Literatura académica sobre el tema "High-resolution Regional Climate Model"

Les épisodes méditerranéens qui concernent chaque automne le nord-ouest de la Méditerranée sont des événements de précipitation extrême à fort impact. L'étude de leur évolution aux échelles climatiques revêt donc une grande importance et constitue encore un défi pour la communauté de modélisation du climat. Depuis quelques années, il est possible d’utiliser des modèles régionaux de climat à résolution kilométrique (Convection-Permitting Regional Climate Models ou CP-RCMs, 1-3 km) dont la convection profonde est simulée de manière explicite. Ces modèles permettent de se rapprocher des échelles spatio-temporelles en jeu et ouvrent de nouvelles perspectives en termes d'analyse. Cette thèse a pour objectif de déterminer la réponse des épisodes méditerranéens du nord-ouest de la Méditerranée au changement climatique d’origine humaine, en s'appuyant sur cette nouvelle génération de modèles de climat associée à une approche orientée objet. Le suivi des systèmes fortement précipitants est appliqué à la fois aux jeux de données d'observation, aux simulations réalisées avec le CP-RCM CNRM-AROME et aux simulations du premier ensemble de CP-RCMs disponible dans le cadre du programme international CORDEX FPS Convection sur un domaine commun couvrant le nord-ouest de la Méditerranée. La première partie de cette thèse est consacrée à l’évaluation des performances des CP-RCMs par comparaison à des données d’observation de référence à haute résolution. La valeur ajoutée des CP-RCMs par rapport aux modèles régionaux à résolution plus grossière (12-15 km) est démontrée pour les extrêmes de précipitation, en particulier au pas de temps horaire. L’approche objet permet également de montrer que, malgré quelques biais résiduels, les CP-RCMs sont capables de représenter correctement les principales propriétés des systèmes fortement précipitants, que ce soit en termes de nombre et de positionnement sur l'ensemble du domaine, ou en termes de durée, d'intensité, de surface, de volume, de vitesse et de sévérité sur le pourtour méditerranéen français où les observations permettent d’évaluer finement ces propriétés. La bonne performance de ces modèles permet de renforcer la confiance en leurs projections futures. La deuxième partie s’intéresse à l’évolution future des épisodes méditerranéens avec l’approche objet appliquée aux simulations milieu et fin de siècle de l’ensemble de CP-RCMs en mode scénario pour étudier les changements des propriétés des systèmes fortement précipitants dans un climat plus chaud. En fin de siècle et selon un scénario de fortes émissions, certains changements se retrouvent dans la plupart des simulations et peuvent être qualifiés de robustes. Ainsi, une augmentation de la fréquence des systèmes fortement précipitants de l’automne dans une grande partie du domaine, en particulier du centre de l'Italie au nord des Balkans, accompagne un doublement des zones touchées par ces événements. Sur la région méditerranéenne française, les modèles s’accordent sur une augmentation d'intensité, de surface et de volume des systèmes précipitants. Toutefois, même avec cette nouvelle génération de modèles, d'importantes incertitudes persistent, notamment pour les changements de fréquence dans le sud-est de la France, probablement dus à des différences dans les conditions synoptiques imposées par les modèles forceurs des CP-RCMs. De même, l’ensemble projette une large gamme de changements possibles dans les propriétés des systèmes, en particulier pour les plus intenses et y compris lorsqu’on normalise par le réchauffement régional correspondant. Si les CP-RCMs sont les outils de modélisation adaptés pour l'étude des extrêmes de précipitation, les efforts doivent être poursuivis pour produire des ensembles plus larges, mieux construits et probablement complétés par des méthodes d'apprentissage machine, afin d’apporter des informations climatiques utiles aux échelles pertinentes pour les politiques d’adaptation<br>The Mediterranean Heavy Precipitation Events (HPEs) that affect the northwestern Mediterranean every fall are high-impact weather events. The study of their evolution on climate scales is therefore of great importance and remains a challenge for the climate modelling community. For some years now, it has been possible to use kilometre-scale regional climate models (Convection- Permitting Regional Climate Models or CP-RCMs, 1-3 km) in which deep convection is explicitly simulated. These models make it possible to get closer to the spatio-temporal scales involved and open up new perspectives in terms of analysis. The aim of this thesis is to determine the response of northwestern Mediterranean HPEs to human-induced climate change, using this new generation of climate models combined with an object-oriented approach. The tracking of heavy precipitation systems is applied to observational datasets, to simulations carried out with the CNRM-AROME CP-RCM and to simulations of the first ensemble of CP-RCMs available as part of the international CORDEX FPS Convection programme over a common domain covering the north-western Mediterranean. The first part of this thesis is devoted to evaluating the performance of CP-RCMs in comparison with high-resolution reference observation data. The added value of CP-RCM compared with regional models with coarser resolution (12-15 km) is demonstrated for precipitation extremes, particularly at hourly time steps. The object-oriented approach also shows that, despite a few residual biases, CP-RCMs are capable of correctly representing the principal properties of heavy precipitation systems, both in terms of number and position over the entire domain, and in terms of duration, intensity, surface area, volume, speed and severity over the French Mediterranean, where observations enable