o (PCP), que reflete o per?odo no qual a precipita??o est? mais concentrada. Para a realiza??o dos agrupamentos foram escolhidas as vari?veis: PCD, PCP, m?dias da precipita??es anuais e m?dias das precipita??es mensais. Posteriormente, foi aplicada a an?lise de agrupamento para obter grupos com caracter?sticas similares. Os resultados mostraram que as precipita??es s?o melhor distribu?das na regi?o leste do estado, neste caso, os meses mais chuvosos s?o de maio a agosto. Os munic?pios localizados nessa ?rea possuem dois picos de chuvas, devido ? atua??o de dois sistemas: Perturba??es Ondulat?rias dos Al?sios (POA s) e Zona de Converg?ncia Intertropical (ZCIT). Nas regi?es localizadas a oeste os meses que possuem maior concentra??o de chuvas s?o mar?o e abril, neste caso temos apenas um pico de precipita??o, devido a atua??o da ZCIT. A identifica??o de ?reas homog?neas favorece o planejamento adequado de acordo com as caracter?sticas de cada grupo formado e o RN pode foi dividido em 4 (quatro) regi?es homog?neas. As t?cnicas de agrupamento utilizadas apresentaram resultados semelhantes, por?m, sugere-se o uso de mais de uma t?cnica para que se possa analisar qual delas reflete melhor a realidade local. O estudo da variabilidade de precipita??o, atrav?s dos ?ndices estudados e do agrupamento realizado, s?o ferramentas adequadas ao planejamento ambiental e econ?mico

Thesis (M.Sc. (Statistics)) -- University of Limpopo, 2020 / Several studies indicated a growing trend in terms of frequency and severity of extreme events. Extreme rainfall could cause disasters that lead to loss of property and life. The aim of the study was to model the monthly rainfall variability in selected provinces of South Africa using extreme value distributions. This study investigated the best-fit probability distributions in the five provinces of South Africa. Five probability distributions: gamma, Gumbel, log-normal, Pareto and Weibull, were fitted and the best was selected from the five distributions for each province. Parameters of these distributions were estimated by the method of maximum likelihood estimators. Based on the Akaike information criteria (AIC) and Bayesian information criteria (BIC), the Weibull distribution was found to be the best-fit probability distribution for Eastern Cape, KwaZulu-Natal, Limpopo and Mpumalanga, while in Gauteng the best-fit probability distribution was found to be the gamma distribution. Monthly rainfall trends detected using the Mann–Kendall test revealed significant monotonic decreasing long-term trend for Eastern Cape, Gauteng and KwaZulu-Natal, and insignificant monotonic decreasing longterm trends for Limpopo and Mpumalanga. Non-stationary generalised extreme value distribution (GEVD) and non-stationary generalized Pareto distribution (GPD) were applied to model monthly rainfall data. The deviance statistic and likelihood ratio test (LRT) were used to select the most appropriate model. Model fitting supported stationary GEVD model for Eastern Cape, Gauteng and KwaZulu-Natal. On the other hand, model fitting supported non-stationary GEVD models for maximum monthly rainfall with nonlinear quadratic trend in the location parameter and a linear trend in the scale parameter for Limpopo, while in Mpumalanga the non-stationary GEVD model, which has a nonlinear quadratic trend in the scale parameter and no variation in the location parameter fitted well to the maximum monthly rainfall data. Results from the non-stationary GPD models showed that inclusion of the time covariate in our models was not significant for Eastern Cape, hence the bestfit model was the stationary GPD model. Furthermore, the non-stationary GPD model with a linear trend in the scale parameter provided the best-fit for KwaZulu-Natal and Mpumalanga, while in Gauteng and Limpopo the nonstationary GPD model with a nonlinear quadratic trend in the scale parameter fitted well to the monthly rainfall data. Lastly, GPD with time-varying thresholds was applied to model monthly rainfall excesses, where a penalised regression cubic smoothing spline was used as a time-varying threshold and the GPD model was fitted to cluster maxima. The estimate of the shape parameter showed that the Weibull family of distributions is appropriate in modelling the upper tail of the distribution for Limpopo and Mpumalanga, while for Eastern Cape, Gauteng and KwaZulu-Natal, the exponential family of distributions was found to be appropriate in modelling the upper tail of the distribution. The dissertation contributes positively to the body of knowledge in extreme value theory application to rainfall data and makes recommendations to the government agencies on the long-term rainfall variability and their negative impact on the economy.

