Identifying erosive periods by using RUSLE factors in mountain fields of the Central Spanish Pyrenees

The Mediterranean environment is characterized by strong temporal variations in rainfall volume and in- tensity, soil moisture and vegetation cover along the year. These factors play a key role on soil erosion. The aim of this work is to identify different erosive periods in func- tion of the temporal changes in rainfall and runoff charac- teristics (erosivity, maximum intensity and number of ero- sive events), soil properties (soil erodibility in relation to freeze-thaw processes and soil moisture content) and current tillage practices in a set of agricultural fields in a mountain- ous area of the Central Pyrenees in NE Spain. To this pur- pose the rainfall and runoff erosivity (R), the soil erodibil- ity (K) and the cover-management (C) factors of the empir- ical RUSLE soil loss model were used. The R, K and C factors were calculated at monthly scale. The first erosive period extends from July to October and presents the high- est values of erosivity (87.8 MJ mm ha 1 h 1 ), maximum rainfall intensity (22.3 mm h 1 ) and monthly soil erosion (0.25 Mg ha 1 month 1 ) with the minimum values of dura- tion of erosive storms, freeze-thaw cycles, soil moisture con- tent and soil erodibility (0.007 Mg h MJ 1 mm 1 ). This pe- riod includes the harvesting and the plowing tillage practices. The second erosive period has a duration of two months, from May to June, and presents the lowest total and monthly soil losses (0.10 Mg ha 1 month 1 ) that correspond to the maximum protection of the soil by the crop-cover (C fac- tor = 0.05) due to the maximum stage of the growing season and intermediate values of rainfall and runoff erosivity, max- imum rainfall intensity and soil erodibility. The third erosive period extends from November to April and has the mini- mum values of rainfall erosivity (17.5 MJ mm ha 1 h 1 ) and maximum rainfall intensity (6.0 mm h 1 ) with the highest number of freeze-thaw cycles, soil moisture content and soil erodibility (0.021 Mg h MJ 1 mm 1 ) that explain the high value of monthly soil loss (0.24 Mg ha 1 month 1 ). The in- teractions between the rainfall erosivity, soil erodibility, and cover-management factors explain the similar predicted soil losses for the first and the third erosive periods in spite of the strong temporal differences in the values of the three RUSLE factors. The estimated value of annual soil loss with the RUSLE model (3.34 Mg ha 1 yr 1 ) was lower than the mea- sured value with 137 Cs (5.38 Mg ha 1 yr 1 ) due to the low values of precipitation recorded during the studied period. To optimize agricultural practices and to promote sustain- able strategies for the preservation of fragile Mediterranean agrosystems it is necessary to delay plowing till October, es- pecially in dryland agriculture regions. Thus, the protective role of the crop residues will extend until September when the greatest rainfall occurs together with the highest runoff erosivity and soil losses.


