Journal of Arid Environments (2001) 48: 373}395 doi:10.1006/jare.2000.0755, available online at http://www.idealibrary.com on Analysis of a 30-year rainfall record (1967+1997) in semi+arid SE Spain for implications on vegetation R. LaH zaro*, F. S. Rodrigo-, L. GutieH rrez*, F. Domingo* & J. PuigdefaH bregas* *EstacioH n Experimental de Zonas ADridas, Consejo Superior de Investigaciones CientnHficas, 04001 AlmernHa, Spain -Departamento de FnHsica Aplicada, Universidad de AlmernH a, 04120 Almern& a, Spain (Received 11 January 2000, accepted 24 October 2000 ) In order to understand the behaviour of ecosystems in semi-arid areas, rainfall must be analysed over time. For this reason, statistical methods were applied to a rainfall time-series (1967–1997) in a typical Mediterranean semi-arid area in SE Spain of great ecological interest. Annual, seasonal and monthly time scales were studied, including rainfall volume, number of rain-days and one-day maximum rainfalls. Results showed neither trends nor abrupt changes in the series, although three periods, averaging 301, 183 and 266 mm year!1 respectively were observed from fluctuation in rainfall. Modal values of annual and monthly rainfalls were lower than average. Inter-annual variability was 36% and intra-annual variability up to 207%. Although there was often a rainfall maximum in autumn and a minimum in July, the only certainty was a summer drought, which marking strong annual cycles. The estimated return periods for events of more than 50, 70 and 100 mm day!1 were over 5, 11 and 30 years respectively; the absolute maximum 1-day rainfall recorded was 98 mm. Results suggest that vegetation is not only adapted to the amount of precipitation, but also to its timing. All types of rainfall, in terms of volume and timing, would have consequences for vegetation. ( 2001 Academic Press Keywords: time-series analysis; semi-arid; Tabernas; SE Spain; rainfall record; extreme rainfall event; plant ecology Introduction The Mediterranean climate has a limited capacity to buffer the environmental consequences of changes in land use (Palutikof et al., 1996). This capacity is still lower in semi-arid zones, where precipitation is highly variable in time, space, amount and duration (Noy-Meir, 1973) and water is often the most limiting factor of biological activity. In such environments, relatively small variation in water input can have a strong Address correspondence to: Dr R. LaH zaro, EstacioH n Experimental de Zonas AD ridas (C.S.I.C.), General Segura, 1, AlmermH a 04001, Spain. Fax: #34 950 27 71 00. E-mail: lazaro@eeza.csic.es 0140-1963/01/070373#23 $35.00/0 ( 2001 Academic Press 374 R. LAD ZARO ET AL. effect on the vegetation and the entire ecosystem. In this context, changes in land use in the Mediterranean Basin over the last few decades, such as abandoning of agricultural fields or over-exploitation of water resources, have raised concern about the risks of land degradation and desertification (Brandt & Thornes, 1996). The analysis of rainfall records for long periods provides information about rainfall variability and helps understand the ‘natural’ behaviour of the vegetation so that the impact of changes in land use on ecological processes can be assessed, and effective management strategies developed (Sutherland et al., 1991). Noy-Meir (1973) suggests that different time scales be considered when rainfall series are analysed for ecological purposes. Several models have been developed to predict vegetation productivity or plant behaviour as a function of annual, seasonal or monthly rainfall, or of rainfall-related indices (Lieth, 1976; Floret et al., 1982; Lane et al., 1984; Le HoueH rou, 1984). According to Woodward (1987), to study relationships between precipitation patterns and vegetation-related processes, the time-scales of the independent and dependent variables must be of the same order. Climatic fluctuation having a greater return period than the response time of any vegetation variable does not produce changes in that variable. On an annual scale, rainfall patterns affect floristic composition and grassland biomass (Sala et al., 1988; Silvertown et al., 1994), productivity (Rutherford, 1980; Naveh, 1982; Le HoueH rou, 1984, 1988), spatial distribution pattern and cover (Woodell et al., 1969). The seasonal variation of rainfall affects phenology and growth (Ackerman & Bamberg, 1974; Beatley, 1974). Germination and annual plant survival (Beatley, 1967; Ackerman, 1979), as well as phenological development (Bertiller et al., 1991) and leaf area (Specht & Specht, 1989) have often been related to monthly rainfall. Recently, there have been attempts to quantify the integrated effect of climatic patterns on erosion, hydrological and vegetation processes in semiarid environments (Mulligan, 1998). Long rainfall records also provide opportunities to validate other time series derived from proxy data such as tree rings or lake sediments, and to assess the potential sustainability of agricultural systems that are heavily dependent on climatic variability (Pickup & Stafford Smith, 1993; PuigdefaH bregas, 1998). The main objective of this paper is to analyse from the point of view of plant ecology the 1967}1997 rainfall records from the weather station in Tabernas (AlmermH a province, SE Spain), as a basis for a future study of relationships between rainfall patterns and vegetation-related processes. Among the driest areas in Europe (Capel-Molina, 1986), and featuring some of the most extensive badlands in Spain (Calvo & Harvey, 1996), the Tabernas region is of great ecological interest. Since 1989, the Arid Zone Research Station of the Spanish National Council for Scientific Research (CSIC) has run two instrumented field sites in the area near Tabernas known as ‘Rambla Honda’ and ‘El Cautivo’. Research at the two field sites focuses on geo-ecological experiments within the framework of national and international research projects. The approach to the statistical analysis of this rainfall series was similar to those of other authors, such as Sutherland et al. (1991) and, in Spain, Esteban-Parra et al. (1997, 1998). The latter served as a reference for comparing our results. Geographical setting The Tabernas weather station (37303@10@@N; 2323@27@@ W) is 32 km north of AlmermH a (Spain), and 490 m a.s.l. in the Tabernas basin. Tabernas is partially surrounded by the Betic cordillera and leeward of the Sierra de los Filabres, Sierra Nevada and Sierra de GaH dor ranges, all of which are over 2000 m a.s.l. (Fig. 1). Rainfall events in the area are produced by rain-bearing fronts coming in from the west from the Atlantic Ocean, mainly in the cold season. This geographic location makes the regional climate in the SE Iberian Peninsula markedly semi-arid due both to the rainfall shadow effect of the THIRTY-YEARS RAINFALL IN SE SPAIN 375 Figure 1. Location of the Tabernas weather station. main Betic ranges and its proximity to northern Africa (Geiger, 1973). After summer and during autumn, rainfall is associated with fronts coming from the Mediterranean Sea, which sometimes results in storms and torrential rainfall. Precipitation in the AlmermH a region is influenced both by the December North Atlantic Oscillation and by the October Southern Oscillation (Rodriguez-Puebla et al., 1998). The average annual temperature in Tabernas is 17)83C and the average total annual rainfall is 235 mm (1967–1997). The Tabernas basin materials are