Deliverable D2.3 - Toolbox fate & transport modelling of PMTs in the environment

Project ID N°: 101036449 Call: H2020-LC-GD-2020-3 Topic: LC-GD-8-1-2020 - Innovative, systemic zero-pollution solutions to protect health, environment, and natural resources from persistent and mobile chemicals Preventing Recalcitrant Organic Mobile Industrial chemicalS for Circular Economy in the soil-sediment-water SystemS Start date of the project: 1st November 2021 Duration: 42 months Main authors: Hans Groot (DELTARES), Anita Sosnowska (QSARLab), Pim Wassenaar (RIVM), Joris Meesters (RIVM), Arjen Wintersen (RIVM), Veronika Zhiteneva (KWB), Nicolas Devau (BRGM), Johan Valstar (DELTARES), Laura del Val Alonso (EURECAT), Steffen Kittlaus (TU Wien), Meiqi Liu (TU Wien), Jos van Gils (DELTARES), Franziska Knoche (KWB), Arjen Markus (DELTARES), Thomas James Oudega (TU Wien), Gijs Janssen (DELTARES), Christoph Sprenger (KWB), Dwight Baldwin (KWB), Matthias Zessner (TU Wien) Lead Beneficiary: DELTARES Type of delivery: DEM Dissemination Level: PU Filename and version: PROMISCES_D2-3_Toolbox-Fate-Transport-modelling (version 3) Website: https://promisces.eu/Results.html Due date: 31/10/2025 Date of revision: 27/10/2025 D2.3 – Toolbox fate & transport modelling of PMTs in the environment D2.3 – Toolbox fate & transport modelling 2 © European Union, 2025 No third-party textual or artistic material included on the publication without the copyright holder’s prior consent to further dissemination by other third parties. Reproduction is authorized provided the source is acknowledged Disclaimer The information and views set out in this report are those of the author(s) and do not necessarily reflect the official opinion of the European Union. Neither the European Union institutions and bodies nor any person acting on their behalf may be held responsible for the use which may be made of the information contained therein. D2.3 – Toolbox fate & transport modelling 3 Document History This document has been through the following revisions: Authorisation Distribution This document has been distributed to: Name Title Version issued Date of issue DELTARES, BRGM, KWB, QSAR Lab, BWB, EURECAT, RIVM, CSIC, TU WIEN, BAFG, BUWW WP2 partners Version 1 Version 2 Version 3 03/03/2025 17/10/2025 28/10/2025 Version date Author/Reviewer Description 0.1 30/01/2025 Hans Groot Draft ready for reviewer check 0.2 06/02/2025 Hans Groot Draft ready for Quality Control 0.3 27/02/2025 Hans Groot Draft corrected for Quality Control 0.4 27/02/2025 Floriane Sermondadaz Quality control 1.0 28/02/2025 Hans Groot Final version for distribution 2.0 11/10/2025 Hans Groot History of changes: - Revision of the Executive summary to include the conclusions and outlooks. - Added chapter 8: Conclusions and outlook. - Revision of the list of Abbreviations. - Added structure to chapter 1. 3.0 27/10/2025 Hans Groot History of changes: - Revised chapter 1: included “requirements and limitations”. - Added reference to “requirements and limitations” in chapters 2 through 7. - Revised chapter 8: included text on requirements and limitations of the models. Authorisation Name Status Date Review Willie Peijnenburg External reviewer 03/02/2025 Validation Petra Hogervorst WP Leader 06/02/2025 Quality Control Floriane Administrative and financial manager 27/02/2025 Approval Julie Lions Project coordinator V1. 28/02/2025 V2. 17/10/2025 V3. 28/10/2025 D2.3 – Toolbox fate & transport modelling 4 Executive Summary The "Toolbox Fate & Transport Modelling of PMTs in the Environment" is a key deliverable from the H2020 PROMISCES project. This toolbox is a demonstrator that includes a collection of models developed in the PROMISCES project which are designed to assess the fate and transport of persistent, mobile, and toxic substances (PMTs) across various scales (local, regional) and conditions (e.g., urban run-off, bank filtration, unsaturated zone, groundwater). This toolbox presents the basic information with links to the software and model input files with which the models can be run. This deliverable is intended for qualified modellers. It is complementary with the Guidance document, deliverable D2.4 (Zessner et al., 2025) which describes how to apply modelling tools in a tiered way as part of predictive risk assessment. This report provides an overview of innovative approaches for fate, transport, and exposure to persistent, mobile, and potentially toxic (PM(T)) chemicals, presenting model results from the PROMISCES project. It includes: • Models for identifying PM(T) properties using in silico approaches (QSPR/QSAR, Artificial Intelligence, machine learning). • Screening level models for PM(T) exposure assessment. • Models for soil-groundwater interaction. • Fate and transport models for bank filtration sites. • Fate and transport models in urban contexts. • Emission models on a catchment scale. Each model includes a technical description, followed by an explanation of how the models were improved and links to executable files and input data. An overview of the model improvements is presented below. Models for identifying PM(T) properties using in silico approaches. Significant improvements were made to the predictive QSPR models designed to assess the properties and environmental behavior of PFAS compounds, enhancing their accuracy, applicability, and predictive power: • 2 new QSPR models for predicting n-octanol-water partition coefficient (LogKow), and Bioconcentration Factor (BCF)); • 2 improved QSPR models (for predicting Water Solubility (logSW) and Vapor Pressure (logVP)) expanding the applicability domain for PFAS compounds; • 1 model for predicting Melting Point (MP) was adapted conducting a detailed applicability analysis which confirmed its use for predicting the melting points of PFAS compounds.) As part of in silico predictive tools (other than QSPR/QSAR) for identification of PM(T) a new predictive model has been developed. The final developed model is based on XGboost on a large diverse dataset of chemicals and species. The innovative character is in its ability to learn patterns from this wide range of chemicals and species, as well as the model architecture which closely resembles current risk assessment practices. D2.3 – Toolbox fate & transport modelling 5 Screening level models for PM(T) exposure assessment. In comparison to the earlier SimpleBox (4.01) version, the following improvements were added to the SimpleBox - Aquatic Persistence dashboard (SB-AP): • Additional model routines to express aquatic persistence of emitted substances across surface water bodies at regional, continental and global scales. • A macro function has been developed to enable the user to insert input values as ranges instead of single fixed values. As such, the SB-AP Dashboard is able to present the level of uncertainty inserted in the input values -emission volumes and physicochemical properties of the substanceas well as resultant ranges of uncertainty in the model outcomes. • An option is added for the user to perform probabilistic sensitivity analyses to investigate the relationship between physicochemical properties of the substance and aquatic persistence of dissolved and sorbed species. The user is as such served with a new tool to evaluate the extent to which a persistent and mobile substance resides in surface waters and displaces to different spatial scale. • The applicability of the SB-AP Dashboard as a tool for dedicated environmental fate studies is demonstrated with a case study in which the aquatic persistences have been expressed for eight different PFASs and how these compare to the current criteria for (very) persistent and (very) mobile substances (Zessner et al., 2025). The operational approach for deriving generic risk limit in a leaching situation makes use of existing models Hydrus 1D and Modflow 6. Whereas these models are commonly applied to predict leaching and transport at local scale, in this application the models are parametrized in order to underpin generic risk limits for leaching. In the Guidance document, (deliverable D2.4, chapter 2.3, Zessner et al. 2025), sample calculations for the ‘Dutch situation’ have been included as well as pointers to decide on parametrization of the models in different areas. The probabilistic human health risk assessment for four reuse pathways (HHEA) is built on Bayesian principles, which enable assessment of risk under conditions of low data availability and high uncertainty. This is particularly useful for evaluation of substances such as PFAS and other industrial persistent, mobile and potentially toxic (iPMT) substances, the removal of which in treatment processes is not yet well studied in literature. To date, Bayesian principles have been applied for assessing human health risks from microbial contaminants in water, but this framework has not yet been successfully adapted for chemical substances, due to their generally more chronic human health effects in comparison to acute effects from microbial contaminants. Therefore, the Bayesian principles were applied in a new model to enable assessment of literature, site specific, and modelled data to quantify the exposure risk for human health. Models for soil-groundwater interaction. In order to develop a model train for the soil-groundwater interaction the existing model codes of Hydrus-1D, Modflow and MT3DMS were coupled to model transport of PFAS in the unsaturated and saturated zone. The Hydrus-1D and MODFLOW are fully coupled considering the interaction between the unsaturated and saturated zone for water flow. For the coupling of the mass transport a post processing tool was developed to process the time-variant water and mass fluxes at the water table interface using input files for MT3DMS. D2.3 – Toolbox fate & transport modelling 6 The presented approach is the combination of two complementary already-existing subsurface modelling codes: MODFLOW/MT3D and HYDRUS1D. The coupling consist in a single concatenation in the execution of both models. Therefore, no addition or modification to the original