This study was conducted in the Borana rangelands, southern Oromia, Ethiopia with the aim of investigating land-use/cover changes, rangeland condition and indentifying major change drivers in the rangelands. Satellite image scenes, ground survey and people’s perceptions were assessed to identify changes on herbaceous and woody species composition. Ordination techniques were used for correlation of grass and woody species into environmental variables. Piospheric effects on vegetation composition around patch resources were also examined. These assessments were designed to bring out both spatial and seasonal variation in vegetation parameters. Rangeland condition was determined using the ecological index (EIM) and the weighted palatability composition (WPC) methods. Grazing capacity assessment was largely based on rain-use efficiency, range condition, density of woody plants. Woody cover in the Borana increased from 11.3% in the 1970s to 49.26% in 2000s, while, grassland cover declined from 58% to 32% during the same period. The cultivated/built up area also increased gradually over the years though the extent of increment was less compared to the woody cover changes. The calculated NDVI values for the 2000s were low relative to the 1970s. Ground survey results demonstrated that herbaceous biomass production and woody plant density varied significantly (P < 0.01) for the different sites in Borana. Nevertheless, herbaceous plant diversity and evenness did not differ significantly (P>0.05) across the different sites and around patch resource areas. This confirms the resilience of the Borana rangeland to the effects of grazing pressure and climatic variability. Overall, the density of woody plants varied from moderate to sever encroachment, which corresponds to the rangeland condition classes from very poor to fair. The results of the present study showed that the nutrient contents of herbaceous plants (CP, NDF, ADF, ADL and ash) were greatly influenced by species, stage of maturity, site and season. Herbaceous species with high crude protein content based on laboratory results were also ranked as the top important species by the pastoralists. Similarly, herbaceous forage species with a high structural fiber were considered inferior. Pastoralists’ knowledge and laboratory results on the nutritive value of key herbaceous species complemented each other. The stocking density of livestock units was higher than the grazing capacity for the Borana rangelands. Stocking density for the Borana rangeland using rain-use efficiency was 1.43 ha/TLU. There was no significant difference in the grazing capacity across the different sites in the study area under the existing management. The majority of pastoralists believed that the rangelands in Borana have been degraded, as consequence of recurrent drought (83%), population pressure (65%), poor management and inappropriate government policy (38%). The majority of the respondents (94%) also asserted that the traditional coping mechanisms of Borana pastoralists have declined. There is therefore a need for designing appropriate pastoral land-use policies that fit the ecological potential of the region to ensure sustainable ecosystem functioning. We suggest a comprehensive and adaptive range management for the implementation of appropriate land-use systems for the different livestock species. There is need to adjust stocking rates based on seasonal availability of forage. This would foster economic feasibility and ecological sustainability of the Borana pastoral production system. / Thesis (PhD)--University of Pretoria, 2013. / gm2013 / Plant Production and Soil Science / Unrestricted

Simulating the processes of rainfall and runoff are at the core of hydrologic modeling. Geomorphologic features, rainfall variability, soil types, and water table depths strongly influence hydrological process in Florida ecosystems. Topographic characteristics of the terrain define the stream paths and landscape. Alteration of these characteristics as a result of urban and/or agricultural developments, for example, can highly influence wetlands and river basin response. There are two predominant landforms in Florida: wetlands, where Variable Saturated Areas form near streams causing saturation excess runoff, and uplands where runoff is mainly generated by infiltration excess. The objective of this work is to analyze the impacts of geomorphologic and hydrologic characteristics on runoff mechanisms in humid environments such as Florida. In general, most research at the hillslope scale use hypothetical values of rainfall, sometimes non-realistic values, and single slope forms to explain the geomorphic and hydrologic process on Variable Saturated Areas. In this thesis, the complexity of hillslope processes on actual Florida topography is assessed by coupling a Digital Elevation Model with a two-dimensional variable saturated-unsaturated flow model called HYDRUS-2D. Actual rainfall records and soil parameters from the Characterization Data for Selected Florida Soils, Soil Survey were used to evaluate hydrologic impacts. A commercial software package, River Tools was used to display and extract topographic information from the Digital Elevation Models. Results show that when inflitration excess runoff is dominant, infiltration and runoff are very sensitive to time resolution, especially for convective storms. When saturation excess occurs, runoff is not affected by rainfall intensity. However, saturated hydraulic conductivity, depth to the water table, slope and curvature highly influence the extent of Variable Saturated Areas. Results indicate runoff in shallow water table environments is produced mainly by subsurface storm runoff, running below the surface, except in hillslopes with concave curvature and mild slopes. Additionally, concave hillslopes generate more saturation excess runoff than straight and convex hillslopes.