mainly Tortonian age marine marls with poorly developed pedogenic soil horizons. Natural vegetation is dominated by sparse dwarf shrubs adapted to severe drought and includes elements of North African flora. Frequently, numerous annual plant species can be found among the shrubs. Both physiognomy and life-form structures of vegetation are more similar to those of North Africa than Europe. Richness of species and endemic plant rates are high. Methods The Tabernas weather station, which belongs to the Spanish National Institute of Meteorology (INM), was established in 1967, and therefore provides a data series long enough for use as an area reference. The Tabernas data used here consists of total monthly rainfall, the number of days with rainfall per month, and the maximum 24-h rainfall recorded each month. As there were no rainfall data from the Tabernas station for 35 months, mainly from 1989 to 1991, prior to statistical analyses, the missing monthly values were replaced by statistical estimations based on 15 other INM stations nearby, all of them in the Province of AlmermH a between the main Betic Cordillera and the Mediterranean Sea. Table 1 shows their main specifications. The stations were selected for their comparable geographic location, total rainfall records and recording period (LaH zaro & Rey, 1991) and for each a significant correlation (a(0)001) could be established with the Tabernas data. For each month missing from the Tabernas database, a value was estimated from each station using the corresponding regression equation between it and the Tabernas. Each missing value was then replaced with the weighted average of the estimates obtained from each of the support stations weighted with the correlation coefficient for the station. Seasonal and annual rainfall values were calculated from monthly values taken from the completed Tabernas database. Databases of rain-days per month and maximum rainfall in 24 h month~1 were obtained by the same method. Annual and seasonal averages were made for ‘hydrological years’ from the 1 September to 31 August, as suggested by Le Houerou (1984) as being more useful for ecological studies in Mediterranean climates in the northern hemisphere. 376 Table 1. List of the Spanish INM weather stations used to complete the Tabernas’ monthly data base, showing the station name, longitude, latitude, altitude and average annual rainfall (mm). All stations have records from 1967 to 1998, except M06321O, M06325A and M06325O which started in 1987 INM reference Latitude (N) Altitude (m a.s.l.) 02339@02@@ 02337@17@@ 02334@07@@ 02332@17@@ 02317@42@@ 02321@55@@ 02327@37@@ 02324@43@@ 02323@17@@ 02317@27@@ 02312@17@@ 02312@02@@ 01356@27@@ 02308@52@@ 01356@27@@ 37306@30@@ 37302@00@@ 36357@25@@ 37307@10@@ 37305@10@@ 37305@33@@ 36356@40@@ 36353@44@@ 36350@35@@ 36351@13@@ 36358@00@@ 37302@25@@ 37310@10@@ 37323@20@@ 37323@20@@ 595 460 520 758 515 503 127 131 21 70 356 550 120 420 270 Station name Nacimiento Alboloduy Alhama de AlmermH a GeH rgal Tabernas a Sorbas CPC Tabernas Planta Solar Rioja Viator Base Militar AlmermH a Aeropuerto El AlquiaH n Los Trancos NmH jar Lucainena de la Torres Los Gallardos Albox HueH rcal Overa Rainfall (mm) 224 233 255 296 245 285 188 211 237 202 272 302 351 304 273 R. LAD ZARO ET AL. M06305 M06306 M06314 M06315 M06321 M06321O M06324 M06325A M06325O M06326 M06327 M06331 M06339 M06364 M06367 Longitude (W) THIRTY-YEARS RAINFALL IN SE SPAIN 377 Statistical analysis of annual and seasonal rainfall Thom test of homogeneity A numerical series with variations in a given climatological factor is said to be ‘homogeneous’ when such variations are caused only by fluctuations in weather and climate. The most important causes of non-homogeneity are changes in instrumentation, exposure and measurement techniques, changes in station location, observation times and station environment, in particular urbanisation (Jones, 1995). To test the series for homogeneity, the Thom test, which studies variation in the series with regard to the median, was used as recommended by the World Meteorological Organisation (WMO), in spite of being a non-parametric test (Peinado-Serna, 1985). In this test, considering xi, with i"1, 2,2, n, a time-series of length n and xmed the series median, a code called ‘a’ is assigned for any value xj'xmed , a code ‘b’ for any value xj(xmed, and the element xk is rejected when xk"xmed . The result is a series made up of ‘a’ and/or ‘b’ codes from the original series. Each uninterrupted series of ‘a’ or ‘b’ codes is called a ‘run’. For N*25, if the series is homogeneous, the distribution of the number of runs (R) approximates a normal distribution with the following average (E) and variance (Var): N#2 E(R)" 2 N (N!2) Var (R)" 4 (N!1) [1] The Z statistic is defined as: R!E (R) Z" JVar (R) [2] For a significance of a"0)01, the null hypothesis of homogeneity is verified when DZD)2)58. Test of normality The annual and seasonal series were tested for normality by skewness and kurtosis coefficients and the s2 test (Table 2). The use of non-parametric methods is largely justified since the median and percentiles play an important role in analysing the series (Sneyers, 1990). Rainfall was considered ‘normal’ between the first (P25) and third quartile (P75), under P25 as ‘dryer’ than normal and over P75 ‘wetter’ than normal (Table 2). The 10th (P10) and 90th (P90) percentiles were considered thresholds for extreme values. This criteria is commonly followed to characterise ‘very dry’, ‘dry’, ‘normal’, ‘rainy’ and ‘very rainy’ climatic conditions (GarcmH a de Pedraza & GarcmH a-Vega, 1989). Analysis of climatic changes during the time period Three basic forms of climatic changes can be distinguished: (1) Non-abrupt change: when a change is caused by normal climatic fluctuations; (2) Abrupt change: when a change might imply a non-homogeneous series; however, other abrupt discontinuities due to natural causes are also possible, for example, a volcanic eruption. (3) Trend: when there is an increase or decrease throughout the series. A rainfall series frequently includes both wet and dry periods, which together form a fluctuation, since they are shorter than the entire series. A series with a trend might R. LAD ZARO ET AL. 378 Table 2. Annual and seasonal statistical summary from the Tabernas’ monthly rainfall series (1967–1997) Parameter Annual Autumn Winter Spring Average (mm) Variance (mm2) SD (mm) VC (%) Minimum(mm) Percentile 10 (mm) Percentile 25 (mm) Percentile 50 (mm) Percentile 75 (mm) Percentile 90 (mm) Maximum (mm) Skewness Kurtosis Thom’s Z s2 235)4 7153 84)6 36 124)4 143)3 174)5 217)6 290)1 347)2 466)6 2)38 0)93 !1)86 4)363 79)7 3536 59)5 75 0)0 24)6 35)5 72)0 105)3 152)9 255)6 2)89 1)95 !0)56 2)468 77)7 1851 43)0 55 18)4 23)1 38)7 79)2 100)6 137)0 188)1 1)15 !0)15 !1)30 1)313 63)2 1246 35)3 56 1)2 13)4 34)5 70)0 98)9 107)0 125)7 !0)15 !1)18 !0)19 3)384 Summer 17)4 528 22)9 135 0)0 1)5 3)6 10)0 24)5 28)2 100)7 5)48 6)82 2)04 25)359 Critical value for Thom’s test was 2)58 at a"0)01 and, for the s2 test was 2)167 at a"0)05 constitute only a progression into another longer series that may or may not have a trend. Conversely, a period starting with a non-abrupt change may produce a trend into a shorter series. Non-abrupt changes To investigate the possible existence of non-abrupt changes in the series and differentiate wet and dry periods, the cumulative sums of deviations were calculated according to Eqn 3 and