codes has been performed. The objective of this python-based model train is to provide a simple but effective modelling approach to estimate transport of PFAS in soil and groundwater, which can be used by operators as a first attempt to characterise PFAS polluted sites with scarce field data. Fate and transport models for bank filtration sites. The smalland large-scale 3D numerical (iMOD/MT3DMS) models employ pre-existing, wellestablished codes for groundwater flow and reactive solute transport. The novelty of the approaches with respect to their application to PMT substances is embedded in the use of experimentally derived sorption rates. The functionality of the chosen reactive transport code (MT3DMS) to deal with the occurring sorption kinetics was already satisfactory, and the field monitoring data did not warrant even more complex kinetics. The generic bank filtration model translates a one-dimensional analytical model for calculating the sorption and degradation of chemicals traveling through groundwater into the R language. It also offers two visualization options that can be used with minimal R knowledge. This model enhances the usability and accessibility of generic analytical model approaches for PMT substances and other chemicals traveling through groundwater. Fate and transport models in urban contexts. The existing emission model in an urban context (urban mass balance/load model for Berlin) has been adapted with some updates and new features. In particular, the model was adapted to include 31 PMTand PFAS-substances. Furthermore, the model's temporal resolution has been improved from an annual to a monthly scale, allowing for a more precise representation of pollutant loads by accounting for the seasonality of rainfall. In addition, the Berlin-specific input data have been updated ensuring a more accurate representation of the current hydrological and pollution dynamics in Berlin's surface waters. The suite of programs that is used for the Berlin urban water system and development of the DELWAQ/SOLUTIONS model, consists of several pre-existing programs. The novelty is that these programs with a different background have been combined into a program suite for a complex surface water system by combining: • An estimate of the loads from urban sources, using stochastic approaches. • A hydraulic model system (water balance model) that provides the flow field and the geometrical information. • A general water quality model that can combine the information on loads and water balance into a program suite for a complex surface water system. The surface water system is first modelled via a hydrodynamic or hydraulic program that has been developed in close cooperation with the water quality program. In this case, the most important step was the estimation of the loads from the various wastewater treatment plants and the incorporation into the set-up for water quality calculations. The water system was schematised on the basis of the pre-existing water balance model. In a complex water system like that of Berlin where actually data on water quality are scarce, it is advantageous to have a tool that can be adapted to the situation with relatively little effort. The model suite was applied to a more or less generic PFAS type, as no information was available on the actual compounds. D2.3 – Toolbox fate & transport modelling 7 Emission models on a catchment scale. The existing Modelling of Regionalized Emissions (MoRE) model was adapted for modelling of PFAS. Additional emission pathways were implemented, which might be significant for PFAS and other pathways with less significance for this substance group were simplified and grouped together. The model now contains the following pathways: point sources and diffuse pathways. Three model variants for the current state were implemented to represent the uncertainty in the model input data: base variant / best-case / worst-case. The PROMISCES watershed model for PM substances (PPM) model is an application of a pre-existing set of interlinked tools to a new application domain. The novelty lies in the application to a larger group of 10 PFAS substances. These substances constitute a significant share of the 24 PFAS substances that are included in the proposed Environmental Quality Standard under the Water Framework Directive. As such, the results have a direct relevance for chemicals policy and risk assessment. Another novelty of the current implementation is the simulation of a precursor-end product combination (PFOS and precursor N-EtFOSAA) at the scale of a large watershed. Here, two separate emission models (one for the precursor and one for the end-product) have been set up to feed into a combined aquatic transport and fate model. This set-up presumes that the transformation of the precursor proceeds only (to a significant degree) in the surface water compartment. It leads to a noticeably better agreement between the simulated and observed concentrations of the endproduct. Conclusion and outlook. While significant progress has been made in developing and improving innovative modelling approaches for fate, transport and exposure through improved prediction of iPMT properties, it should be emphasized that these improvements relate to only 24 PFAS (modelled) and 71 PFAS (modelled properties) compared to the 12 000 considered. In terms of future needs for improved exposure assessment, it is recommended to improve the identification and harmonized inventor of contaminated sites at the national and international (EU) levels and to ensure reliable, openly available information on the production, import, export and therefore the volumes of PFAS use at the national and EU levels. The main scientific advances are improvements of knowledge regarding the fate of PFAS in the environment. These include a better understanding of the extent of local groundwater pollution due to the application of fire-fighting foams, and potential of (old) municipal landfills. There is also better knowledge of PFAS partitioning between different phases (air, water, solids). D2.3 – Toolbox fate & transport modelling 8 Table of contents 1. Introduction ...................................................................................................................................... 12 2. Models for identification of PM(T) properties using in silico approaches ................................... 14 2.1. QSPR models for PM properties identification ...................................................................... 14 2.2. In silico predictive tools (other than QSPR/QSAR) for identification of PM(T) properties of classes of substances ............................................................................................................................ 19 3. Screening level models for the assessment of exposure to PM(T)s ............................................. 21 3.1. SimpleBox model for PM(T)s ................................................................................................... 21 3.2. Operational approach for deriving generic risk limits in a leaching situation ..................... 22 3.3. Probabilistic human health risk assessment for four reuse pathways ................................. 24 4. Models for soil – groundwater interaction ..................................................................................... 26 4.1. Model train Hydrus/MODFLOW/MT3DMS (1D/2D/3D) ........................................................ 26 4.2. Python-based model train ....................................................................................................... 29 5. Fate and transport models for bank filtration sites ....................................................................... 31 5.1. Smalland large-scale 3D numerical (iMOD/MT3DMS) models ........................................... 31 5.2. Generic bank filtration model ................................................................................................. 33 6. Fate & transport models in an urban context ................................................................................ 35 6.1. Emission model in an urban context ....................................................................................... 35 6.2. DELWAQ/SOLUTIONS model ................................................................................................... 37 7. Emission-models on catchment scale ............................................................................................. 39 7.1. Modelling of Regionalized Emissions (MoRE) model ............................................................. 39 7.2. PROMISCES watershed model for PM substances (PPM) model.......................................... 41 8. Conclusions and outlook .................................................................................................................. 44 8.1. Conclusions ............................................................................................................................... 44 8.2. Outlook...................................................................................................................................... 47 9. References......................................................................................................................................... 49 D2.3 – Toolbox fate & transport modelling 9 List of figures Figure 1: Examples of XGBoost SSDs for 10 randomly chosen PFAS chemicals. .................................. 20 Figure 2: Sketch of the model train developed to simulate fate and transport of PFAS in soilgroundwater continuum. ......................................................................................................................... 27 Figure 3: Map of bank filtration sites at Budapest. Left: ground level [m a.s.l.], Right: detailed map of Site 1 (Tahi I-5) and Site 2 (Surany 12). Extraction wells in red, monitoring wells (MW) in orange, unused monitoring wells in green. .......................................................................................................... 