NAPSEA Calibrated models

DELIVERABLE 3.2 CALIBRATED MODELS Work Package 3 Measures & Pathways 31.03.2024 Page 2 of 20 Deliverable D3.2 Grant Agreement number 101060418 Project title NAPSEA: the effectiveness of Nitrogen And Phosphorus load reduction measures from Source to sEA, considering the effects of climate change Project DOI 10.3030/101060418 Deliverable title Calibrated models Deliverable number D3.2 Deliverable version 1 Contractual date of delivery 31.03.2024 Actual date of delivery 31.03.2024 Document status Prepared Document version 1 Online access Yes Diffusion Public (PU) Nature of deliverable DEM Work Package WP3: Measures & Pathways Partner responsible UFZ Contributing Partners Deltares Author(s) Andreas Musolff, José Ledesma Editor Joachim Rozemeijer, Xiaochen Liu Approved by van der Heijden, L.H. Project Officer Christel Millet Abstract Calibrated and validated models in demonstrator basins Rhine and Elbe, selected sub-catchments within the basins and for the Hunze catchments are presented that capture today’s concentrations and exports of nitrogen and phosphorous. Keywords Elbe, Rhine, Hunze, mQM, CnANDY, hydrological model, water quality, Nitrogen, Phosphorus, diffuse source, point source, mitigation measure. Page 3 of 20 Deliverable D3.2 Contents 1. Executive summary ....................................................................................................... 4 2. Methods ........................................................................................................................ 5 2.1 mQM model description and calibration ............................................................... 5 2.2 CnANDY model description and calibration ......................................................... 7 3. Model results ................................................................................................................. 9 3.1 mQM model performance and N simulations ....................................................... 9 3.2 CnANDY model performance and P-simulations ............................................... 16 4. Next steps ................................................................................................................... 19 5. References .................................................................................................................. 20 Page 4 of 20 Deliverable D3.2 1. Executive summary D3.2 reports on the implementation of the mQM model which addresses nitrogen fluxes and concentrations, and the CnANDY model, which focuses on phosphorous fluxes and concentrations. These implementations are applied to the Rhine and Elbe rivers, their selected sub-catchments and the Dutch Hunze catchments. Today`s water quality and quantity conditions serve as a foundation for the model calibration. The optimized model parameters obtained through this calibration process will be used for the upcoming tasks of scenario implementation and quantification of efficiencies. This report explains the model structure, calibration procedure, the model performance and summarizes the main modelling results. Page 5 of 20 Deliverable D3.2 2. Methods 2.1 mQM model description and calibration Model description The multiscale water quality model mQM (Nguyent et al. 2022) is a spatially lumped model for Nitrogen (N) input (soil N surplus), transport and fate considering the catchment as a single entity. The model aims at a parsimonious representation of N-cycling and legacies in the nearsurface soil zone and the subsurface compartments of a given catchment with annual time step. The conceptual biogeochemical model consists of two major compartments: a soil zone and a subsurface zone where transformation processes such as denitrification occur. The subsurface zone provides an integrated representation of both the unsaturated and saturated (groundwater) zone. Transport through the subsurface is based on the principle of water travel time and by that accounts for the sometimes long time lags between N inputs to the soils and N outputs to the river network (Ehrhardt et al. 2021). In addition, an in-stream compartment is accounted for to allow for the representation of transport in the stream network and uptake reactions taking place therein. Figure 1. mQM model scheme following Ngyen et al. (2022). ON is (active (a) or passive (p) organic nitrogen, IN is mobile inorganic nitrogen. Model calibration For this deliverable D3.2, the model is applied to the case-studies the Elbe, Rhine, and the Hunze basins and their sub-catchments within these basins where data availability is good (see deliverable D3.1), meaning that continuous and long time series of discharge and NitrateN observations exist. This allows for a robust parameterization of the model. More specifically, the model uses observed annual averaged discharge as an input and the flow-weighted (WRTDS method, Hirsch et al. 2010) annual nitrate-N concentration observed at the catchment outlet as the calibration target. The calibration procedure varies all model parameters in a Monte-Carlo approach and selects the model best capturing the available annual nitrate-N observations as described in Nguyen et al. (2022). As a selection criterion we use the minimum root mean square error (RMSE) between model and observations for each catchment over time. Model output is an annual concentration time series and an individual parameter set of the best-fitting model. We note that the