Journal of Hydrology (2008) 356, 119– 130

available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/jhydrol

Analyzing temporal changes in climate erosivity
using a simplified rainfall erosivity model
in Basilicata (southern Italy)
D. Capolongo a, N. Diodato b, C.M. Mannaerts
R.O. Strobl c,d

c,*

, M. Piccarreta a,

a

` di Bari, Via Orabona 4, Bari, Italy
Dipartimento di Geologia e Geofisica, Universita
Monte Pino Research Observatory, TEMS Network (FAO, United Nations), 82100 Benevento, Italy
c
Department of Water Resources, International Institute of Geo Information Sciences and Earth Observation,
ITC, P.O. Box 6, 7500 AA Enschede, Netherlands
d
Institute of Environment and Sustainability, European Commission Joint Research Centre, 21202 Ispra, Italy
b

Received 30 May 2007; received in revised form 14 March 2008; accepted 3 April 2008

KEYWORDS
Rainfall;
Erosivity;
Time trends;
Spatial variation;
Southern Italy

Summary The objective of this work was to investigate the magnitude, frequency and
trends of rainfall erosivity in Basilicata (southern Italy), as an initial step in the study of
the consequences of climatic change on soil environments in the central Mediterranean
region. First, we derived and calibrated an equation to predict the RUSLE rainfall erosivity
(EI30) from daily rainfall, using a continuous 5-year series of 549 daily rainfall events,
recorded in 20-min time intervals at five stations located across the region. After validation, this rainfall erosivity model was then applied to the long term daily rainfall series of
53 gauging stations, distributed over the Basilicata territory, yielding time series of annual
and seasonal rainfall erosivity for the period 1951–2000. The Mann–Kendall non-parametric test statistic was used to detect time trends in the rainfall erosivity time series. Results
indicated that more than half of the stations did not show a statistical trend. Four stations
showed an upward trend and 12 a downward trend in seasonal and annual erosivity, the
trends becoming stronger during the last 30-year normal (1971–2000), when compared
to the overall 50-year period. This analysis of rainfall erosion index values and its regional
variability clearly showed the strong local spatial and temporal variation of erosivity in
this Mediterranean region. The results further indicated that the decrease in annual rainfall depths observed in a number of locations is compensated by an increase in single

* Corresponding author. Tel.: +31 53 4874210; fax: +31 53 4874336.
E-mail addresses: capolongo@geo.uniba.it (D. Capolongo), nazdiod@tin.it (N. Diodato), mannaerts@itc.nl (C.M. Mannaerts),
marcopiccarreta@geo.uniba.it (M. Piccarreta), robert.strobl@jrc.it (R.O. Strobl).
0022-1694/$ - see front matter ª 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2008.04.002

mudflows. In this respect. and letters indicate stations with short duration data used in model calibration and validation (B: Basento Freatimetro. . such as soil erosion. Capolongo et al. Numbers refer to station list in Table 1.. P: Potenza and S: Stigliano). ª 2008 Elsevier B. induce an erratic pattern in rainfall and its associated erosivity.V. I: Irsina. especially in a changing climate. All rights reserved. In particular the erosive power of climate depends on how rainfall and runoff are able to produce ‘‘geomorphic work’’. such as removing topsoil. Introduction Weather and climate extremes. reworking land surface materials and modifying the Figure 1 Geographical location of the SAL (Servizio Agrometeorologico Lucano) and SI (Servizio Idrologico Nazionale) meteorological stations in Basilicata region.120 D. which could have a higher impact on the environment than the more-often-cited risk of global warming (Sauerborn et al. L: Lagonegro. 1999. storm rainfall intensity to produce near stationary long term values of annual rainfall erosivity in the region. flash floods and leaching. 2002). Allen and Ingram. rainfall erosivity constitutes an important factor for understanding the multitude of hydrological and geomorphologic processes taking place in a landscape.