32 Figure 4: Schematic representation of source-oriented approach towards emission modelling and key pathways (source: European Commission, 2012). ................................................................................. 41 Figure 5: PPM model domain. .................................................................................................................. 42 List of tables Table 1: Equations of the developed predictive models. ....................................................................... 14 Table 2: Equations of the developed predictive models (continued). .................................................. 15 Table 3: Equations of the developed predictive models (continued). .................................................. 16 D2.3 – Toolbox fate & transport modelling 16 Table 3: Equations of the developed predictive models (continued). Endpoint Model equation and statistics References Melting Point (MP) MP= - 209.04 (±139.9) + 132.95 (±22.63) x ACC + 3.04 (±0.97) x F02[C-F] - 18.97 (±7.98) x C-013 +209.04 (±139.9) x RBF ACC - 'information indices' which characterizes mean information index on atomic composition F02[C-F] - frequency of C-F at topological distance 2 C-013 - carbon connected to at least three electronegative atoms (X) CRX3 RBF - rotable bond fraction R2 = 0.81 RMSEC = 0.40 Q2LOO = 0.78 RMSEEXT= 0.27 Q2F1 = 0.83 Q2F2 = 0.43 Q 2 F3 = 0.93 Publication: Bhhatarai, B. et al., 2011. CADASTER QSPR Models for Predictions of Melting and Boiling Points of Perfluorinated Chemicals. Mol. Inform. 30, 189–204 (2011). The predictive models developed within the PROMISCES project for physicochemical properties were implemented using Python 3.9, in conjunction with the sklearn package. The descriptors incorporated into the models were calculated using the AlvaDesc program, which employs SMILES codes for all compounds. In instances where experimental data for properties such as water solubility, vapor pressure, and the n-octanol/water partition coefficient were scarce or unavailable, these values were estimated through the COSMO-RS method. The performance of the QSPR models was assessed using required metrics: the coefficient of determination (R²) and root mean squared error (RMSECV). A 1:X data split was applied to each model. External validation was carried out using Q²EXT (Q2F1, F2, F3) coefficients, RMSEEXT, and the concordance correlation coefficient (CCCEXT). To confirm the models' applicability domains, a Williams Plot was used to visualize standardized residuals from leverage. All models developed, improved, or adopted in PROMISCES are presented in the QSAR Model Reporting Format (QMRF), which includes essential information and validation results, as detailed below. Additionally, a toolbox for the physicochemical properties of PFAS has been developed which is reported as deliverable D2.1 of the PROMISCES-project (Sosnowska et al., 2024). The toolbox is a summary of the in silico models that were improved, newly developed, and adopted for predicting the physicochemical properties and aquatic toxicity of PFAS. The toolbox includes the requirements and limitations for application of the models. The toolbox in which the models for physicochemical properties are implemented is available as a browser application. This app allows users to predict and visualize properties by inputting descriptors generated using the Ochem software. The app is freely available online at https://physchempfas.streamlit.app/. D2.3 – Toolbox fate & transport modelling 17 2.1.2. Model improvements The following significant improvements were made to the predictive models designed to assess the properties and environmental behavior of PFAS compounds, enhancing their accuracy, applicability, and predictive power. 2 new QSPR models for predicting n-octanol-water partition coefficient (LogKow), and Bioconcentration Factor (BCF): - The objective for logKow was to address gaps in PFAS bioaccumulation data by combining quantum-chemical based methods with data-driven models. COSMO-RS was used to predict n-octanol/water partition coefficients (logKOW) for over 4,000 PFAS compounds, resulting in a highly accurate QSPR model. This model divided the compounds into OECD categories and confirmed the role of fluorine atoms in bioaccumulation. Additionally, the study predicted other critical physicochemical properties such as Henry’s Law constant (kH), air-water partition coefficient (KAW), octanol-air coefficient (KOA), and soil adsorption coefficient (KOC). The addition of over 4,000 PFAS compounds from the NORMAN database extended the model's coverage. The integration of COSMO-RS calculations helped fill experimental data gaps, enabling predictions for compounds with limited data. The model achieved high accuracy with an R² of 0.95 and predictive power (Q²LOO = 0.93). Along with logKOW, the model predicted other important properties like logSw, logVP, kH, KAW, KOA, and KOC, which are essential for assessing the transport and dispersion of PFAS in the environment. This effort resulted in the development of a new QSPR model specifically for predicting PFAS bioaccumulation potential (logKOW), utilizing the NORMAN database to expand the scope of PFAS predictions. - A new QSPR model was created to predict the log BCF for fish based on experimental data from 33 representative PFAS compounds. This model was then applied to a larger dataset of 2,209 PFAS compounds. The model was also able to classify PFAS compounds into bioaccumulative, non-bioaccumulative, and very bioaccumulative categories, a feature does not present in previous tools. The new model advanced predictions by predicting log BCF for a broader dataset and classifying compounds based on their bioaccumulation potential. The model showed high predictive accuracy with an R² of 0.844, correlating well with experimental BCF values for 13 compounds, demonstrating the reliability of the model in predicting bioaccumulation potential. 2 improved QSPR models for predicting Water Solubility (logSW) and Vapor Pressure (logVP): - The main improvement focused on expanding the applicability domain (AD) of two existing QSPR models for predicting water solubility (logSw) and vapor pressure (logVP) of PFAS compounds. This was achieved by incorporating quantum-chemical calculations (COSMO-RS) to fill gaps in experimental data, thereby improving prediction reliability. Additionally, external validation, which was previously absent, was conducted to ensure the improved models robustness. The improvements included extending the applicability domain by broadening the range of compounds the models could predict, incorporating COSMO-RS calculations to address data gaps, and applying external validation to ensure the models reliability for a wider range of D2.3 – Toolbox fate & transport modelling 18 compounds. The models were also applied to a larger dataset of 4,519 PFAS compounds from the NORMAN Database, reducing the number of compounds outside the domain. In terms of development, a new extension strategy was applied by integrating COSMO-RS calculations to extend the applicability domain. Tools were also developed to expand the training set using structural descriptors, which enhanced the model’s accuracy. 1 adapted model for predicting Melting Point (MP): - For the MP model, instead of developing a new one, a detailed applicability analysis (AD) was conducted on the existing QSPR model. This analysis confirmed that the existing model could reliably predict the melting points of PFAS compounds. In summary, these developments significantly enhanced the predictive tools for assessing the environmental behavior of PFAS, expanding the applicability and improving the accuracy of predictions for a broader range of compounds. 2.1.3. Link to executables The toolbox improved in silico models for identification of PMT properties (deliverable D2.1) is available on Zenodo: https://zenodo.org/records/14800915 The models, along with data from the training and validation sets, can be accessed on Zenodo https://doi.org/10.5281/zenodo.13330024. This data includes endpoints such as water solubility, vapor pressure, logP, melting point, and bioconcentration factor. The app is available on site: https://physchempfas.streamlit.app/ D2.3 – Toolbox fate & transport modelling 19 2.2. In silico predictive tools (other than QSPR/QSAR) for identification of PM(T) properties of classes of substances 2.2.1. Technical description of the model The AI/ML-based model developed for prediction of aquatic toxicity takes account of the limited number of toxicity data typically available for PFAS compounds. One of the key issues in this respect is to use the data as efficiently as possible, and to supplement this data with additional information on the impact of chemical structure on toxicity (transfer learning). By far most data are available on PFOA and PFOS with relatively little information available on other PFAS compounds. It is therefore essential to use the scarcely available information on other PFAS compounds as efficiently as possible. The first step in AI/ML-based modelling was the development and validation of a classification as well as a quantification model for the prediction of aquatic toxicity of organic compounds in general, including PFAS chemicals. The classification model is suited to classify PFAS compounds in a set of toxicity classes in line with the requirements for classification and labelling within REACH. The initially developed quantitative ML model can predict the aquatic toxicity of PFAS compounds for different aquatic species. Amongst others, the options for species-species interpolation were exploited and it was specifically considered how in the future it can be demonstrated that models based on AI/ML approaches have an added value as compared to other in silico models. A manuscript has been published on this activity (Viljanen et al., 2024). In addition to the initially developed model, the best-performing model was developed using quantitative and qualitative transfer learning approaches (including Random