La migration latérale est reconnue comme un processus majeur à préserver pour le bon fonctionnement des écosystèmes aquatiques. Ce travail explore sa variabilité spatiale dans le Sud-Est de la France en appui à la mise en œuvre de la Directive Cadre sur l’Eau.Une première partie développe un cadre méthodologique et des outils géomatiques pour la caractérisation géomorphologique et l’analyse multi-scalaire des systèmes fluviaux à partir de données disponibles à large échelle. Un premier article présente la démarche générale basée sur l’agrégation et la désagrégation spatiale d’objets vectoriels. Un second article aborde spécifiquement le problème de l’agrégation spatiale et vise à comparer des techniques statistiques pour la sectorisation d’un continuum en tronçons homogènes. Des exemples illustrent l’intérêt et les limites des outils développés.Une seconde partie porte sur la régionalisation de la dynamique latérale des lits fluviaux. Un troisième article caractérise et modélise son intensité à partir de variables simples à extraire à l'échelle régionale. Il montre qu’elle est structurée spatialement et positivement corrélée à la puissance fluviale. Néanmoins, des limites inhérentes à la modélisation empirique et la nécessaire prise en compte de variables complémentaires à l’échelle régionale émergent (apport sédimentaire notamment). Un quatrième article porte spécifiquement sur les lits à tresses et montre que près de la moitié des 1200 kilomètres recensés avant la construction des grandes infrastructures ont disparu du fait de pressions anthropiques directes et indirectes. Il révèle par ailleurs la grande diversité géomorphologique de ce style fluvial dans le bassin rhodanien. / The channel migration is a key process for preserving the aquatic ecosystem functioning. This work explores its spatial variability throughout the South-East French hydrographic network for the implementation of the European Water Framework Directive.A first part develops a methodological framework and geomatic tools for the characterization and analysis of the fluvial systems for a range of scales from spatially continuous data available at the large scale. A first paper introduces the methodological framework based on the spatial disaggregation and aggregation of geographical objects. A second paper focuses specifically on the aggregating problem and aims to compare statistical techniques for delineating homogeneous reaches along a continuum. Examples illustrate the potentialities and limits of the tools we developed. A second part focuses on the regionalization of the channel migration. A third paper characterizes and models the migration rates from simple variables that can be extracted at the large scale. We show that the migration rate is spatially organized and positively controlled by the gross stream power. Nevertheless, limits of the regional-scale empirical modeling emerge, as well as additional variables that should be integrated (particularly the sediment supply). A fourth paper focuses specifically on braided rivers and shows that near the half of the 1200 kilometers censed prior to the major infrastructure construction disappeared due to direct and indirect human impacts. It also reveals the high geomorphic diversity of the contemporary braided rivers in the Rhône basin.

This work was based in methods of classic descriptive statistics, methods of geostatistics and geoprocessing, in the identification of the size and the structure of the spatial variability of the physical and chemical attributes of the soil in area of Paraguay Tea forestry. The area of study, localized in the Tupian Farm, in the municipality of New Silver, RS, where were raised samples from January to March, 2005, embracing the Latosoil humic dystrophic soil. It was accomplished systematic sample with grid of regular spacing among the one hundred meter points, totalizing thirty-six sample points and the six hundred and thirty pairs of data, in an area of thirty-six hectares. Were collected samples of soil in situ for analysis in laboratory of the physic attributes of the bulky sand (BS), thin sand (TS) , silt (SIL), argil (ARG), soil density (SD), particle density (PD) and whole porosity (WP) and chemical attributes of argil, texture, pH (H2O), phosphorus (P), potassium(P), organic material (OM), aluminum (AL), calcium (Ca), magnesium (Mg), exchangeable aluminum (H + Al), cations real exchange capacity (CTCe), cations exchange capacity to pH7 (CTCpH7) and saturation of basis (V%). The magnitudes of spatial variabilities were obtained by the variation of coefficient (CV%), with confidence level of 95%, through Microsoft Office Excel 2003 program, while the structure was identified by semivariogrames, in applications geostatistics establishing the necessary parameters to the krigagem . All