model is calibrated to the observed Nitrate-N concentration as this is the best available parameter for N-export across all catchments (71% of all available catchments in Germany (GER) only provide nitrate-N). By Page 6 of 20 Deliverable D3.2 that we do not account for other species of inorganic N (nitrite, ammonia) or organic N. In the German catchments, where dissolved inorganic N and total N (TN) data were available, 92.4% of the dissolved inorganic N was present as nitrate-N. A fraction of 81.4% of the observed TN was present as nitrate-N, which means that on average our analysis misses 18.6% of the exported N loads that are present in other N forms (mainly NH4-N and organic-N). This needs to be considered in the communication of the modelling results. The WFD-targets for surface water are for TN in The Netherlands (NL) (3-year summer average concentrations) and for Nitrate-N (90th percentile within a year) and Ammonia-N (1-year average) in GER. Nitrate-N is however the dominant N-fraction in agricultural drainage and responds most directly to land use changes. Table. 1: Parameters of mQM used for calibration and their parameter range. Parameter Description Range beta_pc Protection coefficient [-] 0.00E+00 7.50E-01 beta_toIN fraction of N surplus to soil IN pool [-] 0.00E+00 0.50E+00 beta_den_soil rate constant of denitrification in soil [1/timestep] 1.00E-01 1.00E+00 beta_min_oa rate constants of mineralization of soil active organic N [1/timestep] 5.00E-02 7.50E-01 beta_min_op rate constants of mineralization of soil passive organic N 6.10E-05 4.50E-04 ka first parameter of the SAS function [-] 0.00E+00 1.00E+00 b second parameter of the SAS function [-] 1.00E+00 1.00E+01 half_life subsurface denitrification [1/timestep] 1.00E-02 3.00E-01 vf stream uptake velocity [m/d] 1.00E-02 3.00E-01 Page 7 of 20 Deliverable D3.2 2.2 CnANDY model description and calibration Model description The river-network model CnANDY (Yang et al. 2021) is designed to capture the competition of pelagic and benthic algae for energy (light) and one limiting nutrient (P). The model operates in river networks at high spatial resolution (100 m) and tracks diffuse and point-source inputs of P downstream on the principle of surface water travel time in the river reaches. It is able to model the build-up of algae biomass in dependence of light availability in the water and at the riverbed surface and by that models the apportionment of P inputs into total and dissolved fractions. The model does not account for temporal variability but was built to robustly capture average conditions of P fluxes in each river network. The model also does not account for P absorbed by particles (e.g., by iron(hydr)oxides. For this deliverable the model is applied to the entire river networks of the Elbe and Rhine basins. The model builds on the spatially differentiated information of wastewater point sources and a land-use dependent diffuse flux (see deliverable D3.1) at average flow conditions. The lack of point-source data in the upstream Swiss part of the Rhine catchment needed additional data not described in D3.1. Here, we used the Swiss report BAFU (2017) that reports on annual average wastewater P-inputs into Swiss rivers. This data was spatially distributed to the input data scale of CnANDY (100 m) by weighing the data with the population density. Figure 2. CnANDY model scheme Yang et al. (2021). P is dissolved phosphorous, B is benthic algae, A is pelagic algae. Model calibration For the model calibration averaged observational data of total P (TP) are used. More specifically, we selected parameter sets that best match the observed values splitting the data into calibration and validation stations. This calibration was done for Elbe and Rhine basin separately. The CnANDY model was built for the scale of river networks spanning multiple river orders making the application in smaller catchments with few river sections and river orders such as the Hunze critical. However, we also applied the parameterization from the Elbe and Weser basins to the stream sections in the seven sub-basins of the Hunze. Page 8 of 20 Deliverable D3.2 Table 2: Parameters used for the model calibration and their initial values. Symbo l Definition Value c Phosphorous to carbon quota of algae [mg P/mg C] 0.02 vf,i Phosphorous uptake velocity at a reach i [m/day] 0.17 mL,B Loss rate of benthic algae [day-1] 0.5 mL,A Loss rate of pelagic algae [day-1] 0.6 ms Scouring effect rate for benthic algae [day-1] 0.01 mp Maximum production rate [day-1] 1 m Half saturation constant for nutrient-limited production [mg P/m3] 5 BK Maximum concentration for benthic algae under neither lightnor nutrient limitations [mg C/m2] 15000 Io Light intensity at the surface water [mmol photons m-2s-1] 300 h Half saturation constant for light-limited production [mmol photons m-2s-1] 25 k Light attenuation coefficient of algae [m2/mg C] 0.000 3 kbg Background light attenuation coefficient [m-1] 2.5 Page 9 of 20 Deliverable D3.2 3. Model results 3.1 mQM model performance and N simulations Elbe The model demonstrates strong performance at both the Elbe outlet and within its subcatchments, enabling the quantification of future scenarios. The temporal dynamics of both concentrations and loads are accurately represented, and the spatial distribution aligns well with observations. The multi-year trend and the timing of the trend-reversal matches very well with the