Case study catchments
Three medium-sized coastal catchments situated in Finland, Sweden and Poland were selected for focused modelling studies and comparisons.
The Vantaanjoki (101 km) in southern Finland drains a 1688 km2 catchment, and flows into the Gulf of Finland (Fig. 1). The climate is characteristic of the boreal zone and belongs to the Dfb (warm-summer humid continental) class according to the Köppen-Geiger system (Beck et al. 2018). Mean precipitation is 660 mm year−1 and mean water flow is 11.5 m3 s−1, ranging between 0.7 and 125 m3 s−1. The elevation ranges between 0 and 150 m a.s.l. The highest parts and reliefs of the catchment occur in the north and are mostly covered by forest. Agricultural land and clay soils are predominant in the middle and southern part of the catchment which is flat or slightly hilly. Land use includes forests (53%), agriculture (25%), and urban areas (18%), and water covers 2% of its area. The city of Helsinki with approximately 655 000 inhabitants is located in the south-eastern corner of the catchment.
The Fyrisån catchment (1982 km2) is located in the south-eastern part of Sweden. The approximately 80 km long Fyrisån is part of the wider Norrström drainage basin, and discharges into Lake Mälaren which has its outlet through Stockholm into the Baltic Sea. The climate and vegetation are characteristic of the boreal zone like in the Vantaanjoki catchment, with climate also classified as Dfb. Mean annual precipitation is approximately 550–600 mm year−1. Land use in the catchment is distributed between forests (60%), agriculture (32%), wetlands (4%), lakes (2%), and urban areas (2%). Soils in the catchment are primarily glacial tills on forest land, whereas clay soils dominate in agricultural regions. Altitudes range from 1 to 117 m a.s.l, and mean water flow in the Fyrisån is approximately 13 m3 s−1, ranging between 1 and 140 m3 s−1. The urban area is dominated by the city of Uppsala (population of approximately 170 000). The Fyrisån flows through the city just before reaching Lake Mälaren.
The Słupia catchment (1623 km2) is a coastal river basin in northern Poland. It drains into the southern Baltic Sea through the 138 km long Słupia River (Fig. 1). Climatically it belongs to Dfb class (warm-summer humid continental). Mean annual precipitation is 850 mm a−1 and mean (2000–2016) water flow at the outlet of the catchment is 17 m3 s−1, ranging from 8.4 to 53 m3 s−1. The relief is diverse and characterised by a mosaic of several morainic uplands and sandy outwash plains dissected with a network of eroding and tunnel valleys. The altitude ranges between 0 and 267 m a.s.l., with the highest parts in the south-east of the basin. The most typical soil types are sands and loamy sand. The middle part of the catchment is mostly covered by forest, with a considerable share of urban areas, while agricultural land is predominant in the southern and northern parts. Agricultural land and forest represent 49% and 44% of the catchment, respectively. Urban areas have been growing in size over recent years, and currently constitute approximately 5%, in majority covered by the city of Słupsk (approximately, 90 000 inhabitants).
Identification of similar coastal catchments
Water quality modelling efforts often focus on small areas, predominantly on single catchments due to both systematic (catchment as the basic hydrological unit) and practical (data availability, time and labour input) reasons. On the other hand, regional studies are of great importance, especially in the case of evaluation of impacts related to policy actions (e.g. WFD or BSAP). Our study, covering three middle-sized coastal catchments in three different countries of the BSR, exemplifies such a situation. As in many other hydrological ‘case studies’, this raises the question: to which other geographical areas could conclusions of this study be transferred? One possible way of addressing this question is through identification of catchments that are ‘similar’ to those investigated in the present study. Catchment similarity can be analysed from different perspectives. Here we are mainly interested in factors affecting nutrient load generation process.
It should be emphasised that catchments geographically neighbouring on the three analysed catchments do not necessarily have to share similar behaviours and responses (He et al. 2011). This study adopted a catchment regionalisation perspective where three selected catchments serve as ‘donors’, and the goal is to derive a set of ‘target’ catchments based on certain similarity principles (McIntyre et al. 2005; Zhang and Chiew 2009). More specifically, we used the hydrological distance approach proposed by He et al. (2011). It had been previously applied for another purpose (model parameter transfer problem) in two large river basins in BSR, namely the Vistula and Odra by Piniewski et al. (2017). The method employs a set of physiographic characteristics to evaluate the physical similarity of catchments. In this study, the method was applied in a consistent manner, but separately for each country.
The analysis was based on four groups of catchment characteristics related to land use (percent of different types: urban, agriculture, forest, wetlands, and waters), soils (percent of gravel, sand, silt, and clay in topsoil and in subsoil, and percent of peat areas), topography (slope of agricultural land), and size. This set of characteristics is believed to describe the underlying physical processes leading to nutrient load generation from the landscape. We intentionally ignored other potential characteristics that are related to human activities, e.g. fertilizer rates, wastewater treatment plant (WWTP) loads and dams.
All characteristics were calculated by means of the available GIS data for a group of coastal catchments in each country. We excluded catchments smaller than 300 km2 and larger than three-fold area of a given catchment. We also focused on catchments draining directly to the Baltic Proper, Gulf of Finland, and Bothnian Bay. Climate was also taken into account by restricting the analysis to the Dfb zone, represented by all the catchments.
SWAT is a process-based, semi-distributed, continuous-time model that simulates the movement of water, sediment, N and P compounds within a catchment with a daily time step (Arnold et al. 1998). SWAT requires specific information on weather, soil, topography, vegetation, and land management, and computes the processes associated with water and sediment movement, plant growth, nutrient cycling, etc. based on such input data. SWAT uses basic spatial units called HRUs (hydrological response units) combining land use, soil, and slope within each sub-basin. The land-phase total suspended solid (TSS) and nutrient transport components, namely land erosion, nutrient fluxes and water balance, are computed separately for each HRU, aggregated at the sub-basin level, and then routed through the river network to the main outlet.