models of semivariogrames presented zones of anisotropic influence, having its spatial variability the greatest in the perpendicular sense to these areas the declivity. The chemical attribute that presented greatest variability was phosphorus (P), with CV%=127,73, followed by the aluminum attribute (Al), com CV%= 99,23 and the attribute of least variability was the pH (H2O), com CV%=0,0013. To the physical attributes in the distribution of particles size (%), the attribute of greatest variability was bulky sand, with CV%=36,39, while the statistics made to the density attributes of soil, what present the greatest variability was the attribute of whole porosity, with CV%=95,49 and they least variability was verified was the argil, with CV%=12,32. IN the analysis of the structure of the spatial variability through geostatistics, the chemical attribute Ca presented IDE (%)=64,42 and the physical attribute ARG with IDE (%)=62,50, getting the greatest rates. To the accomplishment of agreement in the program VARIOWIN® 2.21- Software for Spatial Data Analysis , the method used was the visual, named the feeling , where the Spherical model was what better was adapted to the studied attributes, indicated in 55% of the variogrames. The Gaussian s model to the attribute of texture got the most overtaking with a (m)=421. A crusade validation with the usage of the program GSLIB90 Geostatistical Software Library pointed out accuracy in the agreement of the variographic models, having the attributes Ca, Al and ARG with R² (%) of 0,841; 0,705 and 0,760, respectively. The usual krigagem of the studied attributes permitted the detailed of the distribution of these through the maps of isolineas. / Este trabalho utilizou-se de métodos de estatística descritiva clássica, métodos de geoestatística e de geoprocessamento, na identificação do tamanho e da estrutura da variabilidade espacial de atributos físico-químicos do solo em área de florestamento de erva-mate. A área de estudo, localizada na Fazenda Tupi, no município de Nova Prata, RS, foi levantada amostras nos meses de janeiro a março de 2005, compreendendo a classe de solo Latossolo Húmico Distrófico Álico. Realizou-se amostragem sistemática com grid de espaçamento regular entre os pontos de 100 metros, totalizando 36 pontos amostrais e 630 pares de dados, em uma área de 36 hectares. Foram coletadas amostras de solo in situ para análise em laboratório dos atributos físicos areia grossa (AG), areia fina (AF), silte (SIL), argila (ARG), densidade de solo (DS), densidade de partícula (DP) e porosidade total (PoT) e atributos químicos argila, textura, pH (H2O), fósforo (P), potássio (K), matéria orgânica (M.O.), alumínio (Al), cálcio (Ca), magnésio (Mg), alumínio trocável (H+Al), capacidade de troca de cátions efetiva (CTCe), capacidade de troca de cátions à pH7 (CTCpH7) e saturação de bases (V%). As magnitudes das variabilidades espaciais foram obtidas pelo coeficiente de variação (CV%), com nível de confiança de 95,0%, através do programa Microsoft Office Excel 2003, enquanto que a estrutura foi identificada por meio de semivariogramas, em aplicativos geoestatísticos, definindose os parâmetros necessários para a krigagem. Todos os modelos de semivariogramas apresentaram zonas de influência anisotrópicas, tendo sua variabilidade espacial maior no sentido perpendicular a declividade destas áreas. O atributo químico que apresentou maior variabilidade foi o fósforo (P), com CV%=127,73, seguido do atributo alumínio (Al), com CV%=99,23 e o atributo de menor variabilidade, foi o pH(H2O), com CV%=13,79. Para os atributos físicos, na distribuição do tamanho de partículas (%), o atributo de maior variabilidade foi areia grossa, com CV%=36,39, enquanto a estatística realizada para os atributos densidade do solo, o que apresentou maior variabilidade foi o atributo porosidade total, com CV%=95,49 e a menor variabilidade foi constatada a argila, com CV%=12,32. Na análise de estrutura da variabilidade espacial através da geoestatística, o atributo químico Ca apresentou IDE(%)=64,42 e o atributo físico ARG com IDE(%)=62,50, obtendo os maiores índices. Para realização do ajuste no programa VARIOWIN® 2.21 Software for Spatial Data Analysis , o método utilizado foi o visual, denominado a sentimento , onde o modelo Esférico foi o que melhor se ajustou aos atributos estudados, indicado em 55% dos semivariogramas. O modelo Gaussiano para o atributo textura obteve o maior alcance com a(m)=421. A validação cruzada, com o uso do programa GSLIB 90 Geostatistical Software Library mostrou acuracidade no ajuste dos modelos variográficos, tendo os atributos Ca, Al e ARG com R²(%) de 0,841; 0,705 e 0,760, respectivamente. A krigagem ordinária dos atributos estudados permitiu o detalhamento da distribuição destas a partir dos mapas de isolinhas.

