Extreme Precipitation Modelling Using Geostatistics and Machine Learning Algorithms

The paper presents an approach for mapping of precipitation data. The main goal is to perform spatial predictions and simulations of precipitation fields using geostatistical methods (ordinary kriging, kriging with external drift) as well as machine learning algorithms (neural networks). More practically, the objective is to reproduce simultaneously both the spatial patterns and the extreme values. This objective is best reached by models integrating geostatistics and machine learning algorithms. To demonstrate how such models work, two case studies have been considered: first, a 2-day accumulation of heavy precipitation and second, a 6-day accumulation of extreme orographic precipitation. The first example is used to compare the performance of two optimization algorithms (conjugate gradients and Levenberg-Marquardt) of a neural network for the reproduction of extreme values. Hybrid models, which combine geostatistical and machine learning algorithms, are also treated in this context. The second dataset is used to analyze the contribution of radar Doppler imagery when used as external drift or as input in the models (kriging with external drift and neural networks). Model assessment is carried out by comparing independent validation errors as well as analyzing data patterns.

Predicting spatial patterns of plant biodiversity : from species to communities

Understanding the distribution and composition of species assemblages and being able to predict them in space and time are highly important tasks io investigate the fate of biodiversity in the current global changes context. Species distribution models are tools that have proven useful to predict the potential distribution of species by relating their occurrences to environmental variables. Species assemblages can then be predicted by combining the prediction of individual species models. In the first part of my thesis, I tested the importance of new environmental predictors to improve species distribution prediction. I showed that edaphic variables, above all soil pH and nitrogen content could be important in species distribution models. In a second chapter, I tested the influence of different resolution of predictors on the predictive ability of species distribution models. I showed that fine resolution predictors could ameliorate the models for some species by giving a better estimation of the micro-topographic condition that species tolerate, but that fine resolution predictors for climatic factors still need to be ameliorated. The second goal of my thesis was to test the ability of empirical models to predict species assemblages' characteristics such as species richness or functional attributes. I showed that species richness could be modelled efficiently and that the resulting prediction gave a more realistic estimate of the number of species than when obtaining it by stacking outputs of single species distribution models. Regarding the prediction of functional characteristics (plant height, leaf surface, seed mass) of plant assemblages, mean and extreme values of functional traits were better predictable than indices reflecting the diversity of traits in the community. This approach proved interesting to understand which environmental conditions influence particular aspects of the vegetation functioning. It could also be useful to predict climate change impacts on the vegetation. In the last part of my thesis, I studied the capacity of stacked species distribution models to predict the plant assemblages. I showed that this method tended to over-predict the number of species and that the composition of the community was not predicted exactly either. Finally, I combined the results of macro- ecological models obtained in the preceding chapters with stacked species distribution models and showed that this approach reduced significantly the number of species predicted and that the prediction of the composition is also ameliorated in some cases. These results showed that this method is promising. It needs now to be tested on further data sets. - Comprendre la manière dont les plantes se répartissent dans l'environnement et s'organisent en communauté est une question primordiale dans le contexte actuel de changements globaux. Cette connaissance peut nous aider à sauvegarder la diversité des espèces et les écosystèmes. Des méthodes statistiques nous permettent de prédire la distribution des espèces de plantes dans l'espace géographique et dans le temps. Ces modèles de distribution d'espèces, relient les occurrences d'une espèce avec des variables environnementales pour décrire sa distribution potentielle. Cette méthode a fait ses preuves pour ce qui est de la prédiction d'espèces individuelles. Plus récemment plusieurs tentatives de cumul de modèles d'espèces individuelles ont été réalisées afin de prédire la composition des communautés végétales. Le premier objectif de mon travail est d'améliorer les modèles de distribution en testant l'importance de nouvelles variables prédictives. Parmi différentes variables édaphiques, le pH et la teneur en azote du sol se sont avérés des facteurs non négligeables pour prédire la distribution des plantes. Je démontre aussi dans un second chapitre que les prédicteurs environnementaux à fine résolution permettent de refléter les conditions micro-topographiques subies par les plantes mais qu'ils doivent encore être améliorés avant de pouvoir être employés de manière efficace dans les modèles. Le deuxième objectif de ce travail consistait à étudier le développement de modèles prédictifs pour des attributs des communautés végétales tels que, par exemple, la richesse en espèces rencontrée à chaque point. Je démontre qu'il est possible de prédire par ce biais des valeurs de richesse spécifiques plus réalistes qu'en sommant les prédictions obtenues précédemment pour des espèces individuelles. J'ai également prédit dans l'espace et dans le temps des caractéristiques de la végétation telles que sa hauteur moyenne, minimale et maximale. Cette approche peut être utile pour comprendre quels facteurs environnementaux promeuvent différents types de végétation ainsi que pour évaluer les changements à attendre au niveau de la végétation dans le futur sous différents régimes de changements climatiques. Dans une troisième partie de ma thèse, j'ai exploré la possibilité de prédire les assemblages de plantes premièrement en cumulant les prédictions obtenues à partir de modèles individuels pour chaque espèce. Cette méthode a le défaut de prédire trop d'espèces par rapport à ce qui est observé en réalité. J'ai finalement employé le modèle de richesse en espèce développé précédemment pour contraindre les résultats du modèle d'assemblage de plantes. Cela a permis l'amélioration des modèles en réduisant la sur-prédiction et en améliorant la prédiction de la composition en espèces. Cette méthode semble prometteuse mais de nouveaux tests sont nécessaires pour bien évaluer ses capacités.