47 530.33 9.00 763. their mean rainfall erosivity values.98 603.06 873.73 1005.02 981.23 1480.63 3074.07 1340.43 3716.80 685.69 627.56 585.06 10.63 651.24 627.47 1453.14 681.19 1233.60 787.47 1007.67 4526.58 948.36 714.14 12.98 7.53 1032.78 683.37 12.49 705.96 523.41 635.32 990.49 9.33 596.16 1356.90 646.27 886.51 544.94 9.88 634.89 748.78 962.83 880.83 949. climate normals.16 826.92 8.33 1266.98 595.06 11.36 1355.99 666.76 929.65 765.69 1077.03 630.69 12.08 999.48 1094.26 1987.04 888.74 9.12 8.22 1618.88 2115.45 767.19 1254.09 7.53 7.45 9.79 963.75 614.95 1710.18 629.39 1095.87 640.04 1488.00 684.54 940.84 904.05 697.74 711.33 574.3 1471.50 875.74 917.52 686.68 9.62 838.94 665.88 1921.33 3455.44 408.16 1940.27 635.93 602.77 492.93 496.00 1030.34 447.01 684.68 986.03 405.87 768.99 7.62 1005.66 1976.06 556.92 1358.90 687.58 1353.78 1019.02 1290.63 586.50 3418.49 622.01 601.95 1358.76 585.81 992.32 8.98 1915.87 10.95 607.32 1046.08 723.99 1102.88 1447.61 1099.35 889.72 1032.57 942.45 744.30 561.73 8.47 6.83 830.55 11.64 918.45 700.91 799.68 811.39 788.46 9.84 714.79 842.47 403.15 646.42 828.82 1701.29 8.75 11.87 983.19 999.54 367.17 990.25 865.10 809.63 567.55 10.08 3531.04 4161.56 1427.02 2790.71 755.73 11.38 913.65 1715.86 651.96 10.38 2010.11 795.12 697.79 611.66 1360.96 653.21 8.52 1034.52 636.53 6.64 2243.43 7.35 10.82 850.25 7.01 1358.77 8.96 1168.32 915.72 3676.62 1644.26 733.20 718.86 810.80 757.17 476.31 1150.16 572.28 1078.01 2347.97 4090.32 1014.35 849.47 1157.86 549.53 883.16 479.07 11.39 1208.95 17.76 885.45 1635.34 1245.35 600.89 1799.26 924.22 8.06 748.33 1978.31 910.04 1324.13 1132.59 594.29 10.53 1017.10 8.53 1346.12 1981.29 1146.34 1099.37 902.82 1033.95 466.25 875.70 608.59 1487.92 722.14 3083.65 965.14 604.45 2462.27 8.11 1737.13 805.02 898.91 814.95 722.84 526.96 7.46 621.89 676.2 1877.30 858.47 3491.66 1256.22 1143.50 1019.21 1027.67 622.86 1130.53 964.14 907.16 715.34 1055.10 7.73 548.40 1247.02 1668. total annual precipitation and simple daily intensity from 1951 to 2000 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 Station Erosivity (1951–2000) (MJ mm h1 ha1) Normal 1951–1980 Normal 1961–1990 Normal 1971–2000 PRCPTOT (1951–2000) (mm) SDII (1951–2000) (mm/rainday) Acquafredda Agromonte Aliano Armento Atella Avigliano Balvano Bernalda Calvello Cancellara Carbone Castel Lagopesole Cersosimo Cognato Forenza Grassano Grottole Irsina Lagonegro Lauria Inferiore Lavello Maratea Matera Melfi Missanello Monticchio Bagni Noepoli Nova Siri Nova Siri Scalo Pescopagno Picerno Pisticci Policoro Pomarico Potenza Ripacandida Roccanova San Fele San Giorgio Lucano San Martino d’Agri San Mauro Forte San Nicola d’Avigliano San Severiao Lucano Senise Stigliano Tolve Tramutola Trecchina Tricarico Tursi Vaglio di Lucania Valsinni Venosa 1775.13 751.50 9.62 657.55 1115.90 1839.39 843.07 540.10 592.57 1109.95 793.69 11.62 671.65 611.96 870.30 699.74 1168.43 .99 847.06 1212.68 628.50 8.22 1582.72 15.74 7.23 1051.11 1496.41 10.46 1251.25 9.99 478.43 15.14 2366.71 584.69 953.02 1579.34 578.18 8.62 829.67 916.43 8.17 648.89 7.Analyzing temporal changes in climate erosivity using a simplified rainfall erosivity model 121 Table 1 List of the Servizio Idrologico Nazionale (SI) stations.91 692.55 1256.85 665.94 1755.84 588.52 831.41 1430.

