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.