5
Soil erosion in agricultural areas has been studied intensively throughout the last decades and rates have been measured at continuous and event scales. Moreover, temporal variations in soil losses are usually studied at long-term scale due to changes in land use (Navas et al., 2005;Wei et al., 2007) or changing climatic conditions during the past and future predictions (Zhang, 2006). However, it is widely accepted that 10 most soil erosion and sediment yield is triggered by intense rainfall and runoff events (Lecce et al., 2006) and the percentage of precipitation that produces the greatest erosion is very low. In addition, the dominating erosion process depends on rainfall intensity either for low intensity events when splash erosion dominates on the interrill areas or during high intensity events when considerable runoff volumes and rill ero- 15 sion dominates (Kuhnert et al., 2007). On the other hand, temporal variations in soil erodibility during concentrated flow can be mainly explained by variations in soil moisture content (Knapen et al., 2007). Moreover, winter conditions with seasonally frozen soils may have strong effects on aggregate stability, soil structure and erodibility, and consequently in runoff and erosion. Erosion risk maps and soil prediction models that 20 include these factors increase the accuracy of their predictions (Kvaernø and Øygardena, 2006).
Most studies attribute the effect of crops in reducing soil erosion to the effects of the above-ground biomass (Gyssels et al., 2006). However, soil redistribution by conventional tillage practices has been recognised as a process of intense landscape Introduction EGU biomass temporally disappears after the harvest and collection of crop residues and then concentrated flow erosion occurs. Low roughness generated by tillage and bare soil after harvest promotes an increase in soil erosion in agricultural lands (Gómez and Nearing, 2005). Hence, the agricultural practices play a strong control in triggering erosional processes. In addition, seasonal variations in soil erodibility under different 5 tillage practices have been identified by Knapen et al. (2007). The Mediterranean environment is characterized by a contrasted climate with irregular but frequent and intense rain events, low vegetation cover and poor soil characteristics. Soils of Mediterranean agrosystems are particularly vulnerable to changes in such parameters and erosion rates are very high in some areas (Arnaez et al., 2007). 10 Moreover, climate change is increasing the temperature, changing the temporal and spatial distribution of rainfall along the year (Meehl et al., 2005), and increasing the frequency of extreme events, especially in Mediterranean areas (e.g. Tapiador et al., 2007). An increase of extreme daily rainfall in spite of decrease in total values has been recorded in Spain and other Mediterranean countries (Alpert et al., 2002). Therefore, 15 there is great interest in determining the temporal pattern of soil erosion and sediment delivery at seasonal and monthly scales (e.g. Mathys et al., 2007).
In mountainous areas of northeastern Spain the precipitation regime is characterised by a bi-modal annual distribution, with one main maximum in autumn and a secondary peak in spring. Convective storms are frequent in this area during summer with intense 20 precipitation and high values of maximum intensity (Sánchez et al., 2003) and explain the greatest part of the sediment load exported to reservoirs. Changes in the frequency of extreme floods have been identified in mountain areas of the Iberian Range (Machín et al., 2005) and of extreme dry-spell in the middle Ebro Valley (NE Spain) (Vicente-Serrano and Beguería-Portugués, 2003 EGU areas in the north-central Spanish Pyrenees presents also field evidences of seasonal variations (Nadal-Romero et al., 2007). The call for erosion control measures adapted to local farming practices is stressed. Nonetheless, the assessment of monthly and seasonal variations of erosion rates in cultivated fields is still an outstanding question that needs a quick answer due to the strong inter-annual variability of rainfall charac-5 teristics in Mediterranean areas. This work aims to identify different erosive periods in relation to temporal changes in rainfall characteristics (erosivity, maximum intensity and number of erosive events), soil properties (soil erodibility in relation the freeze-thaw processes and soil moisture content) and tillage practices. For this purpose the rainfall and runoff erosivity (R), 10 soil erodibility (K) and cover-management (C) factors of the RUSLE model (Renard et al., 1997) were used. The RUSLE model is widely used in Mediterranean areas (e.g. Ramos and Porta, 1994). The monthly values of the R, K and C factors were calculated in a set of agricultural fields in NE Spain in a mountainous area of the Central Pyrenees. The results of this study could be used for best management practices 15 (BMPs) that are highly recommended within the agrarian policy of the European Union. The information gained will provide data of interest to promote effective measures to avoid soil degradation in the high-productive dryland fields of Mediterranean countries.

Study area 20
A farmland area surrounding the Estaña lakes was selected to carry out this study located in the province of Huesca, Spain (Fig. 1a). This area is located between the Cinca and the Noguera Ribagorzana rivers, in the southern limit of the External Ranges of the Central Pyrenees, close to the northern boundary of the Ebro basin. The study area was selected as representative of rainfed agricultural areas in Mediter-Introduction EGU is underlayed by limestones affected by diapirs largely composed of gypsiferous marls, dolostones, limestones and occasionally salt deposits (Riera et al., 2006). The elevation of the selected fields ranges from 677 to 729 m a.s.l. (López-Vicente and Navas, 2005) with a mean slope of 10.3 % (López-Vicente et al., 2006b). Field evidence of gully erosion has been observed in the steepest fields.