Over the last 2000 years, climatic and anthropogenic factors have influenced vegetation in Madagascar, but the contribution of these two factors has been the subject of intense debate, a debate hampered by the scarcity of palaeoecological studies on the island. Southwest Madagascar is semi-arid and comprises fragmented tropical dry forests where human subsistence strategies are diverse. Therefore, it provides a perfect setting to generate new palaeoecological records and investigate vegetation response to changes in human land-use and rainfall variability. The aim of this thesis is to understand how and when land-use changed, and rainfall variability impacted the landscape in the southwest region, using dendroclimatological and palaeoecological approaches. Carbon isotopes in the rings of four baobab trees (Adansonia spp.), were compared with pre-existing palaeoclimate data to produce rainfall records for the past 1700 years. Pollen, carbon isotopes, and charcoal in sediment cores from two lakes namely Lake Longiza and Lake Tsizavatsy (located in the northern and southern sites respectively in southwest Madagascar) were analysed to produce vegetation and fire records during the late Holocene in the region. Combination of the four baobab isotope records yields a new 700-year record for the southwest, which suggests an aridity trend over time, associated with a long-term reduction and increase in the duration of wet and dry periods respectively. Drying was more pronounced for the southern site than the northern site. A comparison with a high-resolution record from the northwest region allowed the rainfall of the southwest to be extended back to the last 1700 years as follows: from AD 300-500, the southwest region experienced a dry period which was followed by a wet period until AD 1000. Subsequently, there was a dry period from AD 1000-1250, followed by a wet period in the southwest. The period from AD 1300-1400 represented the wettest period in the record, followed by a decreasing wet period until AD 1600. The period between AD 1600 and 1800 represented the driest period, peaking around AD 1700. This was followed by a relatively wet period of about 50 years and another 100-year dry period. The last period assessed, from AD 1985 to 2000, was a relatively wet period. Such variability of the rainfall might have affected ecosystems and human land-use in the region. The tropical dry forest biome of southwest Madagascar has taxa from the dry forest, riparian forest and savanna woodland in the pollen records of both sites. In the core from the northern site (Lake Longiza), the pollen record suggests a heterogeneous mosaic of dry forest and riparian forest that was present over the last 2400 years. In the earliest part of the record, the community was dominated by trees from both the dry forest and riparian ecosystems. The onset of changes was recorded around AD 420, with a decrease in dry forest and riparian trees and an increase in grasses and xerophytics, possibly driven by dry conditions. This was followed by a short recovery of trees and C3 plants around AD 870 punctuated by a decrease in dry forest taxa around AD 980, possibly associated with the suggested expansion of pastoralism. After this period, the landscape became more open and grassier, as indicated by the dominance of C4 plants in the stable isotope record. Similar patterns of change in the pollen record, with a further increase in grasses and pioneer taxa, were recorded around AD 1900. Charcoal influx also started to increase drastically at this time, suggesting fire and forest clearance associated with a shift to agriculture. These large shifts in human land-use (probably a combination of both pastoralism and agriculture) coincided with the decline of floral diversity of the landscape, as indicated by pollen rarefaction. However, the diversity of the floral community gradually recovered, because of the persistent heterogeneity of the landscape. The core from the southern site (Lake Tsizavatsy) had a basal date of approximately 700 years BP but presented a hiatus of about 500 years from AD 1420-1910. The preceding period of AD 1300-1420 was marked by a decrease in the abundance of trees during the wettest period in the region, which was most likely because of human activities (foraging and pastoralism), as inferred by the increase in charcoal influx and pioneer taxa. During the second period, from AD 1910- 2010, there was an increase in xerophytic taxa, which suggests a long dry climate, recorded prior to this period. In addition, from AD 1950, trees decreased while pioneer taxa increased, despite the stable influx of charcoal recorded during this period. This possibly indicates the effect of human activities that did not involve the use of fire, probably conducted by ethnicities other than the forager communities, which are still present in the area today. This thesis contributes significantly to the understanding of palaeoclimate, palaeoecology and the history of human subsistence in a biodiverse region of Madagascar, where no other record is currently available. Results from stable isotope analysis from baobabs showed a drying trend over the past 700 years, which has interacted