graphically represented as a function of k, with k"1, 2,2, n. k Sk" + di [3] i"1 A climatic change in the series can be detected by a change in the slope of the progression of points representing Sk . k for the maximum DSk D represents a point of change (BaH rdossy & Caspary, 1991). This method has been useful for analysing rainfall and runoff series in which strong variability may hide possible changes (Mitchell, 1966). To verify whether the periods delimited by changes are significantly different, average values before and after the possible changing point were compared with a t-test. The advantage of this test is that it can be used for data not having a normal distribution, while their variances are similar (Mitchell, 1966). Abrupt changes and trends To investigate the possible existence of abrupt changes and trends, the sequential version of the Mann–Kendall test proposed by Sneyers (1992) was used. This test seems to be the most appropriate and powerful method for analysis of the climatic variations in a climatological time series (Goossens & Berger, 1986). The Mann–Kendall test is THIRTY-YEARS RAINFALL IN SE SPAIN 379 essentially a non-parametric test of the monotonous dependence of two variables (Lebart et al., 1985). In the case of a time-series, one of the variables (time) is known, simplifying application of the test, which is based on the following procedure. Considering a time-series of length n (xi , where i"1,2 , n), for each xi, the number mi of xj(xi with j(i is computed. Then, it is calculated the sum: N dN" + mi [4] i"1 This sum, under the null hypothesis (stable climate is characterised by a simple random series), presents a normal distribution with an average (E ) and variance (Var ): N (N!1) E (dN )" 4 N (N!1 ) (2N#5) Var (dN )" 72 [5] Then the u(dN ) statistic is calculated as: dN!E (dN ) u (dN )" JVar (dN ) [6] and u(dN ) is compared with a standard normal distribution at the required level of significance, which is usually a"0)05 (Goossens & Berger, 1986; Sneyers, 1992). The null hypothesis is rejected when Du(d )D'1)96. u(dN )'0 indicates an increase and N u(dN)(0 indicates a decrease. The sequential version consists of a graphic representation of u(di) on the time axis, u denoting the series and u@ the retrograde series. The retrograde series is calculated for each xi, the number (m@i ) of terms xj , with j'i and xj(xi. A trend is significant when curve u passes through the 5% significance level from inside outwards. If, in addition to the above condition, curves u and u@ clearly cross each other (not overlap) between the critical values at the 5% level, an abrupt change exists, and the intersection point represents the beginning of that change. To find out the influence of seasonal rainfall on annual rainfall, the Spearman and Pearson correlation coefficients were also calculated between annual and seasonal series (Table 3). Monthly rainfall statistical analysis Tables of frequencies A histogram of frequencies was plotted, with the classes limited to 20 according to Shaw (1983). Monthly rainfall series are useful for calculating the probabilities of monthly rainfall (Dunne & Leopold, 1978); the 10th, 25th, 50th, 75th and 90th percentiles of monthly rainfall were also calculated. A statistical summary for each month and for the entire monthly database is shown in Table 4. Table 3. Correlation coefficients between annual and seasonal series Pearson Spearman Autumn Winter Spring Summer 0)58* 0)52* 0)54* 0)45* 0)46* 0)47* 0)43* 0)10 * Statistical significance at a"0)05 380 Table 4. Statistical summary of monthly rainfall data from Tabernas (1967– 1997) Parameter Oct. 18)0 30)7 831 1250 28)8 35)4 160 115 0)0 0)0 0)0 0)1 0)0 4)5 6)0 22)5 20)7 46)7 58)0 66)6 109)6 174)2 5)12 5)38 5)27 9)29 Nov. Dec. 30)9 24)6 947 1265 30)8 35)6 99 145 0)0 0)0 0)1 0)0 5)0 1)6 19)1 13)5 56)2 35)0 81)0 57)5 96)3 171)1 1)98 6)12 !0)69 10)27 Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Total 27)3 27)1 20)3 23)2 19)7 11)2 2)1 4)1 19)9 827 766 474 413 428 474 19 39 705 28)8 27)7 21)8 20)3 20)7 21)8 4)4 6)2 26)5 105 102 107 87 105 195 207 150 134 0)0 0)0 0)0 0)0 0)0 0)0 0)0 0)0 0)0 1)0 1)3 0)4 2)8 0)1 0)0 0)0 0)0 0)0 5)6 7)6 2)7 6)9 2)5 0)1 0)0 0)0 1)0 15)8 21)5 11)2 17)6 15)0 3)8 0)0 0)5 9)6 42)0 38)0 34)1 38)9 30)0 10)0 2)9 6)0 27)7 64)2 56)8 57)6 51)0 44)1 24)5 4)5 14)5 56)0 113)8 127)8 78)0 76)9 94)4 100)0 18)2 21)1 174)2 2)95 4)26 2)58 2)39 3)98 7)31 6)20 3)14 17)69 1)63 5)35 0)43 0)51 4)93 12)11 8)53 0)73 26)71 R. LAD ZARO ET AL. Average (mm) Variance (mm2) SD (mm) VC (%) Minimum (mm) Percentile 10 (mm) Percentile 25 (mm) Percentile 50 (mm) Percentile 75 (mm) Percentile 90 (mm) Maximum (mm) Skewness Kurtosis Sep. THIRTY-YEARS RAINFALL IN SE SPAIN 381 Analysis of periodicity by sample autocorrelation coefficients To detect the existence of periodicity in a time series, autocorrelation coefficients, which measure the correlation between observations at different distances in time are useful. The autocorrelation function r , which varies between $1, was calculated as L follows: cL rL" 2 s [7] 1 n!L + (xi!xN ) (xi#L!xN ) cL" n!L i"1 [8] n 1 + (xi!xN )2 s2" n!1 i"1 [9] where: and: cL being the auto-covariance, n the number of elements, L the number of lags, xi the value at ith time, and xN the average and s2 the variance. Autocorrelation analysis determines the average extent of mutual dependence of adjacent observations. For instance, when L"2, x1 is correlated to x3 , x2 with x4, etc. An autocorrelogram is useful for interpreting a set of autocorrelations, for which rL is plotted against L. Significance was tested at a"0)05. The autocorrelation analysis was applied to the annual series, the seasonal series, the four series each one of which containing the data from only one season, and to the monthly series. Annual and monthly rain-days A day was considered a ‘rain-day’ when rainfall volume was *1 mm. To examine time patterns, rain-day series were plotted over annual rainfall series, using double-ordinate axes. Annual and monthly rain-day averages were calculated. Non-parametric regressions between the number of rain-days and both annual and monthly rainfall volumes were obtained. Extreme rainfall events Analyses followed two approaches to the time distribution of extreme rainfall values. The first one, on a month-long scale, refers to relative high and low extreme values, using the percentiles for each month as thresholds: P75 for the upper extreme and P25 for the lower. The use of percentiles rather than only one upper and one lower threshold for the set of months is because a rainfall value can be extreme in one particular month but not in others. We preferred to use the P75 and P25 percentiles rather than P90 and P10 because by including more rainy (or dryer) months than normal rather than only the very rainy (or very dry) a larger part of related vegetation processes are likely to be collected, since the very rainy or dry months are occasional and less able to generate plant adaptation. The number of months per year with precipitation higher or lower than their corresponding thresholds were counted. 