The worlds second largest mercury mine operates in the Idrijca catchment. For such a specific case it has been possible to budget the total loss from mining operations - 37,000 tonnes, with a current annual river transport to the coast reduced to 1.5 tonnes into the Gulf of Trieste. Evaluation of the processes of mobilization has highlighted two principal routes, erosion of contaminated soil (also found by TOROS on the Rio Tinto – Elbaz-Poulichet et al., 1961-1973), and methylation. Options for reducing the problem include soil erosion control, with reduction of organic loadings to rivers a possible means of restricting the methylation process (Salomons, 2004).

Methylation is an issue researched by a number of ELOISE projects. This process will be considered in its own right in section 5. Atmospheric pathways: from sea to land. Of relevance to this section on riverine contaminant transport, brief attention is given to the role of riverine organic material in promoting methylation. The MOE project found that mercury in particulates can be deposited on coastal shelves after riverine transport, and that mercury is often strongly associated with humic substances, subsequent methylation increasing the potential for bioaccumulation (Gårdfeldt et al., 2001). Such binding of mercury influences different aspects biogeochemical cycling. As well as bioavailability increasing, evasion from seawater decreases compared to freshwater (Gårdfeldt et al., 2001). In the Black Sea, the deposit of organic material in proximity to river mouths is seen to be a recent feature (Galimov et al., 2002) which increases the possibility for methylation. Understanding gained from nutrient projects, for example DOMAINE examined the transport of dissolved organic matter by rivers to the coast. It was found that the amount of dissolved organic matter exported is dependent on land use and vegetation in the catchment of origin. That the phenomenon in the Black Sea is a recent one is an indication that organic loading is related to economic factors, and the driver-pressure descriptors in the DP-SIR classification could be readily applied if wished. In common with the Idrijca, the organic loading - methylation characteristic, and the knowledge gained from other projects as to underlying controls, open avenues to control, according to political desire and cost.

The scale of river inputs of contaminants vis-a-vis other transport routes has been estimated by some ELOISE projects, as in Table 2 (previous page). One contribution of ELOISE /IMPACTS in this respect has been a significant contribution to appreciation of non-riverine transport routes for pollutants. For substances such as nitrogen (not a ‘contaminant’ in the context of this overview), the restricted contribution of river inputs has long been noted – the seas delivering less than 60% of inputs to European northern regional seas. The ANICE project has recently confirmed such estimates (De Leeuw et al., 2003). Evidence from TOROS, POPCYCLING and MERCYMS amongst other projects indicates still lesser riverine contributions to coastal water inputs for metals and persistent organics. From the estimates in Table 2 (previous page) from TOROS of metal supplies to the western Mediterranean, it can be seen that rivers supply only around 50% of inputs. More markedly, POPCYCLING estimated riverine lindane supply at 15% of total input to the Baltic Sea, and assumed y -HCH to be representative for persistent organic compounds.

The hydrological processes behind contaminant transport allow for understanding of their supply by rivers. Processes operating will change with transport along the river, e.g. from stream driven turbulence and air-water exchange at higher reaches, and wind driven turbulence and exchange at lower estuarine reaches (Middelburg et al., 2002). One project addressing the representation of these was INCA. Although primarily concerned with nitrogen, the project did develop fluvial and hillslope hydrological modelling approaches. Elements such as modelling pre-existing soil moisture content and river bank storage as factors in runoff generation (Butturini et al., 2002), and estimating the hydrologically effective rainfall amounts (Limbrick, 2002) are factors which could improve management where runoff (as with possibly the Rio Tinto and Idrijca) and ground leaching were perceived as issues. The potential relevance of these investigations goes across projects and topics (Salomons, 2004). Naturally, these tools would require tailoring to the contaminant in question given the differences in physical and chemical behaviour of components, but as this has already been undertaken on one occasion (extension from nitrogen to phosphorus–Wade et al., 2002) the concept is not new. Specific site calibration is also always necessary, as INCA is not entirely generic.

Hydrological modelling was also undertaken by the RANR project. Although also concerned with nitrogen flows rather than contaminants as defined here, attention was given to understanding hydrology, and to separating the flow into baseflow, throughflow and runoff components at the catchment/hillslope stage (Andersen et al., 2001). It was found that in sandy soils, baseflow dominates the hydrography, whilst in loamy soils throughflow and surface runoff were predominant. The speed of response, and the propensity to either surface erosion or increased solute loading of flow are factors of relevance to managing contaminated catchments. The lumped rainfall-runoff model used, NAM, is an empirical tool which requires calibrating to the particular catchment in order to obtain the hydrological characteristics of the chosen location, this being a standard procedure. If the ability to predict chemical flows was also required, the chemical characteristics of the contaminant of interest would need to be substituted.

The extent of penetration of river inputs into the receiving sea was evaluated by various projects. The dispersal of metal laden waters in the Bay of Cadiz from the Rio Tinto was examined by TOROS. In the Black Sea EROS-21 noted the widely observed phenomenon of deposit of particulate loads in the near estuary and shelf region, with transport of dissolved components into the open sea (Wijsman et al., 1999). Prior general circulation modelling indicated extreme stratification, and limited deep water mixing. Modelling of 90SR tracers arising from the Chernobyl accident had been able to mimic the penetration of the halocline in the Bosphorus only using time and space averaged data (Stanev et al., 1999). Individual events with short memory could not be simulated without higher resolution and higher quality function representation.