Introduction

The stream–riparian interface is a relevant control point for fluxes and retention of nutrients between terrestrial and aquatic systems (Dahm and others 1998). The stream–riparian interface is the zone connecting the stream and the catchment and it has been defined as a spatially fluctuating ecotone between the surface stream and the deep groundwater (Boulton and others 1998). Consequently, several field studies have investigated the stream–riparian interface in regulating stream hydrology (McGlynn and McDonnell 2003; Butturini and others 2002) or biogeochemistry (Schindler and Krabbenhoft 1998; Vidon and Hill 2004; Wigington and others 2003).

In Mediterranean regions, streams are frequently characterized by strong seasonality with a humid period followed by a drought, normally in summer (Gasith and Resh 1999). In drought there is a disruption in hydrological connectivity that ranges from flow reduction to loss of hydrologic connectivity between surface water, groundwater and the riparian zone (Lake 2003). Streamflow recovers with the arrival of autumn rains. During this transition period between dry and wet conditions the hydrology of the stream–riparian interface is highly dynamic due to: (1) abrupt changes in groundwater levels; (2) occurrence of reverse fluxes in the subsurface stream–catchment interface; (3) rapid expansion and shrinkage of the boundary of the stream–catchment interface; (4) longer riparian groundwater flow-paths (Butturini and others 2003).

Recent studies in the Mediterranean adduce evidence of a severe alteration of precipitation and hydrological regimes and hypothesize an increase in the frequency of extreme events particularly in summer (Christensen and Christensen 2003; Schröter and others 2005). In this context, it is essential to assess how the stream–riparian interface controls the transport and fate of nutrients that flow in Mediterranean freshwater ecosystems under the extreme effects of dry–wet hydrological shifts that characterize the transitional period between summer and autumn.

Within this framework, the dissolved organic carbon (DOC) transported in streams is essential to secondary production in freshwater and marine ecosystems. The effects of temperature, atmospheric CO2, or hydrology on DOC export to oceans are still unclear and a matter of debate (Freeman and others 2001a,b, 2004 Tranvik and Jansson 2002; Clark and others 2005). The quantity and the bioavailability of DOC in streams is the combined result of flushing from the watershed, leaching of leaves and branches from the riparian vegetation, and in-stream primary production (Sachse and others 2005) and its temporal dynamics are usually governed by discharge regime (Butturini and others 2005; Neal and others 2005) and the occurrence of drought periods (Bernal and others 2002). Nevertheless, there is little information about the effect of the abrupt hydrological processes that take place in the stream–riparian interface during the transition between dry and wet conditions on reactivity and transport of DOC in streams. Albeit, some initial studies have been undertaken in Mediterranean streams covering DOC transport and bioavailability after drought periods (Romaní and others 2005).

Our main objective is to explore, in situ and under natural conditions, the effect of the antecedent hydrological conditions on temporal dynamics and the fate of total DOC (TDOC) and discrete dissolved organic molecular fractions transported across the stream–riparian interface to improve our knowledge of the possible effect of altering the discharge pattern on stream DOC dynamics. For this purpose, data from a forested Mediterranean stream were collected during two autumnal periods characterized by contrasting antecedent hydrological conditions. In 2003, the stream was dry in summer owing to few, small intensity, rain episodes in the precedent spring–summer period, in 2004 the streamflow was permanent throughout summer due to high levels of precipitation in the spring–summer period and, consequently, the antecedent conditions were wet and the hydrological transition usually observed at the end of the summer was nonexistent.

Study Site

Fuirosos is a third-order stream that drains a forested granitic catchment of 10.5 km2, near Barcelona (NE Spain, 41°42′ N, 2° 34′ W, 50–770 a.s.l.). The climate is typically Mediterranean, with monthly mean temperatures ranging from 3°C in January to 24°C in August. Precipitation mostly falls in autumn and spring with occasional summer storms. Average annual mean precipitation for this region is 750 mm (Ninyerola and others 2000).

