ANALYSING THE EFFECT OF LAND USE AND VEGETATION COVER ON SOIL INFILTRATION IN THREE CONTRASTING ENVIRONMENTS IN NORTHEAST SPAIN

This study presents a joint analysis of the information from 195 field infiltration experiments, using double ring devices. The experiments were carried out in 20 contrasting types of land use, distributed across three geographic contexts (coast of NE Catalonia, an area of gentle relief in the central Ebro Valley and mid-height mountains from the southern range of the Central Pyrenees). The objective of this research was to determine the most important factors explaining infiltration variability: land use, type of vegetation cover, soil and bedrock characteristics, soil moisture and altitude. Data analysis was performed by comparing variables using statistical methods: bivariate lineal correlation, ANOVA and Bonferroni multiple comparison tests. Results show that infiltration variability is the most important factor and is mainly linked to land use, followed by vegetation type. In contrast, soil moisture did not show any relation with infiltration. The interpretation of these results suggests that the characteristics of the study areas are more decisive than temporal variations of soil water content, although humidity can influence land use to a greater or lesser degree. The validity of the results obtained in this study is supported by the wide range of land use and land cover analysed, located in areas with different geographical and geological characteristics. Análisis del efecto de los usos del suelo y de la cubierta vegetal en las tasas de infiltración en tres áreas contrastadas del Nordeste de España RESUMEN. Este estudio presenta el análisis conjunto de la información obtenida a partir de 195 ensayos de infiltración en el campo, que fueron realizados mediante dispositivos de doble anillo. Los experimentos se realizaron en 20 situaciones contrastadas de usos del suelo, los cuales se encuentran distribuidos en tres


Introduction
The variation of saturated hydraulic conductivity (K sat ) and infiltration rates associated with vegetation cover/land use changes were examined in previous research (Zimmermann et al., 2006;Zimmermann and Elsenbeer, 2008) and linked to the influence of alterations in the physical properties of soil, which can affect both the spatial (Sisson and Wierenga, 1981;Hopmans et al., 1988;Mallants et al., 1996;Strock et al., 2001) and temporal (Moret and Arrúe, 2007;Bormann and Klaassen, 2008;Zhou et al., 2008;Li et al., 2009) variability of hydraulic properties.However, the temporal variability of hydraulic soil properties is less well documented than its spatial counterpart (Mishar et al., 2003;Hu et al., 2009).Many studies have described a decrease in infiltrability with depth in most land uses The effect of land use and vegetation cover on soil infiltration Cuadernos de Investigación Geográfica 43 (1), 2017, pp.141-169 143 (Bonell et al., 1983;Malmer, 1996;Godsey et al., 2004;Zimmerman et al., 2006) except for pastures, where this variable is more stable or may increase (Malmer, 1996;Noguchi et al., 2003).However, the decrease in infiltrability was generally more pronounced in scenarios strongly affected by changes in land use and vegetation cover (Giertz and Diekkrueger, 2003;Martinez and Zinck, 2004;Ziegler et al., 2004;Moraes et al., 2006;Zimmermann et al., 2006) or affected by variations in tillage practices (Wang et al., 2016).
Since the middle of the 20th century, a large area of agricultural land in Spain has been affected by the abandonment of crops (Romero-Díaz et al., 2007;Cerdà et al., 2012;Alonso-Sarriá et al., 2016), especially in mountain areas (Lasanta, 1988;Arnáez et al., 2011), resulting in natural vegetation colonisation and the growth of scrublands and forests (Molinillo et al., 1997;Poyatos et al., 2003;Vicente-Serrano et al., 2004).Another consequence is the reduction of subalpine grasslands (Lasanta and Vicente-Serrano, 2007), which has been related to the decline of grazing and the ascent of the upper tree line due to global warming (Sanjuán et al., 2013).The reduction of meadows and increase in scrubland in mountain areas is a widespread process in the Pyrenees (Camarero et al., 2015).Gartzia, et al. (2016) showed that, below 2100 m a.s.l., the presence of scrubland in areas previously occupied by grassland has increased significantly during the last few years and that there are significantly more trees in areas previously dominated by scrub.
In addition to the hydrological consequences associated with substantial expansion of forests and scrub, other environmental issues occur, such as the increased risk of fire (García-Ruiz and Lana-Renault, 2011;San Román et al., 2013).The effect of fire on soil properties has also been studied (Badía and Martí, 2003;Pardini et al., 2004b), particularly the hydric effects of an ash layer covering the soil surface, where responses vary according to the severity of fire (Pereira et al., 2013), as well as those deriving from fire on water, and sediment yield on different soil types (Badía et al., 2008).During the period immediately after fire, the ash deposit can increase infiltration capacity rates (Cerdà and Doerr, 2008;Zavala et al., 2009) in addition to causing a water repellent effect (Shakesby and Doerr, 2006;Bodí et al., 2012).
All these arguments justify the interest in improving knowledge on the relationships among land uses, the hydraulic properties of soil and overland flow generation (Germer et al., 2010;Price et al., 2010).Many studies on infiltration have focused on the application of models for comparison or validation using data obtained through laboratory or field infiltrometry tests, including comparisons between different land uses (Mbagwu, 1995;Mishar et al., 2003;Shukla et al., 2003;Ghorbani Dashtaki et al., 2009;Haghighi et al., 2010;Duan et al., 2011;Robertson and Sharp, 2015).Despite the large number of studies on the effect of land use changes on infiltration rates (Malmer and Grip, 1990;Ghuman et al., 1991;Alegre and Cassel, 1996;Lal, 1996;Navar and Synnott, 2000;Martínez and Zinck, 2004;Ziegler et al., 2004;Zimmermann et al., 2006), additional information from studies comparing contrasting areas is still needed.In this study, we analysed a large data set of infiltration rates measured under several land use and vegetation cover situations in three study areas in northern Spain, in order to determine their effect on infiltration variability.

