Search

Over recent decades, regions of West and Central Africa have experienced different and significant changes in climatic patterns, which have significantly impacted hydrological regimes. Such impacts, however, are not fully understood at the regional scale, largely because of scarce hydroclimatic data. Therefore, the aim of this study is to (a) assemble a new, robust, reconstructed streamflow dataset of 152 gauging stations; (b) quantify changes in streamflow over 1950–2005 period, using these newly reconstructed datasets; (c) significantly reveal trends and variability in streamflow over West and Central Africa based on new reconstructions; and (d) assess the robustness of this dataset by comparing the results with those identified in key climatic drivers (e.g. precipitation and temperature) over the region. Gap filling methods applied to monthly time series (1950–2005) yielded robust results (median Kling-Gupta Efficiency >0.75). The study underlines a good agreement between precipitation and streamflow trends and reveals contrasts between western Africa (negative trends) and Central Africa (positive trends) in the 1950s and 1960s. Homogenous dry conditions of the 1970s and 1980s, characterized by reduced significant negative trends resulting from quasi-decadal modulations of the trend, are replaced by wetter conditions in the recent period (1993–2005). The effect of this rainfall recovery (which extends to West and Central Africa) on increased river flows are further amplified by land use change in some Sahelian basins. This is partially offset, however, by higher potential evapotranspiration rates over parts of Niger and Nigeria. Crucially, the new reconstructed streamflow datasets presented here will be available for both the scientific community and water resource managers.

International audience; Datacleaninganddatapreparationhavebeenlong-standingchallenges in data science to avoid incorrect results, biases, and misleading conclusions ob- tained from “dirty” data. For a given dataset and data analytics task, a plethora of data preprocessing techniques and alternative data cleaning strategies are avail- able, but they may lead to dramatically different outputs with unequal result quality performances. For adequate data preparation, the users generally do not know how to start with or which methods to use. Most current work can be classified into two categories: 1) they propose new data cleaning algorithms specific to certain types of data anomalies usually considered in isolation and without a “pipeline vision” of the entire data preprocessing strategy; 2) they develop automated machine learning approaches (AutoML) that can optimize the hyper- parameters of a considered ML model with a list of by-default preprocessing methods. We argue that more efforts should be devoted to proposing a principled and adaptive data preparation approach to help and learn from the user for selecting the optimal sequence of data preparation tasks to obtain the best quality performance of the final result. In this paper, we extend Learn2Clean, a method based on Q-Learning, a model-free reinforcement learning technique that selects, for a given data set, a given ML model, and a pre-selected quality performance metric, the optimal sequence of tasks for preprocessing the data such that the quality metric is maximized. We will discuss some new results of Learn2Clean for semi-automating data preparation with “the human in the loop” using active reward learning and Q-learning.

Trabajo final presentado por Morgan Peix para un Máster de la École Nationale Supérieure de Techniques Avancées Bretagne (ENSTA-Bretagne), realizado bajo la dirección de la Dra. Eulàlia Gràcia i Mont del Institut de Ciències del Mar (ICM-CSIC) y del Dr. Pierre Bosser de la École Nationale Supérieure de Techniques Avancées Bretagne (ENSTA-Bretagne).-- 28 pages, [EN] During the INSIGHT-Leg 1 cruise in which I participated, high-resolution bathymetric data at depths up to 5000m were obtained through the use of an AUV, as well as geological and geophysical data obtained from R/V "Sarmiento de Gamboa". The rest of the internship was dedicated to the processing of the data acquired by the AUV. At first, the bathymetry was processed, the data being very noisy and affected by a problem of timelag as well as by a drift of the position in time. The backscatter was then extracted from the bathymetric data. However, the results obtained by the two tested softwares are strongly affected by artifacts rendering the interpretation of the backscatter impossible. Codes have been developed to process the backscatter. These codes use a reference area to compute a pattern of directivity which is then used to process the data of the area to be processed. These codes made it possible to show the importance of taking into account the bathymetric slope during the treatment of the backscatter. For the best results, all parameters other than directivity and reflectivity must be compensated using the sonar equation, [FR] Pendant la campagne INSIGHT-Leg 1 à laquelle j’ai participé, ont été obtenues des données bathymétriques de haute résolution à des profondeurs allant jusqu’à 5000m grâce à l’utilisation d’un AUV ainsi que des données géologiques et géophysiques à l’aide du R/V "Sarmiento de Gamboa". Le reste du stage a été dédié au traitement des données acquises par l’AUV. Dans un premier temps, la bathymétrie a été traitée, celle-ci étant très bruitée et affectée par un problème de décalage de temps ainsi que par une dérive du positionnement dans le temps. Le backscatter a ensuite été extrait à partir des données bathymétriques. Cependant les résultats obtenus par les deux logiciels testés sont fortement affectés par des artefacts rendant l’interprétation du backscatter impossible. Des codes ont donc été développés pour traiter le backscatter. Ces codes utilisent une zone de référence pour calculer un pattern de directivité qui est ensuite utilisé pour traiter les données de la zone à traiter. Ces codes ont permis de montrer l’importance de prendre en compte la pente bathymétrique lors du traitement du backscatter. Pour un meilleur résultat, tous les paramètres autres que la directivité et la réflectivité doivent être compensés en utilisant l’équation du sonar, I also acknowledge the financial support from Ministerio Economia y Competividad through national project INSIGHT “ImagiNg large SeismogenIc and tsunamiGenic structures of the Gulf of Cadiz with ultra-High resolution Technologies (INSIGHT-Leg1)” (CGL2011-30005-C02-02, PIs: Roger Urgeles and Eulàlia Gràcia)