The catchment is covered mostly by perennial cork oak (Quercus suber) and pine tree (Pinus halepensis) with one or two layers of shrubs (for example, Rhamnus alaternus, Viburnum tinus, Arbutus unedo, Prunus spinosa) and lianas (Lonicera implexa, Smilax aspera). Deciduous woodland of chestnut (Castanea sativa), hazel (Corylus avellana) and oak (Quercus pubescens) predominate in the valley head. The soils are poorly developed with an A horizon always less than 5 cm. Soils are sand (46%) and fine sand (24%), with smaller amounts of silt and clay (15% each) (Sala 1983). Traditional land uses comprise periodic harvesting of bark from cork trees and partial clearing of pines and shrubs. Agricultural fields occupy less than 10% of the catchment area.

According to the description by Gasith and Resh (1999), Fuirosos stream exhibits a typical Mediterranean-type discharge regime and several biogeochemical studies used it as a model of Mediterranean catchments to perform comparisons across climate gradients (Sabater and others 2003; Burt and others 2002; Hefting and others 2004; Wade and other 2004).

The stream channel is 2–5 m wide and is characterized by steep-pool morphology with cobbles and boulders. The mean flow is 7–20 l s−1. Discharge is intermittent. The flow is interrupted by a long dry period in summer followed by an abrupt recharge period in late summer–early autumn. The subsequent humid period lasts until late spring. The stream–riparian porous media is characterized by relatively high hydraulic conductivity (4.8–19 m day−1) and variable specific discharges (0.03–1.5 m day−1). During the stream recharge period, the groundwater level increases abruptly and the stream water can infiltrate 10 m into the near-stream riparian zone (Butturini and others 2003). After this recharge, near-stream riparian groundwater level fluctuations followed the same pattern as the stream water, which reflects the hydrological connection between the two water bodies (Butturini and others 2003).

A well-developed riparian forest flanks the 10–20 m wide stream channel, consisting mainly of plane tree (Platanus × hispanica) and alders (Alnus glutinosa). The riparian soil is poorly developed and plane leaf litter tends to accumulate on the forest floor because of extremely low decomposition rates (Bernal and others 2004). In summer, during the dry period, groundwater levels fall from 1.70 to 2.60 m below ground surface causing hydrological stress to the riparian forest area. This results in a high input of leaf litter that accumulates on the streambed and margins (Sabater and others 2001; Acuña and others 2004).

DOC concentrations in stream water at basal discharge conditions are between 2 and 4 mg l−1. However, during the hydrological transition (September–October), between the dry and wet periods, DOC concentrations increase to 10–20 mg l−1 (Bernal and others 2002).

Methods

Sampling Strategy

Field sampling was carried out in two consecutive years, 2003 and 2004. These two years had contrasting precipitation regimes, especially in spring and summer. In 2003, the total precipitation of the hydrological year (September–August) was 630 l m−2, but the accumulated precipitation for the spring and summer period was only 120 l m−2. Consequently, the streamflow was interrupted because of the summer drought. The 2004 hydrologic year was much more humid (total precipitation of 815 l m−2) due to abundant precipitation in spring and summer (300 l m−2) that permitted a permanent streamflow during summer (Table 1).

Table 1. Antecedent Hydrological Conditions for Each Study Period: Total Precipitation for the Spring-Summer Period and Number of Days without Surface Flow.
Full size table

The water samples were collected from September to November of 2003 and 2004, respectively. The 2003 study period (September 5–November 19) refers to dry antecedent hydrological conditions, whereas 2004 (September 6–November 11) refers to wet antecedent conditions. Hence, where the text refers to “study period” or years 2003 and 2004 it refers to that year’s September–November sampling period. During these study periods, the samples were collected from four water bodies whenever it was possible: (1) stream water, (2) groundwater from the near-stream riparian zone, (3) groundwater from a 5 m deep well, located on a forest hill slope, 200 m away from the stream channel, and (4) superficial water from an ephemeral channel that drains the forest soil from a small sub-watershed of 6.54 ha (Figure 1). In 2004, it was not possible to collect samples from this sub-watershed owing to insufficient rainfall episodes to generate superficial runoff in that area.