382 R. LAD ZARO ET AL. The second approach to analysis of extreme rainfall values was on a day-long scale and referred to absolute extreme values. The theoretical extreme distribution most commonly used to analyse annual maximum rainfall on one day is the Gumbel asymptote (FaragoH & Katz, 1990). It estimates the probability that a maximum value will exceed a certain threshold. If xi is the value of the meteorological variable for the i-th time period, then let us call the maximum of xi X m . If the observations of the meteorological variable are distributed with the distribution function F(x)"PMxi)xN, the exact distribution of the maximum is expressed by equation 10. PMXm)xN"[F(x)]m [10] For a sufficiently large parent sample size ‘m’, the probability distribution of the standardized maximum value of the function y"(X!u)/b can be approximated by equation 11. G (y)"e!e !y [11] where u is the location parameter and b the scale parameter, which can be estimated by Eqn 12 from the series yj"!ln (!ln (1!1/Tj )), where Tj"(n#1)/j, and j"1, 2,2, n] pm b" p (y) u"xN !byN [12] where xN and pm are derived from the maximum sample. The empirical values for the reduced variable y depend only on sample size n, instead of the mean and standard deviation of the parent distribution F. These values are usually depicted on Gumbel’s probability graph, the vertical axis of which is a twofold iterated logarithm corresponding to the inverse of the Gumbel distribution function. The sample of ‘extremes’ is ordered on the horizontal axis. If the extreme sample is distributed according to Gumbel’s law, it will lie in a straight line y"(x!u)/b. The correlation coefficient between the reduced variable (y) and the threshold values (x) is r 2"0)98 (a(0)001). A linear equation is then obtained (Eqn 13) that can be used for predictions. y"!1)707#0)060x [13] For any value of x, Eqn 13 yields a value of the reduced variable y and therefore, the return periods (T ) are given by Eqn 14. 1 T" !y 1!e!e [14] Results Annual and seasonal rainfall statistical analysis Table 2 summarizes the statistical analysis of annual and seasonal rainfall values. Figures 2(a) and 3 show the annual and seasonal-scale time series respectively. Annual and seasonal series may be considered ‘homogeneous’ according to Thom’s Z test, which gave values of less than 2)58 (Table 2). Skewness and kurtosis coefficients together with the s2 test indicated that annual and seasonal series (except for winter) do not have a normal distribution (Table 2). Inter-annual distribution of rainfall showed some irregular groups. The cumulative sums of deviations detected three periods (Fig. 2(b)): 1967/1968K1973/1974, THIRTY-YEARS RAINFALL IN SE SPAIN 383 Figure 2. (a) Time-series for total annual rainfall from the Tabernas weather station (1967–1997). The horizontal line shows the average rainfall value. (b) Cumulative sum of deviations from average rainfall. Changes in the slope allow rainy periods to be differentiated. (c) Results of the Mann–Kendall test. The solid line is the u (d ) statistic and the thin line is the * retrograde series u@ (di). 1974/1975K1987/1988 and 1988/1989K1996/1997, which had average rainfalls of 301, 183 and 266 mm, respectively. A t-test found significant (a(0)05) differences between the averages of the first and second and the second and third periods. However, using percentiles, the first period may be classified as ‘wet’, and the other two periods as ‘normal’. Grouping the second and the third periods (1974/1975–1996/1997), the average was 215 mm and an F-test showed that the average of the first period was significantly different (a(0)05) from that of the second plus the third. Application of the sequential version of the Mann–Kendall test to find out whether there were abrupt changes or trends in the annual (Fig. 2(c)) and seasonal series showed a slightly decreasing trend until approximately 1988, and a slight 384 R. LAD ZARO ET AL. Figure 3. Time-series for total seasonal rainfall, illustrating each season. (a) Autumn; (b) Winter; (c) Spring; (d) Summer. increase after that year. Therefore, annual rainfall in Tabernas fluctuated around the average value and, since curve u overlaps curve u@ (Fig. 2(c)), neither abrupt changes nor trends over the entire series were present. Similar behaviour was found for the seasonal series. The slightly decreasing trend detected in the first half of the annual series is caused mainly by winter and spring rainfalls. The main contribution to total annual rainfall came from autumn and winter values. Variation coefficients (VC) indicate high variability in inter- and especially intraannual rainfall, particularly in summer (135%) (Table 2). Spearman and Pearson coefficients showed a significant correlation (a"0)05) between the annual and seasonal series (Table 3). Autocorrelation of annual data showed only significant annual cycles and the analysis of seasonal data showed an approximately sinusoidal curve that offers more evidence of annual cycles (Fig. 4). The four autocorrelations carried out (one per season) Figure 4. Autocorrelation function for the total seasonal rainfall series. Lag is expressed in seasons. Dashed lines mark 95% significant thresholds. THIRTY-YEARS RAINFALL IN SE SPAIN 385 showed that there was no significant autocorrelation for any lag and that therefore one season is not related to the same season of previous years. Monthly rainfall statistical analysis Frequency distribution of monthly rainfall was positively skewed due to the majority of months being dominated by low rainfall values (highest frequency for 0–10 mm) as 34% of the months recorded had values lower than 10 mm and 50% had value (20 mm. There were few cases of high rainfall values: none of the eight frequency classes over 60 mm reached 4%. Frequency was null for classes between 130 and 170 mm and only 0)65% in the highest class (170}180 mm). Monthly distribution of rainfall showed a maximum in October–November and a minimum in July–August, according to the monthly percentiles (Fig. 5(a)). October Figure 5. The percentiles of monthly rainfall are plotted as (a) probability 10, ( ); 25, ( ); ); 75, ( ); and 90% , ( ) of monthly rainfall being less than that specified by the 50, ( ). (b) Average monthly rainfall values corresponding ordinate. Average monthly rainfall ( ), corresponding to the percentiles for rainfall each month in autumn and winter September: ( October ( ), November ( ), December ( ), January ( ), February ( ). 386 R. LAD ZARO ET AL. Figure 6. Autocorrelogram with the monthly rainfall series. Lag is expressed in months. Dashed lines mark 95% significant thresholds. and November, for which the 75th and 90th percentiles had the highest rainfall volumes, have a greater probability of having the highest total values. The ‘normal’ values (between the 25th and 75th percentiles) for November were between 5 and 56 mm, and for October between 4 and 47 mm. In summer, the ‘normal’ values were between 0 and 3 mm for July, and between 0 and 6 mm for August. Autumn and winter months showed a greater probability of more precipitation. Figure 5(b) shows the corresponding percentiles for monthly rainfall values in autumn and winter months. Statistical summaries for each month and for the complete monthly data set are shown in Table 4. Monthly rainfall distribution (Fig. 5(a)) was asymmetric and positively skewed, indicating that the majority of the months had only light rainfall interspersed with a few high values. This caused the median to be lower than the average in all cases. Monthly data (Fig. 6) are significantly autocorrelated (a(0)05) at 2 and 7-month lags (intra-annual or seasonal cycles) and at 11, 12 and 23-month lags (annual cycles). The atmospheric conditions, wet or dry, during two consecutive months are often