1991. Data Figure 2 Daily variation in f(daily) function for a typically rainy year in Italy. Mean or average values of water erosion indices derived from standard rainfall data. (1) or EI = aPb. as typically used in a modified Fournier index analysis (Arnoldus. Torri et al. Therefore the two main variables affecting rainfall erosivity show opposite trends and this implies that space or time trends of erosivity cannot be deduced indirectly.e. 1994). To overcome this problem. 1983). 1987). A 5-year continuous record length (60 months from 2001 to 2005) consisting of 549 recorded storms recorded on a 20-min interval basis was used in the analysis. rainfall erosivity or R-factor of the Universal Soil Loss Equation or USLE (Wischmeier and Smith. Yu et al. The newly derived daily erosivity rainfall index was then subsequently applied to the 50-year (1951–2000) daily record series of 53 rainfall gauging stations in the Basilicata region. As shown in Fig. 2001). kinetic energy and maximum rainfall intensity and erosivity was investigated using 20-min rainfall data from five meteorological stations. Bagarello and D’Asaro (1994) tested the following power relationship: EI ¼ a  P b ð1Þ P is the daily rainfall amount for all events with P > 13 mm day1. Foster. 1978). 2004). a further attempt was made to quantify principally the rainfall erosivity temporal pattern in the Basilicata region (southern Italy) using a more complex form of Eq. II). 1960). the most known and widely used parameter to predict the erosive potential of raindrop impact and to reflect the amount and rate of runoff generated by erosive storms is the rainfall erosion index.2 (reference in Bagarello and D’Asaro.5 to 2. 1994). Besides the development of a new rainfall erosion index equation applicable to the region.. Capolongo et al. whereas empirical approaches gave values ranging from 1.54. 1978) to derive the Rfactor of the RUSLE. in order to generate and evaluate extended historical time series of annual rainfall erosivity for the whole region. . The aim was also to detect and explain possible time trends in rainfall erosivity. 2004) have pointed out a decreasing trend of precipitation amounts and a weak increasing trend of rainfall intensity. the second objective of this paper was to analyze the annual rainfall series and rainfall intensity time trends from 1951 to 2000. Today.. Petrovsˇek and Mikosˇ (2004). 1997. while the exponent b could be considered as a process parameter and is relatively more constant.. no missing data... Bagarello and D’Asaro (1994) derived a single value for Sicily equal to 1. A preliminary inspection revealed a similar rainfall pattern although with differences in magnitude. Several studies about the rainfall regime in Southern Italy (Brunetti et al. which usually shows both temporal and spatial variability (Richardson et al. The relationship between daily rainfall variables. 2005).. 2003). Lagonegro. One of the main factors affecting the erosion intensity or potential erosion on slopes is related to the frequency and different magnitudes of rainfall events in combination with geomorphic thresholds (Baker. Short duration rainfall intensity data are obtained today by either digital pluviographs or tipping-bucket technology (discrete rainfall rates) but. 1977). Potenza and Stigliano) were selected based on their location and coverage of the whole region (Fig.122 D. However. different indirect models have been proposed for estimating rainfall erosivity at longer time scales (Renard and Freimund. 1) as well as on the quality of the data records i. The only attempts to a more resolute and compatible RUSLE erosivity-factor application in the Mediterranean area were made by Diodato (2004) for individual yearly time step. a similar approach as proposed by Mannaerts and Gabriels (2000b) for the Cape Verde islands has been used. The value of the exponent b. Irsina. A special analysis was carried out by Bagarello and D’Asaro (1994) on a daily time scale. 1983). In this study. and no data inconsistencies were found. The five stations (Basento Freatimetro. extending it to account for seasonal variations in rainfall intensity (Davison et al. erosivity is used as an integral part of the Revised Universal Soil Loss Equation (RUSLE) (Renard et al. application of such models to provide annual or long term erosion hazard estimates for a site depend on knowledge of hourly or sub-hourly distribution of rainfall intensities.. obtained by a theoretically based approach approximates 2 (Brown and Foster. 2002. 1.. the full database was split into two data sets (I. 1977) or in simple relationships of proportionality to rainfall amount (Grimm et al. Using data from 32 Sicilian locations. in order to evaluate which variables mostly affect the rainfall erosivity in this Mediterranean region. 2001. and by Loureiro and Couthino (2001). Data and methods Rainfall data recorded with a temporal resolution of 20-min were obtained from five electronic rain-gauge stations of the Italian National Hydrographic Service operating in Basilicata. In particular. 2006). a can be interpreted as a scale factor. in many parts of the world (including the Mediterranean area) records of this type are still limited in time (Yu et al... (2001) and Diodato (2005) for a monthly time step. surface features of the landscape (Wolman and Miller. For the estimation of the daily rainfall erosivity. Recorded rainfall intensities also showed similar behaviour for all five stations. In the literature. are not a good predictor of erosivity (De Ploey et al. and a and b are the Richardson’s coefficients (Richardson et al. This method is considered to be more favourable to our region than the above-mentioned procedures in that it can utilize shorter monitoring data time scales than those commonly prescribed (Wischmeier and Smith.

Data set II (149 records) was used for evaluating model estimates (validation data set). Figure 4 Scatterplots between estimated (EIdaily) and measured rainfall erosivity (EI30-RUSLE) values for the calibration (a). along with the local station names and numbering given in Table 1. 1. a daily rainfall series of 53 stations distributed across the Basilicata region from 1951 to 2000 was used. Finally. The distribution of the stations can be found in Fig.Analyzing temporal changes in climate erosivity using a simplified rainfall erosivity model set I (400 records) was used to determine model parameters (calibration data sets). The calibration and validation data sets were selected in such a way to represent different geographical characteristics of the study area. and validation (b) data set. . These station network data were monitored by the former Italian National Hydrographic Service (SI). Daily rainfall amounts versus kinetic energy and regression line fit. for evaluating the long term tempo- Figure 3 123 ral pattern.