with land-use to affect vegetation structure and composition over time. The pollen and charcoal results suggest the northern site, where vegetation was a mosaic of dry forest and riparian forest, experienced an impact of human activities through a shift to agriculture especially in the last 100 years. The savanna woodland of the southern site, however, was less affected by humans, probably as occupants were subsistence foragers, but the vegetation had a higher response to aridity. The results show that two distinctive human subsistence (pastoralism and foraging) were present simultaneously in the region until modern times. The northern site has evolved possibly from foraging into extensive agriculture, probably related to the fertility of the alluvial soil in the area, while the southern community remained dominated by foragers, while adopting today a seasonal practice of agriculture. From a conservation perspective, strategies of conservation for each ecosystem investigated here are proposed. In the northern site, monitoring and reducing fire-use within the dry forest ecosystem would allow tree recovery. In addition, restoring and establishing protected areas within the riparian forest would allow these ecosystems to act as refugia for regional biodiversity. Such measures will likely reduce the pressure on these ecosystems, where agriculture is a threat due to the availability of both water and fertile soil in their surroundings. Alternative livelihoods are required for the northern populace, for example through the exploitation of invasive aquatic plants such as Typha, which can be used in making of handcrafted artefacts, to reduce pressure on forest ecosystems through agricultural practices. For the southern site, maintaining the resilience of the savanna woodland through reforestation of functional species is also important to allow sustainability of services provided by these ecosystems. These strategies are applicable locally for Madagascar and for worldwide tropical dry forests, one of the globally most threatened vegetation types due to anthropogenic pressure and climate change.

付記する学位プログラム名: グローバル生存学大学院連携プログラム / 京都大学 / 0048 / 新制・課程博士 / 博士(工学) / 甲第19975号 / 工博第4219号 / 新制||工||1653(附属図書館) / 33071 / 京都大学大学院工学研究科社会基盤工学専攻 / (主査)教授 立川 康人, 教授 寶 馨, 教授 堀 智晴 / 学位規則第4条第1項該当 / Doctor of Philosophy (Engineering) / Kyoto University / DFAM

Les multifractals universels (UM) sont un outil puissant et abondement utilisé d'analyse et de simulation de champs géophysiques, comme la pluie, extrêmement variables sur une large gamme d'échelle. Ils sont basés sur le concept de cascade multiplicative qui repose sur la notion physique d'invariance d'échelle pour explorer le phénomène fondamental qu'est l'intermittence. Dans ce cadre, toute la variabilité du champ est caractérisée à l'aide de simplement trois paramètres qui ont en plus une interprétation physique. Dans cette thèse on utilise ce cadre théorique pour quantifier l'impact de la variabilité à petite échelle de la pluie en hydrologie urbaine. La première étape consiste à analyser la variabilité spatio-temporelle de données radar de précipitation à l'aide d'un modèle multifractal anisotrope simple. Divers évènements pluvieux sont analysés. Un comportement scalant a été observé sur deux gammes d'échelles séparées par une rupture à 16 km qui est discutée. Ces données sont globalement en accord avec un modèle spatio-temporel simple reposant un exposant d'anisotropie entre l'espace et de temps. Les résultats suggèrent une possible universalité des paramètres UM pour les précipitations. Cette thèse aborde également un autre aspect de l'intermittence, particulièrement important pour les longues séries temporelles pluviométriques, que sont les nombreuses mesures nulles de la pluie (c'est-à-dire un pixel où aucune pluie n'est relevée), i.e. les longues périodes sèches. L'ancienne question de la source de cette intermittence, et notamment la nécessité d'un modèle dédié, est revisitée. D'abord les effets d'un seuil sur un champ multifractal sont analysés et ensuite un « toy model » qui introduit des zéros au sein du processus de cascade et conditionnellement aux valeurs du champ est développé. Cela permet d'expliquer la plupart des comportements observés, e.g. les différences entre les statistiques évènementielles et globales. L'impact de la variabilité de la pluie est analysé à travers l'étude de la sensibilité de modèles d'hydrologie/hydraulique urbaine à la donnée de pluie. Deux bassins versants essentiellement urbains (un de 3 400 ha en Seine-Saint-Denis à proximité de Paris, et un de 900 ha à Londres) modélisés avec des modèles opérationnels semi-distribués sont pris comme cas d'études. Par ailleurs le modèle distribué Multi-Hydro (en développement au LEESU) est testé sur une portion de 145 ha du cas d'étude parisien. L'impact de la variabilité à petites échelles non mesurée des précipitations (i.e. se produisant à des échelles plus petites que 1 km en espace et 5 min en temps qui sont disponibles avec les données radar à bande C) est d'abord évalué. Ceci est réalisé par la génération d'un ensemble de pluie réaliste désagrégée en continuant stochastiquement le processus sous-jacent de cascade