Ce travail documente la variabilité spatiale et temporelle du rayonnement solaire à la surface sur le sud-ouest de l'océan Indien (SOOI) et l'île de La Réunion à l'aide de deux modèles régionaux de climat (MRC) : les modèles RegCM et WRF. La première partie de ce travail est dédiée à l'analyse de la variabilité temporelle du rayonnement solaire à l'aide du modèle RegCM sur le SOOI avec une résolution spatiale modérée (50km). S'agissant du premier travail sur la modélisation régionale du climat pour l'étude du rayonnement solaire dans le SOOI, une première série de tests pour illustrer les performances du modèle et sa sensibilité au choix des paramétrisations physiques (transfert radiatif, convection), à la taille du modèle, et à la résolution spatiale, est effectuée. Le schéma radiatif par défaut, le schéma CCM, et le schéma convectif mixte : Grell sur les terres et Emanuel sur les océans, donnent les résultats les plus satisfaisants pour la région, comparés aux autres options disponibles. La variabilité climatique interannuelle, intrasaisonnière et jour-à-jour est ensuite examinée sur la base des indices climatiques. Dans un premier temps, plusieurs paramètres (vent horizontal, température, humidité relative) issus des réanalyses ERA-Interim et utilisés comme paramètres d'entrée pour le modèle RegCM, sont analysés en lien avec ceux correspondant fournis en sortie du modèle, pour vérifier l'aptitude du modèle à maintenir les signaux ENSO (El-Nino Southern Oscillation), IOD (Indian Ocean Dipole), MJO (Madden-Julian Oscillation) et les Talwegs Tropicaux-Tempérés (TTT). Dans un second temps, le rayonnement solaire à la surface simulé par le modèle RegCM est mis en lien avec ces différents modes de variabilité. La seconde partie du travail est consacrée à l'analyse de la variabilité spatiale du rayonnement solaire à la surface à La Réunion à l'aide du modèle WRF à très haute résolution spatiale (750m) pour différentes échelles de temps : interannuelle, intrasaisonnière, jour-à-jour. Une classification est appliquée sur les sorties de rayonnement produites par WRF, et le lien avec la circulation atmosphérique de grande échelle est analysé dans chacune des classes. Les résultats de la modélisation sont validés à l'aide des données d'observations du réseau Météo France et des produits satellite CM SAF. Les résultats indiquent que les MRC ont la capacité de représenter la variabilité temporelle et spatiale du rayonnement solaire à La Réunion. / This work documents the temporal and spatial variability of surface solar radiation (SSR) over the southwest Indian Ocean (SWIO) and Reunion Island using two complementary Regional Climate Models (RCMs): RegCM4 and WRF. The first part of the work is dedicated to the analysis of the temporal variability of SSR based on RegCM4 over the SWIO at a moderate spatial resolution (50km). Because RegCM4 is the first RCM that focuses on the solar radiation research over the SWIO region, a first series of test experiments with this model to illustrate the model performance and its sensitivity to the choice of the physical parameterizations (radiation, convection), the domain size, and the spatial resolution, are performed. The default CCM radiative and the mixed convective scheme: Grell scheme over land and Emanuel scheme over ocean, give better performance over the SWIO compared to the other available options. The interannual, intraseasonal and synoptic climate variability is then examined through the climate indices and several ERA-Interim parameters (U, V, T and RH) are firstly analyzed along with the corresponding RegCM4 output data to check whether the RegCM4 model forced by ERA-Interim reanalyses is able to maintain the El-Nino Southern Oscillation (ENSO), the Indian Ocean Dipole (IOD), the Madden-Julian Oscillation (MJO) and the Tropical Temperate Trough (TTT) signals. Secondly, simulated SSR in association with the different modes of variability is examined. In the second part, SSR spatial variability over Reunion Island is analyzed based on WRF simulations at very fine resolution (750m) for seasonal, intraseasonal, and daily time scales. Clustering classification is applied to WRF simulated SSR over Reunion and the effect from the atmospheric circulation is checked together. Météo France observations and CM SAF are used to validate the results of the model. The results indicate that regional climate models have the ability to present the temporal and spatial variability of SSR over Reunion.