Measuring ground surface displacement is essential for studying and understanding many phenomena related to both natural hazards (volcanism, seismicity and landslides) and anthropogenic hazards (subsidence due to mining operations or fluid extraction/injection). The significance of these phenomena in terms of risks – to both people and infrastructures – is considerable; the resulting surface disturbances must therefore be surveyed in order to prevent future incidents (in the case of anthropogenic hazards, surveillance may even be required by the regulations in force).

In Europe, more than one thousand of laboratories intensively collect data to measure various properties of the Earth. Scientists observe environmental conditions, ecosystems and biological species. The ability to understand complex phenomena (e. g., global warming) and predict trends from spatio-temporal data becomes a major issue in observational science. However, theoretical and technical advances in multi-scale data integration are necessary to achieve this. This paper will present some challenging research directions for integrating such massive multi-scale scientific data.

Basic principles of operating Terrestrial Laser Scanners (TLS) are presented in this paper, as well as errors sources, recording modes and data processing methods. Terminology and processing aspects of TLS are defined. A classification of different types of scanners currently available is proposed. In order to assess the accuracy of each point recorded by a laser scanner, we consider all features that may affect the quality of the data produced by a laser scanner. The knowledge of the different errors sources combined to the understanding of the relations between those errors is necessary to quantify data accuracy. Aspects of pre-processing point clouds, direct or indirect georeferencing, registration, segmentation, modelling and texturing are finally presented., Cet article a pour objectif de définir les principes fondamentaux de fonctionnement des Scanners Laser Terrestres (SLT), les sources d 'erreurs, les modes d 'acquisition et les méthodes de traitement des données. Après avoir défini le vocabulaire spécifique et le processus de balayage laser terrestre, une classification détaillée des différents types de scanners disponibles est proposée. Pour estimer la précision de chaque point d 'un levé effectué par un capteur laser, l'ensemble des caractéristiques qui peuvent influencer la qualité des données produites par les appareils utilisés est à considérer. Une bonne connaissance des différentes sources d'erreurs et une compréhension des relations qui les lient est nécessaire pour quantifier la précision des données. Le traitement des données laser terrestre abordent ensuite les étapes de prétraitement, de géoréférencement direct ou indirect, de consolidation ou de recalage, de segmentation, de modélisation et de texturage., Grussenmeyer Pierre, Landes Tania. Lasergrammétrie terrestre. In: Collection EDYTEM. Cahiers de géographie, numéro 12, 2011. Images et modèles 3D en milieux naturels. pp. 11-28.

