Event importance in terms of SS and TP loss
With reference to the river Celone basin, according to De Girolamo et al. (
2018), annual sediment load varies considerably not only depending on hydro-climatic variables, but also on the estimation method; values from 2.6 to 5.2 t ha
−1 year
−1 have been calculated, an interval similar to that found by other authors (e.g., Estrany et al.
2011; Liquete et al
2009) with reference to other Mediterranean basins.
Considering the 2010–2011 monitoring campaign, during the 11 flood events examined (approximately 4.4% of the whole period), a sediment load of 27,226 t has been measured. This value corresponds to 73.1% of the higher limit of the above-mentioned estimated range and even to 146.2% of the lower limit.
During the same 11 flood events, also 15,242.6 kg of TP passed through the section of the river where the gauging station is located. This amount corresponds to 82.9% of the TP load calculated with reference to the whole period 2010–2011.
It has moreover to be considered that the above-mentioned percentages of SS and TP load, in itself already rather high, refer only to 11 out of the 21 events occurred in the same period. These 11 events are those regarded most significant and are for this reason, the only ones discussed with reference to the first of the two monitoring campaigns considered by the present paper.
Discharge, SS and TP variability during the events
The evidence that in case of multiple peak events, the first peak is usually lower than the second, can be explained considering that most likely, the soil was drier before the rainfall responsible for the first peak. This caused greater infiltration and consequently lower runoff coefficient, in the first peak compared to the second and in case, to the third peak.
Examining sub-hourly rain values, it was observed that, among the events with multiple peaks hydrographs, only event 7 is not compatible with this hypothesis because it rained a lot both in Faeto and in Troia on March 1st, and albeit much less, also on March 2nd and 3rd. In this event, by far the longest one, more than 15 h passed between the first and the second flow rate peak. It is likely that during so a long-time interval, during which on the basin adverse weather conditions persisted, it rained in areas other than Faeto and Troia, where gauging stations do not exist and where for this reason, it has not been possible to consider rainfall data. These “unverifiable” rains could have contributed to the second Q peak, higher than the first.
In "
Results" it was reported that considering single peaks, the graphs in Fig.
6 show that in 15 cases out of 19, |SS| and |TP| peaks precede the discharge peak or coincide with it, while in the remaining 4 cases follow it. |SS| and |TP| peaks preceding the corresponding Q peak are a sign of source-limited events and reveal an exhaustion of SS and TP which can take place for example, when one of the sources of the eroded material is the channel itself (riverbed and/or bank erosion) as reported by Soler et al. (
2008). This behavior is favored in hydrologic conditions like that of the study area where, considerable amounts of sediment and phosphorus can settle on the riverbed during the long-lasting periods of normal and low flow. Another case of early exhaustion of SS and TP occurs when easily erodible soil and erodible sediment-bound P are scarce and therefore run out, before Q reaches its maximum (Zabaleta et al.
2007). The occurrence of source-limited events is corroborated by De Girolamo et al. (
2015) that, in a study performed with reference to the same area, when plotting |SS| versus Q, find clockwise hysteresis loops either with reference to the first peak of multiple peak events, or with reference to events generated by very intense rainfall.
Such a high number of source-limited events might seem strange, given the high amount of eroded sediment mentioned earlier with reference to the Mediterranean area and Italy. However, this behavior can be explained by considering first of all that the above remark concerns the Mediterranean basin as a whole and does not consider the extreme variability of the soils in the area and consequently of their erodibility. Secondly, it should also be kept in mind that in addition to the erodibility factor, soil erosion also depends on other factors, such as rainfall erosivity, slope, crop management and support practices (Panagos et al.
2012). Furthermore, another factor that plays an important role in reducing erosion, is the protection effect of surface stone cover. According to Panagos et al. (
2014), stoniness, which is high in the Mediterranean basin and in particular in the study area, can protect soil by reducing its erodibility by up to 40%, with an effect of soil loss limitation that can be even greater than that of vegetation.