Several empirical laws have been proposed to link rainfall kinetic energy to its intensity.93** 1.40 1.36 3. 1998): E ¼ 0:29½1  0:72eð0:05Ir Þ  ð2Þ where E denotes the rainfall kinetic energy and has units of Mega joule per hectare per millimetre of rain or MJ ha1 mm1.40* 3. where a and b are coefficients estimated for a convergent solution of Eq. found to be well approximated by a cosine function based on the Julian day number (J).90 4.61** 1.25 3.87** 3.80 2.76** 2.35* 2.28 1.14 1.29 0.29** 0.06 2.45** 2.58 0.19* 1.45** 2.17* 3.78 0.31** 2.48** 0.28 Bold numbers indicate a significant trend and the asterix (*).31 0.25** 0.21 1.10* 2.47* 2.49 2.88 2.26 2.80 0.03* 0. and f(h) is to modulate hourly rainfall intensity throughout the year.77** 0.49 2.87** 1. the threshold daily rainfall amount.29 0.35** 0.76 0. Wischmeier and Smith.41 3.46 1.58** 0.01 0.11 2. Rainfall erosivity calculation procedure Quantitative evaluation of rainstorm erosivity (EI30) requires rainfall kinetic energy and intensity values associated to 30-min observation time intervals. Cerro et al.28 0.63 0. 1980).27 3. Afterwards.29 0.98 1.45 3.66 2. equivalent to five times the long term annual erosivity was detected.17* 1. Subsequently. .61 0.96 1.17* 1.15 0.15* 3.98 2.63 1.48 1.20 2.07** 1.06 0.50 1. expressed in MJ ha1 mm1.45** 4. and Ir is rainfall intensity in mm h1. (3). One rainstorm with an unusually high I30 intensity (88 mm h1) and with a disproportional erosivity amount of 3276 MJ mm ha1 h1 on a single day.64 1. This equation type was tested and found adequate for Mediterranean and Southern European conditions by several authors (Coutinho and Tomas.24** 0. Zanchi and Torri.41 1. 2005): pffiffiffiffiffi    J  c Pk fðhÞ ¼ a þ b  cos 2p þg ð4Þ 365 in which a.74 1.58 0.87** 1. A thorough overview and analysis of the functional relationship between kinetic energy of rain and intensity can be found in Salles et al.08 1.52 0.51 0.30** 2.58 1.70 2.0 mm.45* 1.49** 0.83** 1.45** 0.33 0.11 2.32 0.09** 4.50 2. the maximum 30-min rainfall interval was then selected for each rainfall day or event and used in the EI30 evaluation. mean yearly precipitation and simple daily intensity for all stations of Basilicata over the period 1951–2000 Acquafredda Agromonte Aliano Armento Atella Avigliano Balvano Bernalda Calvello Cancellara Carbone Castel Lagopesole Cersosimo Cognato Forenza Grassano Grottole Irsina Lagonegro Lauria Inferiore Lavello Maratea Matera Melfi Missanello Monticchio Bagni Noepoli Nova Siri Nova Siri Scalo Pescopagano Picerno Pisticci Policoro Pomarico Potenza Ripacandida Roccanova San Fele San Giorgio Lucano San Martino d’Agri San Mauro Forte San Nicola d’Avigliano San Severino Lucano Senise Stigliano Tolve Tramutola Trecchina Tricarico Tursi Vaglio di Lucania Valsmni Venosa Erosivity PRCPTOT SDII 1.34* 0.52* 0.98** 1. 1995.79** 2. the more recent equation of Brown and Foster (1987).77** 0. 1997) was adopted to derive the kinetic energies of all the rainfall events. Capolongo et al.57** 2.18 5.53 2.38 3..18* 0. used within the RUSLE erosion modelling framework (Renard et al.58** 1.52* 4.65** 0. Development and evaluation of a daily rainfall erosivity model The resulting EI30-value and respective rainfall quantities were summarised to daily totals.49** 1... An Excel spreadsheet was constructed for this purpose.63 0.35 1.36 1. From the various relationships found in the literature (e.47** 1.26 2.40 2.75 1.59 0.72** 2.13 2. In a preliminary and exploratory data analysis.52 1.95 2.16** 3.63 2. 1978.22* 3.43 0.g.56 3.13* 0. This value was therefore removed from the calibration dataset regression equation.53 2.32** 1.19 1.12* 3. A linear interpolation was employed to derive 30-min rainfall values from the continuous 20-min observation time series.73 0. g are coefficients estimated using an iterative processes according to cyclical observations.75 3. (2002). and in MJ ha1 in case of an individual rainfall event.58 0.49 0.83 3. c. a close positive non-linear relationship between daily rainfall and daily EI30 was found for the given data set and could be well described by a power law of the form: EIdaily ¼ a  ðPk  fðhÞ Þb ð3Þ when Pk > P0 with P0 = 10.68 1. In this Table 2 Results of the Mann–Kendall test statistic to the mean yearly rainfall erosivity.69** 0.124 D.80** 1.23 2.53 3.61 0. b.20 0.19** 0.71** 0.61** 3.48 0.94** 3.95 3.41 2.90 1.47* 5. one aberrant value for 11 November 2004 was detected for the rain-gauge station at Basento.43 0.53* 2.97* 0.03 1. the confidence interval (* 95% and ** 99%).24** 2.60’’ 0. as follows (Davison et al.51 0.03 1.33 1.49 4. The kinetic energy is usually expressed as the amount of energy expended per unit area and per unit depth of rain.77** 1.33* 3.09 0.65 2.