- Urban climate and air pollution in big cities provide a new field of research for geographers interested in climate studies. In order to efficiently contribute to research works with strong social implications, they have to collect fine-scale meteorological data and adapt their traditional knowledge in meteorological data processing for new kinds of data. Mapping the urban climate or air pollution requires specific methods of spatial interpolation, using statistical relationships between atmospheric parameters and urban descriptors ; in the near future, models of physical and chemical processes in the urban atmosphere will probably become the most useful tool in that purpose., Résumé. - Le climat urbain et la qualité de l'air en ville constituent un «nouveau territoire » de recherche pour les climatologues géographes. Pour contribuer efficacement à des recherches qui répondent à une forte demande de la société, il leur est indispensable de pratiquer la collecte de données météorologiques à échelle fine et de s'adapter au traitement et à l'interprétation de données très différentes des mesures météorologiques traditionnelles. La cartographie du climat urbain ou de la pollution atmosphérique requiert des méthodes spécifiques d'interpolation spatiale, prenant en compte les relations statistiques entre paramètres atmosphériques et descripteurs synthétiques du tissu urbain ; à terme, la modélisation spatialisée des processus physiques et chimiques devrait se substituer aux méthodes empiriques de cartographie., Kergomard Claude, Bigot S., Charabi Yécine, Leriche Pierre. Climat urbain et qualité de l'air : approches géographiques et collaborations pluridisciplinaires (Urban climate and air pollution : geographical approaches and pluridisciplinary collaborations). In: Bulletin de l'Association de géographes français, 79e année, 2002-4 ( décembre). Villes chinoises. Géoclimatologie. pp. 462-473.

Summer frosts occur frequently on the Bolivian Altiplano (4000 m) and cause severe damage to potato and quinua crops. A regional statistical study of frost risks must combine meteorological data with satellite data. In this work, a new method is developed to use NOAA satellite temperatures together with local minimum air temperature data. This method enables point meteorological data (with important historical records) to be combined with satellite data (spatially extended but with poor temporal coverage). The method involves several stages including a classification of the Altiplano based on the meteorological stations, a calculation of minimum temperatures over the whole Altiplano for several years of data, and finally the computation of minimum temperatures and frost risk percentage maps. The precision obtained in the resulting synthesis images is 0.8°C for the average minimum temperatures and 9% for frost risk maps.

Four biotechnological advances have been accomplished of the last decade: i) sequencing of whole genomes giving chance to the discovery of thousands of genes, ii) functional genomics using high-throughput DNA microarrays to measure expression of each of these genes in multiple physiological and environmental conditions, iii) scaling of proteins using Proteome to map all the proteins produced by a genome, and iv) the dynamics of these genes and proteins in a network of interactions that gives life to any biological activity and phenotype. These major breakthroughs resulted in massive collection of data in the field of Life Sciences. Considerable efforts have been made to sort out, curate and integrate every relevant piece of information from multiple information sources in order to understand complex biological phenomena.

Quantitative Data Cleaning (QDC) is the use of statistical and other analytical techniques to detect, quantify, and correct data quality problems (or glitches). Current QDC approaches focus on addressing each category of data glitch individually. However, in real-world data, different types of data glitches co-occur in complex patterns. These patterns and interactions between glitches offer valuable clues for developing effective domain-specific quantitative cleaning strategies. In this paper, we address the shortcomings of the extant QDC methods by proposing a novel framework, the DEC (Detect-Explore-Clean) framework. It is a comprehensive approach for the definition, detection and cleaning of complex, multi-type data glitches. We exploit the distributions and interactions of different types of glitches to develop data-driven cleaning strategies that may offer significant advantages over blind strategies. The DEC framework is a statistically rigorous methodology for evaluating and scoring glitches and selecting the quantitative cleaning strategies that result in cleaned data sets that are statistically proximal to user specifications. We demonstrate the efficacy and scalability of the DEC framework on very large real-world and synthetic data sets.

A partir de variables elaboradas para dar cuenta del medio ambiente urbano, es decir de los modos de produccion y de consumo de los bienes comunes en Quito, se realizaron analisis estadisticos - Analisis en Componentes Principales (ACP) y nubes dinamicas principalmente - que permitieron destacar perfiles ambientales que expresan una verdadera estructura ambiental de la ciudad. Tal estructura se construye primeramente con base en una diferenciacion del modo de produccion y de consumo del suelo entre la zona central y el area periferica. Los mayores desafios de la cuestion ambiental en Quito remiten, por una parte, al manejo de la centralidad urbana y, por otra, al desarrollo urbano de los espacios residenciales recientes. Se trata en especial de crear nuevos modos de producir y de consumir los bienes comunes, particularmente el suelo y la movilidad. (Résumé d'auteur)