Figure 1.
figure 1

Fuirosos catchment with sampling sites marked and main land uses in different grey shades.

Full size image

Samples from the riparian groundwater and the well on the forest hill slope were collected using an ISCO field peristaltic pump and, in 2004, with a Sigma 900 Max Sampler, from a well (2.5 m deep, 15 cm diameter) located 3 m from the stream channel. In 2003, eight samples were collected from the stream and riparian ground waters. As the well in the forest hill slope and the ephemeral channel were dry at the beginning of the sampling period no samples could be collected until a rainstorm on the 15 October. In 2004, eight more samples were collected from each water body from the beginning of the study period, except from the ephemeral channel, which was dry.

Samples were preserved at 4°C for storage prior to analysis. Samples were analyzed for DOC and its fractions (ultrafiltration). Chloride and silica were also analyzed and used as conservative tracers to discern the hydrological origin of stream water and riparian groundwater (Hill and others 1998; Hornberger and others 2001; McGlynn and others 2004).

Chloride was estimated using capillary electrophoresis (Waters CIA Quanta 4000) (Romano and Krol 1993) whereas silica and sodium were measured with an induced coupled plasma-atomic emission spectrometer (Thermo Jarrell Ash Iris Advantage ER/S).

Hydrology and Dynamics of Conservative Solutes

Water levels in the stream, ephemeral channel and riparian groundwater were monitored constantly by water pressure sensors (Campbell CS401) connected to a data logger (Campbell CR10X). In the riparian area, 24 wells were placed in a regular grid (4 rows and 6 lines). The data from the groundwater levels of the entire riparian plot allowed us to determine the direction of the groundwater flow and the extension of water exchange at the stream–riparian interface (For additional details see Butturini and others 2003). In this study, the groundwater level in the riparian plot was monitored manually three times in 2003 (September 12, October 13, November 22) in each well and once in 2004 (September 15). Stream water discharges were measured by the chloride “slug” addition method (Gordon and others 1992). Then, continuous discharges were estimated using an empirical relationship between measured discharges and the corresponding stream water levels.

DOC Characterization

Water samples were filtered through pre-combusted Whatman GF/F and 0.2 μm porous size nitrocellulose filters (Whatman) to eliminate particles and bacteria before ultrafiltration. In 2004, samples were ultrafiltrated immediately after sampling (maximum 24 h), so these were only filtered with pre-combusted Whatman GF/F filters. From each filtered sample, subsamples of 20 ml, for TDOC determination, and subsamples of 330 ml (in 2004), for conservative solutes content analysis, were collected in glass flasks. These flasks were previously heated for 4 h at 400°C to prevent sample contamination. Each filtered sample was ultrafiltrated with Prep-Scale TFFI cartridges (Millipore). Typically, the initial volume for ultrafiltration was 10 l, but on occasions when not enough water could be collected the initial volume was 5 l. The following molecular weight fractions were obtained for each sample:

  1. (1)

    > 100 kDa (very high molecular weight or VHMW fraction),

  2. (2)

    10–100 kDa (high molecular weight or HMW fraction),

  3. (3)

    1–10 kDa (medium molecular weight or MMW fraction) and

  4. (4)

    <1 kDa (low molecular weight or LMW fraction).

DOC samples were measured using a Skalar 12 SK TOC Analyser with UV-promoted persulfate oxidation.

There were three replicates for each molecular weight fraction of each sample (20 ml). DOC analysis of samples that were not ultrafiltrated provided the measurement of total DOC (TDOC, three replicates).

Results

Hydrology

Stream hydrology during 2003 and 2004 showed important differences in antecedent hydrological conditions, discharge magnitudes and dynamics (Table 1).