these properties to be assessed in detail. The good performance of these models lends greater confidence to their future projections. The second part focuses on the future evolution of Mediterranean HPEs using the object-oriented approach applied to mid and end-of-century simulations of the CP-RCMs ensemble in scenario mode to study changes in the properties of heavy precipitation systems in a warmer climate. At the end of the century, and according to a scenario of high emissions, certain changes are found in most of the simulations and can be described as robust. For example, an increase in the frequency of fall heavy precipitation systems over a large part of the domain, particularly from central Italy to the northern Balkans, is accompanied by a doubling of the areas affected by these events. Over the French Mediterranean region, the models agree on an increase in the intensity, surface area and volume of precipitating systems. However, even with this new generation of models, significant uncertainties remain, particularly for changes in frequency over southeastern France, probably due to differences in the synoptic conditions imposed by the CP-RCMs driving models. Similarly, the ensemble projects a wide range of possible changes in the properties of systems, particularly for the most intense ones and even when standardised by the corresponding regional warming. While CP-RCMs are the appropriate modelling tools for studying precipitation extremes, efforts must be continued to produce larger, better constructed ensembles, probably supplemented by machine learning methods, in order to provide useful climate information at scales relevant to adaptation policies

The field of urban hydrology is in need of high temporal resolution data series in order to effectively model and analyse existing and future trends in extreme precipitation. When high resolution data sets are, for any number of reasons, not available for a given location, the technique of disaggregation using a random cascade model can be applied. Previous studies have demonstrated the relevance of random cascades in the context of rainfall data disaggregation with temporal resolutions usually down to 1 hour. In this study, an attempt at disaggregation to a resolution of 1 minute was made. Using newly disaggregated rainfall data for different regions in Sweden, the possibility of clustering rain events into separate regional hyetographs was investigated. The random cascade model was calibrated using existing municipal rainfall data with a temporal resolution of 1 minute, in order to disaggregate continuous 15 minutes data series provided by the Swedish Meteorological and Hydrological Institute (SMHI). The disaggregation process was then performed in multiple stochastic realisations, in order to correct the uncertainties inherent to the random cascade model. The disaggregation results were assessed by comparing them with calibration data: two main rainfall parameters, EV and ED, were analysed by determining their behaviours and distribution. The possibility of transfering calibration parameters from one station to another was also assessed in a similar manner, again by studying EV &amp; ED for different scenarios. Finally, hyetographs were clustered, compared and contrasted, in order to ascertain previously theorized differences between regions. This research showed the feasibility of applying a random cascade model to very high temporal resolutions in Sweden, while replicating rainfall characteristics from the calibration data quite well. The analysis of the spatial transferability of calibration parameters yielded inconclusive results, as rainfall characteristics were preserved in some cases but failed in others. Lastly, distinct regional differences in hyetographs were noted, but no clear conclusions could be drawn owing to the delimitations of this study.<br>Inom småskalig hydrologisk modellering finns det idag ett behov av dataserier med hög tidsupplösning för att effektivt kunna modellera och analysera både aktuella och kommande trender hos extrema regnhändelser. När högupplösta dataserier är otillgängliga vid en önskad mätplats kan disaggregering med hjälp av en slumpmässig kaskadmodell tillämpas. Tidigare forskning har visat att kaskadmodeller är användbara för disaggregering av regndata med en tidsupplösning av 1 timme. I denna studie disaggregerades dataserier med syftet att uppnå en tidsupplösningav av 1 minut. För att kunna analysera eventuella skillnader mellan regioner klustrades även hyetografer med de framtagna dataserierna. Den slumpmässiga kaskadmodellen kalibrerades med befintlig kommunal data med en tidsupplösning på 1 minut, för att sedan kunna disaggregera 15 minuters data från SMHIs databaser. Disaggregeringen genomfördes i ett antal olika stokastiska realisationer för att kunna ta hänsyn till, och korrigera, de inneboende osäkerheterna i den slumpmässiga kaskadmodellen. Disaggregeringsresultaten bedömdes genom en jämförelse med kalibreringsdata: två regnegenskaper, regnvaraktighet (ED) och regnvolym (EV), analyserades för att kunna bestämma derasfördelningar och beteenden. Kalibreringsparametrarnas överförbarhet analyserades också med hjälp av ED &amp; EV för olika scenarier. Slutligen klustrades hyetografer för att fastställa potentiella skillnader mellan regioner. Studien påvisade möjligheten att använda en slumpmässig kaskadmodell till höga tidsupplösningar i Sverige. Modellen lyckades återskapa regnegenskaper från kalibreringsdata vid disaggregeringen. Möjligheten att överföra kalibreringsparametrar från en station till en annan visade sig dock inte vara helt övertygande: regnegenskaper återskapades endast i vissa fall, men inte i samtliga. Slutligen konstaterades regionala skillnader i hyetografer, men tydliga slutsatser kunde inte dras på grund av underliggande begränsningar med studien.