5
This area has a continental Mediterranean climate with mean annual precipitation of 665, 563 and 464 mm at the weather stations of Benabarre, Camporrélls and Canelles, respectively (López-Vicente et al., 2005). These weather stations located NW, SW and SE of the study area at a distance of around 10 km have an elevation of 740, 628 and 508 m a.s.l., respectively (Fig. 1b). In spite of the short distance between the weather 10 stations the differences in the annual precipitation are explained by their geographical situation, between the semiarid areas of the Ebro valley to the south (Camporrélls and Canelles) and the humid areas of the Pyrenees to the north (Benabarre). (López-Vicente et al., 2005) estimated an annual precipitation of 595 mm for the study area as well as the monthly values of minimum and maximum temperature (Fig. 2a). 15

Rainfall and runoff erosivity factor (R)
Soil loss in agricultural fields is associated with the product of the total storm energy (E, MJ ha −1 ) and the maximum intensity in 30 min (I 30 , mm h −1 ). The result of this product is the EI 30 index or storm erosivity index (MJ mm ha −1 h −1 ) that reflects the combined effect of soil detachment and runoff transport capacity to produce net soil 20 erosion. Renard et al. (1997) defined the rainfall factor R (MJ mm ha −1 h −1 ) as the sum of the EI 30 values for the whole year according to the equations:

EGU
where: j is the number of erosive events for the n number of years; k is the temporal interval; m is the number of temporal intervals established for each storm event; e r (MJ ha −1 mm −1 ) is the kinetic energy of a storm for the r period; and ∆V r (mm) is the volume of rainfall registered during the r period. When n=1 the calculated R value is 5 the rainfall erosivity for one specific year. The kinetic energy is assessed in the RUSLE model following the approach of Brown and Foster (1987) such as: (3) where: 10 i r (mm h −1 ) is the rainfall intensity for the r period; and ∆t r (min) is the duration of the r period.
Soil erosion rates in the rill and interrill areas as well as the rates of sediment yield in the deposition areas are mainly controlled by storm events with medium and high values of intensity and rainfall volume. Hence, the erosivity factor in the RUSLE model is 15 calculated from erosive storm events with values of rainfall volume higher than 12.7 mm or with a value of intensity higher than 6.35 mm in 15 min. The guide of the RUSLE model established a period of six hours with a rainfall volume lower than 1.27 mm to distinguish between two different storm events.
The R-RUSLE factor assesses the effect of the rainfall impact on the soil surface as 20 well as the magnitude of runoff. However, it does not account the water supplies from snow melting neither the water from irrigated areas nor the effect of rainfall impact over frozen soil.

Soil erodibility factor (K)
Soil erodibility is a complex property and is thought of as the ease with which the soil 25 is detached by splash during rainfall or by runoff or both. In the RUSLE model the soil 2117 Introduction EGU erodibility factor (K, Mg h MJ −1 mm −1 ) is the rate of soil loss per rainfall erosion index unit as measured on a unit plot that is 22.1 m long, 1.83 m width and has a 9% slope. The K factor is a lumped parameter that represents an integrated average annual value of the soil profile reaction to the processes of soil detachment and transport by raindrop impact and surface flow, localized deposition due to topography and tillage-induced 5 roughness, and rainwater infiltration into the soil profile (Renard et al., 1997). This factor can be assessed as a function of five soil parameters: percentage of organic matter (OM, %), percentages of modified silt (2-100 µm) and sand (100-2000 µm), and classes of aggregates structure (s) and soil permeability (p). For those cases where the silt fraction does not exceed 70% the following equation is used to calculate 10 the K factor: where M is the product of the percentages of modified silt and sand. The RUSLE model established four different soil structure classes (Table 1) and six permeability classes ( Table 2) that were taken from the National Soils Handbook No. 430 (USDA, 1983). 15 This handbook defined the permeability classes according to the soil texture, though this parameter can also be assessed by field estimation of the saturated hydraulic conductivity (K f s , mm day −1 ). The approach of Rawls et al. (1982) is used in the RUSLE model to estimate the different permeability classes (Table 2) according to K f s values.

Soils with rock fragments
Surface rock fragments reduce significantly the splash detachment rates in a manner similar to the crop residues that protect the soil surface from raindrop impact. However, in coarse textured soils surface and subsurface rock fragments affect infiltration and thus runoff by reducing the soil void space and soil hydraulic conductivity and in- EGU the soil in the same area, rocks appear in the soil profile as a frame, especially in interrill areas, where runoff cannot move them. Moreover, rock fragments larger than 2 mm were excluded when K-factor values were estimated in Eq. (5). To account the effect of rocks in soil erodibility the RUSLE model includes the following approach: where K b (mm day −1 ) is the modified saturated hydraulic conductivity after accounting the effect of rock fragments, and R W (%) is the weight percentage of coarse fragments.