The population genetics of clonal and partially clonal diploids.

The consequences of variable rates of clonal reproduction on the population genetics of neutral markers are explored in diploid organisms within a subdivided population (island model). We use both analytical and stochastic simulation approaches. High rates of clonal reproduction will positively affect heterozygosity. As a consequence, nearly twice as many alleles per locus can be maintained and population differentiation estimated as F(ST) value is strongly decreased in purely clonal populations as compared to purely sexual ones. With increasing clonal reproduction, effective population size first slowly increases and then points toward extreme values when the reproductive system tends toward strict clonality. This reflects the fact that polymorphism is protected within individuals due to fixed heterozygosity. Contrarily, genotypic diversity smoothly decreases with increasing rates of clonal reproduction. Asexual populations thus maintain higher genetic diversity at each single locus but a lower number of different genotypes. Mixed clonal/sexual reproduction is nearly indistinguishable from strict sexual reproduction as long as the proportion of clonal reproduction is not strongly predominant for all quantities investigated, except for genotypic diversities (both at individual loci and over multiple loci).

Analysis and spatial modeling of the indoor radon Swiss data

The present research deals with an important public health threat, which is the pollution created by radon gas accumulation inside dwellings. The spatial modeling of indoor radon in Switzerland is particularly complex and challenging because of many influencing factors that should be taken into account. Indoor radon data analysis must be addressed from both a statistical and a spatial point of view. As a multivariate process, it was important at first to define the influence of each factor. In particular, it was important to define the influence of geology as being closely associated to indoor radon. This association was indeed observed for the Swiss data but not probed to be the sole determinant for the spatial modeling. The statistical analysis of data, both at univariate and multivariate level, was followed by an exploratory spatial analysis. Many tools proposed in the literature were tested and adapted, including fractality, declustering and moving windows methods. The use of Quan-tité Morisita Index (QMI) as a procedure to evaluate data clustering in function of the radon level was proposed. The existing methods of declustering were revised and applied in an attempt to approach the global histogram parameters. The exploratory phase comes along with the definition of multiple scales of interest for indoor radon mapping in Switzerland. The analysis was done with a top-to-down resolution approach, from regional to local lev¬els in order to find the appropriate scales for modeling. In this sense, data partition was optimized in order to cope with stationary conditions of geostatistical models. Common methods of spatial modeling such as Κ Nearest Neighbors (KNN), variography and General Regression Neural Networks (GRNN) were proposed as exploratory tools. In the following section, different spatial interpolation methods were applied for a par-ticular dataset. A bottom to top method complexity approach was adopted and the results were analyzed together in order to find common definitions of continuity and neighborhood parameters. Additionally, a data filter based on cross-validation was tested with the purpose of reducing noise at local scale (the CVMF). At the end of the chapter, a series of test for data consistency and methods robustness were performed. This lead to conclude about the importance of data splitting and the limitation of generalization methods for reproducing statistical distributions. The last section was dedicated to modeling methods with probabilistic interpretations. Data transformation and simulations thus allowed the use of multigaussian models and helped take the indoor radon pollution data uncertainty into consideration. The catego-rization transform was presented as a solution for extreme values modeling through clas-sification. Simulation scenarios were proposed, including an alternative proposal for the reproduction of the global histogram based on the sampling domain. The sequential Gaussian simulation (SGS) was presented as the method giving the most complete information, while classification performed in a more robust way. An error measure was defined in relation to the decision function for data classification hardening. Within the classification methods, probabilistic neural networks (PNN) show to be better adapted for modeling of high threshold categorization and for automation. Support vector machines (SVM) on the contrary performed well under balanced category conditions. In general, it was concluded that a particular prediction or estimation method is not better under all conditions of scale and neighborhood definitions. Simulations should be the basis, while other methods can provide complementary information to accomplish an efficient indoor radon decision making.