A SWAT model application was developed for each catchment. The sources of data used in modelling of the three case study catchments are presented in Table S1 of the Electronic Supplementary Material (ESM). Despite some obvious differences in terms of data sources between countries, an attempt was made to develop SWAT models for each catchment in a consistent way. The catchments were delineated into 93, 51, and 76 sub-catchments, and 1708, 1760 and 1764 HRUs for Fyrisån, Vantaanjoki, and Słupia, respectively. Rigorous model calibration and validation for river discharge, sediment, and nutrient loads was performed for each case (see Piniewski et al. 2019 for the Vantaanjoki case). Basic characteristics of flow gauges and water quality monitoring points can be found in Table S2 (cf. Figure 1 for their location). Automatic calibration and uncertainty analysis with SWAT-CUP software (SUFI-2 programme) were carried out (Abbaspour et al. 2004). Kling–Gupta efficiency (KGE) number was adopted as the goodness-of-fit criterion (Gupta et al. 2009). However, we considered other important model performance measures such as R2 and percent bias (PBIAS) (cf. Table S6). The calibrated/validated model applications were used as ‘Baseline’ scenarios.
Trends in the catchments
Collecting historical data were necessary for the analysis of current trends in the catchments’ characteristics influencing water quality, and for the development of a model scenario that would include near-future developments in the catchments. The historical data were collected in 6 thematic areas: land use, agricultural area, livestock, fertiliser use, WWTP loads, and demography (Table S2). We collected data at a spatial scale feasible for a given feature, ranging from catchments to administrative units such as municipalities, counties, provinces, or even countries. Although we used the data sources that to our knowledge are the best for each case, we acknowledge the fact that differences in spatial levels of aggregation may have an impact on trend analysis, particularly in the case when data for a larger unit (e.g. a province in Poland or a production area in Sweden) are taken to estimate trends in a catchment that occupies its small fraction. Annual data for the longest available sub-period within the period 2000-2018 were acquired for all characteristics except land use. For land use, we used the Corine Land Cover data base for years 2000, 2006, 2012, and 2018.
Assessment of RBMPs
RBMPs are the basic planning documents required by WFD and the national Water Law for designing and enforcing actions aimed at the improvement or maintenance of good status of water bodies. The formal way of RBMP design and implementation is the discretion of Water Authority of a given Member State, but in all cases, the design of the Programme of Measures (PoM) requires the identification of the gap between the water’s current status and good status, followed by planning of a possibly cost-effective combination of measures.
Due to the differences in the RBMP development process in different countries, we collected the related country-specific information in Table S3. We selected measures feasible for SWAT modelling from each RBMP. In short, the Finnish RBMP contains primarily agri-environmental measures: buffer zones, constructed wetlands, winter time vegetation cover of fields, and perennial grass cultivation. The Swedish plans include the same first two measures as the Finnish ones, followed by reduced P leakage from spreading of manure, and stormwater ponds in urban areas. In contrast, the Polish RBMP only includes measures related to the wastewater sector: WWTPs modernization/extension, constructing new on-site WWTPs, and constructing new septic tanks.
Targeting of measures
WFD implementation has triggered a shift towards a targeted approach to the placement of measures that could significantly improve their environmental efficiency and cost efficiency (Doody et al., 2012). As revealed by our analysis of RBMPs, they often specify the quantity or spatial extent of particular measures (e.g. X constructed wetlands, Y ha of buffer zones), but rarely their precise location within the river basin. In Finland, targeting of measures is recommended in the current programming period, but in practice it has remained rather limited.
From the modelling perspective, the placement of measures is one of the key decisions in scenario design. In the case of no evidence or suggestions regarding the placement of measures, they can be placed randomly. Another approach is spatial targeting of measures at certain areas within the model setup (in the case of SWAT they would be sub-basins or HRUs). Here, we have applied simple common-sense rules towards targeting, summarised as follows:
winter time vegetation cover and buffer zones are targeted at agricultural HRUs with highest slopes;
constructed wetlands are primarily targeted at sub-basins with a high proportion of agricultural land;
stormwater ponds are targeted at sub-basins with a high proportion of urban land;
Implementation of scenarios in SWAT
This study involved the development of a consistent set of four scenarios per catchment: (i) Baseline, describing the current status of nutrient loads (years 2002–2016) based on calibrated models, and three scenarios looking into the mid-term (~ 15 year) development of nutrient loads: (ii) Business-As-Usual (BAU), extrapolating historical trends in key variables affecting water quality into the future, (iii) BAU + RBMP, adding currently implemented and recommended measures to achieve the WFD goals for pollution reduction to BAU scenario, and (iv) BAU + RBMP + targ, either adding the same measures that were already planned under RBMPs, but targeting them at potential nutrient load generation hot spots, as described in “Targeting of measures
WFD” section (Vantaanjoki and Fyrisån), or adding new measures (from outside RBMPs) and also targeting them at hot spot areas (Słupia).
Table 2 shows how particular factors of BAU scenario and measures included in BAU + RBMP and BAU + RBMP + targ scenarios were implemented in SWAT. For BAU scenario, the situation was straightforward—all three factors, i.e. land use change, fertilisation change and change in load from WWTPs, were analysed in each catchment. Measures included in RBMPs, however, differed between catchments, and so did the implementation of these measures in SWAT.