Krause et al. responding to different aspects of model behaviour (Mulligan and Wainwright. because only the dispersion is quantified by r2. However. 1999. The Nash and Sutcliffe (1970) model-efficiency (ModEff) measure is most recommended for this purpose. 1. Fig. when the majority of intense storms occur in Italy. (2005) suggest not to use the coefficient of determination or (r2). Krause et al.0 is in agreement with the theory and between experimental values as suggested by Richardson et al. (Erosivity expressed in MJ ha1 mm h1. As a result. 2004). the f(h) function is compatible with higher kinetic energy (KE) per mm of rain throughout the late spring. when adopting a common threshold of 10 mm for the daily rainfall depth. Total monthly and annual erosivity is obtained as the sum of all single daily rainfall erosivity values in a year as EIannual ¼ J X ðEIdaily Þj 125 ð6Þ j¼1 Model evaluation requires some measurements of how well the modelled system values (Mi) represent observed system values (Oi).. 1991). and Bagarello and D’Asaro (1994). larger values in a time series are strongly overestimated whereas lower values are neglected (Legates and McCabe. annual precipitation (b) and simple daily rainfall intensity (c) in Basilicata for the period 1951–2000.40–2. with values ranging from approximately 6 in winter to 12 in July. Figure 6 Annual rainfall erosivity time series and 5-year moving average (in bold) for Carbone (a).81. 2 shows the daily variation of f(h). too. (1983). It also reacts to peak values.) . Missanello (b) and Vaglio di Lucania (c) stations from 1951 to 2000. To solve this gap. 1. in the form: ModEff ¼ MedjOi  Oj  MedjMi  Oi j MedjOi  Oj ð7Þ Figure 5 Spatial distribution of time trends in observed long term annual rainfall erosivity (a). the largest disadvantage of the efficiency is the fact that the differences between the observed and predicted values are calculated as squared values. 2005). It is recognized that there are different goodness-of-fit measures. and this represents one of the major drawbacks of this coefficient when it is considered alone. The complete and explanatory equation to predict daily total erosivity (EI daily in MJ mm ha1 h1) as a function of daily total rainfall (Pk) and of its simulated daily intensity (f(h)) becomes therefore EIdaily ¼ 0:000900 pffiffiffiffiffi     2 J  3 Pk þ 3:0  Pk  9:0 þ 3:0  cos 2p 365 ð5Þ The value of the exponent b = 2.20. that is equally sensitive to both low and high values. (1996). and with less intense but more prolonged rainfall events in the remaining periods of the year (Bazzoffi and Pellegrini. summer and early autumn.Analyzing temporal changes in climate erosivity using a simplified rainfall erosivity model way. a modified non-parametric Nash–Sutcliffe efficiency was suggested by Zacharias et al.

(Erosivity expressed in MJ ha1 mm h1. According to the World Meteorological Organization (WMO). Grassano (d). then the ModEff = 0. When a ModEff > 0. Bernalda (b). Pomarico (j).126 The range of ModEff lies between –1 and 1. If the error is of the same magnitude as the observed variance. the erosivity equation was applied to the daily series of 53 rainfall gauging stations distributed over the whole region (Fig. 1). Maratea (f). Cancellara (c).5 (O is achieved. The daily series were homogenized using the MASH v.) . and the observed mean values  i ) are well represented by the model. For every year and for each station. 1. Matera (g). the total precipitation amount (PRCPTOT in mm/year) and the simple daily rainfall intensity index (SDII in mm/rainday) were also calculated. Application of the daily rainfall erosivity model to the long term daily precipitation time series of the Basilicata region After deriving a daily rainfall erosion index equation from the short duration rainfall intensity data of 549 daily events recorded at the three selected sample stations. The precipitation data analyzed in this paper were collected from the former National Hydrographic Service (SI) and cover a time span of 50 years (1951–2000).. Capolongo et al. single and multiple linear regressions were performed using the available data to cover the data gaps (Piccarreta et al. (5) and (6) to the daily time series. The annual rainfall erosivity for every rainfall station was then calculated by applying Eqs. When missing data were present.1 software (Szentimrey.0 (perfect fit). 2004). Lauria Inferiore (e).0. for the period 1951–2000 (Table 1). Pisticci (i). Nepoli (h). 1999. 2000) in order to eliminate natural non-homogeneities. then the model can be considered to be performing satisfactorily. the normal precipitation relative to a given station can be assumed as the mean of the annual precipitation relative to 30 Figure 7 Annual rainfall erosivity time series and 5-year moving average (in bold) for Balvano (a). Stigliano (k) and Trecchina (l) stations from 1951 to 2000. D.

the kinetic energy of rainfall. corresponds to quantitative values. given by EðtÞ ¼ l ¼ nðn  1Þ 4 ð8Þ varðtÞ ¼ r2 ¼ nðn  1Þð2n þ 5Þ 72 ð9Þ The standard normal distribution u(t) is given by ½t  EðtÞ uðtÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi varðtÞ ð10Þ The distribution of the test statistic u(t) is compared with a standard normal distribution at the certain level of significance. 1990). 3. for each element xi (i = 1–n).. precipitation amount and simple daily intensity index. which is usually a = 0. In the absence of any trend (H0 or null hypothesis). registered at the five investigated stations. Results As shown in Fig. observed in other parts of the world e. The no trend or H0 hypothesis is rejected for high values of |u(t)| > 1. The Mann–Kendall rank correlation test statistic is given by t = Ri(ni). the distribution function of t can be assumed to be asymptotically Gaussian. For all the parameters considered. 1961–1990 and 1971–2000.05. the Mann–Kendall non-parametric test. In the Mann–Kendall test. annual erosivity. the series under study has been subdivided into three normals: 1951–1980.Analyzing temporal changes in climate erosivity using a simplified rainfall erosivity model 127 Figure 7 (continued) years. These 30-year normals were calculated in order to analyze climate erosivity anomalies. the number of lower elements xj (xj < xi) preceding it (j < i) is calculated. the Philippines (White. 1988) and Cape Verde . with the expected value of E(t) and variance var(t). Accordingly. of all the 549 precipitation events.96. was applied in order to detect possible trends. as described in Helsel and Hirsch (1992).g.. Zambia (Pauwelyn et al.

