Spatio-temporal Near-land-surface Temperature Change Trend Over Mainland Spain

Climate change is a very hot and popular research topic and has attracted widely attentions from all over the world.Climate change has made a lot of negative effects on ecosystem,food,agriculture,forest,ocean,human being and so on.Thus,some rules have been promulgated from government to prevent or decrease the negative effects from climate change.A number of studies has indicated that although the climate of the world present a significant increasing trend in recent decades or recent centuries,different local areas or different research time intervals lead to different trend detection results.Thus,the identification of systematic small-scale and intermediate-scale temperature changes(trends)in a time series is a significant issue in the analysis of negative effects from climate change.Mainland Spain,located on the Iberian Peninsula in southwest Europe,is a mountainous country which dominated by high plateaus and mountain chains and has frequently affected by Mediterranean Sea and Atlantic Ocean.Thus,the climate of mainland Spain has attracted lots attentions from international researchers.In order to obtain a reliable trend in temperature changes,it is necessary to consider two important aspects,one is the correctness of the data(temperature)measurements or estimation,the other is the choosing of a reasonable trend detection method.Considering these two aspects,this PhD thesis mainly focuses on:Access spatio-temporal changes in temperature trend from 1950 to 2011 at a relative large area(mainland Spain covering an area of around half a million square kilometres).Firstly,the thesis proposed a novel clustering method,constrained spatial clustering method,based on an objective function and a contiguity constraint restricts the sets of allowable solutions to make the correlation between altitude and temperature as strong as possible.Secondly,we validate the constrained spatial clustering by assessing the semivariogram of the residual of the regression between altitude and temperature and testing the significance of the Pearson's correlation coefficient.Thirdly,an improved regression kriging estimator,which is optimally determined by the cluster analysis,is used to estimate the optimal areal values of near-land-surface temperature at grid and average level on mainland Spain.Finally,the statistical tests are used to test for temperature trends in the regional temperature time series.Compare and explore nine commonly used tests for detecting trend in time series.We firstly design a simulation study to compare the nine tests.We do so by considering the influence of four factors: the amount of data,the type of random field,the amount of spatial or temporal correlation and parametric drifts.In order to validate the results in simulation study,the statistical tests are also applied to mean annual temperature measured at 13 weather stations located in the Valencia region(Eastern Spain).Compare two commonly used tests for detecting trend in spatial statistics.We firstly design a simulation study using two-dimensional data to compare the two methods.Secondly,we check the feasibility of the two methods when they applied to time series analysis with one-dimensional simulated data(we have used the Mann-Kendal test with and without prewhitening as a benchmark for comparison purposes)then compare the trend detection ability of two tests in time series analysis.Finally,the two methods are applied to real case studies to validate the results of simulation study.