Forest of Fontainebleau, Ile de France Region, France : a visual analysis on a 22 of September color composite brings to a mas of Equal Appearance Units which is used to build up a network of sampling points for ground data collection. With the first half sample a landscape nomenclature is established, while the second one tests its validity. The results allow to, frame a statistic cartography of the foregoing forest-space., Un traitement analogique sur composition colorée de l'image du 22 septembre 1979 de la forêt de Fontainebleau (Région Ile-de-France) débouche sur l'établissement d'une carte d'Unités d'Egale Apparence qui sert de base au calage d'un échantillon de points de collecte de données de terrain établi à partir d'un sondage stratifié représentatif aléatoire. Un premier demi-échantillon est utilisé pour l'établissement d'une nomenclature paysagère qui prend en compte les caractères spécifiques de la "vision spatiale" du radiomètre, le second pour tester la validité de cette nomenclature, Les résultats font l'objet d'une cartographie statistique de l'espace forestier considéré., Amat Jean-Paul, Hotyat Micheline. Etude du couvert forestier par les images Landsat. Bilan et évaluation sur la forêt domaniale de Fontainebleau. In: Travaux de l'Institut Géographique de Reims, n°57-58, 1984. La télédétection. Aspects techniques, thématiques et exemples régionaux, sous la direction de Jean-Paul Amat. pp. 103-135.

Los métodos de medición de distancia con taquímetros electrónicos y las técnicas de posicionamiento relativo con satélites tienen aplicaciones cada vez mas importantes en geofísica para estudiar las deformaciones de la corteza y los movimientos tectónicos globales. Este trabajo presenta una rápida comparación de estos métodos desde el punto de vista de la precisión ,el alcance y las condiciones de operación, así como la discusión posterior de los procedimientos y el tratamiento de los datos que adoptamos para hacer mediciones de distancias largas(hasta 93 km) con taquímetros electrónicos. Con estas técnicas llevamos a cabo dos levantamientos, en 1982 y en 1986, de una red geodésica de 11 puntos y 150 km de extensión que construimos a través del Golfo de California, en su porción central. Los resultados de estas mediciones dan estimaciones de desplazamiento de Baja California con respecto a Sonora entre 6 y 8 cm/año, valores consistentes con la velocidad relativa de las placas Pacifica Y Norteamericana. Entre la Isla Ángel de la Guarda y Baja California, se observo un movimiento de cizalla lateral derecho con desplazamiento de 17 cm en 4 años. Más al sur las deformaciones son mas pequeñas que los errores estimados, lo que sugiere que esta parte de la frontera entre las placas esta bloqueada; esta observación concuerda con la presencia de un hueco sísmico en esta zona. En la parte oriental de la red, entre las islas centrales del golfo y la costa de Sonora, se evidenciaron también fuertes deformaciones de cizallamiento lateral derecho, con desplazamiento de hasta 6+3 cm/año. Estos resultados muestran que la acumulación de deformaciones se produce en una región amplia alrededor de la frontera entre placas.

CityShark, a station dedicated to ambient noise (microtremor) recording, was designed and built jointly by the French Institute for Research for Development (IRD) and LEAS, a French company dedicated to geophysical data acquisition. Because portable seismic stations are designed to be installed for long time periods (days to months) and to detect earthquakes, operators using them for microtremor recording face several obstacles that considerably slow and complicate the data acquisition process. Seismological stations are not easy to use in experiments for which time of recording is on the order of tens of minutes (site studies) or for repetitive recording of several minutes at interval times on the order of hours (building response studies). Running seismological stations requires setting a series of parameters ( e.g., STA, LTA, trigger level, pre-event memory), usually by communicating with the station with the help of an external computer. The data recorded are then downloaded using a portable computer or by manipulation of a removable high-capacity storage medium. In order to maximize the station autonomy, data are often compressed and power consumption is minimized. The use of a GPS receiver is needed for time accuracy. All the equipment must be packaged in a rugged box to be able to function in extreme environments ( e.g., rain, sun, snow, wind, and dust). Finally, these stations need to be installed and run by well trained people. These stations are made to record earthquakes and avoid recording ambient noise, i.e., the opposite goal of microtremor recording. Hence, the use of portable seismological stations for microtremor recording considerably slows data acquisition, and data processing is needlessly complicated. Running microtremor experiments for site studies in Pujili, Ecuador (Gueguen et al., 1998) and Nazca, Peru (Chatelain et al., 1997) and encountering the kind of problems mentioned above led us to the idea of building …