Predicting current and future spatial community patterns of plant functional traits

Community-level patterns of functional traits relate to community assembly and ecosystem functioning. By modelling the changes of different indices describing such patterns - trait means, extremes and diversity in communities - as a function of abiotic gradients, we could understand their drivers and build projections of the impact of global change on the functional components of biodiversity. We used five plant functional traits (vegetative height, specific leaf area, leaf dry matter content, leaf nitrogen content and seed mass) and non-woody vegetation plots to model several indices depicting community-level patterns of functional traits from a set of abiotic environmental variables (topographic, climatic and edaphic) over contrasting environmental conditions in a mountainous landscape. We performed a variation partitioning analysis to assess the relative importance of these variables for predicting patterns of functional traits in communities, and projected the best models under several climate change scenarios to examine future potential changes in vegetation functional properties. Not all indices of trait patterns within communities could be modelled with the same level of accuracy: the models for mean and extreme values of functional traits provided substantially better predictive accuracy than the models calibrated for diversity indices. Topographic and climatic factors were more important predictors of functional trait patterns within communities than edaphic predictors. Overall, model projections forecast an increase in mean vegetation height and in mean specific leaf area following climate warming. This trend was important at mid elevation particularly between 1000 and 2000 m asl. With this study we showed that topographic, climatic and edaphic variables can successfully model descriptors of community-level patterns of plant functional traits such as mean and extreme trait values. However, which factors determine the diversity of functional traits in plant communities remains unclear and requires more investigations.

Robust parametric indirect estimates of the expected cost of a hospital stay with covariates and censored data.

We consider the problem of estimating the mean hospital cost of stays of a class of patients (e.g., a diagnosis-related group) as a function of patient characteristics. The statistical analysis is complicated by the asymmetry of the cost distribution, the possibility of censoring on the cost variable, and the occurrence of outliers. These problems have often been treated separately in the literature, and a method offering a joint solution to all of them is still missing. Indirect procedures have been proposed, combining an estimate of the duration distribution with an estimate of the conditional cost for a given duration. We propose a parametric version of this approach, allowing for asymmetry and censoring in the cost distribution and providing a mean cost estimator that is robust in the presence of extreme values. In addition, the new method takes covariate information into account.

Stable isotope geochemistry and formation mechanisms of quartz veins; Extreme paleoaltitudes of the Central Alps in the Neogene

Quartz veins ranging in size from less than 50 cm length and 5 cm width to greater than 10 m in length and 5 m in width are found throughout the Central Swiss Alps. In some cases, the veins are completely filled with milky quartz, while in others, sometimes spectacular void-filling quartz crystals are found. The style of vein filling and size is controlled by host rock composition and deformation history. Temperatures of vein formation, estimated using stable isotope thermometry and mineral equilibria, cover a range of 450 degrees C down to 150 degrees C. Vein formation started at 18 to 20 Ma and continued for over 10 My. The oxygen isotope values of quartz veins range from 10 to 20 permil, and in almost all cases are equal to those of the hosting lithology. The strongly rock-buffered veins imply a low fluid/rock ratio and minimal fluid flow. In order to explain massive, nearly morromineralic quartz formation without exceptionally large fluid fluxes, a mechanism of differential pressure and silica diffusion, combined with pressure solution, is proposed for early vein formation. Fluid inclusions and hydrous minerals in late-formed veins have extremely low delta D values, consistent with meteoric water infiltration. The change from rock-buffered, static fluid to infiltration from above can be explained in terms of changes in the large-scale deformation style occurring between 20 and 15 Ma. The rapid cooling of the Central Alps identified in previous studies may be explained in part, by infiltration of cold meteoric waters along fracture systems down to depths of 10 km or more. An average water flux of 0.15 cm 3 cm(-2)yr(-1) entering the rock and reemerging heated by 40 degrees C is sufficient to cool rock at 10 km depth by 100 degrees C in 5 million years. The very negative delta D values of < -130 permil for the late stage fluids are well below the annual average values measured in meteoric water in the region today. The low fossil delta D values indicate that the Central Alps were at a higher elevation in the Neogene. Such a conclusion is supported by an earlier work, where a paleoaltitude of 5000 meters was proposed on the basis of large erratic boulders found at low elevations far from their origin.