au-delà de l'échelle d'observation, puis la simulation de l'ensemble correspondant d'hydrographes. Il apparaît que la variabilité à petites échelles de la pluie engendre une variabilité hydrologique qui ne doit pas être négligée. De plus le modèle Multi-Hydro génère une variabilité plus importante et pas seulement au niveau du pic de débit, i.e. même pour les pluies modérées. Ces résultats mettent en lumière la nécessité d'installer des radars en bande X (dont la résolution est hectométrique) en milieu urbain. Dans un deuxième temps les outils multifractals sont employés sur les pluies et les débits simulés qui présentent aussi un comportement scalant. Il apparaît que le réseau d'assainissement transmet simplement la variabilité des précipitations sans l'atténuer, au moins en termes de statistiques multifractals / The Universal Multifractals (UM) are a powerful tool which has been extensively used to analyze and simulate geophysical fields, such as rainfall, that are extremely variable over wide range of scales. It is based on the concept of cascade phenomenology that relies on the physical notion of scale invariance to explore the fundamental phenomenon of intermittency. In this framework the whole variability of a field is characterized with the help of only three parameters that are furthermore physically meaningful. In this PhD thesis we use this theoretical framework to quantify the impacts of small scale rainfall variability in urban hydrology. The first step consists in analysing radar rainfall space-time variability with the help of a simple anisotropic multifractal model. A variety of rainfall events are analyzed. It appears that a scaling behaviour was observed on two distinct ranges of scales separated by a break at roughly 16 km that is discussed. These data sets are in overall agreement with a simple space-time scaling model relying on single anisotropy exponent between space and time. The results hint at a possible universality of the UM parameters for rainfall. This thesis also explores another facet of intermittency, which is particularly important for long time series of precipitation, that of numerous zero rainfall measurements (a pixel or a time step with no recorded rainfall), i. e. long “dry” periods. We revisit the long lasting discussion on the source of this intermittency, e.g. whether it requires a specific modelling. First the effects of a threshold on a universal multifractal field are investigated and second a toy model that introduces some zeros within the cascade process conditioned by the field value is developed. This enables to explain most of the observed behaviour, e.g. the difference between event statistics and overall statistics. The impact of rainfall variability is investigated through the analysis of the sensitivity to the rainfall input of urban hydrologic-hydraulic models. Two predominantly urban catchments (a 3 400 ha one in Seine-Saint-Denis near Paris, and a 900 ha one in London) modelled with the help of operational semi-distributed models are used as case studies. The fully distributed model Multi-Hydro (under development at LEESU) is also tested on a 147 ha portion of the Paris case study. First the impact of unmeasured small scale rainfall variability (i.e. occurring at scales smaller than 1 km in space and 5 min in time which are available with C-band radar data) is evaluated. This is achieved by generating an ensemble of realistic downscaled rainfall fields by continuing the stochastic cascade process below the observation scale and then simulating the corresponding ensemble of hydrographs. It appears that the small scale rainfall variability generates significant hydrological variability that should not be neglected. Furthermore the Multi-Hydro model generates a larger variability not only during the peak flow, but during the whole event, i.e. for moderate rain rates. These findings highlight the need to implement X-band radars (whose resolution is hectometric) in urban areas. In a second part multifractal tools are used on both rainfall and simulated discharges that also exhibit a scaling behaviour. It appears that the rainfall drainage system basically transmits the rainfall variability without damping it, at least in terms of multifractal statistics

Com os significativos índices de desmatamento na região Amazônica, grande é a preocupação sobre os efeitos que essa mudança na cobertura vegetal pode ter sobre a dinâmica climática e pluviométrica dessa região, como também em outras localidades do Brasil. Dentro desse contexto, o presente trabalho teve como objetivo verificar se os atuais índices de desmatamento em Rondônia, ocorridos nos últimos 30 anos, são suficientes para contribuir para a modificação do regime hídrico e pluviométrico. Os dados utilizados no presente estudo foram obtidos juntos à Agência Nacional de Águas ANA e são provenientes de 29 postos pluviométricos instalados no estado de Rondônia. Essa é uma região de pequena disponibilidade de dados meteorológicos devido à colonização recente, sendo que as séries dificilmente passam de 30 anos de dados. Além disso, essas séries são comprometidas por falhas na coleta de dados, que em algumas localidades chegam a apresentar