estimated from the data. Missanello (Fig. in order to verify time trends during the observation period.5 in. The coefficient of determination of r2 = 0. Missanello. The Mann–Kendall test statistic was applied to the erosivity and precipitation variables for all the investigated rainfall stations. The RUSLE erosivity or the EI30-RUSLE value of every erosive rainfall event was then evaluated by multiplying the rainfall kinetic energy E with the maximum 30-min rainfall intensity I30 for all the storms. 1975 and 1993. Fig. (Mannaerts and Gabriels. Capolongo et al.. An analysis Long-term mean annual rainfall erosivity (isoerodent) map of Basilicata with erosivity expressed in MJ ha1 mm h1.55 of the validation dataset is lower than that for the calibration data set (r2 = 0. The EI30RUSLE erosivity values were then used to verify the daily erosivity model. Carbone (Fig. or 12. 5). 6) of the three stations showing an upward erosivity trend is coherent with the values of the different climate normals (Table 1). 2000a. More than half of the stations do not show statistically significant trends in annual erosivity. 4 shows measured versus estimated daily erosivity for both calibration and validation data sets. match the EIdaily  EI30 (RUSLE) data for the calibration data set.7 mm (Renard et al. apart from interannual differences. Fig. Change points were derived by visual inspection of the homogenized erosivity time series.824) and validation (ModEff = 0. 6c) three change points are detectable respectively in 1966. (4) or (5) was judged to be adequate.67). 1997). 6b) shows two principal change points (1968 and 1987). confirming the good performance of the model when estimating more than 99% of the erosivity distribution. However.748) tests. which was correlated to a statistically significant upward trend in daily rainfall intensity (Fig. being close to the commonly accepted threshold for erosive rainfall of 0. Vaglio di Lucania) out of 53 showed an upward trend in rainfall erosivity. 6 furthermore reveals that. Therefore. The rainfall series was truncated at 10 mm rainfall. The parameter values appearing in Eq. In addition. Only three stations (Carbone. An analysis of the 5-year moving averages (Fig. Figure 8 The long term mean annual rainfall erosivity. the modelefficiency is high for both calibration (ModEff = 0. . Eleven stations showed a statistically significant downward trend in rainfall erosivity. the errors are reduced when estimated erosivity data are aggregated in a monthly or annual analysis. mean annual precipitation and simple daily rainfall intensity were derived for all 53 stations in Basilicata and are shown in Table 1. 6a) station displays two main change points in 1963 and 1993. The Mann–Kendall statistics are shown in Table 2.128 D.b). (5). and correlated to a statistically downward trend in annual precipitation. the reliability to estimate rainfall erosivity using Eq. while for Vaglio di Lucania (Fig. and this can be explained by several erosivity outliers present in the validation dataset.