The transport of elements by streams from headwater regions to the sea is influenced by landscape characteristics. This thesis focuses on the influence of landscape characteristics (e.g. proportion of wetland/forest coverage) on temporal and spatial variations of Fe, Mn, S and trace elements (As, Co, Pb) in streams located in northern Sweden, a boreal region characterized by coniferous forests and peat wetlands. Water samples from a network of 15 streams revealed a different hydrogeochemistry in forested catchments compared to wetland catchments. The temporal variation was dominated by spring flood, when concentrations of Fe, Mn and trace elements increased in forested headwaters. However, in streams of wetland catchments concentrations decreased, but Pb concentrations were higher in comparison to other streams. Both Fe and Pb showed positive correlations with wetland area, while Co correlated with forest coverage. The anthropogenic contribution of As and Pb appear to be larger than the supply from natural sources. During spring flood SO42- decreased in most streams, although concentrations increased in streams of wetland catchments. Concentrations of SO42- were higher in streams of forested catchments than in wetland dominated streams, the former being net exporters of S and the latter net accumulators. Isotope values of stream water SO42- (δ34SSO4) were close to that of precipitation during spring flood, indicating that the major source of S is from deposition. The results show that, although emissions of anthropogenic S have been reduced, there is still a strong influence of past and current S deposition on runoff in this region. In conclusion, wetlands are key areas for the hydrogeochemistry in this boreal landscape. The findings emphasize the importance of understanding stream water chemistry and element cycling from a landscape perspective. This may be important for predicting how boreal regions respond to environmental disturbances such as climate change.

The Mississippi River Valley Alluvial Aquifer (MRVAA) is one of the most productive agricultural regions in the United States. The main objectives of this research are to identify long term trends and change points in hydrological variables (streamflow and rainfall), to assess the relationship between hydrological variables, and to evaluate the influence of global climate indices on hydrological variables. Non-parametric tests, MMK and Pettitt’s tests were used to analyze trend and change points. PCC and Streamflow elasticity analysis were used to analyze the relationship between streamflow and rainfall and the sensitivity of streamflow to rainfall changes. PCC and MLR analysis were used to evaluate the relationship between climate indices and hydrological variables and the combined effect of climate indices with hydrological variables. The results of the trend analysis indicated spatial variability within the aquifer, increase in streamflow and rainfall in the Northern region of the aquifer, while a decrease was observed in the southern region of the aquifer. Change point analysis of annual maximum, annual mean streamflow and annual precipitation revealed that statistically decreasing shifts occurred in 2001, 1998 and 1995, respectively. Results of PCC analysis indicated that streamflow and rainfall has a strong positive relationship between them with PCC values more than 0.6 in most of the locations within the basin. Results of the streamflow elasticity for the locations ranged from 0.987 to 2.33 for the various locations in the basin. Results of the PCC analysis for monthly maximum and mean streamflow showed significant maximum positive correlation coefficient for Nino 3.4. Monthly maximum rainfall showed a maximum significant positive correlation coefficient for PNA and Nino3.4 and the monthly mean rainfall showed a maximum significant positive correlation coefficient of 0.18 for Nino3.4. Results of the MLR analysis showed a maximum significant positive correlation coefficient of 0.31 for monthly maximum and mean streamflow of 0.21 and 0.23 for monthly maximum and mean rainfall, respectively. Overall, results from this research will help in understanding the impacts of global climate indices on rainfall and subsequently on streamflow discharge, so as to mitigate and manage water resource availability in the MRVAA underlying the LMRB.

The primary objective of this research was to quantify the improvement in service life of a flexible pavement constructed using full-depth reclamation (FDR) in conjunction with cement stabilization when specified reductions in the spatial variability of specific construction-related parameters are achieved. This study analyzed pavement data obtained through field and laboratory testing of a reconstruction project in northern Utah. Data analyses included multivariate regression, Monte Carlo simulation, and mechanistic-empirical analyses of a model pavement structure. The results of the research show a steadily increasing trend in 28-day unconfined compressive strength of the cement-treated base (CTB) layer with increasing reductions in variability for cement content, moisture content, and reclaimed asphalt pavement (RAP) content across each of five different reliability levels. The most significant increases in CTB strength occurred with reductions in the standard deviations of moisture content and RAP content. Decreasing the variability of cement content did not provide significant additional strength to the CTB layer. Therefore, when involved on FDR projects, members of the pavement industry should focus energy on reducing the variability of both moisture content and RAP content, which both significantly impact pavement life, to achieve high-quality, long-lasting pavements.