Always considering the high number of source-limited events, it is important to note first of all, that as reported in the introduction, given that large amount of the annual sediment and P loads originate from few storm events, the 27,226 t SS found, while being only the amount of sediment transported during the 11 most significant events considered, it is very likely to represent a good percent of the amount of the soil eroded in the basin during a whole year, which is therefore not very high.
Secondly, even if not particularly high, this quantity of eroded soil can, however, damage the water quality of a stream, due to the high concentration of P bound to eroded soil particles (Arias et al.
2016) and has, for this reason, to be taken in due account. In this regard, the OECD considers 0.03–0.10 mg l
−1 to be the critical range of |TP| for eutrophication to develop (OECD
1982), and this range was exceeded not only during the 15 flood events examined (line: “discharge-weighted concentration TP mg l
−1” in Table
2), but also by the average value (0.27 mg l
−1) of a series of manual samplings, carried out under normal flow conditions, between 2009 and 2010 (De Girolamo et al.
2014).
The cases in which |SS| and |TP| peaks follow discharge peaks (anticlockwise hysteresis loops) can instead be explained by bank collapse during the descending limb of the hydrographs (De Girolamo et al.
2015) or by abundant precipitations (responsible for the peak flow) occurred earlier on less erodible areas and later, by less abundant precipitations occurring on areas richer in sediment and phosphorus, due for example, to recent tillage and/or fertilizations.
In addition to the variability of the moment in which they occur with respect to the corresponding peak flow (clockwise and anticlockwise hysteresis loops just considered), not always higher values of maximum discharge correspond to higher |SS| and |TP| peaks values. Event 7 for example, presents a maximum flow rate of 30.2 m3 s−1, with corresponding |SS| and |TP| peaks of 7.1 g l−1 e 5.2 mg l−1, respectively. Instead during event 13, maximum |SS| and |TP| values of 20.2 g l−1 and 20.7 mg l−1 were measured in correspondence of a streamflow peak of only 6.1 m3 s−1.
An explanation for this evidence could be that while event 7 occurred in March, when soil is protected from erosion by the emergence of wheat plants, event 13 took place in January when, on large parts of the basin, the soil is still completely bare. The fact that event 7 is characterized by SS and TP loads that are much higher than those of all the other events, on the other hand, depends on its extremely long duration (almost 6 days).
Plotting |SS| against Q, also mixed shaped loops have been observed. Such a variable trend of the sediment-flow curves confirms the multiplicity and complexity of the factors that govern the erosion phenomena in the study area.
De Girolamo et al. (
2015) observed that in the study area, during the occurrence of multiple peak events, there is no exhaustion of |SS| availability and explained this behavior hypothesizing that multiple sediment source areas in the basin can contribute sediment to the water body during floods.
Even with reference to |TP|, no exhaustion has been monitored during multiple peak episodes. This represents further evidence that phosphorus mainly moves bound to the eroded soil and that for this reason, different sediment sources (e.g., soil erosion, riverbed erosion, riverbank collapse) potentially represent as many phosphorus sources.
The only exception to this SS and TP behavior, during multiple peak events, is represented by the two discharge peaks of event 4, during the second of which, a considerable exhaustion of |TP| is observed, without a corresponding |SS| exhaustion. A likely explanation could be, in this case, the spreading on the ground of manure coming from one of the numerous animal farms present in the basin, just before the rain responsible for the first peak of flow. In correspondence with the second discharge peak (higher than the first), a slightly higher |SS| value and a decidedly lower |TP| value are observed, compared to those of the first peak; this probably due to the removal of much of the manure present on the ground, already with the rain responsible for the first peak of flow.
Correlations of |TP|, |SS| with one another and with Q during the events
Considering the
R2-weighted averages in the last line of Table
1, the fact that 0.67 > 0.62 > 0.48 indicates that while TP and SS are more directly related to each other, PP represents in fact, in most of cases the dominant form of P lost (Shields et al.
