The comparison between the mass-balance method based solely on load-oriented approach and results obtained from the Makromodel DNS/SWAT for the medium-sized catchment, which has been presented in this study, has demonstrated discrepancies in the total nutrient loads and in the contribution of individual sources.
The river flow measurements and their translation into the model can be identified as the first source of these discrepancies. For the purposes of this study, the flow rate monitoring data (2017) has been taken from the Kowanówko gauge station (IMGW-PIB), where the water level is measured continuously (Fig. 4). The flow data from the very same profile, and from two others (Pruśce and Ryczywół) has also been used to calibrate, verify and validate the Wełna River model for the period of 2001-2018, which covers the entire flow cycle (low, medium and high river flows). Despite the very good or good model performance with regard to river flow monitoring data (Supplementary Table S2 and S3), generally an underestimation of the peak flows could be observed (Fig. 4). This phenomenon has been reported quite frequently and related to the curve number method used in the model adopted to estimate direct runoffs from rainfall events.
To calculate total loads in the mass-balance method (eq. 1) TN and TP concentrations obtained from the SEM at the Oborniki monitoring station (2017), localised approx. 5 km downstream from the gauge station (Kowanówko), have been used. This difference has been neglected since no other tributary discharges into the main river at this distance. Since the quality monitoring in 2017 in Oborniki was performed with a monthly frequency, the linear interpolation method (eq. 2) was used to obtain the daily concentration values (Fig. 5). To confront these values with nutrient concentrations from the modelling approach, the TN and TP values were calculated from the simulated loads. This comparison (Fig. 5). clearly shows that daily variability of nutrient concentrations is high and not considered in the mass-balance method. The uncertainty of measurements performed under the SEM system were estimated at 19% and 16% for TN and TP concentrations, respectively. Taking into consideration that the uncertainty of flow measurements has been estimated at 5%, the total uncertainty for mass-balance loads could be estimated (based on the root mean square error propagation method) at 19.6% and 16.7% for TN and TP loads, respectively. Whereas the uncertainty of nutrient loads in the modelling approach has been estimated at 12.5% and 20% for TN and TP loads, respectively.
The distribution of total TN and TP loads among the particular sources (Table 1) depicts anthropogenic activity related to municipal and industrial wastewater production as the second highest source of nutrients. It contributes to 11.0 and 36.0% according to the mass-balance method (MWS, SCS, INS), and 8.8 and 25.9% according to the modelling approach (A1), for TN and TP respectively. Due to the mainly agricultural character of studied catchment, the contribution of industrial sources (INS) can be considered as marginal, while substantial nutrient loads are discharged from the wastewater treatment facilities (Fig. 6). In both approaches, the discharges from different size WWTPs are regarded as point sources, releasing wastewater directly into the river bed phase. However, unlike in the mass-balance method, the modelling approach takes into account a number of processes, such as transport, biodegradation, transformation, dilution, diffusion, deposition and accumulation, taking place in this phase, therefore modelled nutrient loads at the discussed profile can be expected to be much lower. The relatively low level of urbanization in the analysed catchment, and hence the sparsity of households, impacts wastewater management in this area, which is still largely based on septic tanks. The key factor for the calculation of the contribution of this source (SCS) to the total loads in both methods, is related to nutrient retention in the land phase. Significant differences between the adopted retention estimation methods lead to considerable differences between TN and TP load values. In the case of the mass-balance method, retention is calculated separately for nitrogen and phosphorus and based on the comparison between river loads and the sum of inventoried emissions in the catchment (source-oriented approach). Since in the studied catchment the nitrogen river load exceeded inventoried emissions, zero retention was assumed for this nutrient. As for phosphorus, the river load was much smaller than the inventoried emissions, therefore 99% of the retention was adopted and attributed solely to the diffuse sources, according to methodology described in 2.3.3. In the case of the modelling method, the retention of TN and TP resulting from anaerobic processes in septic tanks and transport of nutrients between land and river bed phases was included in the group of point sources.
In the case of sources where nutrient loads depend on the form of land use, the mass-balance method relied on an assessment based on outflow relative to the respective area of the catchment and TN and TP concentrations, measured or adopted for the particular land use form. In the modelling method, nutrient loads for these sources have been simulated based on land features embedded in the HRU system (slopes, soil type, land use) and combined with the meteorological and hydrological data. Eventually, the SWAT module simulates five different forms of nitrogen and six different forms of phosphorus, taking into account different pathways of nutrients delivery into the catchment and their removal through e.g. uptake by plants, volatilization, erosion, surface runoff, etc. The differences between the values of nutrient loads for diffuse sources in both methods have been shown in Table 1 and Fig. 6. In the case of urban areas (URB), which account for approx. 0.5% of the Wełna catchment, this form of land use has been omitted in the mass-balance method, due to the lack of pertinent data on nutrient concentrations. In the modelling approach (A2), three categories of urban areas depending on the intensity of development (continuous, discontinuous, industrial) characterized by the different share of impervious surfaces have been taken into consideration. The resulting nutrient loads were relatively low, however the noticeably higher value for TP should be highlighted. Many published records indicate that indeed runoff from urban areas is rich in phosphorus compounds due to industrial activities, dust particles from roads and construction sites, burning of fossil fuels, fertilizers and biogenic particles from green areas.
Since approx. 75% of the total Wełna catchment area represents agricultural land use (AGS), this source plays the dominant role in both methods, contributing to approx. 82 and 54-61% of the TN and TP loads, respectively. Although relative contributions in both methods are at a similar level, the absolute loads are lower in the modelling method (A3) by 30% for both nutrients. To adapt the mass-balance method to the Wełna catchment conditions, the measured nitrate and phosphate concentrations in shallow groundwater in this area and specific coefficients have been used (section 2.3.2). As for nutrient retention, zero nitrogen retention and 99% phosphorus retention has been implemented, as for other diffuse sources. The Macromodel DNS/SWAT, in turn, is a tool with a very extensive agricultural module and takes into account numerous data related to each type of the crop and agricultural practices (e.g., agrotechnical treatments, fertilization, crop rotation), therefore its results should be considered as more detailed.
The remaining 20% of the land use covered by forestry (MFS) delivered a minimal contribution to TN and TP loads according to the mass-balance method, which seems to be an artefact of the adopted low concentrations representing this type of land use. However, the high ability of forest soils to retain nutrients through different biotic and abiotic mechanisms is also a well-known phenomenon. The estimated loads (A4) according to the modelling approach were noticeably higher, although still represented the lowest contribution to the nutrient content in relation to their surface area. Nutrient loads from the two remaining sources, natural background (NBS) and atmospheric deposition (ATS), were discharged both into the land and into the river bed phases. Although estimations according to both methods were based on a similar concept, the differences reached 30% and 80% for TN and TP, respectively. In the case of NBS, an attempt was made to recreate nutrient concentrations from the pre-human period, using coefficients on the natural background in undisturbed catchments according to the mass-balance method, and creating a variant scenario in a modelling approach where the entire catchment area was covered by forestry without any human activities. As for ATS, which was combined in a single scenario along with NBS in the modelling simulations (A5), the data on pollutant atmospheric deposition was taken into consideration. However, phosphorus has not been included in the modelling method due to the limitations of the SWAT module, although both dry and wet deposition is minimal in Poland and does not exceed 0.34 kgPha−1y−1 (Supplementary Table S1).