similar but not during a third month; and approximately each half-year alternates a wet period (autumn and winter) with a dry one. Due to this alternation, rL was negative for 2and 7-month lags while it was positive for 11-, 12- and 23-month lags because the pattern was similar for all years. Any month was likely to be related to the previous one due to the monthly lag rL being 0)097, value very near to the significance threshold, which varies from 0)10 for lag"1 to 0)11 for lag"23. (This variation is too small to be distinguished in Fig. 6.) Annual and monthly rain-days The annual average of rain-days in Tabernas was 36. The number of rain-days was significantly correlated to total annual rainfall (Spearman r"#0)79; n"30; p(0)001), but not to time (Spearman r"#0)21; a'0)05). These two results tend to indicate that the number of rain-days per year did not present a significant global trend (Fig. 7). Moreover, in most of the years, a decrement in rainfall volume implied a decrease in number of rain-days. But in dry periods the correlation between annual rainfall and annual rain-days cannot be significant, as in the longest dry period THIRTY-YEARS RAINFALL IN SE SPAIN 387 Figure 7. Relationship between annual rainfall ( ) and annual rain-days ( ). 1975–1988 with 43% of the years (1977, 1978, 1979, 1984, 1985 and 1987), the number of rain-days remained high in spite of the low precipitation. The regression between rain-days and monthly rainfall was also significantly positive (Spearman r"#0)86; a(0)001). The monthly distribution of rain-days was similar to monthly rainfall, November having the highest value (4)5 days) and July the lowest (0)8). The mean value was 6)6 mm rain-day~1 for the whole series and only slightly lower, 5)5 mm rain-day!1, during the dry period 1974/1975–1987/1988. Extreme months and extreme rainfall events Monthly rainfall varied considerably during the 30 years studied, having both very high and very low values. For example, 174 mm was registered in October 1969 (49% of that year’s total annual rainfall), 171 mm in December 1971 (37%) and 100 mm in June 1972 (21%). With regard to low values, every month registered 0 mm at least once. Figure 8(a) shows the number of months per year with heavy rainfall. Rainfall in a given month was considered ‘heavy’ when it exceeded the corresponding P75 percentile. Which season corresponded to each month was also distinguished. Figure 8(a) reflects pattern similar to annual rainfall (Fig. 2(a)), with higher-than-median values at the beginning and at the end of the series. All seasons showed the same relative monthly frequency (25%). Figure 8(b) shows the number of months per year with light rainfall (when rainfall was lower than the P25 percentile for that month), distinguishing again the season for each month. Figure 8(b) allows differences between the first ‘normal’ period mentioned above (1974/1975–1987/1988) with many dry autumn months, and the other ‘normal’ period (1988/1989–1996/1997) with fewer dry autumn months and some irrelevant dry summer months, to be observed. The seasons with the most dry months were winter and spring (22 cases for each, 68% of the months). The amount of maximum daily rainfall for each month, previously grouped into 10-mm classes, is associated with monthly (a"0)05) and yearly rainfall (a"0)001). The wettest year, 1972, had the three most intense 1-day rainfalls: 40–50, 60–70 and 90–100 mm. In contrast, that year also had few medium values of monthly maximum precipitation in one day, i.e. only 2 days between 10 and 30 mm. On the other hand, a wet year like 1989 had more days with medium rainfall (6 days between 10 and 30 mm) and fewer days with heavy rainfall (only one 40–50 mm). 388 R. LAD ZARO ET AL. Figure 8. (a) Number of months per year with extremely heavy rainfall. (b) Number of months per year with extreme light rainfall. Different thresholds were used for each month, coincidental with the corresponding values of its P and P percentiles. Different shadings in the 75 25 bars indicate the season to which the extreme month corresponds; Summer ( ), Spring ( ), winter ( ), autumn ( ). Among the extreme 24-h annual maximum rainfall values those of June 1972 (98 mm), January 1992 (65)5 mm), December 1971 (63 mm) and February 1993 (61 mm), may be pointed out, each of which alone exceeds the 90th percentile for the corresponding month (Table 4). Figure 9 shows the frequency of annual maximum 1-day rainfall for each month. The majority of these rainfalls are concentrated in autumn and winter. When drier-than-average Figure 9. For each month, the number of times that 1-day annual maximum rainfall occurred in that month (in %) were calculated by considering the whole series ( ), only the wetter-thanaverage years ( ) and only the drier-than-average years ( ). THIRTY-YEARS RAINFALL IN SE SPAIN 389 Figure 10. The return period of extreme 1-day annual maximum rainfall events. The fitted equation is: T"0)477 * e0)045 *x (r 2"0)9644). and wetter-than-average years are represented separately, important differences in the time of year of the maximum events can be seen (February and October, respectively). The greatest frequency of relatively high 1-day annual maximum rainfall volumes (30–60 mm) was in autumn, but there were high values on the order of 60–70 mm in winter and the heaviest rain (90–100 mm) curiously enough, fell in June. Monthly rainfall is positively associated with monthly 1-day maximum rainfall (Spearman’s r"0)974; a(0)001); and similarly, annual rainfall is positively associated with the annual one-day maximum rainfall (Spearman’s r"0)587; a(0)001). Figure 10 shows the application of the Gumbel method to determine the theoretical extreme distribution of 1-day annual maximum rainfall. Rainfall of up to 20 mm day!1 can be expected for every year analysed. The estimated return period for events of over 50, 70 and 100 mm day!1 were longer than 5, 11 and 30 years respectively. Absolute maximum one-day rainfall recorded was 98 mm. Discussion and conclusions Tabernas rainfall features Rainfall in Tabernas fluctuates, showing neither a definite trend nor abrupt changes. This result is consistent with the analysis of the climate of AlmermH a from 1911 to 1991 (Esteban-Parra et al., 1997) and with the analysis of the Mediterranean climate of the Iberian Peninsula (Esteban-Parra et al., 1998). The use of percentiles detected two main periods: a ‘wet’ period from the beginning of the series until 1973/1974, followed by a ‘normal’ period with rainfall increasing slightly from 1988/1989 to present (Fig. 2(a, b)). On a seasonal scale, this pattern was clearer for winter (Fig. 3). But the ‘normal’ period can be divided into two parts, a first dryer period to 1988 and another relatively wetter period afterwards, according to their significantly different averages and their differences in number of dry months per year, particularly in autumn (Fig. 8(b)). This type of irregular sequence of dry and wet groups of years (Fig. 2(a)) is typical for many semi-arid regions (Geiger, 1973). The monthly rainfall series in Tabernas and the AlmermH a airport (one of the stations used to complete the Tabernas series, Table 1) were a very significantly correlated (Spearman’s r"0)8312; a;0)001). Compared to the AlmermH a series (1911–1991), which includes data from the AlmermH a airport (Esteban-Parra et al., 1997), the Tabernas dry period was the longest and driest in the last 80 years. The first