Spatial and time behaviours of rainfall shortage and excess are analysed for Catalonia (NE Spain) using a database obtained from 99 rain gauges with monthly totals collected from 1961 to 1990. The distribution of monthly amounts for each rain gauge is modelled by means of the gamma or Poisson-gamma distributions. Then, using an equiprobable transformation, monthly amounts described with these distributions are substituted by values given by the Standardized Precipitation Index (SPI), which follows a standardized normal distribution and provides a unique pluviometric scale. After that, principal component analysis (PCA) is applied to the set of monthly SPIs. A double regionalization of the 99 rain gauges, distinguishing between episodes of rainfall shortage and excess, is achieved by taking into account the rotated factor loadings (RFL) correlating rain gauges and principal components (PC). A time classification of rainfall shortage and excess episodes is also established, considering in this case the factor scores (FS) obtained after a PCA of variables based on monthly SPIs. The spatial regionalization achieved becomes a rough picture of the different topographic domains (Pyrenees, Pre-Pyrenees, Central Basin, Littoral and Pre-Littoral chains and Mediterranean coast), the climatic diversity of Catalonia being enhanced by these results. The time clustering suggests a quite complex behaviour of the rainfall shortage and excess episodes. Moreover, the spatial distribution of these time clusters is very disperse, in such a way that monthly shortage and excess sometimes affect the whole of Catalonia and sometimes just a small area. Besides results obtained from PCA and clustering algorithms, it is worth noticing that the severity of the episodes increases remarkably only for rainfall shortage. In addition, an analysis of the number of rain gauges affected by monthly shortage and excess shows an interesting fact: whereas the number of rain gauges associated with a shortage has an increasing tendency, a significant decreasing tendency for excess is detected in the period 1961-1990.

ABSTRACT:Daily maximum and minimum temperatures, T max and T min , and diurnal temperature range, DTR, from 37 temperature stations in Catalonia (NE Spain) are analyzed to detect significant daily time trends for the period . The homogeneity of the series is tested by means of the Standard Normal Homogeneity, the Buishand range and the Pettitt tests. The lack of randomness of the series, suggesting time trends, is also investigated by means of the Von Neumann ratio test. The daily time trends obtained and their spatial and temporal patterns are mainly in agreement with overall time trends recently derived for the Northern Hemisphere. The results indicate generalized increasing annual trends of daily T max and T min (0.5°C/decade), especially relevant in spring and summer, with values reaching 0.8-0.9°C/decade, and also remarkable for T max in winter (0.7°C/decade). In autumn, however, average trends point at a decrease in T max (−0.5°C/decade). As a result, an average annual decreasing trend of DTR is found, particularly relevant in autumn (−0.9°C/decade). Several periods with an outstanding number of stations showing significant positive time trends are detected and analyzed during the spring and summer seasons both for daily T max and T min . The only period with a relevant number of significant negative trends is detected in February for daily T min , thus implying a significant increasing trend of DTR during this short winter period. Comparisons are established with large-and regional-scale temperature trends, paying attention to the west Mediterranean atmospheric dynamics change.

Spatial and temporal patterns in the daily rainfall regime of Catalonia (northeastern Spain) recorded for the 1950-2000 period are analysed from several points of view, including the irregularity of the time series in terms of entropy, the Mann-Kendall test for time trends, a principal component analysis (PCA), an average linkage (AL) clustering algorithm and, finally, a power spectrum analysis, which includes a comparison of white-noise and Markovian red-noise hypotheses. The analyses are based on three monthly variables derived from the amounts recorded on a daily basis: the average daily rainfall and the standard deviation of the daily rainfall for each month, together with the corresponding coefficient of variation. The joint spatial-temporal variability is manifested by the irregularity index, which is characterized by relevant values in all cases and gradients from the north (Pyrenees and Pre-Pyrenees mountain ranges) to the south (Ebro Valley) and to the Mediterranean coast. The interpretation of the factor scores derived from the PCA and of the clusters obtained from the AL algorithm also describes the complex spatial distribution of the daily rainfall regime, given that the effects of atmospheric circulation patterns on rainfall regimes are conditioned by the complex orography of Catalonia and its proximity to the Mediterranean Sea. The factor loadings associated with the PCA also suggest a distinction between hot, cold and mild seasons. Finally, it is worth noting that monthly series are usually accompanied by white background noise and, in a few cases, signs of Markovian behaviour and some significant periodicities, which are generally of less than 10 months and which change from one cluster to another.

Statistical distributions of annual extreme and long dry spells for the Iberian Peninsula are investigated by using a daily database compiled from 43 rain gauges, with the recording period extending from 1951 to 1990 and with a minor lack of data. Dry spell lengths are derived for three different daily rainfall thresholds of 0.1, 1.0 and 5.0 mm/day. On one hand, the generalised extreme value (GEV) and generalised Pareto (GP) distributions are considered for modelling the series of annual extreme (AE) dry spells. On the other hand, both theoretical distributions are assumed for the partial duration (PD) series, which are derived from the dry spell lengths exceeding the 95th percentile. In both cases, a robust estimation of the three parameters of the GEV and GP distributions is obtained by L-moments. The fit between empirical and theoretical distributions is evaluated by using the 95% confidence bands of the Kolmogorov-Smirnov test and the L-skewness-kurtosis distance. Even though AE spells are quite well fitted by the GEV model, the GP distribution is a better option for some rain gauges. The PD series are usually better fitted by the GP distribution, only a few cases being better modelled by the GEV distribution. The basis for climatic drought risk assessment in the Iberian Peninsula is then established for dry spell lengths associated with return periods of 2, 5, 10, 25 and 50 years and accurately reviewed by comparing with results deduced from the AE and PD sampling strategies. As a general feature, both the spatial distribution of the statistical parameters and the dry spell lengths for the different return periods depict a north to south gradient. Some local deviations of this behaviour could be due to the vicinity to the Mediterranean Sea and the Atlantic Ocean.

Daily maximum and minimum temperatures recorded without interruption at Fabra Observatory (Barcelona) from 1917 to 1998 are analysed studying their homogeneity, randomness, possible trends and their statistical significance, and time irregularities detected by means of concepts of entropy and spectral power analysis. The homogeneity of the series is tested on a monthly scale using the adaptive Kolmogorov-Zurbenko filter. With respect to the randomness of the time series, the von Neumann ratio test is applied to standardized values of extreme temperatures in four different time-scales (daily, monthly, seasonal and annual). The statistical significance of trends is quantified by applying the Spearman and Mann-Kendall tests to daily, monthly and seasonal data. The Mann-Kendall sequential test also leads to the detection of sharp changes in the time series when monthly data is analysed. The quantification of irregularities through entropy is investigated for standardized temperatures on daily, monthly and seasonal scales, based on the concept of mathematical information theory. Periodicities derived from spectral power analyses are checked with the hypothesis of white-noise and Markov red-noise stochastic processes. The most notable features, common to maximum and minimum temperatures, are the lack of randomness of the series for all the time-scales considered and the different trends obtained for the periods 1917-1980 and 1917-1998, which are confirmed by the Spearman and sequential Mann-Kendall tests. Nevertheless, the maximum and minimum temperature series show quite a different behaviour from the point of view of results concerning time irregularities in terms of entropy and periodicities. The main features of the results are discussed by comparing them with those obtained for other areas of the Mediterranean domain.