Study area
The experimental areas were located in the north-eastern sector of the Iberian Peninsula (Fig. 1) which includes a wide variety of land uses and vegetation covers (Table 1): (1) In the central-western Spanish Pyrenees (close to the city of Jaca), the sampled area contained low to medium-height mountains (700-1350 m a.s.l.) characterised by the presence of many abandoned fields in various stages of plant colonization (García-Ruiz and Lasanta, 1990), alternating with natural and artificial open and dense forest.The study area involves bare badlands, badlands covered with grass, south-facing pasture, north-facing meadows, south-facing slopes with natural forest (Quercus faginea) and north-facing slopes with natural forest (Pinus sylvestris and Fagus sylvatica).
(2) In the Ebro valley the sampled area is located in the Zuera structural platform (450-700 m a.s.l.), that contains a mixed forest (Castejón area) of Pinus halepensis, and scrubland (Remolinos area), including Genista Scorpius, Retama sphaerocarpa, Rosmarinus officinalis and Quercus coccifera.This area was burned in 2008 and 2009 (León et al., 2013).The experimental plots were selected based on a combination of three factors: vegetation cover, lithology of the bedrock and the occurrence/non-occurrence of fire.Thus, seven land uses were selected, including burned forest (south and north aspects), burned trees and scrubs (limestone and gypsum), burned scrub (gypsum), trees and scrubs (gypsum) and scrubs (gypsum).
(3) In the north-eastern coast of Spain (Cap de Creus, 90-260 m a.s.l.), which displays heavy human disturbance of the landscape, several types of land use, and vegetation cover associated with land abandonment due to reduced agricultural activity (Dunjó et al., 2003).The study was carried out on seven scenarios: north-facing pasture, scrubland (Erica and Cistus), pine and cork tree plantations, olive groves and vineyards.The dominant climatic conditions in the three areas are Mediterranean, although with some variations derived from the distance to the coast and altitude.The northeastern coastal area (Cap de Creus) shows the most marked characteristics of the Mediterranean climate, with strong prevailing winds from the north (Tramuntana) that reaches 90 km h -1 .The mean annual temperature is about 16ºC and the mean precipitation 450 mm yr -1 .In the Ebro Valley area (Zuera), the climate is Mediterranean with continental features and a very high water deficit (Herrero and Synder, 1997).The mean annual temperature is significantly lower (12ºC) and the mean annual precipitation is approximately 400 mm yr -1 .Climate characteristics in the sampled area of the central-western Pyrenees have been defined as sub-Mediterranean mountain, with distinct Mediterranean and, especially, Atlantic influences (Creus, 1983).Mean annual temperature is 10ºC and mean precipitation of approximately 900 mm yr -1 .Soils in the north-eastern coastal area (Cap de Creus) are shallow Lithic Xerorthent, and bedrock is composed of granodiorite and schists (Enram et al., 2012).Bedrock in the Ebro valley area (Zuera) is formed by Miocene limestones and gypsum, and soils are Rendzic Phaeozem, Leptic and Haplic Gypsisol (Badía et al., 2013;León et al., 2015a).The greatest soil diversity is found in the Pyrenees area (Jaca), with Kastanozem, Phaeozem, calcic Regosol, Leptosol or Cambisol, and bedrock is composed of Eocene Flysch turbidites in the Pyrenean mountain range and Eocene marls in the Inner Pyrenean Depression (García-Ruiz et al., 2008;Lana-Renault et al., 2011).Table 2 shows some descriptive characteristics of the soil types in each study site.