Includes bibliographical references<br>The main aim of this thesis is to investigate the potential benefits of increasing resolution in regional climate models in the simulation of climate variability and change over East Africa. This study is based on two high resolution regional climate simulations with a horizontal resolution of 50km and 10km, respectively. These represent present day climate and a projection of future climate change over East Africa. The regional climate model (RCM) used here is HIRHAM5, which is driven by the global circulation model (ECHAM5). Downscaled ECHAM5 output is used to drive the 50km HIRHAM5 simulation for the period 1950-2100, and output from this simulation is used to drive the 10km simulation for three time slices: 1980-1999, representative for present-day climate and two time slices for near future (2046-2065) and far future (2080- 2099), respectively. HIRHAM5 is evaluated with respect to the observed mean climatologies of rainfall, surface temperature and surface winds over East Africa, and representations of the observed annual cycles and inter-annual variability of rainfall and surface temperature. This study utilizes reanalysis and observational datasets: a hindcast of HIRHAM5 forced with ERA Interim, as well as two observation datasets for temperature and rainfall. Since reanalyses aim to make "best use" of all available observations by making a physically consistent representation continuous in time and space, and since there is a paucity of observations over many parts of Africa, the ERAI reanalysis is also used as a best estimate for model evaluation. Additionally, for evaluation of the bimodal nature of East Africa's rainfall, especially over Tanzania, three stations run by the Tanzania Meteorological Agency were used. The model data used in th is evaluation ranges from 1980 to 2006 iv HIRHAM5 demonstrates reasonable skill in the reproduction of observed patterns of mean climatology of rainfall, surface temperature and winds over East Africa. Moreover, the patterns of annual cycles of rainfall and surface temperature in the bimodal nature of East Africa are well represented. Furthermore, the model showed reasonable skill in the representation of the inter- annual variability and ENSO signals as suggested by the observation. Despite these strengths, HIRHAM5 shows some shortcomings. One weakness of the model is the simulation of the magnitude of a given variable over a specific region. For example, HIRHAM5 driven by ERAI underestimates rainfall and overestimates surface temperature over the entire domain of East Africa. The higher resolution HIRHAM5 (10km resolution) overestimates rainfall over high ground. The model bias could be due in part to the inadequacy of the observation networks in East Africa, represented in this thesis by the CRU and FEWS datasets. However, these two datasets draw on some different sources and neither do they have the same resolution. FEWS is a high resolution data (0.1 o ) gridded satellite-derived precipitation estimate covering the entire African continent while CRU datasets is a relatively low resolution (0.5 o ) dataset based on rain gauge monthly precipitation only; in addition , near surface temperature is also available. As no reliable wind observations exist, wind data was taken from the ERA-Interim reanalysis. The different observational datasets do not agree particularly well, which impedes evaluating the quality of the HIRHAM5 simulations, in particular the high resolution one. So while the higher resolution HIRHAM5 appears to be generally reliable, caution must be exercised in formulating conclusions from the results, especially over high ground and remote areas without adequate observation data. Under these constraints, the results suggest HIRHAM5 may be useful for assessing climate variability and change over East Africa. A weakness of the analysis presented here is that only one combination of GCM and RCM could be investigated in depth due to computer and time constraints. Therefore the results presented here, if used in application for climate change adaptation, should be considered in conjunction with a broader suite of data, such from the CORDEX programme. This has potential to increase the reliability of information about climate variability and change at a regional to local level necessary for impact assessment.