379–386. Nanni. The results of the study led to the following conclusions: – Due to the large local spatial variation in rainfall. Baker. Ghent. References Agnese. 2001. Storm erosivity using idealized intensity distributions. Only one single station (Vaglio di Lucania) showed a significant increase in rainfall erosivity during winter and spring. A simple daily rainfall erosivity equation was derived from a short duration rainfall intensity data set of storm events.... Evoluzione climatica e modelli previsionali. total annual rainfall or single day storm intensity) plays the dominant role in the decrease or increase of rainfall erosivity in the region. spring. This is clearly illustrated by the annual isoerodent map of Basilicata (Fig. Ingram.. 1991. This reveals that a significant downward trend in annual rainfall erosivity is observed around 1970 in the 12 stations. Stigliano (Fig. 1994. Colacino. . 22–26 September. H. Allen.. a negative time trend in annual rainfall erosivity was detected.e. D’Asaro. According Piccarreta et al. M.. It seems. as being confirmed now also in other parts of Southern Italy (Agnese et al. 2002. 785–791. pp. summer and autumn). Geological Society of America Bulletin 88. 224–231. 39–51. FAO Soils Bulletin 34. annual erosivity values also show a high variation. For all other stations. the total annual precipitations and single day rainfall intensities show opposite trends. P.. Acknowledgements Conclusions This study presents a contribution to the quantification of rainfall erosivity and behaviour of long term climate erosivity in the Mediterranean region. 8). Examining long-term series of rainfall erosivity in Sicily.. 2004). In: Proceeding of Symposium ‘25 Years of Assessment of Erosion’. could be correlated with the decrease in average winter rainfall. 161–182. D’Asaro. Brown... statistically significant trends were mostly found for some stations during the winter periods.Analyzing temporal changes in climate erosivity using a simplified rainfall erosivity model of the 5-year moving averages for the stations with a downward erosivity trend (Fig. At 11 locations. 1977. several significant trends could only be confirmed during winter months in some stations..R. 129 – From the 53 rainfall data sets of the investigated rainfall point locations in Basilicata. Table 1 lists the long term mean annual erosivity values for 53 locations in Basilicata. respectively eleven. For the stations with an upward trend. when using the Mann–Kendall test statistic. S.J. ranging from a low 479 MJ ha1 mm h1 for Matera station with a long term mean annual precipitation of 477 mm. Brunetti. M.. 7d). 2003. L. the Lauria Inferiore (Fig. Methodology used to determine the maximum potential average annual soil loss due to sheet and rill erosion in Morocco. used in the erosivity map.. 1057–1071.. (2004). 7k) stations show an evident tendency towards an increase in annual rainfall erosivity. F.. Missanello and Vaglio di Lucania) showing a positive tendency. L.C. no time trends were significant at the 95% or 99% levels of confidence. W. D’Agostino. Arnoldus.. Consequently. Transaction of the ASAE 30. Bagarello. The stations with a negative trend showed distinct change points of decreasing erosivity around 1970. This research was supported by the PRIN-COFIN 2005 (Coordinator: Federico Boenzi) grant to Domenico Capolongo. International Journal of Climatology 21. 7) could also be confirmed by the values of the different climate normals (Table 1). Transaction of the ASAE 37 (3).M. V.. The equation permits the use of daily storm rainfall larger or equal than 10 mm to obtain the erosivity of a single erosive event. 2004). At the seasonal time scale. Bazzoffi. – Concerning seasonal tendencies. V. 299–316. Annali dell’Istituto Sperimentale Studio e Difesa del Suolo 22. while for the other seasons few stations showed any statistically significant trends. Caratteristiche delle piogge influenti sui processi erosivi nel periodo 1964–1990 in un ambiente della valle dell’Era (Toscana). V. M. For the four seasons (winter. trends and anomalies in annual and seasonal rainfall erosivity for the whole Basilicata region of Southern Italy. 1987. F.R. C.. Maugeri. A closer inspection of Table 2 reveals that for the majority of the stations which are characterised by a non-significant trend in annual erosivity. the remaining three stations (Carbone. with examples from central Texas. Foster. 7j) and Trecchina (Fig. Pellegrini. the decrease in winter rainfall erosivity in the region. in 39 locations no significant time trend in annual rainfall erosivity was observed. temporally close to those necessary for the computation of the RUSLE rainfall erosion index or R-factor. T. notable during the last 30-year normal period (1971–2000). Bagarello. Belgium.. two or three main change points (from 1965 to 1975 and around 1990) towards an increase in rainfall erosivity were detectable. Application of the daily rainfall erosion index equation to the long term daily time series of precipitation of 53 rainfall stations in the Basilicata region permitted us to generate and analyze long term tendencies. four. 1977. 2004.. Estimation single storm erosion index. to a local extreme high value of 4090 MJ ha1 mm h1 for Trecchina with a long term mean annual precipitation of 1839 mm.. Constraints on future changes in climate and the hydrologic cycle. Trends in the daily intensity of precipitation in Italy from 1951 to 1996. M.J. During the last two decades. three and six stations are characterised by a negative rainfall erosivity time trend. that the decrease in annual rainfall depths is somewhat compensated by an increase of single storm rainfall intensities to produce stationary long term values of annual rainfall erosivity in the region. Nature 419 (12). it remains difficult to define which precipitation variable (i. Stream channel response to floods. – From this dataset. G.R. we were yet unable to determine whether the annual precipitation totals or simple daily rainfall intensities played the major role for explaining the observed time trends in rainfall erosivity. This datum is in agreement with the downward trend in the rainfall regime recorded for the Basilicata region (Piccarreta et al.