Methodology
Data acquisition was performed by double ring infiltrometer tests in the Pyrenees and Cap de Creus study sites, using devices of similar design, with an inner cylinder diameter of 29.5 cm (external cylinder diameter 49.5 cm).For the tests performed in the Ebro Valley, a simple ring infiltrometer, with a smaller inner cylinder (15 cm) diameter was used.In order to analyze differences between these two types of infiltrometers, several tests were performed simultaneously with both devices in the Ebro valley experimental area to obtain a correction coefficient.The simple ring infiltrometer returned rather higher mean infiltration rates (2.74 times); therefore, this coefficient was applied to all results obtained from the Ebro Valley tests.
Different scenarios of land use and vegetation cover were selected for the experimental plots, taking the most representative environments of each study site (Pyrenees, Ebro Valley and Cap de Creus).
Experiments were carried out under different soil moisture conditions, including situations of high, medium and low humidity.Former infiltration studies enabled an average value of saturated hydraulic conductivity (K sat ) to be calculated, corresponding to the final value of infiltration rates obtained in each test (infiltration is the term that will be used throughout the paper).Data from all tests were analysed using simple descriptive statistics (median, average, standard deviation, maximum, minimum and percentiles), bivariate linear correlations, ANOVA and Bonferroni tests for multiple comparisons (IBM SPSS Statistics 22, 2013), which were applied to the total data set and to the average values obtained for each type of land use and vegetation cover.
The variables included in the study were: study area, land use, type of vegetation cover, altitude (m a.s.l.), median and average infiltration (mm h -1 ) and soil moisture (percentage value) measured through sampling at 0-5 cm soil depth.3, including land use (20 types), vegetation cover (6 types), number of tests performed for each land use, as well as average, median and standard deviation of moisture and infiltration rates.Land uses were classified from lower to higher infiltration rates.A total number of 195 tests were performed: 20 were carried out in the Pyrenees (Jaca), 64 in the Ebro valley (Zuera) and 111 in the northeastern coast (Cap de Creus) areas.The distribution of the tests was: 3 in bare soils, 25 in grass, 47 in scrub, 62 in forests, 8 in burned scrub and 50 in burned forest.

Infiltration variability in relation to the study sites
Figure 2 shows average, median and standard deviation values from the 195 infiltration tests.Land uses were classified by ascending average infiltration rates to analyze whether the distribution showed any interesting relationship with land uses, study sites and type of vegetation cover.Median and, especially, standard deviations were not in proportion to the average infiltration rate, indicating that its variance is not proportional to its value.However, the most important observation from the classification of land use was that distribution seemed to be fairly well related to the study site.Thus, the lowest values corresponded to the Pyrenees (Jaca), the intermediates were on the northeastern coast (Cap de Creus) and the highest were in the Ebro valley (Zuera).This result must be statistically explored to assess whether this factor should be taken into account in data analysis.Linear correlations between quantitative variables (average infiltration, altitude and moisture), showed a good correlation of altitude with moisture and average infiltration, although it was somewhat lower in this case, and no correlation was found between average infiltration rates and moisture (Table 4).This result suggests that the average infiltration rates may be more strongly linked to the characteristics of the study sites than to soil water status.It seems logical to assume that infiltration rates were conditioned by the lithology of the substrate and the type of soil, since they showed significant differences in texture, average organic matter and soil depth (Table 2), although only soil depth showed a significant correlation with average infiltration rates (R: 0.568; σ: 0.009; N: 20).Therefore, when analyzing the variance of infiltration rates with respect to type of land use and vegetation cover, the depth of the soil is implicitly taken into account.The effect of the study site on infiltration rates and soil moisture was further analysed by statistical descriptors and by applying statistical comparative methods.Figure 3 shows that infiltration rates differed between study sites, but soil moisture presented a similar range of values.Results from the ANOVA and Bonferroni tests confirmed that infiltration rates were statistically different in all study sites (Table 5).Results of a moisture comparison from the ANOVA test showed differences between the sites, but the Bonferroni test confirmed that only the northeastern coast (Cap de Creus) differed from the Pyrenees (Jaca) and Ebro valley (Zuera), which showed similar moisture values.Land use and types of vegetation cover were more heterogeneous in the Pyrenees (Jaca) and northeastern coast (Cap de Creus) than in the Ebro valley (Zuera), where experimental conditions always took place in a forest or scrub environment.In this case, the only differential factors in defining land use were the effects of fire, the bedrock (limestone or gypsum) and its respective type of soil.Thus, these results confirm that the highest infiltration rates and least dispersion of soil moisture values occur in the environment affected by fire, in contrast with the lower infiltration and more dispersed soil moisture values observed in the Pyrenees area (Jaca), while in the northeastern coast (Cap de Creus) the values were intermediate in comparison with the other two sites (Figure 3).It should also be noted that the difference in infiltration rates observed among the three study sites could be partly due to variations in the experimental procedures, differences between the devices used, or even because tests were performed by different operators in each area.