wet period could be an 390 R. LAD ZARO ET AL. extension of a longer wet period starting in the mid-1940s and lasting around 30 years. It is therefore possible that a 30-year series is not long enough to be representative in a semi-arid climate. Intra-annual variability is characterized by a maximum, often in November and a minimum in July, with clear alternation of dry and wet seasons (Fig. 5(a)). Palutikof et al. (1996) pointed out in a 20-year series that rainfall in Tabernas seems to have a lightly bimodal annual pattern, with a winter drought and a relatively wet spring. This study has not found such bimodality (Fig. 5(a)). In spite of all years showing a similar rainfall pattern on a whole, inter-annual variability is high (36%, Table 2). This is typical of semi-arid climates and variability is similar, for example, in Turkey (TuK rkes, 1996). The Tabernas climate is Mediterranean rather than desert-like, because it rains every year and, according to the results of the autocorrelation analysis (Fig. 6), there is a significant annual cycle based on a strong summer drought. This drought contrasts with one (or several) rainfall maximums at an unknown time although more frequently in autumn or winter. Autocorrelation showed other less extensive season-related cycles. Annual and seasonal periodicity has been found in other Mediterranean countries, e.g. Dalezios & Bartzokas (1995) found a periodicity of 26, 55, 122 and 365 days in Greece. But the Tabernas climate also has low total rainfall and strong inter and intra-annual variation, which features are half Mediterranean and half desert-like and are typical of SE Spain. This is due to its geographic location sheltered from the Atlantic fronts by the Betic cordillera and affected by the Mediterranean Sea (Castillo-Requena, 1989). And in some areas, such as Tabernas, there is an additional feature: numerous rain-days with very little precipitation, and many months (34%) with less than 10 mm rainfall. Highest monthly totals are more probable in October and November (Fig. 5(a)), as reported for southern Spain by other authors (Geiger, 1973). Figure 5(a) shows the peculiarity of September, in which there is a long interval between P75 and P90: ‘normal’ September rainfall volume is low (between 0 and 20 mm), but occasionally it is the rainiest month of the year. At the other extreme is July, for which the absolute maximum was only over 10 mm once (18 mm, Table 4). It may also be observed in Table 4 that P25 is only 0)1 mm in June and 0 mm in July, August and September, and therefore, it is ‘normal’ for there to be no rainfall during these months. Monthly variation in precipitation within a single year is very frequently over 100% (Table 4, Fig. 5(a)). This, together with the considerable inter-annual variability (see above) implies that it is practically impossible to predict how much rain will fall in a month, although for July and August, even a very large relative error in the prediction would not be important. The dry and wet years are associated with an increase in dry and wet months. Figure 8(a) is similar in shape to Fig. 2(a), and regression analysis showed a significant correlation between annual rainfall and the number of wet months (Spearman’s r"0)628; a(0)001). The average of 36 rain-days per year found for Tabernas is well within the 10–50-day range suggested by Noy-Meir (1973) for arid zones. The more rain-days in a year or month, the greater the annual and monthly-scale rainfall volumes (both a(0)001) (Fig. 7). But the increase in volume is not due only to an increase in the number of rain-days, but to the increase in the amount of rainfall in each event (see extreme month and event analysis results). There is no general trend in the number of rain-days per month, which closely follows the variation in annual rainfall, except in some specific dry years (Fig. 7). It seems that rain-days, though having considerable inter-annual oscillation, tend to be constant, with shorter, more regular oscillation than annual rainfall. Intense rainstorms in some years can represent up to 30–50% of the annual rainfall in this area, although they are not as frequent as in other Mediterranean areas of Spain. THIRTY-YEARS RAINFALL IN SE SPAIN 391 Implications of rainfall for vegetation Even though there is no trend, rainfall fluctuation is important since the results of an ecological study could be different depending on the time at which the study is carried out. A longer climatic series is required to find out whether the short period recorded in a project is typical or anomalous. Both the lack of water throughout a year and the lack of water in autumn can be a limiting factor, whereas a lack of water in summer is practically irrelevant for plants (GutieH rrez, 2000). Thus, the ‘normal’ period from 1975 must be divided in two, before and after 1988. The majority of annual and monthly precipitation was below average (Figs 2(a) and 5(a)). This is important because plants are limited by dry years, and not by the average, which is an abstraction that cannot generate plant adaptation. Long inter-annual droughts cause serious problems for germination or survival during the first years of development in many species. Their success requires a continuous series of wet years, which promotes episodic recruitment, very frequent in arid and semiarid environments (Barbour, 1969), and resulting in the absence of some age classes. GutieH rrez (2000) has demonstrated that episodic recruitment does occur in the Tabernas area. On the other hand, when water is limiting, initial growth is often devoted to underground biomass that guarantees deep, wet soil layers will be reached (Went, 1948). Regardless of the amount, rain falling in rainy seasons (autumn and winter) will strongly influence vegetation and, in general, ecological processes. In semi-arid environments, these processes are frequently pulse-like (Noy-Meir, 1973) and associated with higher-than-threshold precipitation. Autumn is doubly important, because it is the rainiest season and because is the first wet season able to replenish soil water deficit accumulated after the long summer drought (PuigdefaH bregas et al., 1998). Some processes, such as erosion, runoff and floods are more probable during this season. Beatley (1974) indicated that biological success depends on autumn rainfall. The certainty of summer drought is a noteworthy feature of this climate that has important implications for floristic and life-form composition of the vegetation. The unpredictability as to how much it will rain in specific months generates some uncertainty in plant phenology. For example, in some years there may be up to two generations of annual plants in this area, as was the case in 1992 (LaH zaro, 1995). To understand the activity of vegetation it is important to distinguish in the month that rainfall volume falls, since different species or different life forms have different thresholds of previous rainfall for germinating or starting their annual growth cycles (Beatley, 1974; Ackerman, 1974; Turner & Randall, 1989; Bertiller et al., 1991). Moreover, due to the importance of vegetation density for rainfall interception (Domingo et al., 1998), erosion will be relatively greater if rainfall of sufficient volume and/or intensity occurs in September or October, due to the lower vegetation. The average number of rain-days is spatially highly variable, even on a local scale. In Cabo de Gata, only 40 km SSE from Tabernas, there were 17 rain-days (Neumann, 1961); 15 rain-days (with 251 mm year!1) in Zurgena, 45 km NE, and 40 in Rioja, 15 Km SSW from Tabernas (Capel-Molina, 1986). And, for example, the type of erosion process dominant is distributed according to the number of rain-days (Calvo & Harvey, 1996). Monthly rain-days do not properly indicate rainfall intensity, because it is too coarse a measurement (even if a specific day accounts for all the rainfall in a month, that day can still have a low intensity rainfall if it falls long enough). But the monthly rainfall/monthly rain-day ratio provides some information on the time of rainfall concentration (Sutherland et al., 1991), and explains some hydrological/erosion (see above) or biotic (such as lichen cover, see below) processes, because when the volume/rain-day ratio is low, it does not necessarily imply light intensity, but heavy intensity would probably be of very short duration. 