In 2003, the stream was dry from 30 June until 4 September. The streamflow was re-established on 5 September after several rain events. Stream hydrology was characterized by two contrasting sub-periods (Figure 2a). The first was relatively dry and lasted from 5 September to 14 October and was characterized by an intermittent and low discharge regime (less than 4 l s−1). The second sub-period was much more humid and started on 15 October after a severe rain episode (total precipitation = 186 mm) that generated a storm peak of 2,500 l s−1 and a basal discharge up to 15 l s−1. During the dry sub-period, the stream water discharged into the riparian groundwater up to 8 m within the riparian area and generated extended groundwater flow paths (Figure 3a, b) at the stream–riparian interface. Subsequently, the stream water and riparian groundwater levels followed the same pattern over the entire study period (Figure 2b) and the stream water infiltrated only the first 2–4 m of the riparian strip (Figure 3c). The well located on the forest hill slope and the ephemeral channel remained dry during the first sub-period. The runoff in the ephemeral channel was permanent from 15 October to 8 November and followed the same discharge pattern as that of the stream channel (Figure 2c).

Figure 2.
figure 2

Hydrological characterization during 2003 (dry antecedent conditions) and 2004 (wet antecedent conditions) study periods: a) stream water discharge in 2003; b) groundwater level in the riparian zone (thin line) and in stream (thick line); c) discharge of runoff in the ephemeral channel (2003); and d) stream water discharge in 2004.

Full size image
Figure 3.
figure 3

Groundwater level in the riparian plot surface during three dates in 2003. a) September 12 2003; b) October 13 2003; c) November 22 2003. Arrows show the groundwater flow direction. Cross dot shows the location of the riparian groundwater well used for water sampling.

Full size image

In contrast to 2003, the stream was permanent throughout 2004 due to abundant rains in late spring. However, no rain episode occurred from September to November and the discharge was steady at about 4.4 l s−1 (Figure 2d). The lack of hydrological transition greatly reduced the extent of the water exchange between the stream and riparian ground waters, and the stream water infiltrated only the first 2–4 m of the riparian strip and the groundwater level surface was similar to that observed in November 2003. The hill slope groundwater was monitored throughout the study period, but the ephemeral subcatchment remained dry. The level of the riparian groundwater was coupled to the stream water level.

Conservative Solutes

In 2003, chloride concentrations in stream and riparian groundwater followed the same pattern (r = 0.95, n = 12, p < 0.001) and their temporal dynamics reflect the two contrasting hydrological sub-periods. During the dry sub-period, chloride concentrations were high (from 60 to 33 mg l−1) and decreased drastically at high discharges, during the wetter sub period, to 18 mg l−1. During this dry sub-period, the chloride concentration was higher in the riparian groundwater than in the stream (t-test, t = 7.2, df = 5, p < 0.01), but during the wetter sub period the concentrations were identical (paired t-test, t = 1.31, df = 5, n.s.). Throughout the entire 2003 study period, chloride concentrations in the hill slope groundwater (mean value of 32.3 ± 10.82 mg l−1) were significantly higher than in the stream and riparian ground waters (paired t- test, t = 4.8, df = 5, p < 0.01 for stream water and t = 4.68, df = 5, p < 0.01 for riparian groundwater) (Figure 4a). Silica concentrations in the stream and the riparian ground waters were identical (mean value= 9 mg l−1) and remained constant throughout the sampling period. In the hill slope groundwater silica concentration averaged 15.8 mg l−1 and was significantly higher than in the stream (t-test, t = 54.01 df = 5, p < 0.01) and the riparian groundwater (t-test, t = 44.53, df = 5, p < 0.05). Neither the ephemeral channel (t-test, t = 1.45, df = 3, n.s.) nor the riparian groundwater (t-test, t = 0.11, df = 3, n.s.) presented significant differences in chloride content with the stream. However, the silica content was slightly higher but not as much as in the hill slope groundwater (t-test, t = 2.24, df = 3, p < 0.05) (Figure 4c).