Seasonal variations in soil erodibility
K values are difficult to estimate mainly because of seasonal variations in soil properties that are primarily related to three factors: soil freezing, antecedent soil-water and soil-surface conditions (soil texture and structure). The greater the number of freeze-thaw cycles, the longer the erosion resistance of a soil is at a minimum. Freezethaw cycles reduce bulk density, stability and cohesion of the soil leading the soil to its maximum value of soil erodibility (K max ) at the beginning of the free-freezing period. Moreover, high soil-water content can delay infiltration and water movement into the 15 soil profile. Hence, soil during the thawing period is extremely susceptible to erosion caused by splash and runoff. On the other hand, during the free-freezing period soil erodibility decreases exponentially reaching its lowest value (K min ) at the end of this period. Although the time span between the maximum and minimum values of soil erodibility varies with location and soil type, a value of 6 months or less appears to be 20 reasonable in most areas and scenarios. EGU period of maximum soil erodibility (t max , day) can be calculated as follows: These equations were established according to the U.S. customary units therefore con-5 version from SI units must be done.

Cover-management factor (C)
The cover-management factor of the RUSLE reflects the effect of cropping and management practices on erosion rates. The C factor is the most commonly used to compare the relative impacts of management options on conservation policy. This factor 10 allows estimating how the conservation policy will affect the average annual soil loss. The soil loss ratio (SLR) is an estimate of the ratio of soil loss under actual conditions to losses experienced under reference conditions (clean-tilled continuous-fallow). An individual SLR i (0-1) value is thus calculated for each time period i , as: 15 where the sub-factors for each time period i are the prior land (PLU i ), the canopy cover (CC i ), the surface roughness (SR i ), the surface cover (SC i ), and the antecedent soil moisture (SM i ). The prior land use sub-factor expresses the influence on soil erosion of subsurface residual effects from previous crops and the effect of previous tillage practices on soil consolidation. The canopy cover sub-factor expresses the effectiveness 20 of vegetative canopy in reducing the energy of rainfall striking the soil surface. The surface roughness sub-factor measures how depressions and barriers trap sediment and water, during a rainfall event, causing rough surfaces to erode at lower rates than do smooth surfaces under similar conditions. The surface cover sub-factor estimates 2120 Introduction EGU how crop residues, rocks, and other nonerodible material reduce the transport capacity of runoff. Finally, antecedent soil moisture is an inherent component of continuos-tilled fallow plots, and these effects are reflected in the soil erodibility factor. Hence, no adjustment is made for changes in soil moisture to calculate the C factor. Each SLR i value is then weighted by the fraction of rainfall and runoff erosivity (EI 30i , 5 %) associated with the corresponding time period, and these weighted values are combined into an overall C factor value as: where EI 30t (%) is sum of EI 30i percentages for the entire time period, n is the total number of time period i . The values of C factor ranges from 0 (total control of the 10 erosion) to 1 (no effectiveness of cover-management practices).