mais de um ano sem registro. Para que as análises fossem possíveis, foi realizado o preenchimento das falhas nas séries de dados através do gerador de dados CLIMGEN. Em seguida, os dados foram analisados pelo método de Duplas Massas para que eventuais falhas e inconsistências nas séries preenchidas pelo CLIMGEN fossem detectadas. Como passo seguinte, os dados foram submetidos ao ajuste de distribuição de probabilidade pela função Gama, para que fossem identificadas regiões homogêneas quanto à distribuição de probabilidade da chuva em Rondônia. Acompanhando essa análise foi utilizada a análise de agrupamento para que fossem identificas as localidades e os meses com precipitação semelhante. Finalmente, os dados de precipitação foram submetidos ao teste de tendência Mann- Kendall, assim como as variáveis do balanço hídrico seqüencial das localidades de Ouro Preto dOeste e Porto Velho. Os resultados mostram que o CLIMGEN é uma ferramenta que pode ser utilizada no preenchimento de falhas em séries de dados em Rondônia, porém o método de Duplas Massas identificou inconsistências em duas localidades, sendo que as mesmas foram descartadas nas análises seguintes. Quanto à função gama foi verificado um menor grau de assimetria na distribuição de probabilidade da precipitação nos meses mais chuvosos. Nos meses secos, junho, julho e agosto, a distribuição de probabilidade da precipitação é mais assimétrica, sendo semelhante entre todas as localidades. Acompanhando a distribuição gama, a análise de agrupamento mostrou uma heterogeneidade espacial na distribuição da precipitação. Finalmente, as análises de tendência não identificaram mudanças sistemáticas que indiquem que o estado de Rondônia esteja passando por mudanças no regime pluviométrico no período analisado, sendo as tendências observadas possivelmente decorrentes da variabilidade climática natural. As séries de temperatura máxima e mínima de Porto Velho e Ouro Preto dOeste apresentaram tendências, com indícios de aumento principalmente da temperatura mínima. Quanto às análises de tendência nas variáveis do balanço hídrico foram verificadas tendências de aumento principalmente da ETR e ETP. Os resultados permitiram concluir que apesar do significativo aumento da área desmatada em Rondônia, isso ainda não foi suficiente para que mudanças no clima fossem identificadas, sendo necessárias análises de séries mais longas. / Amazon deforestation rate has been significant, resulting in a great concern about the effect that the reduction of the forest can have on the climatic dynamics in this region, as well as in other brazilian locations. In such context, the present work aimed to verify if the current rate of deforestation in Rondônia, last 30 years has been contributed for changes in the rainfall and balance patterns. Data from 29 rainfall stations in Rondônia State which belongs to the National Water Agency ANA were used for this purpose. This is a region of small availability of weather data due to recent settling, with the data series hardly passing of 30 years. Moreover, those data series are compromised by imperfections in the collection of data, resulting in missing data. To have this study feasible, the through the weather missing data were filled out data generator CLIMGEN. After that, the data was analyzed by the method of Double Masses so that imperfections and inconsistencies in the series filled out with CLIMGEN were detected. As a next step, the data were submitted to the adjustment of the Gamma distribution of probability to identify homogeneous regions in relation to the rainfall distribution of probability in Rondônia. Sveh analysis was applied to identify the localities and months with similar rainfall pattern. Finally, the rainfall data was submitted to Mann-Kendall trend test, as well as the variables of the sequential water balance in Ouro Preto dOeste and Porto Velho locations. The results showed that the CLIMGEN is a tool that can be used to fill out missing data in series of Rondônia; however the method of Double Masses identified inconsistencies in two localities, which were not considered in the analyses. The Gamma distribution function presented a smaller degree of asymmetry in the distribution of rainfall probability in the rainy months. Dry months, like in June, July and August, the distribution of rainfall probability is more asymmetrical, being similar among all evaluated locations. Following Gamma the distribution, the cluster analysis showed spatial heterogeneity in the rainfall distribution. Finally, the tendency analyses did not identified systematic changes which indicated that Rondônia State is having changes in the rainfall pattern during the analyzed period, being verified only oscillations of natural order in the climate. The series of maximum and minimum temperature of Porto Velho and Ouro Preto dOeste presented positive tendency, mainly in the minimum temperature. For the variables of the water balance, positive tendencies were detected for ETR and ETP. The results allowed to conclude that although the significant increase of the deforested area in Rondônia, this was not enough yet to cause changes in the climate, being necessary analyses of longer series.