Forest and XGboost). This was done by using a set of algorithms and 10-fold Cross Validation (CV). The developed model (i.e. the initially developed model + transfer learning) predicted PFAS better after transfer learning, especially when equally weighting PFAS and non-PFAS to get better predictions for PFAS. More details on the statistics of the model, the requirements and limitations, can be found in Deliverable 2.1 (Sonosnowska et al., 2024). Model description The final AI/ML-based model allows to generate predictions of effect levels for PFAS congeners for any of the aquatic species for which toxicity data are available in the database used for model development. Options for endpoints include LC50-, EC50-, and NOEC-values and predictions of the toxicity of PFAS congeners can be generated for any of these endpoints. The model is integrated into an application for environmental risk assessment of PFAS congeners. The basic approach towards chemical risk assessment, for instance advocated in the REACH legislation, is to use toxicity data on different aquatic organisms to generate so-called species-sensitivity distributions (SSDs). The Maximum Permissible Concentration (MPC) for the environment is defined as the concentration which protects at least 95% of the species in an ecosystem, thereby protecting the functioning of the ecosystem. The MPC is also termed ‘HC5’ to indicate the Hazardous Concentration that affects 5 % of the species. The AI/ML-based model is used as the basis for the generation of predictions of HC5values for PFAS chemicals for which toxicity data are available, as well as for non-tested PFAS. This allows, amongst others, for ranking PFAS chemicals according to their predicted HC5-values. Subsequently, these predictions of HC5 can for instance be used for identifying PFAS chemicals that D2.3 – Toolbox fate & transport modelling 20 are Safer by Design than currently commercialized PFAS chemicals. Examples of SSDs generated for a random set of PFAS chemicals are provided in Figure 1. Figure 1: Examples of XGBoost SSDs for 10 randomly chosen PFAS chemicals. 2.2.2. Model improvements As part of this work a new predictive model has been developed. The final developed model is based on XGboost on a large diverse dataset of chemicals and species. The innovative character is in its ability to learn patterns from this wide range of chemicals and species, as well as the model architecture which closely resembles current risk assessment practices. 2.2.3. Link to model input/executables The underlying model code for the calculation of HC5-values has been stored in a repository on GitHub (https://github.com/rivm-syso/predictive_toxicology.git) and can be accessed via Zenodo (https://zenodo.org/records/14929053). This repository also contains the code needed to run a userfriendly interface. Although this repository is currently not accessible to outside collaborators, we aim to make this repository public as soon as all code and documentation is ready. D2.3 – Toolbox fate & transport modelling 21 3. Screening level models for the assessment of exposure to PM(T)s 3.1. SimpleBox model for PM(T)s SimpleBox is a model to predict environmental concentrations of substance across different compartments of air, water, soil and sediment at different spatial scales, e.g. local, regional, continental and global. The model has served as 'regional distribution module' in the European Union System for the Evaluation of Substances (EUSES). It is currently used as part of the CHESAR tool, hosted by the European Chemicals Agency (ECHA) to demonstrate possibilities for 'safe use' of chemicals, which is required for registration of chemical substances under the Registration Evaluation Authorization and Restriction of Chemicals (REACH) framework (ECHA, 2016). SimpleBox also serves as fate module in Life Cycle Impact Assessment models (Rosenbaum et al., 2011). The SimpleBox Aquatic Persistence (SB-AP) Dashboard has been developed within PROMISCES as an additional module to SimpleBox, so that users can screen the extent to which substances persist in surface water compartments at a regional, continental or global scale (Meesters, 2024). 3.1.1. Technical description of the model SimpleBox is a multimedia mass balance model that simulates the environmental fate of chemicals as fluxes (mass flows) between a series of well-mixed boxes of air, water, sediment and soil on regional, continental and global spatial scales. The model does so by simultaneously solving mass balance equations for each environmental compartment box in the model (rivm.nl/simplebox 2024). The input that SimpleBox needs for its simulations refer to physicochemical properties of the substance, landscape characteristics, and emission volumes. The model then delivers exposure concentrations in the environment as output. The SB-AP Dashboard is developed as an additional module to SimpleBox operated as a Microsoft Excel spreadsheet, supported by numerical computations in R, which are linked to the spreadsheet via RExcel (rivm.nl/simplebox 2024). The SBAP Dashboard demands the same input parameter as the SimpleBox model and calculates the aquatic persistence for substance emissions. The dashboard is as such added to the SimpleBox version 4.01 as new worksheets that link with SB’s emission and landscape scenarios, substance data, the simulation of environmental fate processes and the derived chemical mass in the water compartments. Detailed information how to use the SB-AP Dashboard can be found in deliverable 2.6 (Meesters, 2024b). Additional information about the application of the model, the requirements and limitations, can be found in the Guidance document, deliverable 2.4, chapter 2.2 (Zessner et al., 2025). 3.1.2. Model improvements In comparison to the earlier SimpleBox 4.01 version, the SB-AP Dashboard includes additional model routines to express aquatic persistence of emitted substances across surface water bodies at regional, continental and global scales. A macro function has been developed to enable the user to insert input values as ranges instead of single fixed values. As such, the SB-AP Dashboard is able to present the level of uncertainty inserted in the input values -emission volumes and physicochemical properties of the substanceas well as resultant ranges of uncertainty in the model outcomes. A button is added for the user to perform probabilistic sensitivity analyses to investigate the relationship between physicochemical properties of the substance and aquatic persistence of dissolved and sorbed species. The user is as such served with a new tool to evaluate the extent to D2.3 – Toolbox fate & transport modelling 22 which a persistent and mobile substance resides in surface waters and displaces to different spatial scale. The applicability of the SB-AP Dashboard as a tool for dedicated environmental fate studies is demonstrated with a case study in which the aquatic persistence has been expressed for eight different PFASs and how this compare to the current criteria for (very) persistent and (very) mobile substances (Zessner et al., 2025). Moreover, the case study includes before mentioned sensitivity analyses to demonstrate that uncertain degradation rate constants and uncertain octanol-water or organic carbon-water partitioning coefficients yield the largest uncertainties in predicted aquatic persistence of PFASs. 3.1.3. Link to executables The SimpleBox Aquatic Persistence Dashboard is available as an MS Excel Spreadsheet model at Zenodo (https://zenodo.org/records/13752192). 3.1.4. Link to model input The SimpleBox model is available at GitHub at which different versions are available as MS Excel Spreadsheet and R programming script, see https://github.com/rivm-syso/SimpleBox References on the SB-AP Dashboard: ECHA, 2016, Meesters, 2024, Rivm.nl/simplebox, 2024, Rosenbaum et all., 2011. 3.2. Operational approach for deriving generic risk limits in a leaching situation 3.2.1. Technical description of the model Introduction In this study, environmental quality criteria for PFAS in groundwater and surface water, e.g. based on WFD quality targets or drinking water protection, are the starting point for the calculation of corresponding concentrations in the soil or dredged material (sediment) to be used for the derivation of generic risk limits. Conceptual models have been drawn up for two main variants: 1. Groundwater. This main variant assumes that PFAS that are carried along with infiltrating rainwater end up completely in the groundwater. 2. Surface water. This main variant assumes a relatively shallow aquifer in which complete mixing takes place of PFAS originating from the topsoil layer. Modelling approach The groundwater variant is modelled using a Hydrus and Modflow model train whereas the surface water scenario is calculated using a simple dilution calculation based on the assumption of instantaneous mixing in the aquifer and a flux through POC2 that is equal to the net infiltration. For information about the application of the model, the requirements and limitations, the Guidance document, deliverable D2.4, chapter 2.3 (Zessner et al. 2025) should be consulted. Transport of PFAS in the unsaturated zone was performed with the software 1D-HYDRUS. The subsequent transport of PFAS through the aquifer was modelled using MODFLOW 6. A general D2.3 – Toolbox fate & transport modelling 23 description of the software of 1D-Hydrus, MODFLOW and MT3DMS can be found in the manuals: Šimůnek et al., 2013, Harbaugh, 2005 and Zheng and Wang, 1999 respectively. Parametrization of 1D-Hydrus 1. Geometry soil profile and spatial discretization The profile has a length of 5 meters and consists of six materials, corresponding to different soil layers. The materials have unique soil physical and soil chemical properties. The thickness of the material depends on the scenario). For the simulations performed, cells of 1 cm are used. This discretization leads in all cases to minimal errors in the mass balance. 