392 R. LAD ZARO ET AL. A better approximation to rain intensity comes from monthly and annual maximum rainfall in one day. It is important to know the return periods in this environment in order to find the expected frequency of episodic processes with relevant consequences, such as generation of flash flooding, recharging of water tables or land degradation (Boer, 1999; PuigdefaH bregas et al., 1998). Extreme events also explain some biotic processes, such as certain paradoxical evapotranspiration/rainfall relationships: Domingo et al. (2001) show that these events, by recharging deep soil layers, allow shrubs like Retama sphaerocarpa at the bottom of dry ramblas, to achieve real evapotranspiration greater than the precipitation. In semi-arid regions, important effects on the ecosystem come not only from rainstorms. Small rainfall events play a selective effect on the plant life forms, e.g. favouring some grass species with a short response time to frequent rainfalls of less 5 mm (Sala & Lauenroth, 1982; 1985). Terricolous lichens, important in the Tabernas badlands, would also be differentially favoured by frequent light rainfalls, since they are highly sensitive to erosion (LaH zaro et al., 2000). A large proportion of light rainfalls are intercepted by the canopies (Domingo et al., 1998) and water infiltrates only slightly into the soil and is quickly evaporated (Domingo et al., 1999), but as lichen is often spatially segregated from higher plants (LaH zaro et al., 2000) and needs only a small amount of water for its activity (Lange et al., 1970), that is water sufficient. The volume and timing of rainfall control the types of plants that germinate or grow (Turner & Randall, 1989). Moreover, rainfall often has a ‘carryover’ effect on vegetation, increasing the seedlings (Webb et al., 1978), or inducing masting during the next growing season (Haase et al., 1995). Synthesizing the whole set of revised papers a more general idea appears, that vegetation is not only adapted to the amount of precipitation but also to its timing. All types of rainfall, in terms of volume, timing and intensity, would have consequences for the composition, density and distribution of vegetation. This work was carried out as part of the following research projects: MEDALUS III (Mediterranean Desertification and Land Use, III) collaborative project, funded by the EU under its European Environment Programme, contract no ENV4-CT95-0118; PROHIDRADE (Hydrological processes in fragile or degraded Mediterranean environments) co-operative project, funded by the Spanish ‘Plan Nacional de I#D’ (Environment), ref. AMB95-0986-C02-01; and by the Spanish project funded by the CICYT, ref. CLI95-1874. We would also like to thank the INM (Spanish Instituto Nacional de MeteorologmH a) which provided us with the climatic data and Dr Nicole Archer for her helpful comments on an early version of the manuscript. References Ackerman, T.L. (1979). Germination and survival of perennial plant species in the Mojave Desert. The Southwestern Naturalist, 24: 399–408. Ackerman, T.L. & Bamberg, S.A. (1974). Phenological studies in the Mojave Desert at Rock Valley (Nevada test site). In: Lieth, H. (Ed.), Phenology and Seasonality Modeling, pp. 215–226. New York: Springer-Verlag. 444 pp. Barbour, M.G. (1969). Age and space distribution of the desert shrub Larrea divaricata. Ecology, 50: 679–685. BaH rdossy, A. & Caspary, H.J. (1991). Conceptual model for the calculation of the regional hydrologic effects of climate change. Proceedings Symposium on Hydrology for the Water Management of Large River Basins. Vienna: IASO. pp. 114. Beatley, J.C. (1967). Survival of winter annuals in the Northern Mojave Desert. Ecology, 48: 745–750. Beatley, J.C. (1974). Phenological events and their environmental triggers in Mojave Desert ecosystems. Ecology, 55: 856–863. Bertiller, M.B., Beeskow, A.M. & Coronato, F. (1991). Seasonal environmental variation and plant phenology in arid Patagonia (Argentina). Journal of Arid Environments, 21: 1–11. THIRTY-YEARS RAINFALL IN SE SPAIN 393 Boer, M.M. (1999). Dryland degradation: site water balance and vegetation density as indicators of change in moisture-stressed environments. Ph.D. thesis. Department of Physical Geography, Utrecht University, Utrecht, The Netherlands. 291 pp. Brandt, C.J. & Thornes, J.B. (1996). Mediterranean Desertification and Land Use. Chichester, UK: Wiley & Sons. 554 pp. Calvo, A. & Harvey, A.M. (1996). Morphology and development of selected badlands in Southeast Spain: Implications of climatic change. Earth Surface Processes and Landforms, 21: 725–735. Capel-Molina, J.J. (1986). El clima de la provincia de AlmernH a (2nd Edn). AlmermH a: Publicaciones Caja Ahorros AlmermH a. 262 pp. Castillo-Requena, J.M. (1989). El Clima de AndalucnH a. AlmermH a: Instituto de Estudios Almerienses. 293 pp. Dalezios, N.R. & Bartzokas, A. (1995). Daily precipitation variability in semiarid agricultural regions in Macedonia, Greece. Hydrological Sciences — Journal des Sciences Hydrologiques, 40: 569–585. Domingo, F., SaH nchez, G., Moro, M.J., Brenner, A.J. & PuigdefaH bregas, J. (1998). Measurement and modelling of rainfall interception by three semi-arid canopies. Agricultural and Forest Meteorology, 91: 275–292. Domingo, F., VillagarcmH a, L., Brenner, A.J. & PuigdefaH bregas, J. (1999). Evapotranspiration model for semi-arid shrub-lands tested against data from SE Spain. Agricultural and Forest Meteorology, 95: 67–84. Domingo, F., VillagarcmH a, L., Boer, M.M., Alados-Arboledas, L. & PuigdefaH bregas, J. (2001). Evaluating the long-term water balance of arid zone stream bed vegetation using evapotranspiration modelling and hillslope runoff measurements. Journal of Hydrology, 243: 17}30. Dunne, T. & Leopold, L.B. (1978). Water in Environmental Planning. San Francisco, California: W.H. Freeman. 818 pp. Esteban-Parra, M.J., Rodrigo, F.S. & Castro-DmH ez, Y. (1997). Estudio de variaciones climaH ticas en AlmermH a. In: GarcmH a-Rosell, L. & Navarro-Flores, A. (Eds), Recursos Naturales y Medio Ambiente en el Sureste Peninsular, pp. 489–501. AlmermH a: Instituto de Estudios Almerienses and Ayuntamiento de Cuevas de Almanzora. Esteban-Parra, M.J., Rodrigo, F.S. & Castro-DmH ez, Y. (1998). Spatial and temporal patterns of precipitation in Spain for the period 1880-1992. International Journal of Climatology, 18: 1557–1574. FaragoH , T. & Karz, R.W. (1990). Extreme and Design Values in Climatology. Geneva: WMO/TDNo. 386. 