2009; David and Gentry
2000; Vaithiyanathan and Correll
1992), SS and Q are only indirectly related. SS in fact depends not only on erosion, which in turn depends on precipitation which influences Q, but also on other factors such as transport. For this reason, TP depends on Q even less directly (mainly because most of the P reaches the river bound to soil particles). This less direct link between |TP| and Q seems to be confirmed also by the lower value of the sum of
R2-weighted average referred to the |TP|–Q correlations, shown in the last column of Table
1.
Apart from the
R2 values, also the slope coefficients of the correlation straight lines in Table
1 provide information on flood events. Within each of the three correlations |TP|–|SS|, |SS|–Q and |TP|–Q, a considerable slope variability of the corresponding straight lines is observed. For example, for the |TP|–|SS| correlation, the angular coefficient values range from just over 0 to 4.97.
Rodriguez-Blanco et al. (
2010b) to explain this variability hypothesize a difference between the sediment sources. TP–SS lines with a higher slope would be those referring to events that mobilize more sediment from areas with a higher P content.
Significant slope differences can also be observed considering the |SS|–Q correlations. In this case, the differences could be attributed (all other conditions being equal) to a diversity in the erodibility of soils (due to for example, to differences in the spatial location of rainfall). |SS|–Q straight lines with a higher angular coefficient would be those relating to events that affected areas of the basin characterized by greater soil erodibility (e.g., event 13 compared to event 3). On the other hand, in case of events such as 2 and 11, which occurred in November and May, respectively, the greater quantity of sediment transported into the river with the same Q by event 2, is most likely because in November fields are almost bare and therefore, they offer less protection against erosion.
Considerable slope differences can also be observed considering the |TP|–Q correlations (maximum and minimum slopes 2.96 and 0.01 referred to event 13 and 5, respectively). In this case, a possible explanation could be that event 13 mobilized more P from an area better connected to the stream than event 5 did, so that the probability of P redeposition after the detachment from the source, would be much lower for event 13 than for event 5. This hypothesis would also seem to be supported by the fact that even considering the correlation |SS|–Q, the angular coefficient of event 13 is greater than that of event 5.
SS and TP losses during the events
Considering only the hydrological factors on which they depend and the equations with which L, AHL and DWC have been calculated, it appears evident that, L is directly proportional to flow rate, but also to erosivity and therefore to rainfall intensity. AHL is directly proportional to L and inversely proportional to the duration of the event, while DWC is directly proportional to L and inversely proportional to the volume of water that passes, during each event, in the section of the river where the flow gauging station is established.
This is especially true for events 1, 2, and 7, described below in some detail, in which particularly high or low values of runoff and of rainfall quantity and duration from Faeto and Troia, are mainly responsible for the values of L, AHL and DWC reported in Table
2.
Event 1 took place in July and lasted only seven hours, during which the highest precipitation value (73.4 mm from Faeto) compared to that of all the other events was recorded. Because of so a great rain quantity fallen in so a short time interval, also rainfall intensity from Faeto outweighs by far that of all the other events. This, in turn entails rather high values of SS and TP load although, due to the short duration of the event, the measured runoff is the lowest compared to that of all the other events. An effect of so a low runoff and high loads is that the DWCs of SS and TP are the highest of all the events considered.
Event 1 represents a typical flash flood. These sudden episodes are frequent in the Mediterranean area from late summer to early autumn and although typically not entailing particularly high runoff, they are often responsible for the contribution of considerable sediment quantities in water bodies. This happens not only because of the particularly high intensity that allows to mobilize sediment deposited in the riverbed during the previous dry season, but also because of the scarce summer vegetation which favors erosion (De Girolamo et al.
2012,
2015).
Event 7 occurred in March, one of the wettest months and lasted almost 6 days, during which not only the rain from Troia, but also the total precipitation (average of Troia and Faeto) is higher than that of all other events. Consequently, this episode is characterized by the maximum runoff values of the entire study period. Nevertheless, mainly because of its long duration, average rainfall intensity is not particularly high (0.38 mm h−1, from both Faeto and Troia).