Data collection
The EI 30 parameter for the study area has been calculated from the rainfall values recorded at the weather station of Canelles each 15 min because the weather stations of Benabarre and Camporrélls have a worse temporal resolution and only record daily 15 precipitation. The database of Canelles for the period 1997-2006 was obtained from the Water Authorities (SAIH, Confederación Hidrográfica del Ebro). Its rainfall record also includes monthly values of precipitation for the period October 1940-December 1991 that was not considered for calculating the R factor. A field survey was carried out and a total of 60 soil samples were collected in the 20 selected agricultural fields. Samples were air-dried, grinded, homogenized and quartered, to pass through a 2 mm sieve. The general soil properties analysed were: organic matter (OM), coarse fragments (>2 mm, R w ) and soil texture (<2 mm). Analysis of the clay, silt and sand fractions were performed using laser equipment. Organic matter was determined by the Sanerlandt method (Guitian and Carballas, 1976)  EGU 1998) identifying eight soil types (Fig. 2b). López-Vicente et al. (2005) measured the saturated hydraulic conductivity for each soil type obtaining values that range from 9.9 to 2252.5 mm day −1 for Haplic Gypsisols and Haplic Leptosols, respectively. Two types of structure of soil aggregate were identified. Very coarse granular and very coarse prismatic structure (class 4) was associated to Luvic Gleysols, Haplic Gypsisols and 5 Gypsic Regosols and granular and medium crumb and coarse granular structure (class 3) was associated to Haplic Calcisols, Haplic Regosols, Lithic Leptosols, Hypercalcic Calcisols and Haplic Leptosols.
The volumetric soil water content (θ S ) in the upper 8 cm of the soil was measured using a Theta Probe soil moisture device. Soil moisture was controlled in 79 points 10 following a regular grid to obtain a representative database of the soil moisture that was measured in February, May, August and December.
The soil loss ratio, SLR, was calculated for periods of fifteen days. To estimate the prior land use sub-factor, PLU, the data of mass density of live and dead roots and of the incorporated to the surface residue in the upper inch of the soil, and the 15 consolidation of soil surface for barley fields were obtained from the guide of the RUSLE model (Renard et al., 1997). To calculate the canopy cover sub-factor, CC, the values of proportion of land surface covered by canopy, and the distance of raindrops falls after striking the canopy were also obtained from these authors.
The role of rainfall interception by crops on the seasonal variations of soil erosion 20 was analyzed by Castro et al. (2006) in olive orchards in Córdoba (Spain). In this work the rainfall interception of the crop vegetation and residues were added in the assessment of the canopy cover following Morgan (2001). The rainfall interception has a value between 0 and 1 and is defined as the amount of rainfall that remains in the branches and leaves of the canopy and crop residues and returns to the atmosphere 25 by evaporation. In this work, the values of rainfall interception for barley (0.14) and its residues (0.03) were obtained from Eberbach and Pala (2005)  EGU in the study area is mouldboard plow (Renard et al., 1997) were used to calculate the surface roughness and surface cover sub-factors. The percentage of coarse fragments was also used to assess the sub-factor of surface cover.  10 The mean value of I 30 for the study area is higher than that obtained by Usón and Ramos (2001) in vineyards of Barcelona (NE Spain) with a mean value of 10 mm h −1 and a maximum of 103 mm h −1 which is quite similar to the obtained in our study area. September had the maximum mean value of I 30 (26.9 mm h −1 ), whereas the mean values from December to March ranged between 5.4 and 7.1 mm h −1 . This variability 15 was also observed in the EI 30 values, with a mean of 107.1 MJ mm ha −1 h −1 for the May-September period which is higher than the mean registered in the November-April period (26.8 MJ mm ha −1 h −1 ). The highest values of rainfall erosivity were associated with the highest values of maximum intensity. The mean value of EI 30 for the June-August period was 334% higher than that for the January-March period. However, 20 the rainfall was only 19 % higher for the June-August period. The Pearson correlation coefficient between the erosivity and precipitation was low (r=0.47) (Fig. 3a) and it was high (r=0.95) between the erosivity and the maximum intensity of rainfall (Fig. 3b).

Results and discussion
On the other hand, the erosivity presented a high monthly variability. The mean erosivity is higher than its median value in nine months and higher than its 75th-percentile 25 in May (Fig. 3c). This variability is explained due to the high variability in rainfall ero- EGU sivity during the April-October period, especially in September. The 10 most erosive storm events happened in September (6 events), October (2 events), August (1 event) and May (1 event), whereas the 10 highest values of maximum intensity were registered in September (7 values), October (1 value), August (1 value) and May (1 value). Moreover, 31 % of the identified erosive events happened in September and October.

