RT info:eu-repo/semantics/bachelorThesis
T1 Estudio de las causas sinópticas de la precipitación en Canarias mediante modelos climáticos. Presente y futuro
A1 Delgado González, Daniel
AB This is a statistical study of prediction databases and observational rainfall data ofthe Canary Islands using Principal Component Analysis (PCA), clustering techniquesand meteorological knowledge to predict the change of rainfall patterns in the CanaryIslands in the future.First of all, we have to know that we are going to work with databases thatrepresent the observations of precipitation and sea level pressure in the CanaryIslands or their surroundings, and predictions of these same variables. These datacovers three periods of time, one for the recent past (1980-2009) and two for the future(2030-2059 and 2060-2099). The data for the future periods cover two scenarios relatedto global emissions paths (RCP-4.5 and RCP-8.5). The data gathered from the past,corresponds to predictions and real data, and the data from the future, obviouslycorresponds to climate projections.Specifically, the observational precipitation data for the recent past are directlyextracted from the SPREAD database (R. Serrano-Notivoli, Beguerıa y col., 2017 ). Thisis a high-resolution gridded precipitation dataset covering Spain. This wasconstructed by estimating precipitation amounts and their corresponding uncertaintyat each node on a 5x5 km grid. Sea level pressure data around the islands wereextracted from the ERA5 reanalysis (Hersbach y col., 2020).Apart from this, other databases are used, that correspond to regional climatemodels, which predict sea level pressure and precipitations, i.e., they are notobservational data, but simulations of these variables. Specifically, three databases areused, each one associated to the global climate model used for its generation: GFDL,IPSL and MIROC. Both, past and a future simulations, were provided.These models, which are regional climate simulations, have been performed withthe WRF model (Non Hidrostatic Weather and Research Forecasting- WRF/ARWv3.4.1) using a unidirectional triple nesting configuration with grid resolutions of27x27 km, 9x9 km and 3x3 km. These simulations were carried out by the Group ofEarth and Atmospheric Observation (GOTA) of the University of La Laguna (ULL).The used domain is centered in the Northeast Atlantic region and covers a large areato capture the main mesoscale processes affecting the Canary climate, while othermore internal domains are centered in the Canary archipelago. The WRF version andthe physical parameterizations that they used to represent the different subgrid-scaleatmospheric processes were selected by GOTA according to previous work in the same study area (Pérez y col., 2014) (Expósito y col., 2015).Now that the data used in this study have been explained, the methodology isoutlined. First, some statistical methods are applied to the aforementioned databasesto extract some features and information.In this study, among other methods, we use Principal Component Analysis (PCA),which is a mathematical technique to summarize the information contained in a set ofdata by means of other independent parameters; more specifically, it is a rotation ofthe coordinate axis of the original variables to new orthogonal axes, so that these axescoincide with the direction of maximum variance of the data. In this case, the data towhich we apply this method are the daily rainfall values, and the axes correspond toeach of the land pixels of the Canary Islands of the SPREAD database. In this way, wemanage to group the pixels of the islands in different groups in which rainfall iscorrelated.Although with this method we could already have a grouping of pixels with acertain correlation in terms of rainfall, what we do now is, with the axes rotated by thePCA performed, to apply some Clustering technique to group the pixels in differentregions. This should give us a coherence of the regions a little higher than thegroupings that were made with the PCA. Specifically, we use the K-means method todivide the pixels of the Canary Islands in 6 groups.The weather types for each day are determined from the sea level pressure valuesmeasured at certain points of a grid located over the Canary Islands. We use theformulas proposed by Jones y col., 2013. Once we have defined the type of weather(WT) for each day, and the amount of daily precipitation related to each of our pixelgroups (regions), we can elaborate heatmaps representing the percentage of rain andannual mean precipitation or heavy precipitation days related to each WT and region.Once we have each heatmap related to the past and to every RCP scenario of thefuture, we discuss them and extract some features of these heatmaps that we obtainedfrom the aforementioned databases. These heatmaps could throw some light on howthe patterns of rain in the Canary Islands could evolve from now to the next decades.Lastly, we mention some starting points on what next studies related to this subjectcould be based on.
YR 2021
FD 2021
LK http://riull.ull.es/xmlui/handle/915/25735
UL http://riull.ull.es/xmlui/handle/915/25735
LA es
DS Repositorio institucional de la Universidad de La Laguna
RD 11-nov-2025