Efficiency of Gumbel analyses for determining extreme daily precipitation in Switzerland

Gumbel analyses were carried out on rainfall time-series at 151 locations in Switzerland for 4 different periods of 30 years in order to estimate daily extreme precipitation for a return period of 100 years. Those estimations were compared with maximal daily values measured during the last 100 years (1911-2010) to test the efficiency of these analyses. This comparison shows that these analyses provide good results for 50 to 60% locations in this country from rainfall time-series 1961-1990 and 1980-2010. On the other hand, daily precipitation with a return period of 100 years is underestimated at most locations from time-series 1931-1960 and especially 1911-1940. Such underestimation results from the increase of maximal daily precipitation recorded from 1911 to 2010 at 90% locations in Switzerland.

Preservation of an extreme transient geotherm in the Raft River detachment shear zone

Extensional detachment systems separate hot footwalls from cool hanging walls, but the degree to which this thermal gradient is the product of ductile or brittle deformation or a preserved original transient geotherm is unclear. Oxygen isotope thermometry using recrystallized quartz-muscovite pairs indicates a smooth thermal gradient (140 degrees C/100 m) across the gently dipping, quartzite-dominated detachment zone that bounds the Raft River core complex in northwest Utah (United States). Hydrogen isotope values of muscovite (delta D-Ms similar to-100 parts per thousand) and fluid inclusions in quartz (delta D-Fluid similar to-85 parts per thousand) indicate the presence of meteoric fluids during detachment dynamics. Recrystallized grain-shape fabrics and quartz c-axis fabric patterns reveal a large component of coaxial strain (pure shear), consistent with thinning of the detachment section. Therefore, the high thermal gradient preserved in the Raft River detachment reflects the transient geotherm that developed owing to shearing, thinning, and the potentially prominent role of convective flow of surface fluids.

Two-stage, extreme albitization of A-type granites from Rajasthan, NW India

Albitization is a common process during which hydrothermal fluids convert plagioclase and/or K-feldspar into nearly pure albite; however, its specific mechanism in granitoids is not well understood. The c. 1700 Ma A-type metaluminous ferroan granites in the Khetri complex of Rajasthan, NW India, have been albitized to a large extent by two metasomatic fronts, an initial transformation of oligoclase to nearly pure albite and a subsequent replacement of microcline by albite, with sharp contacts between the microcline-bearing and microcline-free zones. Albitization has bleached the original pinkish grey granite and turned it white. The mineralogical changes include transformation of oligoclase (similar to An(12)) and microcline (similar to Or(95)) to almost pure albite (similar to An(0 center dot 5-2)), amphibole from potassian ferropargasite (X-Fe 0 center dot 84-0 center dot 86) to potassic hastingsite (X-Fe 0 center dot 88-0 center dot 97) and actinolite (X-Fe 0 center dot 32-0 center dot 67), and biotite from annite (X-Fe 0 center dot 71-0 center dot 74) to annite (X-Fe 0 center dot 90-0 center dot 91). Whole-rock isocon diagrams show that, during albitization, the granites experienced major hydration, slight gain in Si and major gain in Na, whereas K, Mg, Fe and Ca were lost along with Rb, Ba, Sr, Zn, light rare earth elements and U. Whole-rock Sm-Nd isotope data plot on an apparent isochron of 1419 +/- 98 Ma and reveal significant disturbance and at least partial resetting of the intrusion age. Severe scatter in the whole-rock Rb-Sr isochron plot reflects the extreme Rb loss in the completely albitized samples, effectively freezing Sr-87/Sr-86 ratios in the albite granites at very high values (0 center dot 725-0 center dot 735). This indicates either infiltration of highly radiogenic Sr from the country rock or, more likely, radiogenic ingrowth during a considerable time lag (estimated to be at least 300 Myr) between original intrusion and albitization. The albitization took place at similar to 350-400 degrees C. It was caused by the infiltration of an ascending hydrothermal fluid that had acquired high Na/K and Na/Ca ratios during migration through metamorphic rocks at even lower temperatures in the periphery of the plutons. Oxygen isotope ratios increase from delta O-18 = 7 parts per thousand in the original granite to values of 9-10 parts per thousand in completely albitized samples, suggesting that the fluid had equilibrated with surrounding metamorphosed crust. A metasomatic model, using chromatographic theory of fluid infiltration, explains the process for generating the observed zonation in terms of a leading metasomatic front where oligoclase of the original granite is converted to albite, and a second, trailing front where microcline is also converted to albite. The temperature gradients driving the fluid infiltration may have been produced by the high heat production of the granites themselves. The confinement of the albitized granites along the NE-SW-trending Khetri lineament and the pervasive nature of the albitization suggest that the albitizing fluids possibly originated during reactivation of the lineament. More generally, steady-state temperature gradients induced by the high internal heat production of A-type granites may provide the driving force for similar metasomatic and ore-forming processes in other highly enriched granitoid bodies.