2. Time discretization A simulation time of 500 years was used for this study. The initial time step for all simulations is 9.9E4 days. The minimum and maximum allowed time steps are 9.9E-6 days and 5 days, respectively. The actual time step used is determined by HYDRUS-1D (Rassam, 2018; Šimůnek et al., 2013). 3. Hydrology For the hydrological aspect of the model, an atmospheric boundary condition was used for water input. The same daily precipitation and evaporation data were used for this as in Verschoor et al., (2006) which was based on an average year for the period between 1980 and 2000. The average year was used repeatedly for the simulated period of 500 years. The second boundary condition concerns the groundwater level. Namely, it is assumed that the groundwater level is constant for the simulated period. 4. Soil physical and hydraulic parameters The soil physical properties are based on a standard sandy soil (Verschoor et al., 2006). In this study, no preferential flow paths were assumed and the single porosity Van Genuchten-Mualem model (Van Genuchten, 1980) was used. For hydrodynamic dispersion one tenth of the distance covered by water was assumed. This is defined per layer in HYDRUS-1D.5. Soil chemical parameters In HYDRUS-1D, initial concentrations must be described as total concentrations (mass per volume soil) or pore water concentration (mass per volume pore water). For this study, soil concentrations are set equal to total concentrations and are calculated with total concentrations. A Kd must be entered in HYDRUS-1D. It is therefore necessary to convert a KOC to a Kd. This can be done with the following formula: Kd = KOC x fOC In this formula, fOC gives the fraction of organic carbon in the soil. Because organic matter concentrations are often reported and not organic carbon concentrations, it is necessary to determine what fraction of soil organic matter consists of organic carbon. In this study, it is assumed that 58% of organic material consists of organic carbon. Parametrization of Modflow 6 A simplified schematization was chosen for the model. The model calculates the transport of PFAS in the groundwater that infiltrates during the application to the right-hand side of the model, where the Point of Compliance (POC) is located. Infiltration takes place at the surface of the model. The infiltration is, just like in the 1D modeling, set equal to 300 mm/year. On the location of the application, the infiltration is given a time-dependent concentration, equal to the results of the 1D D2.3 – Toolbox fate & transport modelling 24 modelling. Outside the application, the concentration in the so-called recharge (supply of groundwater) is set equal to 0 ng/l. The flow from left to right is imposed by means of fixed-head cells (fixed boundary conditions) on the left and right sides of the model. The fixed-head cells are set so that the groundwater velocity is 5.1 m/year near the application. This is equal to the median groundwater velocity in groundwater protection areas with a well-permeable soil in the Netherlands. In addition to the imposed gradient, the flow in the model is also influenced by the replenishment of the groundwater, which means that there is no uniform flow velocity in the model. As a result, the flow velocity increases further away from the application. No flow occurs through the bottom of the model. The model includes sorption of PFAS. For the sorption, values were assumed equal to the 1D modelling, namely a KOC equal to the 10th percentile in Dutch conditions and a percentage of organic matter of 0.6%. 3.2.2. Model improvements This study makes use of existing models Hydrus 1D and Modflow 6. The innovative aspect lies in the application at a higher level of abstraction. Whereas these models are commonly applied to predict leaching and substance transport on al local scale, in this application the models are parametrized much more generically in order to underpin generic risk limits for leeching. In the Guidance document, deliverable D2.4, chapter 2.3 (Zessner et al. 2025), sample calculations for the ‘Dutch situation’ have been included as well as pointers to decide on parametrization of the models in different areas. 3.2.3. Link to executables Hydrus 1D is available at: https://www.pc-progress.com/en/Default.aspx?hydrus-1d Modflow 6 is available at: https://www.usgs.gov/software/modflow-6-usgs-modular-hydrologicmodel. 3.2.4. Link to model input Input files for Hydrus 1D and Modflow are available at Zenodo: https://doi.org/10.5281/zenodo.14755349 3.3. Probabilistic human health risk assessment for four reuse pathways 3.3.1. Technical description of the model A risk-based human health exposure assessment (HHEA) was developed to evaluate the exposure for humans in 4 circular economy (CE) routes investigated in 6 of the 7 case studies in the project PROMISCES. The HHEA is a probabilistic tool evaluating the risk posed to human health. The HHEA was applied to the following routes: 1) semi-closed drinking water cycle; 2) groundwater remediation; 3) water reuse for agricultural irrigation; and 4) nutrient recovery. Each of these exposure routes results in a product – drinking water or lettuce – which can be consumed by humans. For some routes, the exposure is purely theoretical, while for others, the entire process chain is investigated in the PROMISCES case study. The HHEA is built on Bayesian principles, which enable assessment of risk under conditions of low data availability and high uncertainty. This is particularly useful for evaluation of substances such as PFAS and other industrial persistent, mobile and D2.3 – Toolbox fate & transport modelling 25 potentially toxic (iPMT) substances, the removal of which in treatment processes is not yet well studied in literature. As the results from the experimental sites are not yet available, the current deliverable (D2.5 – Open source model for probabilistic human health risk assessment (version 0.4)) had to be based purely on removal of substances reported in literature. The deliverable explains the different treatments, environmental matrices, and substances which were the focus of the initial assessment. It describes the construction of the HHEA tool, with explanations of how different data types – literature data, site specific data, and modelled data – are used to update the prior probability of the removal factor for substances in a process. It also describes how non-technical processes, such as mixing or evaporation, have been included into the treatment trains evaluated. Finally, individual reference quotients for the substances are established, which are used to assess the relative risk of the final concentrations in the products which could be consumed by humans. A preliminary discussion of the results of the 4 CE routes is available as of October 2024. However, a detailed evaluation or scenario assessments of the routes will be provided by April 2025, in an updated version of D2.5 once all experimental data is ready. This final version will report on a full evaluation of the routes in the HHEA tool and includes requirements and limitations for application of the model. 3.3.2. Model improvements The HHEA is built on Bayesian principles, which enable assessment of risk under conditions of low data availability and high uncertainty. This is particularly useful for evaluation of substances such as PFAS and other industrial persistent, mobile and potentially toxic (iPMT) substances, the removal of which in treatment processes is not yet well studied in literature. To date, Bayesian principles have been applied for assessing human health risks from microbial contaminants in water, but this framework has not yet been successfully adapted for chemical substances, due to their generally more chronic human health effects in comparison to acute effects from microbial contaminants. Therefore, the Bayesian principles were applied in a new model to enable assessment of literature, site specific, and modelled data to quantify the exposure risk for human health. The HHEA tool stores intermediate results after every treatment step, which enables a transparent assessment of the changes occurring after every treatment step. This has the advantage that there are no restrictions imposed by external standards, however, the files cannot be opened by other applications. These intermediate results are only meant for further analysis (i.e. visualisation, establish a Bayesian networks) in Python. 3.3.3. Link to executables To run the model, users need to download Python (https://www.python.org/). A list of packages and scripts needed to run the model is provided for more streamlined use. 3.3.4. Link to model input The open source model for probabilistic human health risk assessment (PROMISCES deliverable D2.5 - final version) as well as the GitHub repository for the code will become available at: https://doi.org/10.3030/101036449. D2.3 – Toolbox fate & transport modelling 32 Figure 3: Map of bank filtration sites at Budapest. Left: ground level [m a.s.l.], Right: detailed map of Site 1 (Tahi I-5) and Site 2 (Surany 12). Extraction wells in red, monitoring wells (MW) in orange, unused monitoring wells in green. 5.1.2. Model improvements The models employ pre-existing, well-established codes for groundwater flow and reactive solute transport. The novelty of the approaches with respect to their application to PMT substances is embedded in the use of experimentally derived sorption rates. The functionality of the chosen reactive transport code (MT3DMS) to deal with the occurring sorption kinetics was already satisfactory, and the field monitoring data did not warrant even more complex kinetics. 