Infiltration and soil moisture variability in different land uses
Figure 4 shows the statistical descriptors of the average infiltration rates obtained in each land use.The first thing to be noticed was that ranges vary widely and their variability has no relation to the average value of infiltration.In any case, such seemingly obvious differences among land uses must be confirmed by comparative statistical analysis before they can be adequately described.Figure 5 shows the statistical descriptors for the moisture values in each experimental plot.It is evident that there is no relationship between the average rates of infiltration and initial moisture in the experimental plots.This observation is consistent with the absence of linear correlation presented in Table 4.
The variability of infiltration rates between the land uses was analysed by a Bonferroni test (Table 6).The results indicate that land uses in the Ebro Valley (Zuera) show greater difference from the other two study sites, whereas most of the land uses in the Pyrenees (Jaca) and north-eastern coast (Cap de Creus) show statistical similarities.The land use that returns the greatest differences from the others is burned forest limestone, which is only similar to two of the other uses in the same study site.On the contrary, scrub gypsum in the Ebro Valley (Zuera) has similarities with all land uses in the experimental areas of the Pyrenees (Jaca) and north-eastern coast (Cap de Creus), with the only differences showing in two land uses (burned forest limestone and burned forest north) within its own study site.With respect to the land uses of the Pyrenees (Jaca) and north-eastern coast (Cap de Creus), the most outstanding result is observed in vineyards (which display the greatest differences from the others in these study sites, but greater similarities with those of the Ebro Valley).Olive groves and north-facing pasture are the other land uses in the north-eastern coast (Cap de Creus) and Pyrenees (Jaca) that have certain similarity with some uses in the Ebro Valley (Zuera), in addition to the scrub gypsum mentioned above.It is important to emphasize the position listed for grass badland (Pyrenees), which surprisingly is above that of the two pastures (Pyrenees and northeastern coast) and the southern forest (Pyrenees).The explanation is that, in one of the tests carried out in the grass badland, a very high infiltration rate was obtained, the highest of all the tests carried out in the Pyrenees (Jaca) experimental area.The statistical descriptors for Figure 4 give an excellent view of the effect of this possibly anomalous result.Listed at position 5 of land uses, it is easy to appreciate that values are very highly dispersed and the difference between the median and average value is too large, which confirms the occurrence of some irregularity between this result and those obtained in the other tests.
For a more in-depth study of the differences and similarities between land uses within each study site, ANOVA and Bonferroni tests were performed on each one.Table 7 presents the results of the statistical comparisons between land uses in the same experimental site, but in this case the analysis for the Pyrenees (Jaca) study site was done twice, the first included the previously mentioned anomalous result from the grass badland plot, which was excluded in the second.The analyses confirmed that, in the Pyrenees experimental site, the result of the anomalous test conditioned the statistical comparison, so that if included, the ANOVA test indicated that there was no statistically significant difference in infiltration rates for different land uses.However, if this result was excluded, the ANOVA test identified at least one use that showed significant differences, while the Bonferroni test identified the difference as being associated with the north-facing meadow, which only shows similarity with the northern forest.1.000 1.000