Figure 4.
figure 4

Temporal dynamics of conservative solutes (Cl and Si) during the sampling period of 2003 (dry antecedent hydrological conditions, panels a and c) and during the sampling period of 2004 (wet antecedent hydrological conditions, panels b and d). Symbols: (▀) stream water; (△) riparian ground water. (•) hill slope ground water; (□) ephemeral channel.

Full size image

In 2004, concentrations of solutes were constant over time due to the absence of storm events during the sampling period. In fact, the chloride content in stream (27.3 ± 1.18 mg l−1) and riparian ground waters (26.9 ± 1.24 mg l−1) was identical (paired t-test, t = 5.18, df = 6, n.s.), although it was significantly lower (24.35± 1.41 mg l−1) in the hill slope groundwater than in the stream water (paired t-test, t = 0.35, df = 7, p < 0.01). However, in contrast to the previous year, the difference in chloride content between the hill slope and riparian ground waters was not significant (paired t-test, t = 3.14, df = 6, n.s.) (Figure 4b). Silica content in stream water (7.92 ± 0.24 mg l−1) was significantly higher than in the riparian groundwater (7 ± 0.22 mg l−1; paired t-test, t = 17.13, df = 7, p < 0.01) and visibly lower than in hill slope groundwater (15 ± 0.13 mg l−1; paired t-test, t = 3.7, df = 7, p < 0.01) (Figure 4d).

Dynamics of Total DOC (TDOC) and Molecular Weight Fractions

In 2003, the peak in stream water TDOC (10 mg l−1) was measured at the beginning of the sampling period, just after the re-establishment of the streamflow. Afterwards, the concentration decreased until normal base concentrations were reached (2–4 mg l−1), and then increased during the floods of mid-October (until 8 mg l−1) and then decreased again to base values. TDOC in riparian ground and stream waters followed the same trend (r = 0.82, df = 8, p < 0.01) but concentrations in riparian groundwater were lower than in the stream water (paired t-test, t = 5.53, df = 7, p < 0.01). The TDOC concentration in the hill slope groundwater remained similar during the sampling period with a mean value of 1.7 ± 1.07 mg l−1. This concentration was significantly lower than in stream water (t-test, t = 5.54, df = 4, p < 0.01) but similar to the riparian groundwater (t-test, t = 0.83, df = 4, n.s.). Mid-October storms originated superficial runoff in the ephemeral subcatchment and TDOC concentrations were similar to those of the stream water (t-test, t = 1.41, df = 2, n.s.) (Figure 5a).

Figure 5.
figure 5

Temporal dynamics of TDOC and discharge during 2003, panel (a), and 2004, panel (b), study periods. Symbols: see Figure 4.

Full size image

In the 2004 sampling period, TDOC in stream water averaged 2.4 ± 0.5 mg l−1 and was clearly lower than in 2003 (paired t-test, t = 3.92, df = 7, p < 0.0). On the other hand, no differences were observed in DOC concentrations in riparian and hill slope ground waters between the 2 years (riparian, paired t-test, t = 1.51, df = 7, n.s.; hill slope, paired t-test, t = 2.35, n.s.) (Figure 5b). Similar to the previous year, TDOC concentrations in the stream water and riparian groundwater followed the same trend over time (r = 0.93, df = 7, p < 0.01), albeit they were significantly lower (mean value of 1.83 ± 0.68 mg l−1; stream, paired t-test, t = 5.26, df = 7, p < 0.01) in the latter compartment. The mean concentration in the hill slope groundwater was 0.54 ± 0.34 mg l−1, which is significantly lower than in the stream and the riparian groundwater (stream, paired t-test, t = 14.54, df = 7, p < 0.01; riparian groundwater, paired t-test, t = 6.6, df = 7, p < 0.01).