The first objective of this research was to investigate temporal trends in the mechanical properties of base materials stabilized with asphalt emulsion and to assess the rate at which emulsion-treated base (ETB) design properties are achieved. The second objective of this research was to identify construction and environmental factors most correlated to specific mechanical properties of ETB layers and to determine which construction factors exhibit the greatest variability. Additional statistical analysis was performed to determine if significant differences existed between different test sections on a given project. In this research, three experimental sections were established along a pavement reconstruction project near Saratoga Springs, Utah. Field tests were performed to assess the structural properties of the ETB immediately following construction and at 2, 3, 7, and 14 days; 4 months; and 1 year. Measured values were plotted against time to determine trends in ETB strength development. Several statistical analyses were then performed on the collected data. Modulus values were consistently low in all three sections during the first two weeks of testing, increased dramatically by 4 months, and then decreased considerably by 1 year. During the first two weeks following construction, the average ETB structural coefficient was 0.04. Only two of the three sections reached the design structural coefficient of 0.25, which occurred after approximately 3 months; however, the average structural coefficient measured for all three sections after 1 year of curing, which included a winter, was only 47 percent of the design strength. The results of this research show that, while pavement capacity is sufficient at 4 months, it is severely reduced during the first two weeks and at 1 year. Trafficking under these reduced capacities is not recommended. Statistical analysis showed that gradation, binder change during emulsion treatment, and moisture content have the most significant impact on ETB structural properties. Gradation and binder change during emulsion treatment also exhibited significant variability; tighter specifications on material gradations and improved uniformity in emulsion distribution should therefore be considered. Because of the negative impacts of moisture on ETB strength development, construction should not be performed in conditions of excess moisture.

Thesis (M.Sc. (Soil Science)) -- University of Limpopo, 2022 / In South Africa, land use planning has received limited attention in areas perceived as suitable for agricultural production. In the lack of reliable soil type and fertility status information, crop yields remain lower than the land’s potential, with subsequent land degradation. Despite this, studies that focused on land capability and soil suitability to date have not considered the spatial variability of the soil nutrients and factors influencing their variability. However, this information is key for site-specific soil management. Therefore, it is vital to link land capability and soi suitability with the spatial variability of soil nutrients as it opens opportunities for more rational management of the soil resources since soil nutrients directly affect crop growth and consequently yield. To address this issue, a study was conducted on a 12 ha banana plantation portion of the Makuleke farm. The main objectives of this study were to (1) survey, classify and characterise soils in order to derive and map land capability classes of Makuleke farm, (2) quantify the physical and chemical properties of the soils in order to derive and map the soil suitability of Makuleke farm for banana production, (3) assess the spatial variability and structure of soil nutrients across the Makuleke farm and (4) Identify the factors of control of the spatial variability of the soil nutrients across the Makuleke farm. To begin with, a field soil survey was conducted using transect walks complemented by auger observations to sub-divide the 12 ha banana plantation portion of the farm into varied soil mapping units. Thereafter, soil classification was done to group soils based on their morphological properties and pedological processes. During soil classification, a total of 12 representative profile pits (1.5 m × 1.5 m long × 2 m deep/limiting layer) were excavated, studied, described, and sampled. At each profile pit, three replicates samples were collected at 0 – 30 cm depth intervals giving rise to 36 bulk soil samples. From the gathered soil profile information, four soil units were thus delineated and identified across the 12 ha banana plantation. For soil fertility assessment, a grid sampling strategy at 50 × 50 m was adopted to collect the samples across the 12 ha banana plantation. A total of 27 composite samples were collected at the nodes of the grid, and thereafter bagged, labelled, and transported to the laboratory. In the laboratory, all collected samples were air-dried and sieved using