0.340
The effect of land use and vegetation cover on soil infiltration Cuadernos de Investigación Geográfica 43 (1), 2017, pp.141-169 155 The comparative analysis between the land uses in the experimental site of the northeastern coast (Cap de Creus) reveals that the greatest number of differences between uses occurs mainly with vineyards (which are only similar to olive groves), and olive groves (which are only similar to vineyards and pine trees).Lastly, the Ebro Valley (Zuera) experimental site shows the greatest homogeneity of results, especially considering the result from the Pyrenees (Jaca), where the anomalous test was ruled out.The land uses with the largest differences from the others are scrub gypsum (burned forest north and burned forest limestone) and burned forest limestone (scrub gypsum).In general, these results are consistent with those shown in Table 6, but also suggest that the comparison between the results may show some variations if an unreliable outcome is included or excluded.In any case, it is evident that the land uses show some remarkable differences, although probably the ones between experimental sites are more important than those within the experimental sites (see Fig. 3A and 4).

Analysis of results from 6 vegetation types
Data analysis was performed using six groups, each with similar vegetation (Table 8).The main difference from the other analyses was that, in this case, they are mixed infiltration rates obtained from different situations in experimental areas and land use.The distribution of infiltration rates according to vegetation type suggests significant differences, with lower values found on bare soils, followed by grass, scrub and forest, and the higher values recorded in burned environments, such as burned scrub and burned forest (Fig. 6).The high variance in each type of vegetation, except for bare soil, can be partially explained by the large number of different land uses included in each category.This does not occur in the burned scrub group, whose wide dispersion of infiltration rates is linked to a single land use.The average and median values of each vegetation type suggest that the most significant differences observed in The effect of land use and vegetation cover on soil infiltration Cuadernos de Investigación Geográfica 43 (1), 2017, pp.141-169 157 the transition between groups are from grass to scrub, and from forest to burned scrub, so that pairs of groups formed by bare soil-grass, forest-scrub and burned scrub-burned forest seem to show high similarities among each other.In any case, it is necessary to verify these observations through comparative statistical analyses (ANOVA and Bonferroni).Statistical results confirmed that the variance of infiltration rates can be explained by the type of vegetation cover, if only partially (Table 9).Differences were mostly found between burned environments and all other vegetation groups and between grass and forest.However, the coefficient between the grass and scrub groups is very close to the statistical significance (0.059), which is in agreement with the differences between the average and median values of infiltration rates of both groups (Fig. 6), but probably the null hypothesis is rejected by the extreme values included in the grass group.No significant differences were observed between the other vegetation groups.However, these results are somewhat surprising to say the least, since more differences could have been expected between bare soil and grass, scrub and forest groups.Probably the most decisive factor is the location of the study area and the fact that bedrock determines soil characteristics, although the effect of land use can also contribute to a large extent.These results confirm that the vegetation cover type is not a determining factor in explaining the variance in the infiltration process.However, these results suggest that some types of vegetation cover can affect the infiltration process more than others, since the statistical comparison returns significant differences, even if the analysis includes a variety of land uses and soil characteristics.On the other hand, statistical descriptors of infiltration rates provide another option for analyzing differences between vegetation groups, by obtaining their quotients.Figure 7 shows the results of comparing correlative vegetation groups, following the same ascending order as given in Figure 6.This procedure has only been applied between successive groups to show results in a single graph.Descriptive values above 1 indicate an increase over a previous vegetation group and a quotient below 1 means a decrease.The first noteworthy item is that the differences do not affect all descriptors (i.e., average, median, maximum, minimum and standard deviation) equally, as there are large differences in one and very small ones in another.In this respect, it can be observed The effect of land use and vegetation cover on soil infiltration Cuadernos de Investigación Geográfica 43 (1), 2017, pp.141-169 that the difference in maximum rate of infiltration between bare soil and burned forest is practically the same as between bare soil and grass.The same occurs with the standard deviation, which shows the largest variation between bare soil and grass, but varies very little among the following groups until reaching burned forest.However, average, median and minimum values show significant variances between grass and burned scrub, especially in the minimum rate of infiltration.This means that the vegetation cover has little influence on the initial infiltration capacity, but it does have a determining effect on the persistence and stability of the infiltration process.The second is the large difference in minimum infiltration rates between burned scrub and forest, which logically must be greater if compared with the scrub, grass and bare soil groups.The third observation is the great similarity between both scrub and forest, likewise observed between burned scrub and burned forest.This method of comparison coincides quite well with the results of the statistical comparison, but also shows some differences that, without being statistically significant, should not be neglected.