5
The mean value of R was 1000.3 MJ mm ha −1 h −1 y −1 with a wide range of variation between 215.0 and 1969.2 MJ mm ha −1 h −1 yr −1 in 2004 and 1998, respectively (Fig. 3d). The R factor was calculated for a dry period (mean annual rainfall 445.53 mm) because 8 years of the period 1997-2006 had a lower value of rainfall than that measured in the weather station of Canelles for the reference period (1961-1990: 10 519.95 mm). These results agree with values obtained in other Mediterranean areas as central and southern Italy (580-2300 MJ mm h −1 ha −1 yr −1 ) (Diodato, 2004). The mean values of R in NE Spain were between 1049 and 1200 MJ mm ha −1 h −1 yr −1 (Ramos and Porta, 1994).
For better characterizing the storm erosivity in the study area, the ten year fre-15 quency single storm erosivity (10-yr EI 30 , in Renard et al., 1997) was calculated following the generalized Pareto distribution that was successfully applied by Vicente-Serrano and Beguería-Portugués (2003) in a study of extreme hydrological events in the middle Ebro valley (NE Spain). The rainfall erosivity of the ten year frequency was 706.1 MJ mm ha −1 h −1 and the estimated mean volume of precipitation for this rainfall 20 event was 76.3 mm. According to this value there is only one rainfall event with a higher value of rainfall erosivity, that corresponds to the outlier of May (Fig. 3c) and explains the high value of R registered in 1998 (Fig. 3d). The organic matter in the soil samples ranged between 0.7 and 7.5 % with a mean of 2.4%. Almost all of the soil textures were silt-loam and the values of M ranged 25 between 0.3 and 0.9 (Table 3). The mean and maximum percentages of coarse fragments were 21 and 56%, respectively, which are common within Mediterranean areas (Govers et al., 2006). The high stone contents modified the saturated hydraulic conductivity obtaining a mean value of 433.4 mm day −1 . The mean soil erodibility was  (Table 3). The soils with a coarse granular and very coarse prismatic structure and low organic matter contents present higher erodibility than those with a granular and medium crumb structure and high content in organic matter. These results agree with the decrease in soil erodibility calculated by Tejada and Gonzalez (2006) in soils of Sevilla (southern Spain), and suggest the clear role of organic matter on the stability of soil aggregates.
The lowest soil moisture was obtained in August, with a mean content of 10.6%, whereas the means for February, May and December were 13.1, 15.6 and 17.7%, respectively. The highest rates of soil erodibility were obtained in Luvic Gleysols and Haplic Gypsisols due to their low saturated hydraulic conductivity and organic 10 matter and the lowest rates were in Haplic Leptosols and Calcisols. The minimum and maximum values of soil erodibility due to seasonal variations were 0.004 and 0.029 Mg h MJ −1 mm −1 , respectively. The K max /K min ratio was 7.5. This high value was similar to the ratios of 7.4 and 10 obtained by Hussein et al. (2007) in a semi-arid catchment of northern Iraq where the rainfall and runoff erosivity was 900 MJ mm ha −1 h −1 .

15
According to Renard et al. (1997), high K max /K min ratios are expected in regions with low mean seasonal or annual R values and less uniformly distributed monthly R values, such as in our study area. The duration of the period of maximum soil erosivity, t max , was 128 days. From this value, the duration of K min and K was estimated in 50 and 187 days, respectively.