a 2 mm sieve in preparation for soil physical and chemical properties analysis. The land capability assessment of Makuleke farm was done using the concepts and principles of the FAO framework for Land Evaluation (FAO, 1976), but adapted to South African conditions by Smith (2006). Soil suitability assessment was done using the FAO framework for Land Evaluation (FAO, 1976) coupled with the guidelines for rainfed agriculture (FAO, 1983) and the criteria proposed by Sys et al. (1993) and Naidu et al. (2006). To assess the spatial variability and structure of the soil nutrients across the farm, classical and geostatistical techniques were employed respectively. A correlation matrix was employed to identify key factors influencing the spatial variability of soil nutrients across the farm. For interpolation, ordinary kriging was used to generate soil nutrient spatial distribution maps. In this study, four soil forms were identified and classified as Hutton, Westleigh, Glenrosa, and Valsrivier, which are broadly distinguished as Lixisols, Plinthosols, Leptosols, and Cambisols. Land capability results revealed that 17% of the 12 ha portion of the farm has very high arable potential (I), 60% of the farm has medium arable potential (III), 6% has low arable potential (IV) and 17 % is non-arable (VI), which might explain the varied banana yields in the farm. Soil suitability analysis revealed that 12% of the 12 ha farm is highly suitable (S1), 34% is moderately suitable (S2), 38% is marginally suitable (S3) and 16% is permanently not suitable (N2) for banana production. The low arable and marginally suitable portion of the farm was under Valsrivier soils which were limited by its shallow depth, shallow rooting depth, acidic soil pH, low organic carbon (OC), and the fact that it was located on a steeper slope gradient. The non-arable and not suitable portion of the farm for banana production was under Glenrosa and it was limited by its location on a steep slope gradient and was characterised by shallow effective rooting depth, low OC, low clay content, and acidic soil pH. Classical statistical techniques revealed that phosphorus (P), potassium (K), calcium (Ca), zinc (Zn), manganese (Mn), and copper (Cu) content varied highly across the banana plantation, while magnesium (Mg) and total nitrogen (TN) varied moderately. In addition, the geostatistical analysis revealed that spatial dependency was weak (Ca, Cu, and TN), moderate (Mg and Zn), and strong (P, K, and Mn) for the different soil nutrients across the 12 ha banana plantation. Soil nutrients with strong spatial dependency have a good spatial structure and are easily manageable (in terms of fertilisation, liming, and irrigation) across the farm compared to the ones with weak spatial dependency which have a poor structure. This study also found that land attributes, which are soil type and topographic position were the main factors driving the spatial variability of the soil nutrients across the farm. In terms of soil type, soils such as Valsrivier and Glenrosa with 2:1 clay-type smectite were the ones that had nutrient content compared to soils with 1:1 clay-type kaolinite (e.g., Westleigh and Hutton). Higher nutrient contents were also observed in the footslope position compared to the middleslope of the farmland. Correlation analysis revealed that Mn was the key polyvalent cation influencing the spatial variability of P, K, and Zn. Soil pH and effective cation exchanges capacity (ECEC) were the key soil factors driving the spatial variability of Ca, while ECEC was the key factor affecting the spatial variability of Mg. Moreover, the spatial variability of soil Mn and Cu was driven by soil Cu and clay content, respectively. The kriged maps showed that P, Mg, Zn, and Mn were high in the northeast part and low in the northwest part of the farm. Similarly, K and Ca were low in the northwest part, but they were high in the south to the southwest part of the study area. Total nitrogen was high in the west part and low in the east-northeast part, while Cu was evenly distributed across the plantation. This study highlights the importance of prior land use planning (i.e., land capability and soil suitability) and fertility assessment for agricultural production. The research results obtained provide the actual reference state of the capability of the land for arable farming and soil suitability for banana production at Makuleke farm. Moreover, the research results provide the spatial variability and structure of the soil nutrients which have a greater impact on the growth and yield of bananas. The results obtained in this study will be useful for site-specific management of soil nutrients and other soil management practices (e.g., irrigation, fertilisation, liming, etc.), developing appropriate land use plans, and quantifying anthropogenic impacts on the soil system and thus improving land productivity. / National Research Foundation (NRF)

No description available.