Effect of the study site
The first factor is linked to the characteristics of each study area, which includes geographical, topographical and geological aspects determining the type of soil and its physical properties, which are critical for the hydrological response (Rawls et al., 1993) and hydric properties (Seeger et al., 2006).Only infiltration rates showed clearly different ranges in the three study areas, but this was not observed with moisture values (Fig. 3 and Table 5).The effect of moisture has been evaluated in some studies in the opposite way, as a determining factor which can positively affect the infiltration capacity (Kirkby, 1978), since moisture is one variable that can influence the temporal evolution of surface hydraulic conductivity (Zhou et al., 2008).Although, it has been observed that the relationship between these two variables may depend on the degree of development of soil and vegetation cover (Regüés et al., 2012), or even that there may be no relationship (Hu et al., 2009).In this study, the variability of infiltration rates was separate from soil moisture (Table 4).
In any case, it must be noted that the Pyrenees (Jaca) area showed the lowest average infiltration values, at a ratio of 1:2 against the northeastern coast (Cap de Creus), and 1:6 for the Ebro valley (Zuera), and also returned the smallest range of dispersion (Fig. 3A).However, the moisture values indicated that the northeastern coast (Cap de Creus) is the driest area and showed less fluctuation, while the moisture values in the Ebro valley (Zuera) and Pyrenees (Jaca) areas were significantly higher and more similar (Fig. 3B).It seems obvious, even without looking at the statistics, that moisture values do not explain the differences in infiltration rates from the three study areas because of the absence of proportionality among them.

Effect of land use
Land use classification by ascending infiltration rates (Fig. 2) was used to analyse the statistical descriptors of infiltration rates and revealed that the classification is strongly influenced by the study area (Fig. 4), while moisture distribution shows no relation to land use (Fig. 5).Comparative statistics between infiltration rates and land use returned only a few differences but several similarities (Table 6), even when the analysis was done in the respective study sites, although in this case more differences among land uses were detected (Table 7).
These results are partially consistent with findings from previous studies, which have shown that different land uses are associated with different physical, chemical and biological properties of soils (Badía and Martí, 2003;Pardini et al., 2004a) determining their hydric properties and infiltration capacity (Boarmann and Klaassen, 2008;Zhou et al., 2008;Hu et al., 2009).Previous work has shown that both the infiltration capacity and saturated hydraulic conductivity are particularly sensitive to changes in land use (Alegre and Cassel, 1996;Schoenholtz et al., 2000;Zimmerman et al., 2006;Hu et al, 2009).It has also been observed that the dispersion of infiltration rates in a type of land use can be very high and is probably associated with temporal and spatial variations of hydric properties of the soils (Sisson and Wierenga, 1981;Mallants et al., 1996;Moret and Arrúe, 2007;Li et al., 2009).This variability may be even greater than that observed among different land uses (Zahou et al, 2008;Hu et al, 2009).This information agrees with the results from our study, where the variance of infiltration rates is probably due to differences in conditions in the study sites, rather than to those of land use.
Land use classification shows that the order of infiltration rates adheres quite closely to differences among the three study areas (Fig. 3A).It also shows that, with this grouping, land use classification for comparable vegetation cover types tends to be close.Lower infiltration values are generally linked to land use with less developed vegetation cover and soil (Tables 2 and 3).Intermediate infiltration values relate to scattered trees The effect of land use and vegetation cover on soil infiltration Cuadernos de Investigación Geográfica 43 (1), 2017, pp.141-169 161 or scrub, followed by forest, and the highest infiltration values are frequently present in scenarios affected by fires.However, it is surprising to find coastal vineyards in the upper middle part of the classification.The higher infiltration rates found in these vineyards may be because it is a semi-abandoned crop and, as has been observed in previous studies, agricultural practices can result in a significant temporary variation of the infiltration capacity (Zhou et al., 2006).This explains the differences with the infiltration rates of land uses with scrubs, except for the land use burned scrub gypsum in the northeastern coast (Cap de Creus).The infiltration capacity in meadows is reduced by grazing (Godsey and Elsenbeer, 2002;Zimmerman and Elsenbeer, 2008), which agrees with the lower infiltration rates recorded in the south-facing pasture meadow in the Pyrenees (Jaca) or in the north-facing pasture meadow in the northeastern coast (Cap de Creus) compared with north-facing meadow in the Pyrenees (Jaca).It should also be noted that the exposure of the slopes is one of the determining factors explaining infiltration rate variation in land use in the Pyrenees (Jaca) where south-facing slopes have less developed soil and vegetation cover than the north ones (Navas et al., 2008;Regüés et al., 2012), and this is why infiltration rates in south-facing forests and pasture are so low in relation to similar land uses on north-facing slopes.This effect has also been observed among the north and south burned forest of the Ebro valley.