Evaluating the use of ‘‘goodness-of-fit’’ measures in hydrologic and hydroclimatic model validation.. M. Navarra. 1978. D.). 2006. Gabriels. John Wiley. Diodato.. Maugeri. Sutcliffe. 245–261.. Journal of Soil and Water Conservation 56. D. J. 282–290. A. 1983. 1994. River flow forecasting through conceptual models.E. pp.. Comparison of different efficiency criteria for hydrological model assessment. Advances in Geosciences 5. 2004. L. Calculation of erosivity indices from a rainfall databank. 1970.. Yoder D. pp. Prediction of soil erosion by water: a guide to conservation planning with the revised universal soil loss equation (RUSLE). ... Poesen. Budapest. Poesen... C. 2004. 47–54. precipitation and extreme events during the last century in Italy. G. 41. 1991. The Environmentalist 25. 2004. Science of the Total Environment 344.R. Soil Erosion in Europe.. John. 269– 276... Piccarreta. Hungary. De Ploey. Journal of Geology 68. Soil Technology 1 (3). 2000. G. Ba ¨se. In: Wainwright.M.. 49. Nanni. Modelling and model building.. G.. Hungary.. M. Catena 40... Part I – a discussion of principles. 15–25. International Journal of Climatology 24.. Codina. Johnson.. Hydrology and Earth System Sciences 8 (1). T. C.130 Brunetti. 207–212. 7–73. Transaction of the ASAE 39. Ahnert. Sauerborn. Earth surface processes and landforms 16. Wright.. Cerro.M. Global and Planetary Change 40. Z.. Yu.. W. Brunetti. M. Capolongo et al.. M.J. 63–70. F. Zacharias. T. Bech. Kirkby. 256–270. P.S.R. 2004.M. M. Robust quantitative pesticide transport models.. D.C. L. 1992.. 731–735. Lord.G.A. 1960.G. USDA–Agricultural Research Service Oxford. 2005. Botschek. Freimund.M. D. 2001.K. Tomas. Sciences and Education Administration. International Journal of Climatology 22. Richardson. D. Torri. Diodato.. Mulligan. R.. Environmental Modelling. An example from the Philippines.. Guzzetti. 2001. A. Rainfall erosivity in Cape Verde. F.. Hillslope erosion by rainstorms – a magnitude – frequency analysis. T. Agricultural Research. 1998. M. B. 399–409.P. Legates. Journal of Hydrology 157. M. B. J. 537. Coutinho. C.... 233–241.. Toronto. D. Salvador Sanchis. Predicting RUSLE (revised universal soil loss equation) monthly erosivity index from readily available rainfall data in mediterranean area. Estimating RUSLE’s rainfall factor in the part of Italy with a Mediterranean rainfall regime. DC.R. 2000a. Journal of Hydrology 10 (3). 2004. N. S. pp. p. Anthony. D. Predicting rainfall erosion losses – a guide to conservation planning.W. Revised Universal Soil Loss Equation. Agric. C. John Wiley & Sons. N.W. Pauwelyn.J. Wiley and Sons.. A new procedure to estimate the RUSLE EI30 index. Smith. 1455–1471. Journal of Hydrology 257. Maugeri.. M.. Jones. Geoderma 93. Sempere-Torres.. 1980. Wolman. E. In: De Boodt. Hydrological Sciences-Journal 49 (5).. Mannaerts. Luxemburg. WMO. Elsevier Science Publ. NCTD.. J. Rusco. M. Boyle. Nanni. Renard. In: Proceedings of the Second Seminar for Homogenization of Surface Climatological Data. Estimating the R factor from daily rainfall data in the sub-mediterranean climate of southwest Slovenia.. (Eds.D. 1996. J. 1997.. Luxemburg. USDA.G. 2004. Foster. Modelling rain erosivity using distrometric techniques. Statistical methods in water resourcesElsevier Studies in Environmental Sciences. M. Mark (Eds. Capolongo. Washington. Lorente. 101–105.D. 27–46.. Fiji.A.. J. Renard. Soil erosion in Italy: an overview... 287–306.. Ltd. 1990. J. D. 1999. S. 89–97. Wainwright.. McCabe Jr. Magnitude and frequency of forces in geomorphic processes.. Bartolini. In: Proceedings of the IAHS Symposium of Suva. A probabilistic approach for predicting rainfall soil erosion losses in semiarid areas. 54–74. Nash.. Szentimrey. Mulligan. 869– 877. P. Mannaerts.. P. P.. Chichester.P. Droughts and extreme events in regional daily Italian precipitation series. J. P. G. Skowronek. 1995.. F. Zanchi. Hirsch. Soil and Tillage Research 55..L. M. Estimation of erosion index from daily rainfall amounts. Foster.. Szentimrey. 1999. Budapest. 2003. Boenzi. Eusof. Iso – erodent map of Zambia. Salles. Wischmeier. 2005. G. pp. 192. 141–149.L. A. EUR 20677 EN. Handbook 703. J... 403–420.. Krause.. Helsel. 235–251. C... D. P... Mikosˇ. F. Calzolari. Chichester. Mississippi. M..H. Hashim.G. UK.. Borselli. 153–156. Miller. 103–107. Temperature.McCool D. Soil erosion risk in Italy: a revised USLE approach. M. D.R. Part I.... D. Mangianti. Evaluation of rainfall energy in central Italy. Soil Science Society of America Journal 62.. C.. White... 907–922. The influence of tropical cyclones as soil eroding and sediment transporting events. C. Bazzoffi. Office for Official Publications of the European Communities. Version2.R.. The relationship between potentially erosive storm energy and daily rainfall quantity in England and Wales. 2002.. E. Characterisation of raindrop size distributions at the Vale Formosa experimental Erosion Center. Klein.. M.). 259–269. USDA. Water Resource Research 35 (1). P. D. Weesies G.G. K. Future rainfall erosivity derived from large-scale climate models – methods and scenarios for a humid region.. Coakley. Ungaro. 26.).I. J. Gabriels. Torri. 2005. 12–18. Transaction of the ASAE 26 (1).. Estimating the R-factor with limited rainfall data: a case study from peninsular Malaysia.A. D. C.. F.K...V. Loureiro. National Sedimentation Laboratory. J.. User’s reference guide.R. Sakala.. Montanarella.... Journal of Hydrology 250..A. Petkovsˇek. Agriculture Handbook No. pp. T. 1999. S. J. F. 187–197. Couthino. Kinetic energy of rain and its functional relationship with intensity.. M. Heatwole.P.S. In: Boardman. Multiple analysis of series for homogenization (MASH). Multiple analysis of series for homogenization (MASH). WMO. 1988. Portugal. R. vol. 133–142. WCDMP-N. Foster G. Betson.. W.. The Netherlands. (Eds. Grimm. C. L. Gabriels... IAHS Publication No.. Davison.M. N. Hutchins. Buffoni.. J. In: Proceedings of the Third Seminar for Homogenization and quality control in Climatological Databases. D. Lenvain. 2000b. Trend analysis of precipitation and drought in Basilicata from 1923 to 2000 within a Southern Italy context... 2002. 58 pp. Seasonal application of MASH (SAM)... based on monthly rainfall data and applied to the Algarve region. Catena 25... K. 1990. Automatic use of metadata..A. Using monthly precipitation data to estimate the R-factor in the revised USLE... Assessment of Erosion. M.