5.1.3. Link to executables The models can be run by using the iMOD Graphical User Interface (GUI) or by using the iMOD Python package (freely available at https://gitlab.com/deltares/imod/imod-python, documentation at https://deltares.github.io/imod-python/). With the latter, one is able to make quick changes to the models (e.g. changes to hydraulic conductivities, pumping rates, etc.) by editing the runfile directly. The input files for the models are already structured correctly for use with the iMOD Python package. D2.3 – Toolbox fate & transport modelling 33 The output of the models consists of IDF-files with water levels and concentrations of each species, for each layer at every timestep (i.e. daily). These can be read either with the iMOD GUI or programming software like Python. Because the models are computationally taxing (especially on the CPU), a strong computer is advised. More information on how to use iMOD-WQ and iMOD-python can be found in the manual: https://content.oss.deltares.nl/imod/imod56/iMOD_User_Manual_V5_6.pdf 5.1.4. Link to model input Link to model input on Zenodo: https://zenodo.org/records/14931007 Link to model input on TU Wien Research Data Repository: https://researchdata.tuwien.ac.at/records/bank filtration models for PFAS 5.2. Generic bank filtration model 5.2.1. Technical description of the model This Generic bank filtration model is designed as a tool for the initial assessment of PMTs including PFAS contamination at riverbank filtrate sites. The model was developed using an analytical equation for solute transport in porous media. The one-dimensional analytical solution accounts for advection, dispersion, linear isothermal sorption, and first-order decay as described in Bear, 1979, solution given in (West et al., 2007). For more information, the requirements, limitations, and a demonstration, please read chapter 4.4.2 and 4.6.3 of the PROMISCES deliverable D2.4 Guidance document (Zessner et al., 2025). The model is available as an R package. Model use is demonstrated in two R markdown vignettes. The first demonstrates a generic riverbank filtrate scenario, and another presents results from Promisces case study 2 data. It is possible to run your own model simply using the package functions, or by altering parameter values in the vignette itself. The code can be applied to any fate and transport problem fitting at least two criteria: (1) long-term exposure, and (2) travel times of the bank filtrate are known. The model creates data frames used for two plot outputs – substance Breakthrough curves and a Heat map of equilibrium resulting concentrations for a range of Koc and Half-life parameters. Output options for the combined breakthrough curve include the possibility of plotting multiple Koc and halflife values at the same time, and the option to change the frequency of the line breaks on the x axis. For the heatmap plot, it is possible to create a second box inside the plot representing parameters of interest (for example a range of koc and half-life values for a given substance). Along with that, it is possible to plot a line where a given substance attenuation would lie in the heatmap’s parameter space. The model is intended as an early assessment tool for PMTs and PFAS and may be used before development of more detailed fate and transport models. D2.3 – Toolbox fate & transport modelling 34 5.2.2. Model improvements This model translates a one-dimensional analytical model for calculating the sorption and degradation of chemicals traveling through groundwater into the R language. It also offers two visualization options that can be used with minimal R knowledge. This model enhances the usability and accessibility of generic analytical model approaches for PMT substances and other chemicals traveling through groundwater. 5.2.3. Link to software / model input The software and model input files can be found on: Zenodo: https://zenodo.org/records/13767204 Github: https://github.com/KWB-R/kwb.1dbear or use remotes::install_github("KWBR/kwb.1dbear") in an Rstudio / R console. / R console. D2.3 – Toolbox fate & transport modelling 35 6. Fate & transport models in an urban context The model train assesses the rain-related PMT/PFAS pollution of surface waters caused by urban drainage inputs through source-pathway and process descriptions. An emission model calculates monthly pollutant loads entering surface waters via stormwater discharges, combined sewer overflows (CSOs) and wastewater treatment plant (WWTP) effluent, using measured and/or literature data with monthly resolution. These loads are then integrated into a surface water model to simulate the fate and transport of the substances. The surface water model considers the potential adsorption of contaminants to particulate organic matter, which can lead to sedimentation in areas or periods of low flow. Variations in adsorption behaviour between PMTs are also taken into account to ensure more accurate predictions. 6.1. Emission model in an urban context 6.1.1. Technical description of the model The emission model is a tool developed specifically for urban catchments to calculate the loads of specific substances - such as PFAS (perand polyfluoroalkyl substances) and PMT (persistent, mobile and toxic substances) - entering surface waters. This model uses measured and /or literature data to determine monthly substance loads. It focuses on three main pathways of urban water systems: stormwater discharges from separate sewer systems, effluent from wastewater treatment plants and combined sewer overflows. The model's calculations are based on data with a monthly resolution, allowing detailed temporal analysis of substance emissions. By integrating hydrological and chemical data sets, the model estimates the quantities of pollutants discharged to surface waters, providing valuable insights into the contribution of each pathway. Input Data The calculation of pollutant loads requires a variety of input data, integrating both measured and simulated datasets, as well as supplementary literature-based information when local data is unavailable. Key inputs should be in csv format and include: Concentration data; measured concentrations of pollutants in stormwater runoff, wastewater, and treated wastewater (WWTP effluent). Volume data; local data on effluent volumes from WWTPs, stormwater runoff volumes entering surface waters, CSO discharge volumes (with stormand wastewater shares). Calculation methodology The emission model uses a probabilistic approach to deal with uncertainties in the input data. First, the probability distributions of pollutant concentrations and effluent volumes must be determined. A Monte Carlo simulation is then performed, generating, for example, 1000 random data points for each input parameter. These random data points are used to calculate 1000 monthly pollutant loads, with the results summarized in statistical measures such as means, medians and percentiles. This approach ensures that the outputs capture the underlying variability and uncertainty of the input data, providing a more realistic and comprehensive estimate of pollutant loads than deterministic methods. D2.3 – Toolbox fate & transport modelling 36 Results and applications The outputs of the model include both the total monthly discharge volumes and the corresponding pollutant loads to watercourses, broken down by source (stormwater or wastewater) and pathway (separate sewer system, CSO or WWTP effluent). The results are summarized in a list in R and can be exported in various file formats, such as csv or Excel tables. Primary applications include identification of main pollution sources, uncertainty analysis, temporal and spatial analysis, urban drainage and infrastructure planning. For more information, the requirements, limitations and a demonstration, please read chapter 4.3.2 and 4.5.2 of the PROMISCES D2.4 Guidance document (Zessner et al., 2025). 6.1.2. Model improvements The existing emission model (urban mass balance/load model) has been adapted with some updates and new features. In particular, the model was adapted to include 31 PMTand PFAS-substances. Furthermore, the model's temporal resolution has been improved from an annual to a monthly scale, allowing for a more precise representation of pollutant loads by accounting for the seasonality of rainfall. In addition, the Berlin-specific input data have been updated, as the previous input dataset of urban runoff volumes, wastewater volumes and WWTP effluent has not been updated for about 10 years, ensuring a more accurate representation of the current hydrological and pollution dynamics in Berlin's surface waters. 6.1.3. Link to executables The model can be run using the software R or R Studio: https://cran.r-project.org https://posit.co/download/rstudio-desktop/ 6.1.4. Link to model input The emission model and an exemplary model input (from the Berlin case) can be downloaded by using the following links: Zenodo:https://zenodo.org/records/14931210 Github: https://github.com/KWB-R/kwb.promisces-emission-model.git or by using remotes: install_github("KWB-R/kwb.promisces-emission-model") in an Rstudio / R console. D2.3 – Toolbox fate & transport modelling 37 6.2. DELWAQ/SOLUTIONS model 6.2.1. Technical description of the model The DELWAQ model is a general software program that is used to simulate water quality in a wide variety of situations, both with respect to the type of water system and to the type of water quality issues. General information can be found at https://www.deltares.nl/en/software-anddata/products?types=Water%20quality%20and%20ecology. Note that the program is distributed as part of the Delft3D 4 and the Delft3D FM packages and is the computational core of the Delft3DWAQ and D-Water Quality modules. It is also used as part of SOBEK and Delft3D FM 1D2D. As such the program solves coupled advection-diffusion-reaction equations using the finite volume approach. It is flexible in a number of important ways: • The geometry of the water system is described by means of a connectivity table rather than a structured or unstructured grid. This allows DELWAQ to be used in network applications, like for the urban water system of Berlin but also for coastal and estuarine regions. • The user defines what substances to use and what water quality processes to include from a built-in library. There is therefore no need to programme this yourself, though it has a feature that allows the addition of new processes without having to change the program code itself. For the urban context two different schematisations were devised: • A network of channels and lakes to model the Berlin surface water system. • A 1D vertical schematisation to represent the lake “Flughafensee", which is a more or less separate lake in Berlin with no significant connections to the rest of the surface waters, but with noticeable discharges from surrounding activities. In many cases the schematisation and the accompanying hydrodynamic or hydrological data are provided via a hydrodynamic or hydrological model, but in this case there was no suitable model available (that is, a hydrological model that can be coupled directly to DELWAQ, like output from SOBEK-FLOW or D-Flow FM). For this reason, the hydrological information from the BIBER model (Schumacher, 2023) was used to set up an ad hoc model schematisation. This was done via an auxiliary program, called mknetwork. The schematisation for the Flughafensee was set up using the Berlin water atlas, which publishes detailed information about all the urban lakes. For more information, the requirements, limitations, and a demonstration, please read chapter 4.3.3, 4.5.3 and 4.5.4 of the PROMISCES D2.4 Guidance document (Zessner et al., 2025). 