Effect of vegetation cover
The third explanatory factor of infiltration rates is the type of vegetation cover.The infiltration rate classification according to the type of vegetation shows that both, the magnitude and dispersion of the values, are different for each group (Fig. 6).The comparative statistical analysis confirms that vegetation cover has some effect on infiltration rates (Table 9).It is important to note that the infiltration variability range in vegetation groups is higher than for land use (Fig. 4), which results from integrating different study areas and land uses into several vegetation groups, except those of bare soil and burned scrub (Table 8).
Previous studies provide useful information that can be assimilated into the results of this study.Low infiltration rates are recorded in bare soil due to the typical surface sealing in badlands (Regüés and Gallart, 2004;Nadal-Romero and Regüés, 2009), which agreed with the properties of soils developed on marls (Badía et al., 2015).Slope aspect has often been identified as a factor that determines soil degradation, vegetation cover and infiltration rates (Seeger et al., 2006;Regüés et al, 2012;Robertson and Sharp, 2015), especially in areas affected by farmland abandonment (Lana-Renault, 2009).Significant differences were observed between grass and forest, but not between scrub with grass and forest (Table 9), although the difference between grass and scrub is very close to the statistical significance mentioned above (paragraph 3.3).This can also be seen from the statistical descriptors (Fig. 6).Since scrub and forest show similar infiltration rates, our analysis focuses on the differences between grass and forest (which can be comparable to differences between grass and scrub).Differences in infiltration capacity between grasses and forest are usually associated with the hydric properties of forest soils, which are more stable in depth than in other environments Cuadernos de Investigación Geográfica 43 (1), 2017, pp.141-169 (Price et al., 2010).The saturated hydraulic conductivity K sat at depth is also lower in meadow than forest soils (Alegre and Cassel, 1996;Martinez and Zinck, 2004;Price et al., 2010).It has been estimated that, in some cases, infiltration capacity in forest compared to meadow and pasture can be more than one order of magnitude higher (Zimmermann et al., 2006).Our study estimated a difference between forest and grass of about 2.5:1 for the average rate (Fig. 7).It was also observed that deforestation causes a reduction in infiltration capacity (Ghuman et al., 1991) and increases the frequency and duration of the occurrence of perched water tables (Cox et al., 1996;Rochefeller et al., 2004Lana-Renault et al., 2007).In contrast, runoff in forest environments basically depends on sub-surface flows (Freeze, 1977;Beven, 1982;Serrano-Muela et al., 2008) so that groundwater levels fluctuate rapidly and do not reach the surface (Germer et al., 2010;Serrano-Muela et al., 2010).In this regard, it is clear that the expansion of scrub and forest have hydrological consequences (García-Ruiz and Lana-Renault, 2011).
There was a notable increase in infiltration rates between forest and burned scrub, especially for the minimum values, which was not observed between burned scrub and burned forest; this agreed with the slight differences observed between scrub and forest (Fig. 7).These differences could be related to the higher dispersion of the values in the forest group (which may be related to the fact that this group includes forests from the three study areas), whereas burned scrub only includes one land use in one study area (Table 8).In any case, these differences would agree with the results of previous studies, which found that the infiltration capacity of soils affected by fire increases after being covered by ash (Ebel et al., 2012;Leon et al., 2015b).Finally, despite the similarities shown by statistical comparison between the burned scrub and burned forest groups, it is worth mentioning some differences in comparative land uses of burned forest (Table 7 and Fig. 4).Thus, gypsum soils generally have a lower infiltration capacity than calcareous soils, as has been observed in previous studies (Cerdà, 1996;Badía and Martí, 2000), although the aforementioned effect of ash cover on gypsum soils is less evident than on calcareous soils (León et al., 2015b).The small effect of the ash layer on the infiltration rates of gypsum soils is evident from the comparison between gypsum forest and gypsum burned forest.However, this effect is less evident between scrub gypsum and burned scrub gypsum, which show important differences in infiltration rates.