In 2003, the small DOC fractions (LMW and MMW) predominated over the larger fractions (HMW and VHMW) in all the monitored water bodies. For instance, the contribution of LMW+MMW in stream water, riparian groundwater, hill slope and in the ephemeral channel, averaged 68 ± 14%, 85 ± 7%, 81 ± 2% and 62% ± 8 % of TDOC, respectively. In stream water, these two smaller MW fractions had a DOC concentration peak at the beginning of the sampling period and followed the same pattern (df = 6, p < 0.05) as TDOC throughout the study period (Figure 6a, b).

Figure 6.
figure 6

Temporal dynamics of DOC MW fractions (expressed as % of TDOC) during the 2003 (dry antecedent conditions) (panels a,b) and 2004 (wet antecedent conditions) (panels c,d) sampling periods in stream water and riparian groundwater. Symbols: (

figure a
) LMW; (
figure b
) MMW; (
figure c
) HMW; (
figure d
)VHMW.

Full size image

In 2004, the concentration of the DOC fractions remained constant throughout the study period and the contribution of each MW fraction to the total TDOC was more homogeneous. For instance, the LMW and MMW fractions remained the most abundant DOC fractions, but the sum of their contribution decreased to 56 ± 5%, 56 ± 10% and 63 ± 25% in stream water, riparian and hill slope groundwaters respectively. The lack of any temporal DOC pattern in stream and riparian ground waters hinders the detection of any significant relationships between TDOC and MW fractions (Figure 6c,d).

Fate of DOC Across The Stream–Riparian Interface

TDOC concentration in the riparian groundwater was regularly lower than in the stream water for both years. In the sampling period of 2003, the 57 ± 18% of TDOC decreased at the stream–riparian interface, whereas in 2004 the decrease was only 28 ± 7% (Figure 7a).

Figure 7.
figure 7

Comparison of content of TDOC and MW DOC fractions measured in the stream water with that in the riparian groundwater during the two study periods. The solid line shows the 1:1 line. Data located within the grey area indicate that DOC concentration was higher in stream water than in riparian ground water. Symbols: (○)2003; (•) 2004.

Full size image

In 2003, the abrupt increase in water level in the riparian groundwater after the re-establishing of runoff in the stream channel and the data from conservative solutes (Cl and Si) revealed that it was clearly being fed by stream water and the influence of the hill slope groundwater was not apparent. Hence, the decrease of 57 ± 18% of TDOC concentration observed between stream and riparian ground waters was due to in situ retention at the stream–riparian interface.

In 2004, the stream and riparian ground waters showed similar geochemical characteristics suggesting the same conclusion obtained from data from 2003. Stream and riparian ground waters, for instance, had a similar Si content, which was much lower than in the hill slope groundwater. However, chloride content in the stream water, riparian and hill slope ground waters were similar. In addition, the lack of temporal change in the hydrometric data made it impossible to obtain an accurate picture of the water exchanges between the stream and riparian ground waters. Therefore, the dilution between the stream water and hill slope groundwater at the riparian interface cannot be discarded a priori. Hence, during this period, the retention of 28 ± 7% of TDOC at the stream–riparian interface must be interpreted with caution.

In 2003, we observed a direct relationship between DOC size fraction and DOC retention across the stream–riparian interface. In fact, VHMW showed higher DOC retention (78 ± 27%) (Figure 7b) followed by HMW and MMW fractions with a 70 ± 9 and 66 ± 16% retention, respectively (Figure 7c, d). Finally, nil retention of LMW molecules was observed, indicating that it was a nearly conservative fraction (Figure 7e).

In 2004 the selective retention of DOC across the stream–riparian interface according to its molecular size disappeared. In fact, LMW and HMW fractions were virtually conservative with a DOC retention of only 10 ± 22 and 4 ± 32%, respectively (Figure 7c, e). On the other hand, the MMW fraction had a negative retention value (that is, it was released, −34 ± 90%) suggesting that stream–riparian interface acts as a source of this DOC fraction (Figure 7d). Finally, only the VHMW fraction appeared systematically retained (25 ± 27%) across the stream–riparian interface in 2004 (Figure 7b).