A l’échelle mondiale, les prairies fournissent une grande variété de services écosystémiques et sont le support économique de nombreux systèmes d’élevage. Dans un contexte global éminemment changeant, une meilleure compréhension de la structure et du fonctionnement des prairies est incontournable pour proposer à la fois des gestions plus durables des ressources et promouvoir la fourniture de services écosystémiques diversifiés par ces écosystèmes. Les prairies étant des écosystèmes dynamiques et hétérogènes, notre capacité à prédire leur fonctionnement et leurs trajectoires de réponse à un facteur environnemental (climat, gestion) reste un défi scientifique important. Ainsi, dans des prairies pâturées, l’activité de grands herbivores va être facteur d’hétérogénéité des nutriments du sol via l'excrétion. Cependant les effets de ces apports nutritifs en « patchs » et de l'hétérogénéité spatiale du sol sur la structure et les propriétés de la prairie restent peu connus. L’objectif de cette thèse est d’examiner les effets de l'hétérogénéité spatiale de l'azote (N) dans le sol sur l’écosystème prairial, en portant une attention particulière sur les réponses des communautés végétales. Notre démarche a combiné des approches expérimentales et de modélisation pour analyser les impacts d'un certain nombre d'attributs de « patch » (différentes formes d’N, taille et contraste du patch), et leurs interactions possibles avec le régime de pluviométrie ou encore la date des apports en N. Nous montrons que des apports hétérogènes en N augmentent la production des plantes et la variabilité de la biomasse intra-parcelle quel que soit la forme d’N, mais qu’ils ne modifient pas, à court terme, la production à l’échelle de la parcelle prise dans son entier. Néanmoins, des apports hétérogènes d’N-organique favorisent l’asynchronie spatiale et temporelle entre les compartiments plante - sol, avec des implications pour le fonctionnement de la prairie à plus long terme. Contrairement à la production, la structure de la communauté végétale répond significativement à l’hétérogénéité en N, avec une dominance accrue de certaines espèces et un changement dans le rang des espèces subalternes. Contre toute attente, dans cette étude, la quantité de pluie ne modifie pas les effets de l'hétérogénéité sur la production et la structure de la communauté végétale. Des simulations réalisées avec un modèle spatialisé montrent que les effets de l'hétérogénéité sur la production à l’échelle de la parcelle varient selon la taille et le niveau de contraste du patch. Pour un même apport total en N, la production répond positivement à la taille de patch, mais elle diminue dans des conditions de fort contraste en comparaison à des conditions de faible contraste. Nous n’avons pas relevé d’interactions entre la taille de patch, le niveau de contraste de patch ou la date des apports en N sur la production de prairie. D’une manière générale, nos résultats soulignent l'importance de l’hétérogénéité en N pour les processus plante-sol à différentes échelles spatiales et montrent que les effets de l'hétérogénéité varient en fonction des attributs des patchs. Les interactions biotiques (ici la compétition) semblent jouer un rôle relativement plus important que les facteurs abiotiques (ici changements chroniques de pluviométrie) pour les effets d'hétérogénéité. Nous concluons que les impacts de l'hétérogénéité en N sur les processus plante-sol peuvent avoir des conséquences sur les rétroactions plante-sol impliquées dans la régulation des cycles biogéochimiques, et sont à même de fournir des informations utiles pour le développement de pratiques de gestion efficientes dans l’utilisation de l’N. / Grasslands provide a variety of important ecological and economic services worldwide. Improved understanding of grassland structure and function is necessary for the development of sustainable management and maintaining the provision of multiple ecosystem services in a changing environment. However, predicting grassland structure and function is a challenge because grasslands are dynamic, heterogeneous systems. In grazed grasslands, large herbivore activities promote heterogeneity in soil nutrients via excretion, but the effects of patchy nutrient inputs and soil spatial heterogeneity on grassland structure and function remain unclear. This thesis addresses effects of spatial heterogeneity in soil nitrogen (N) for grassland ecosystem structure and function, with particular emphasis on community responses. A combination of experimental and modelling approaches are used to study impacts of a number of different patch attributes (N form, patch size, patch contrast), as well as possible interactions with rainfall regime and timing of N inputs. We find that patchy N inputs enhance within plot-plant production and biomass variability irrespective of N form, but do not modify whole-plot plant production in the short term. Nevertheless, patchy organic N promotes spatial and temporal asynchrony in plant-soil responses, with implications for longer-term grassland function. Unlike plant production, community structure responds significantly to patchy N inputs, with increased community dominance and a shift in the rank of subordinate species. Contrary to expectations, rainfall quantity does not modify heterogeneity effects on either plant production or community structure. Modelling work shows that heterogeneity effects on field-scale production vary depending on patch size and patch contrast. For a fixed total N input, field-scale grassland production responds positively to patch size, but decreases in high- versus low-patch contrast conditions. Patch size does not interact with patch contrast or timing of N inputs on grassland production. Overall, our results highlight the importance of N heterogeneity for plant and soil processes at different spatial scales, and demonstrate that heterogeneity effects vary depending on patch attributes. Biotic interactions (competition) appear to play a relatively greater role than abiotic factors (chronic rainfall changes) for heterogeneity effects. Impacts of N heterogeneity on plant and soil processes may have significant implications on plant-soil feedbacks involved with the regulation of biogeochemical cycling, and provide useful information for the development of efficient N management strategies.