Conclusions
The study of a wide variety of land uses, in 20 scenarios located in three different areas in Spain has enabled differentiating factors that best explain the variability of infiltration capacity.The most decisive ones are related to the characteristics of the study area, including soil type, land use and the type of vegetation cover.There does not seem to be a link with the preceding soil water status, although it might have an effect on a single land use.Therefore, the differences must relate to a combination of several factors, indicating the complexity inherent in carrying out comparative studies that include a wide range of scenarios.
The comparative analysis of infiltration rates linked to a large number of land uses allows accurate evaluations of the cause-effect relationships.This provides more detailed information than that obtained by grouping land uses into types of vegetation cover, although the more general analysis based on types of vegetation cover can turn out to be even more interesting, because it provides general information that could be valuable for land management on larger scales.As an example, the infiltration capacity of forest soils, compared with grass and pastures, results in an increase similar to that observed in previous studies, although significantly lower in our case.This might be a consequence of integrating several study areas from different experimental sites.In fact, an important conclusion from the analysis is the confirmation that variance of infiltration rates associated with land uses can be less than the variance between different study areas.
The results from this study contribute additional insights into the hydric effects from changes in land use, providing information that can be used to assess the implications of changes in vegetation cover in the hydrological balance.The importance of this information is related to the current expansion of forest cover, causing a reduction in flow rates and availability of water for storage.The most valuable result provided by this comparative analysis is probably the integration of data from different areas representative of several environments in the Mediterranean basin.This fact can be considered as a useful reference, both from the scientific and socio-economic point of view.

Aknowledgements
Field infiltration tests were done with a grant from the CETSUS (CGL2007-6644-C04-01/HID) project funded by the "Plan Nacional I+D+I 2004-2007" from the Spanish Ministry of Science and Innovation.The current analysis and all the expenses associated with the publication of this study were funded by ESPAS project Ref.: CGL2015-65569-R (MINECO/FEDER).Estela Nadal-Romero is the recipient of a "Ramón y Cajal" postdoctoral contract (Spanish Ministry of Economy and Competitiveness).

Figure 1 .
Figure 1.Situation of the study areas in the Iberian Peninsula: Pyrenees/Jaca area (red circle), Ebro valley/Zuera (orange circle) and north-eastern coast/Cap de Creus (blue circle).

Figure 2 .
Figure 2. Average, median and standard deviation values of infiltration rates.Land uses are ranked from low to high average infiltration values.Py: Pyrenees/Jaca; C: north-eastern coast/Cap de Creus; Ev: Ebro valley/Zuera.

Figure 6 .
Figure 6.Infiltration rates for each vegetation type group (bare soil, grass, scrub, forest, burned scrub and burned forest).

Figure 7 .
Figure 7. Quotient between the statistic descriptors of infiltration from correlative vegetation groups, following the order shown in Figure 6.

Table 1 .
Main features of the 20 land use/vegetation cover types in the study.

Table 2 .
Descriptive characteristics of the soils regarding its study site, vegetation cover and land use.

Table 3 .
Descriptive values of infiltration test conditions: land use, vegetation type, number of tests performed, average and median values of moisture and infiltration rates (including standard deviation of moisture and infiltration rates).Land uses are classified from lower to higher infiltration rates.

Table 4 .
Results of lineal correlation for each of the 195 infiltration tests, the analysis includes 3 variables (altitude, infiltration rates and soil moisture).

Table 5
. Statistical comparison (ANOVA and Bonferroni test) of the infiltration rates and moisture for the 3 study sites.Significant coefficients are in bold.

Table 6 .
Statistical comparison through Bonferroni test of average infiltration rates from each land use.Py: Pyrenees (Jaca); C: north-eastern coast (Cap de Creus); Ev: Ebro valley (Zuera).Significant coefficients are in bold.

Table 7 .
Comparison (ANOVA and Bonferroni tests)of infiltration rates for land use in the respective study sites: Pyrenees/Jaca (6 land uses, 20 tests); north-eastern coast/Cap de Creus (7 land uses, 111 tests); Ebro valley/Zuera (7 land uses, 64 tests).Only the significant differences are provided.Results of Pyrenees/Jaca marked with an asterisk were obtained without the exceptional value recorded in a grass badland experimental test.

Table 8 .
Summary of vegetation group features.L: low, M: medium; S: South, N: North.Between brackets the number of soil uses included in the group from its respective experimental site.

Table 9 .
Comparison, through ANOVA and Bonferroni tests, ofthe infiltration rates returned for the type of vegetation cover.Significant coefficients are in bold.