78 pp. Floret, C., Pontanier, R. & Rambal, S. (1982). Measurement and modelling of primary production and water use in a south Tunisian steppe. Journal of Arid Environments, 5: 77–90. GarcmH a de Pedraza, L. & GarcmH a-Vega, C. (1989). La sequnH a y el clima de Espan8 a. Calendario MeteoroloH gico, 188–198 pp. Madrid: Instituto Nacional de MeteorologmH a. 223 pp. Geiger, F. (1973). El Sureste espan8 ol y los problemas de la aridez. Revista de Geograf nH a, 7(1–2): 166–209. Goossens, C.H. & Berger, A. (1986). Annual and seasonal climatic variation over the Northern Hemisphere and Europe during the last century. Annales Geophysicae, 4: 385–400. GutieH rrez, L. (2000). Estructura y productividad de la vegetacioH n de estepa mediterraH nea semiaH rida en relacioH n con la variabilidad climaH tica: El sistema de ladera en Rambla Honda (AlmermH a). Ph.D. thesis. University of AlmermH a, Spain. 242 pp. Haase, P., Pugnaire, F.I. & Incoll, L.D. (1995). Seed production and dispersal in the semi-arid tussock grass Stipa tenacissima L. during masting. Journal of Arid Environments, 31: 55–65. Jones, P. (1995). The instrumental record: its accuracy and use in attempts to identify the ‘CO2 signal’. In: Von Storch, H. & Navarra, A. (Eds), Analysis of Climate Variability. Applications of Statistical Techniques, pp. 54–75. Berlin: Springer. 334 pp. Lane, L.J., Romney, E.M. & Hakonson, T.E. (1984). Water balance calculations and net production of perennial vegetation in the Northern Mojave Desert. Journal of Range Management, 37: 12–18. Lange, O.L., Schulze, E.D. & Koch, W. (1970). Experimentell-oK kologische Untersuchunge an Flechten der Negev-WuK ste. II. CO2 Gaswechsel und Wasserhaushalt von Ramalina maciformis (Del.) Bory am natuK rlichen Standort waK hrend der sommerlichen Trockenperiode. Flora, 159: 38–62. LaH zaro, R. (1995). Relaciones entre vegetacioH n y geomorfologmH a en el aH rea acarcavada del Desierto de Tabernas. Ph.D. thesis. University of Valencia, Spain. 241 pp.#attachments. 394 R. LAD ZARO ET AL. LaH zaro, R. & Rey, J.M. (1991). Sobre el clima de la provincia de AlmermH a (SE IbeH rico): primer ensayo de cartografmH a automaH tica de medias anuales de temperatura y precipitacioH n. Suelo y planta, 1: 61–68. LaH zaro, R., Alexander, R.W. & PuigdefaH bregas, J. (2000). Cover distribution patterns of lichens, annuals and shrubs in the Tabernas Desert, AlmermH a, Spain. In: Alexander, R.W. and Millington, A.C. (Eds), Vegetation Mapping: from Patch to Planet, pp. 19}40. Chichester: John Wiley & Sons. 350 pp. Le HoueH rou, H.N. (1984). Rain use efficiency: a unifying concept in arid-land ecology. Journal of Arid Environments, 7: 213–247. Le HoueH rou, H.N. (1988). Relationship between the variability of primary production and the variability of annual precipitation in world arid lands. Journal of Arid Environments, 15: 1–18. Lebart, L., Morineau, A. & FeH nelon, J.P. (1985). Tratamiento estadnH stico de datos. Madrid: Marcombo. 520 pp. Lieth, H. (1976). The use of correlation models to predict primary productivity from precipitation or evapotranspiration. In: Lange, O.L., Kappen, L. & Schulze, E.-D. (Eds), Water and Plant Life: Problems and Modern Approaches, pp. 392–407. Berlin: Springer-Verlag. 536 pp. Mitchell, J.M. (1966). Climatic Change. Geneva: WMO Tech. Note 79. 79 p. Mulligan, M. (1998). Modelling the geomorphological impact of climatic variability and extreme events in a semi-arid environment. Geomorphology, 24: 59–78. Naveh, Z. (1982). The dependence of the productivity of a semi-arid Mediterranean hill pasture ecosystem on climatic fluctuations. Agriculture and Environment, 7: 47–61. Neumann, H. (1961). El clima del sudeste de Espan8 a. Estudios GeograH ficos, 21 (79): 171–209. Noy-Meir, I. (1973). Desert ecosystems: environment and producers. Annual Review of Ecology and Systematics, 4: 25–51. Palutikof, J.P., Conte, M., Casimiro Mendes, J., Goodess C.M., & Espiritu Santo, F. (1996). Climate and Climatic Change. In: Brandt, C.J. & Thornes J.B. (Eds), Mediterranean Desertification and Land Use, pp. 43–86. Chichester: John Wiley & Sons. 547 pp. Peinado-Serna, A. (1985). Lecciones de climatolognH a. Madrid: Instituto Nacional de MeteorologmH a. 111 pp. Pickup, G. & Stafford Smith, D.M. (1993). Problems, prospects and procedures for assessing the sustainability of pastoral land management in arid Australia. Journal of Biogeography, 20: 471–487. PuigdefaH bregas, J. (1998). Variabilidad climaH tica y sus consecuencias sobre la sostenibilidad de los sistemas agrarios. In: Jimenez, R.M. & Lamo, J. (Eds), Agricultura Sostenible, pp. 41–70. Madrid: Agrofuturo, Life & Mundi-Prensa. 616 pp. PuigdefaH bregas, J., del Barrio, G., Boer, M., GutieH rrez, L. & SoleH , A. (1998). Differential responses of hillslope and channel elements to rainfall events in a semi-arid area. Geomorphology, 23: 337–351. RodrmH guez-Puebla, C., Encinas, A.H., Nieto, S. & Garmendia, J. (1998). Spatial and temporal patterns of annual precipitation variability over the Iberian Peninsula. International Journal of Climatology, 18: 299–316. Rutherford, M.C. (1980). Annual plant production-precipitation relations in arid and semi-arid regions. South African Journal of Science, 76: 53–56. Sala, O.E. & Lauenroth, W.K. (1982). Small rainfall events: an ecological role in semiarid regions Oecologia, 53: 301–304. Sala, O.E. & Lauenroth, W.K. (1985). Root profiles and the ecological effect of light rainshowers in arid and semiarid regions. The American Midland Naturalist, 114: 406–408. Sala, O.E., Parton, W.J., Joyce, L.A. & Lauenroth, W.K. (1988). Primary production of the central grassland region of the United States. Ecology, 69(1): 40–45. Shaw, E.M. (1983). Hydrology in Practice. Berkshire, England: Van Nostrand Reinhold (U.K.) Co. 559 pp. Silvertown, J., Dodd, M.E., McConway, K., Potts, J. & Crawley, M. (1994). Rainfall, biomass variation, and community composition in the park grass experiment. Ecology, 75(8): 2430–2437. Sneyers, R. (1990). On the Statistical Analysis of Series of Observations. Geneva: WMO Tech. Note 143. 192 pp. Sneyers, R. (1992). On the use of statistical analysis for the objective determination of climatic change. Meterorol. Zeitschrift. 1: 247–256. Specht, R.L. & Specht, A. (1989). Canopy structure in Eucalyptus-dominated communities in Australia along climatic gradients. Acta Oecologica, 19: 191–213. THIRTY-YEARS RAINFALL IN SE SPAIN 395 Sutherland, R.A., Bryan, R.B. & Wijendes, D.O. (1991). Analysis of the monthly and annual rainfall climate in a semi-arid environment, Kenya. Journal of Arid Environments, 20: 257–275. TuK rkes, M. (1996). Spatial and temporal analysis of annual rainfall variations in Turkey. International Journal of Climatology, 16: 1057–1076. Turner, F.B. & Randall, D.C. (1989). Net production by shrubs and winter annuals in Southern Nevada. Journal of Arid Environments, 17: 23–36. Webb, W., Szarek, S., Lauenroth, W., Kinerson, R. & Smith, M. (1978). Primary productivity and water use in native forest, grassland, and desert ecosystems. Ecology, 59: 1239–1247. Went, F.W. (1948). Ecology of desert plants. I. Observations on germination in the Joshua Tree National Monument, California. Ecology, 29: 242–253. Woodell, R.J., Mooney, H.A. & Hill, A.J. (1969). The behaviour of Larrea divaricata (creosote bush) in response to rainfall in California. Journal of Ecology, 57: 37–44. Woodward, F.I. (1987). Climate and Plant Distribution. Cambridge: Cambridge University Press. 174 pp.