Heritability, determinants and reference values of renal length: a family-based population study.

OBJECTIVES: In this population-based study, reference values were generated for renal length, and the heritability and factors associated with kidney length were assessed. METHODS: Anthropometric parameters and renal ultrasound measurements were assessed in randomly selected nuclear families of European ancestry (Switzerland). The adjusted narrow sense heritability of kidney size parameters was estimated by maximum likelihood assuming multivariate normality after power transformation. Gender-specific reference centiles were generated for renal length according to body height in the subset of non-diabetic non-obese participants with normal renal function. RESULTS: We included 374 men and 419 women (mean ± SD, age 47 ± 18 and 48 ± 17 years, BMI 26.2 ± 4 and 24.5 ± 5 kg/m(2), respectively) from 205 families. Renal length was 11.4 ± 0.8 cm in men and 10.7 ± 0.8 cm in women; there was no difference between right and left renal length. Body height, weight and estimated glomerular filtration rate (eGFR) were positively associated with renal length, kidney function negatively, age quadratically, whereas gender and hypertension were not. The adjusted heritability estimates of renal length and volume were 47.3 ± 8.5 % and 45.5 ± 8.8 %, respectively (P < 0.001). CONCLUSION: The significant heritability of renal length and volume highlights the familial aggregation of this trait, independently of age and body size. Population-based references for renal length provide a useful guide for clinicians. KEY POINTS: • Renal length and volume are heritable traits, independent of age and size. • Based on a European population, gender-specific reference values/percentiles are provided for renal length. • Renal length correlates positively with body length and weight. • There was no difference between right and left renal lengths in this study. • This negates general teaching that the left kidney is larger and longer.

Blood pressure change and outcome in acute ischemic stroke: the impact of baseline values, previous hypertensive disease and previous antihypertensive treatment.

BACKGROUND: Management of blood pressure (BP) in acute ischemic stroke is controversial. The present study aims to explore the association between baseline BP levels and BP change and outcome in the overall stroke population and in specific subgroups with regard to the presence of arterial hypertensive disease and prior antihypertensive treatment. METHODS: All patients registered in the Acute STroke Registry and Analysis of Lausanne (ASTRAL) between 2003 and 2009 were analyzed. Unfavorable outcome was defined as modified Rankin score more than 2. A local polynomial surface algorithm was used to assess the effect of BP values on outcome in the overall population and in predefined subgroups. RESULTS: Up to a certain point, as initial BP was increasing, optimal outcome was seen with a progressively more substantial BP decrease over the next 24-48 h. Patients without hypertensive disease and an initially low BP seemed to benefit from an increase of BP. In patients with hypertensive disease, initial BP and its subsequent changes seemed to have less influence on clinical outcome. Patients who were previously treated with antihypertensives did not tolerate initially low BPs well. CONCLUSION: Optimal outcome in acute ischemic stroke may be determined not only by initial BP levels but also by the direction and magnitude of associated BP change over the first 24-48 h.