All together these features are responsible for the quite high values presented by DWC and AHL, especially as regards SS and this is because evidently, the effect of the exceptionally high load of SS exceeds the dilution effect due to the high runoff.
Again, considering event 7, the AHL and DWC values referred to phosphorus are not as high as those referred to SS. This is most likely due to the occurrence of concentrated erosion caused by the intense rain fallen both in Faeto and Troia. The formation of rills and gullies, which may represent in some cases the major sediment source to the water of a river (Rodriguez-Blanco et al.
2010a) is actually, a frequent phenomenon in the study area. As reported in the introduction in fact, because considerable P amounts are retained in the first 1–5 cm of the soil profile, sediment that reaches the river coming from rills and gullies, whose depth goes well beyond 5 cm (Queensland Government
2013), is characterized by much lower amounts of sediment-bound P.
Event 2 is characterized by particularly low values of numerous variables. The loads of SS and TP (15.9 t and 57.3 kg) are for example lower than those transported in the river during all other events. The measured values of both AHL and DWC referred to SS are also lower than those of all other events. The reason for this characteristic of the event lies in the combined effect of the short duration (less than 7 h) and the low quantity of rain (average of the precipitations from Faeto and Troia higher only than those of event 9).
However, as has already been mentioned, in addition to hydrological factors, numerous other elements affect the supply of sediment and phosphorus to water bodies. These elements take on a leading role in the events other than 1, 2 and 7, just due to the lower weight exerted by factors such as runoff and amount and duration of precipitation from Faeto and Troia.
Among these, for example, rains fallen in areas of the basin more or less rich in P. It is known in fact, that most of the P present in the agricultural land derives from fertilization and that the Italian rural landscape is extremely fragmented. This is also true in the study area where, moving only a few kilometers, it is possible to observe the presence of different crops that require a different P fertilizer supply. The consequence is a notable fragmentation also of the P content of the soil, with the possibility that rains with similar duration and intensity characteristics, fallen within hours of each other, on different areas of the basin, are responsible for the contribution of very different amounts of TP to the river. This is what can be observed by comparing for example, for events 3 and 4, the values of AHL and DWC referred to TP and which is confirmed observing, in Table
1 the slope coefficient of the straight-line TP–SS referred to event 4, greater than that referred to event 3 (0.42 and 0.17, respectively).
A significant contribution of P to the river waters due to rains, even if not very intense, also occurs in case of P fertilizations performed shortly before rainfall occurrence and/or with phosphorus quantities beyond the saturation of the P sorption capacity of the soil, with consequent lack of time, for both crop P uptake and/or for P adsorption on soil particles. An explanation of this type justifies the differences observed by comparing, for example, L, AHL and DWC referred to TP in event 6, with respect to the same parameters referred to event 3.
Another factor that explains, other conditions being similar, the differences between one event and another, in terms of SS and TP transported in the water body, is vegetation cover density. In this regard, for example, Sauer and Ries (
2008) report that a plant cover exceeding 60% can considerably reduce erosion in semiarid environments.
This is very likely, the reason of the differences observed for example, comparing events 11 and 14. Event 14 occurred in February, when vegetation cover is not well developed yet and despite having similar duration, intensity and runoff to those of event 11, it has L, AHL and DWC referring to SS much greater than those of the event. 11, which occurred in May.
The same parameters referring to TP, on the other hand, have very similar values between the two events. This means that for the same amount of sediment that reaches the water body, during event 14 a much lower TP load is transported in the water body than that which reaches the river during event 11. This can be attributed to the occurrence of concentrated erosion because of the intense rain fallen both in Faeto and in Troia in the days immediately preceding event 14.
Finally, it should be underlined that the explanations provided in this paragraph regarding the values of L, AHL and DWC consider the rain that has fallen only in Faeto and Troia, the unique two locations within the basin where there is a rainfall measurement station. Nevertheless, it is very likely that in the days corresponding to the events discussed, it also rained in other non-instrumented areas. These unverifiable rains, despite having contributed to runoff formation and to SS and TP losses, could not be considered among the data reported in Table
2 and in the related discussion.