6.2.2. Model improvements The suite of programs that is used for the Berlin urban water system and development of the DELWAQ/SOLUTIONS model, consists of several pre-existing programs. The novelty is that these programs with a different background have been combined into a program suite for a complex surface water system. This was done combining: • An estimate of the loads from urban sources, using stochastic approaches. • A hydraulic model system (water balance model) that provides the flow field and the geometrical information. D2.3 – Toolbox fate & transport modelling 38 • A general water quality model that can combine the information on loads and water balance into a program suite for a complex surface water system. The surface water system is first modelled via a hydrodynamic or hydraulic program that has been developed in close cooperation with the water quality program. In this case, the most important step was the estimation of the loads from the various wastewater treatment plants and the incorporation into the set-up for water quality calculations. The water system was schematised on the basis of the pre-existing water balance model. In a complex water system like that of Berlin where actually data on water quality are scarce, it is advantageous to have a tool that can be adapted to the situation with relatively little effort. The model suite was applied to a more or less generic PFAS type, as no information was available on the actual compounds. Also, the water system has a short residence time, which means that for such persistent substances as PFAS the accumulation in the sediment layer may be the most important aspect, and not so much the presence in the water itself. 6.2.3. Link to executables The DELWAQ program is available as part of the Delft3D 4 or Delft3D FM suite - https://www.deltares.nl/en/software-anddata/products?types=Water%20quality%20and%20ecology&types=Hydrodynamics%20and%20mor phology The source code is available via the Deltares GitLab repository – https://git.deltares.nl/oss/delft3d. Note: you need to contact Deltares for access. It is part of the complete set of modules that comprise the Delft3D FM suite. The documentation is available via the documentation site - https://content.oss.deltares.nl/ The auxiliary program mknetwork is available via: https://zenodo.org/uploads/14772635 6.2.4. Link to model input The model input for both water quality models and a description of this input is available via: https://zenodo.org/records/14772635 D2.3 – Toolbox fate & transport modelling 39 7. Emission-models on catchment scale 7.1. Modelling of Regionalized Emissions (MoRE) model 7.1.1. Technical description of the model The model system MoRE (Modelling of Regionalized Emissions) was initially developed by the Karlsruhe Institute of Technology (KIT) in cooperation with the German Federal Environment Agency. It is based on the MONERIS model system. MoRE was developed as a tool in an open-source environment for modelling substance emissions into surface waters for a wide range of substances with relevance for water quality (Fuchs et al., 2017). The modelling in MoRE is carried out as a regionalized pathway analysis. The substance emissions are modelled with temporal and spatial differentiation via various emission pathways, as indicated in the EU Guidance Document No 28 (EC, 2012) for tier 3 for establishing an inventory of emissions. The temporal resolution of the model are annual time steps, and the spatial resolution is 526 subcatchments with a size of 354 ± 352 km². In the PROMISCES project the model was adapted for modelling of PFAS, which means additional emission pathways were implemented, which might be significant for PFAS and other pathways with less significance for this substance group were simplified and grouped together. Thus, the model contains now the following pathways: • Point sources: o Municipal wastewater treatment plants o Industrial direct dischargers • Diffuse pathways: o direct atmospheric deposition onto water surface o surface runoff from unsealed areas o soil erosion o Groundwater with contribution from  legacy pollution from PFAS production site (in case of PROMISCES CS#2, the industrial park at Gendorf, Germany)  legacy pollution from aerodromes caused by fire-fighting training activities  legacy pollution from municipal landfills o sewer systems Due to the flexible structure of MoRE, new substances and emission pathways can be integrated at any time, provided that the necessary input data are available, and modelling can be carried out in a reasonable way. In addition, MoRE offers the possibility to modify existing calculation approaches and to test different input data sets by comparing them. For this purpose, different variants can be created. In the PROMISCES project three model variants for the current state were implemented to represent the uncertainty in the model input data: • Base variant: Based on the median evaluation of environmental concentrations this variant should present the most likely model outcome. If more than 80% of the environmental D2.3 – Toolbox fate & transport modelling 40 concentrations were measured as below the analytical limit of quantitation (LOQ), or less than 3 concentrations were observed above the LOQ, half value of the LOQ was used as input data. • Best-Case: This variant is based on the 25th percentile of environmental concentrations and represents a best-case evaluation with rather low pollution. If more than 80% of the environmental concentrations were measured as below the LOQ, or less than 3 concentrations were observed above the LOQ, 0 was used as input data. • Worst-Case: This variant is based on the 75th percentile of environmental concentrations and represents a worst-case evaluation with rather high pollution. If more than 80% of the environmental concentrations were measured as below the LOQ, or less than 3 concentrations were observed above the LOQ, the value of the LOQ was used as input data. The MoRE model system is based on an Open Source PostgreSQL or SQLite database, a generic calculation engine and the MoRE Developer user interface, which can be used to read, modify and extend the contents of the database. All computations are performed by the calculation engine, which is controlled via the user interface. The modelling results can be exported as tables via the MoRE Developer user interface and the results can be used in GIS for mapping. Users can work with MoRE in two different ways: on the basis of multi-user access in a PostgreSQL database via the Internet or as a stand-alone application on the PC (SQLite version). More information on how to use MoRE, including requirements and limitations, can be found in the manual: https://more.iwu.kit.edu/wiki-en-neu. 7.1.2. Model improvements The existing MoRE model was adapted for modelling of PFAS. Additional emission pathways were implemented, which might be significant for PFAS and other pathways with less significance for this substance group were simplified and grouped together. The model now contains the following pathways: point sources and diffuse pathways. Three model variants for the current state were implemented to represent the uncertainty in the model input data: base variant / best-case / worst-case. 7.1.3. Link to executables / input data The modelling guidance document (D2.4) in Chapter 4.2 and 4.6 (Zessner et al., 2025) provides an example of the application of the model in the Upper Danube region. The model itself including all necessary input data is available for download under https://doi.org/10.48436/wg9dy-r9r31.. D2.3 – Toolbox fate & transport modelling 41 7.2. PROMISCES watershed model for PM substances (PPM) model 7.2.1. Technical description of the model The PPM model conceptually relies on the Technical Guidance on the Preparation of an Inventory of Emissions, Discharges and Losses of Priority and Priority Hazardous Substances (European Commission (2012)). This Guidance defines sources as “all processes and activities that are likely to contribute to the input of pollutants into the environment”. Pathways are “the means or routes by which specific substances can migrate or are transported from their various sources to the aquatic environment”. The PPM models follow the “source oriented” approach, see Figure 4. Pathways P1 Atmospheric deposition directly to surface water P7 Storm water outlets and combines sewer overflows + unconnected sewers P2 Erosion P8 Urban wastewater treated P3 Surface runoff from unsealed areas P9 Individual – treated and untreated – household discharges P4 Interflow, drainage and groundwater P10 Industrial wastewater treated P5 Direct discharges and drifting P11 Direct discharges from aquaculture, fisheries and other instream activities P6 Surface runoff from sealed areas P12 Natural background Figure 4: Schematic representation of source-oriented