Discussion

Our findings demonstrate that the antecedent climatic and hydrological conditions strongly affect both the quality and quantity of DOC transport in stream water. Furthermore, the abrupt hydrological transition between the dry and wet period enhance the hydrological exchange at the stream–riparian interface and therefore stimulate DOC processing. From the climate/hydrological perspective our results provide a contrasting picture with respect to Freeman and others (2004) who minimize the importance of alteration of hydrological regime on DOC transport in northern stream-peatland ecosystems. There is a lively debate on the role of hydrology on DOC transport in peatlands (Freeman and others 2001a, b, 2004; Pastor and others 2003; Tranvik and Jansson 2002). Clark and others (2005), for instance, demonstrated that drought years enhance a decrease in DOC concentration in peat waters. Our study in Fuirosos illustrates the opposite trend. Undoubtedly the comparison of data from different climatic regions and hydrological systems requires a much more rigorous investigation. In fact, DOC concentrations in peatlands are usually much higher with far more predictable dynamics than in Mediterranean streams, and droughts are less severe than those illustrated in our study. However, our results are useful in obtaining a more comprehensive perspective regarding the effects of future alterations of the rain regime, particularly during summer (Christensen and Christensen 2003), on DOC transport in running waters in Mediterranean regions.

In addition, we have shown that DOC retention across the stream–riparian interface also varied greatly, from a selective retention of higher molecular weight fractions of DOC in 2003 under dynamic hydrological conditions after dry antecedent conditions to uniformly low, and even the absence of DOC retention under steady hydrological conditions after wet antecedent conditions (2004). Thus, the dynamic hydrological conditions at the stream–riparian interface might facilitate the retention, through microbial heterotrophic biota (Findlay and others 2003), of irregular pulse inputs to the labile DOC pool associated with organic matter leaching during rain episodes.

Hydrological periods characterized by rapid and abrupt hydrological changes are usually avoided in field studies that deal with nutrient processes across stream–hyporheic interfaces and/or riparian ground waters because of the complexity of stream hydrology during storm events. Thus experimental field works are conducted under basal discharge conditions (Butturini and Sabater 1999; Simon and others 2005). Nevertheless, dynamic hydrological conditions are far from anomalous situations. Empirical and theoretical studies (Serrano and Workmann 1998; Butturini and others 2005) indicated that the interactions between stream water and the surrounding stream–riparian interface are amplified, especially during these periods. During the transition period in Fuirosos, for instance, the stream water infiltrated up to 10 m into the riparian sediments (Butturini and others 2003).

Most of the studies focused on the influence of the stream sediment on DOC fractionation, bioavailability and bacterial uptake have been performed under laboratory conditions (Fischer and others 2002; Findlay and others 2003) and/or under controlled experimental conditions at the study site (Tipping and others 1999; Sobczak and Findlay 2002; Freeman and others 2004; Valett and others 2005). Results obtained in the laboratory are essential to study and identify the biochemical processes that regulate the DOC availability in streams and interstitial water. However, field data obtained under natural conditions are necessary to gain information about the effective incidence of these processes in nature. Our results showed that selected hydrological periods characterized by abrupt changes constituted a natural experiment that allowed for exploration in situ and under natural conditions of the fate of DOC transported across the stream–riparian interface and, therefore, provided a better understanding of DOC dynamics in stream ecosystems.

In the dry year sampling period (2003), both the abrupt hydrometric changes observed in the riparian groundwater and the chemical (chloride and silica) data show that the riparian groundwater is recharged by stream water only, a rapid process that started with the re-establishing of runoff in the stream channel (Butturini and others 2003), while the hill slope groundwater had no influence on riparian ground water. Consequently, the identification of the origin of water that flows through the riparian sediments enabled us to attribute the decrease in DOC across the stream–riparian interface to in situ DOC retention rather than to hydrological mixing between stream and hill slope ground waters.