observations. The year-to-year variations are not fully captured as the observations show more variability compared to the model results. However, these year-to-year variations are likely caused by meteorological variability and are not important for multi-year scenario evaluations. Note that the year-to-year variability of the exported loads is very well captured, because mQM uses measured discharges to calculate loads. This shows that the year-to-year variation in loads depend strongly on the discharge variations. At the outlet of the Elbe near the estuary the model performed with a RMSE of 0.65 mg/L NO3-N. Figure 3. mQM model results for nitrate-N concentrations at the Elbe outlet, station 6340110. A comparison of observations (dots) and optimal modelling results (line). Page 16 of 20 Deliverable D3.2 3.2 CnANDY model performance and P-simulations Elbe The CnANDY model is able to capture the spatial patterns of TP concentrations in the Elbe river network for the averaged time period of 2010 to 2020. It demonstrates close alignment with observed concentrations at 243 observation points, indicating suitability for future scenario implementations. The model performs with a RMSE of modelled TP concentrations at these stations of 61 µg/L. Figure 15. CnANDY model results for the spatial distribution of mean TP concentrations (2010-2020) in the Elbe river network in mgP/m3 [µgP/L]. Legend shows the range of mean TP concentrations from 0 to 400 µP/L. Figure 16. CnANDY model performance in the Elbe river network comparing observed (blue) and modelled TP concentrations in mgP/m3 [µgP/L] across the different river orders. Page 17 of 20 Deliverable D3.2 Rhine The CnANDY model is able to capture the spatial patterns of TP concentrations in the Rhine river network for the averaged time period of 2010 to 2020. It demonstrates close alignment with observed concentrations at 317 observation points, indicating suitability for future scenario implementations. The model performs with a RMSE of modelled TP concentrations at these stations of 86 µg/L. Figure 17. CnANDY model results for the spatial distribution of mean TP concentrations (2010-2020) in the Rhine river network in mgP/m3 [µgP/L]. Legend shows the range of mean TP concentrations from 0 to 400 µP/L. Figure 18. CnANDY model performance in the Rhine river network comparing observed (blue) and modelled TP concentrations in mgP/m3 [µgP/L] across the different river orders. Page 18 of 20 Deliverable D3.2 Hunze CnANDY performed without further calibration (see above) with a RMSE of 102 µg/L for TP. The average modelled TP concentration is 616 µg/L with highest concentration in the subcatchment 4206 (Oostermoersevaart) receiving wastewater discharge from WWTP Gieten. Figure 19. Map of the CnANDY model results for average TP concentrations (2000-2022) in the Hunze basin river network. Results per river section. The green triangle depicts the position of the wastewater treatment plant. Page 19 of 20 Deliverable D3.2 4. Next steps The mQM model is applied to all catchments with high-quality data availability, facilitating optimal modelfitting and parameters calibration. In the subsequent phase, we aim to enlarge the set of catchments in the following manner: - Applying the mQM model to sub-catchments that have nitrate-N but lack discharge observations, by utilizing modeled daily discharge (from the mHM-model) instead of observed discharge to derive flow-weighted annual nitrate concentrations. This approach will enlarge the number of modeled catchments approximately 500. - Applying the mQM model to sub-catchments in The Netherlands that do not directly drain into the River Rhine, in order to accurately capture the Dutch contribution of N exports to the Wadden Sea. - Enhancing model performance in the Hunze basin by correcting N input and discharge with the help of local knowledge. These enlarged set of catchments will be used for the scenario evaluation in D3.5. Page 20 of 20 Deliverable D3.2 5. References Batool, M., Sarrazin, F. J., Attinger, S., Basu, N. B., Van Meter, K., & Kumar, R. (2022). Long-term annual soil nitrogen surplus across Europe (1850-2019). Scientific Data, 9(1), ARTN 612. https://doi.org/10.1038/s41597-022-01693-9 BAFU (Hrsg.) 2017: Phosphorflüsse in der Schweiz 2015: Stand, Entwicklungen und Treiber. Bundesamt für Umwelt, Bern. Ehrhardt, S., Ebeling, P., Dupas, R., Kumar, R., Fleckenstein, J. H., & Musolff, A. (2021). Nitrate Transport and Retention in Western European Catchments Are Shaped by Hydroclimate and Subsurface Properties. Water Resources Research, 57(10). https://doi.org/10.1029/2020WR029469 Hirsch, R. M., Moyer, D. L., & Archfield, S. A. (2010). Weighted Regressions on Time, Discharge, and Season (Wrtds), with an Application to Chesapeake Bay River Inputs. Journal of the American Water Resources Association, 46(5), 857-880. https://doi.org/10.1111/j.1752-1688.2010.00482.x Nguyen, T. V., Sarrazin, F. J., Ebeling, P., Musolff, A., Fleckenstein, J. H., & Kumar, R. (2022). Toward Understanding of Long-Term Nitrogen Transport and Retention Dynamics Across German Catchments. Geophysical Research Letters, 49(24). https://doi.org/10.1029/2022GL100278 Yang, S., Bertuzzo, E., Büttner, O., Borchardt, D., & Rao, P. S. C. (2021). Emergent spatial patterns of competing benthic and pelagic algae in a river network: A parsimonious basinscale modeling analysis. Water Research, 193. https://doi.org/10.1016/j.watres.2021.116887