20
The highest mean of soil loss ratio, SLR i , was in the November-April period (0.23), which is much higher than for the rest of the year (0.12) (Fig. 4c). These values are controlled by the schedule of the tillage practices and the phenology of the crops and agree with those obtained by Renschler et al. (1999)  EGU maximum intensity and the duration of a typical rainfall event almost doubles that in EP-I (Fig. 5c). The mean soil erodibility is the highest of the three periods (Table 4) and is three times higher than the rate in the first period because almost all the freeze-thaw cycles are concentrated in EP-III in coincidence with the highest soil moisture content. Moreover, the cover-management is at its highest because crops are at the stages of 5 sowing, tillering and at the early stages of the growing season (Table 4). For a better assessment of the temporal variations in the studied parameters, the total and monthly soil losses were calculated for each erosive period (Table 4) as the product of the rainfall erosivity, soil erodibility and cover-management factors and without considering corrections by topography and support practices (Renard et al., 1997). 10 The lowest rates in total and monthly soil erosion are found in EP-II, whereas EP-I has the highest monthly soil erosion. Nonetheless, EP-III has a monthly soil erosion similar to EP-I and the highest total rate of soil erosion due to its longer duration. These seasonal trends with higher rates at the end of autumn and in summer were also observed in badlands of the south-eastern Pyrenees (Regüés and Gallart, 2004), in north-central Pyrenees (Nadal-Romero et al., 2007), in cultivated fields of Navarra in north Spain (De Santisteban et al., 2006) and in southern French Alps (Mathys et al. 2007). Concerning the role played by frost, Bullock et al. (1988) found that frost only has an effect on moist soils in which the water content exceeds 0.2 g g −1 . Hence, the calculated soil erodibility for winter months will be overestimated under drier and warmer winter conditions. 20 Soil erodibility is one of the most important factors to estimate soil losses. Hence, a more accurate assessment of this property than the made with the RUSLE model will be necessary to account for the chemical and mineralogical composition of the soil such as Tejada and Gonzalez (2006) made in wheat fields of Spain.
Approaches which promote early canopy development may reduce the amount of 25 erosive rainfall by increasing rainfall interception by the crop canopy. Litter cover plays an important role in runoff and on the reduction of soil loss and is also fundamental for the control of erosion during intense rainfall (Bochet et al., 2006 EGU delay will extend the rainfall interception and surface protection by crop residues as well as will reduce the total number of days of bare soil in the year. Another effective measure that could be adopted is to increase the thickness of the crop residues where it may be possible to increase the rainfall interception (Cook et al., 2006) and thus reduce the amount of water that reaches the soil surface. This practice does not require 5 elaborate tillage operations and will increase the percentage of organic matter in the soil reducing the soil erodibility. Finally, planting cover crops (rye or ryegrass) could be a solution to minimize soil erosion and runoff in the period between the harvest and plowing. These strategies agree with those proposed by Martínez-Casasnovas and Sánchez-Bosch (2000) for the prevention of land degradation in agricultural fields 10 under Mediterranean conditions. In spite of the clear differences in the climatic parameters for the identified erosive periods, we also consider as Usón and Ramos (2001) that further research may be done including rainfall values registered each 5 or 10 min. This will allow a more accurate assessment of the R factor. Because soil moisture is a key parameter in soil 15 erosion in Mediterranean environments a soil moisture sub-factor will be of interest to account its effect on the formation of surface crust. These proposed improvements will help to assess the effects of the temporal and spatial variations of rainfall that are expected to happen in Mediterranean areas due to climate change.
The patterns of rainfall distribution in this study are representative of Mediterranean 20 environments. Moreover, the described tillage practices are common in dry farmlands. Therefore, the results obtained can be implemented in runoff and erosion models to improve their predictions in Mediterranean agrosystems.

Conclusions
The monthly values of rainfall and runoff erosivity, soil erodibility and soil loss ratio have 25 shown a strong temporal variability along the year. The first erosive period identified in this work has a duration of four months, from July to October, and is characterized EGU by the highest values of rainfall erosivity, maximum rainfall intensity and monthly soil erosion and the minimum values of erosive storm duration, freeze-thaw cycles, soil moisture content and soil erodibility. The second erosive period is the shortest with a duration of two months, from May to June, and presents the lowest rates of total and monthly soil losses that correspond to the maximum protection of the soil by the crop-5 cover. The third erosive period has a duration of six months, from November to April, and presents the minimum values of rainfall erosivity and maximum rainfall intensity. The erosive storm events associated with this period present the longest duration, and the soil erodibility is the highest value of the three erosive periods in accordance with the high number of freeze-thaw cycles and wettest soil. The monthly soil loss is slightly lower than in the first erosive period though the total soil loss is higher. This work has highlighted that the interactions between the rainfall erosivity, soil erodibility, and cover-management can explain similar predicted soil losses found in the first and the third erosive periods in spite of the strong differences in the values of the three factors. The second erosive period, May and June, is the period with the lowest 15 rates of soil erosion in the year. To promote sustainable strategies for the preservation of the fragile Mediterranean agroecosystems and especially in dryland agriculture it is recommended to delay the plowing practices till October. This delay will extend the protection role by the crop residues in September, as this month concentrates the highest rainfall and runoff erosivity and soil losses.   [209][210][211][212][213][214][215][216][217][218][219][220][221][222][223][224][225]1999. Riera, S., López-Sáez, J. A., and Julià, R.: Lake responses to historical land use changes in northern Spain: The contribution of non-pollen palynomorphs in a multiproxy study, Rev. Palaeobot. Palyno., 141, 127-137, 2006. Sánchez, J. L., Fernández, M. V., Fernández, J. T., Tuduri, E., and Ramis, C.: Analysis of 5 mesoscale convective systems with hail precipitation, Atmos. Res., 67-68, 573-588, 2003. Tapiador, F. J., Sanchez, E., and Gaertner, M. A.: Regional changes in precipitation in Europe under an increased greenhouse emissions scenario, Geophys. Res. Lett., 34 (6)