The lack of a hydrological transition in the 2004 period hindered the characterization of water flow through the stream–riparian interface. In fact the hydrometric data gave no evidence of water exchange between stream and riparian ground waters. Furthermore, the results of chloride data do not rule out the possibility of TDOC dilution in the riparian groundwater by hydrological mixing between stream water and hill slope groundwater.

The chemical data of conservative tracers, TDOC and its MW fractions observed in the ephemeral stream reveal that the water flowing through the forest hill slope during a severe rain episode is similar to the stream water during high discharge conditions. This result substantiates the importance of leaching of allochthonous DOC from hill slope forest soils towards TDOC transport in stream water during these storm episodes.

The TDOC pulse in stream water observed at the beginning of the dry year study period (2003) is a typical phenomenon observed in previous dry years in Fuirosos (Bernal and others 2005) as well in stream-floodplain systems (Valett and others 2005). Its origin is most probably found in the leaching of the abundant leaf litter that accumulated on the streambed in summer (81 g C·m−2, Acuña and others 2004). In fact, leaf accumulation in wet years (that is, 2004) is much lower (6.9 g C·m−2, Acuña and others 2004). Consequently, in 2004, TDOC concentrations are much lower than in the previous study period and solute flushing is absent. These low and constant concentrations of TDOC and of all MW DOC fractions, both in stream water and in riparian groundwater, observed throughout the study period of 2004 prevented the identification of the most reactive and/or recalcitrant DOC molecular fractions across the stream–catchment interface. This problem is absent in data from the dry year (2003). In fact, the high concentrations of TDOC (and MW DOC fractions as well), allowed us to separate the most reactive fractions (HMW and VHMW) from the most recalcitrant ones (LMW) and therefore to identify the selective DOC retention across the stream–catchment interface according to the DOC molecular size (Figure 7). Therefore, our field observations corroborated the size-reactivity conceptual model proposed by Amon and Benner (1996), whereby larger molecules would likely be more labile whereas smaller molecules would be more recalcitrant. Nevertheless, there is no general consensus in the literature about the relationship between bioavailability and DOC molecular size, and more complex patterns have been reported. For instance, Kaiser and others (2004) showed that HMW compounds were generally highly recalcitrant while LMW could be bioavailable or recalcitrant according to its origin and diagenetic state. In laboratory conditions, conversely Fischer and others (2002) measured high DOC retention for HMW and LMW compounds, but low retention for intermediate MW fractions.

Under the wet antecedent condition (that is, the 2004 sampling period) and low DOC concentrations in stream and riparian ground waters, exclusively the larger DOC molecular weight fractions (VHMW) appeared retained across the stream–riparian interface in most of the sampling dates whereas the remaining MW fractions appear to behave conservatively. Sobczak and Findlay (2002) pointed out that under low DOC concentrations stream DOC removal is often negligible and transport may be conservative.

Conclusion

This paper has shown that unstable hydrological periods can provide an excellent opportunity for studying, in situ and under natural conditions, the fate of organic matter in Mediterranean freshwater ecosystems.

Results indicate that the occurrence of a summer drought period favored the transport of DOC with labile behavior (with a retention efficiency of 56% TDOC across the stream–riparian interface) during the successive hydrological transition period, whereas in the absence of a drought period the DOC showed more refractory behavior. Moreover, the efficiency of DOC retention across the stream–riparian interface also varied greatly, from selective retention of higher molecular weight fractions of DOC in 2003 to uniformly low, even loss of, retention under wet antecedent conditions (2004). We hypothesize that a change in discharge regime in Mediterranean regions as a consequence of a warmer climate might favor more severe and dynamic hydrological processes at the stream–riparian interface, which might facilitate the retention of bigger and more labile DOC molecules originated by abrupt and irregular inputs of organic matter leaching during rain episodes.