approach towards emission modelling and key pathways (source: European Commission, 2012). D2.3 – Toolbox fate & transport modelling 48 • Innovative predictive fate & transport models for PM(T) in the environment: - The model train for simulation fate & transport of PFAS from the unsaturated to saturated zone will be used and adapted in the EU project Phishes (https://www.phishes-project.eu/). - The small and large scales bank filtration models have an academic follow-up in PhD research by TU WIEN: Enhanced application at Vienna bank filtration locations in the frame of the FateRiskAqua-Project. - The large-scale emission model at catchment scale (MoRE) will be applied in the Tethys project within the Interreg Danube Regional Program. Enhancement and application of the MoRE model in different catchments across Europe has been applied for the Biodivesa call of Horizon Europe and is selected during the first stage of the decision process, submitted to the second stage. Application of the MoRE model for detailed evaluation of sources and pathways of PFAS-pollution in the transnational Dyje catchment (Czech Republic and Austria) has been further developed to assure a sound coupling of MoRE (Modeling of Regionalized Emissions) and PPM (PROMISCES watershed model for PM substances). - For the PROMISCES watershed model for PM substances (PPM) there have not been any PFAS related applications yet in external projects. We expect such applications in projects for the Dutch government with the aim to compile national balances of PFAS. The building blocks of PPM have been and will be applied in a range of external projects aiming at other (emerging) contaminants. For example, on July 1st a modelling study dealing with balances of Tire and Road Wear Particles will be starting. D2.3 – Toolbox fate & transport modelling 49 9. References Bakker, M., Post, V., Langevin, C.D., Hughes, J.D., White, J.T., Starn, J.J., Fienen, M.N., 2016. Scripting MODFLOW Model Development Using Python and FloPy: Groundwater, v. 54, p. 733–739, https://doi:10.1111/gwat.12413 Bakker, M., Post, V., Hughes, J.D., Langevin, C.D., White, J.T., Leaf, A.T., Paulinski, S.R., Bellino, J.C., Morway, E.D., Toews, M.W., Larsen, J. D., Fienen, M. N., Starn, J. J., Brakenhoff, D. A., Bonelli, W. P., 2024. FloPy v3.10.0.dev1: U.S. Geological Survey Software Release, 20 December 2024, https://doi.org/10.5066/F7BK19FH Bear, J. (1979) Groundwater hydraulics. McGraw-Hill, New York. Beegum, S., Simunek, J., Szymkiewicz, A., Sudheer, K.P., Nambi, I., 2018. Updating the coupling algorithm between Hydrus and MODFLOW in the Hydrus package for MODFLOW. Vadose Zone J. 17:180034. https://doi:10.2136/vzj2018.02.0034 Bhhatarai, B. et al., 2011. CADASTER QSPR Models for Predictions of Melting and Boiling Points of Perfluorinated Chemicals. Mol. Inform. 30, 189–204 (2011). Boisgontier, H., van Gils, J., 2020. Physics-based basin-scale modelling of water quantity and sediment dynamics using wflow. Proceedings of the 22nd IAHR-APD Congress 2020, Sapporo, Japan. Collenteur, R., Brunetti, G., Vremec, M., 2021. Phydrus: Python implementation of the hydrus-1d unsaturated zone model. Phydrus Documentation. Release 0.2.0. 2021 Deltares, 2024. D-Emissions (EM) emission modelling framework, https://content.oss.deltares.nl/delft3d/DEmissions_User_Manual.pdf. Deltares, 2024a. The D-WaterQuality (WQ) water quality modelling framework, https://content.oss.deltares.nl/delft3d4/Delft3D-WAQ_User_Manual.pdf EC, 2012. Common Implementation Strategy for the Water Framework Directive (2000/60/EC) Guidance Document No. 28 Technical Guidance on the Preparation of an Inventory of Emissions, Discharges and Losses of Priority and Priority Hazardous Substances. https://dx.doi.org/10.2779/2764 ECHA, 2016. Guidance on information requirements and chemical safety assessment Chapter R.16: Environmental Exposure Estimation. Version 3.0 February 2016. https://echa.europa.eu/documents/10162/17224/information_requirements_r16_en.pdf/ Fuchs, S., Kaiser, M., Kiemle, L., Kittlaus, S., Rothvoß, S., Toshovski, S., Wagner, A., Wander, R., Weber, T., Ziegler, S., 2017: Modeling of Regionalized Emissions (MoRE) into Water Bodies: An Open-Source River Basin Management System. Water 9:239. https://doi.org/10.3390/w9040239 Harbaugh, A.W., 2005. MODFLOW-2005, the U.S. Geological Survey modular ground-water model -- the Ground-Water Flow Process: U.S. Geological Survey Techniques and Methods 6-A16, 2005. Harris, C.R., Millman, K.J., van der Walt, S.J., 2020. Array programming with NumPy. Nature 585, 357–362 (2020). DOI: 10.1038/s41586-020-2649-2. Hughes, J.D., Langevin, C.D., Paulinski, S.R., Larsen, J.D., Brakenhoff, D., 2023. FloPy Workflows for Creating Structured and Unstructured MODFLOW Models. Groundwater. https://doi.org/10.1111/gwat.13327 Kowalska, D., Sosnowska, A. & Puzyn, T. Structure-property modeling of bioconcentration factor (BCF) for Perand Polyfluoroalkyl Substances (PFAS). Under revision step in Chemosphere, 2024. D2.3 – Toolbox fate & transport modelling 50 Langevin, C.D., Hughes, J.D., Banta, E.R., Provost, A.M., Niswonger, R.G., Panday, S., 2017. MODFLOW 6 Modular Hydrologic Model: U.S. Geological Survey Software, https://doi.org/10.5066/F76Q1VQV Meesters, J.A.J., 2024. Deliverable D2.6 - An improved SimpleBox model for improved environmental risk assessment and life cycle impact assessment. https://promisces.eu/Results/_/D2.6.pdf . Mudlaff, M., Sosnowska, A., Grob, L., Bulawska, N., Jagiello, K., & Puzyn, T., 2024. Environmental impact of PFAS: Filling data gaps using theoretical quantum chemistry and QSPR modeling. Environment International, 185, (2024), 108568. https://doi.org/10.5281/zenodo.11382313 Rassam, D., J. Šimůnek, Mallants, D., van Genuchten, M. Th., 2018. The HYDRUS-1D Software Package for Simulating the One-Dimensional Movement of Water, Heat, and Multiple Solutes in Variably-Saturated Media: Tutorial. Rivm.nl/simplebox., 2024. SimpleBox. Latest modification date 08-03-2024. Rosenbaum, R.K., Huijbregts, M.A.J., Henderson, A.D., Margni, M., McKone, T.E., Van de Meent, D., Hauschild, M.Z., Shaked, S., Sheng Li, D., Gold, L.S., Jolliet, O., 2011. USEtox human exposure and toxicity factors for comparative assessment of toxic emissions in life cycle analysis: sensitivity to key chemical properties. International Journal of Life Cycle Assessment, 2011, 16, 710-727, doi : 10.1007/s11367-011-0316-4 Schumacher, 2023, BIBER (Berechnungsund Informationssystem Berliner Oberflächengewässer) Stationäre Berechnung der Wasserstandsund Durchflussverhältnisse im Berliner Gewässersystem bei niedrigen (MNQ) und mittleren (MQ) Durchflüssen, Ingenieurbüro für Wasser und Umwelt. Šimůnek, J., van Genuchten, M.Th., M. Šejna, 2005. The HYDRUS-1D Software Package for Simulating the Movement of Water, Heat, and Multiple Solutes in Variably Saturated Media, Version 3.0. HYDRUS. Software Series 1, Department of Environmental Sciences, University of California Riverside, Riverside, California, USA, 2005 Šimůnek, M., Šejna, H., Saito, M., Sakai, van Genuchten, M.Th., 2013. The HYDRUS-1D Software Package for Simulating the One-Dimensional Movement of Water, Heat, and Multiple Solutes in Variably-Saturated Media; Version 4.17; Department of Environmental Sciences, University of California Riverside, Riverside, California, USA; June 2013; https://www.pc-progress.com//Downloads/Pgm_Hydrus1D/HYDRUS1D4.17.pdf Smits, J.G.C., van Beek, J.K.L., 2013. ECO: A Generic Eutrophication Model Including Comprehensive Sediment-Water Interaction, .PLoSONE 8(7): e68104. doi:10.1371/journal.pone.0068104 Sosnowska, A., Mudlaff, M., Gorb, L., Bulawska, N., Zdybel, S., Bakker, M., Peijnenburg, W., & Puzyn, T., 2023. Expanding the applicability domain of QSPRs for predicting water solubility and vapor pressure of PFAS. Chemosphere 340, (2023) 139965. Zenodo. https://doi.org/10.5281/zenodo.11446922 Sosnowska, A., Puzyn, T., Peijnenburg, W., Wassenaar, P., 2024. Deliverable D2.1 – Toolbox improved in silico models for identification of PMT properties. https://zenodo.org/records/14800915 Van Genuchten, M.Th., 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44:892–898. doi:10.2136/sssaj1980.03615995004400050002x Verkaik, J., Hughes, J.D., Sutanudjaja, E.H., 2015. A hybrid, Parallel Krylov Sover for MODFLOW using Schwarz Domain Decomposition; in: AGU Fall Meeting Abstracts. Vermeulen, P.M.T., Roelofsen, F., 2024. IMOD User Manual 5.6.1. Deltares, Delft. D2.3 – Toolbox fate & transport modelling 51 Verschoor, A., Lijzen, J., van den Broek, H., Cleven, R., Comans, R., Dijkstra, J., & Vermij, P., 2006. Kritische emissiewaarden voor bouwstoffen. Milieuhygienische onderbouwing en consequenties voor bouwmaterialen. https://www.rivm.nl/bibliotheek/rapporten/711701043.html van Verseveld, W.J., Weerts, A.H., Visser, M., Buitink, J., Imhoff, R.O., Boisgontier, H., Bouaziz, L., Eilander, D., Hegnauer, M., ten Velden, C., Russell, B., 2024. Wflow_sbm v0.7.3, a spatially distributed hydrological model: from global data to local applications. Geoscientific Model Development, 17(8), 3199-3234. DOI: 10.5194/gmd-17-3199-2024. Viljanen, M., Minnema, J., Wassenaar, P.N.H., Rorije, E., Peijnenburg, W., 2023. What is the ecotoxicity of a given chemical for a given aquatic species? Predicting interactions between species and chemicals using recommender system techniques, SAR QSAR Environ. Res., 2023. DOI: 10.1080/1062936X.2023.2254225. West, M.R., Kueper, B.H. and Ungs, M.J. (2007) On the use and error of approximation in the Domenico, 1987 solution. Groundwater 45, 126-135. Zessner, M., Baldwin, D., del Val Alonso, L., Derx, J., Deveau, N., Janssen, G., Jou Claus, S., Kittlaus, S., Knoche, F., Lions, J., Liu, M., Markus, A., Valstar, J., Meesters, J., Meijers, E., Obeid, A., Oudega, T., Pathak, D., Sprenger, C., van Gils, J., Wicke, D., Winstersen, A., Zhiteneva, V., Groot, H, 2025. Deliverable D2.4 – Guidance Document on emission-, fateand transport modelling for exposure assessment of surface and groundwater. [in preparation], https://cordis.europa.eu/project/id/101036449/results Zheng, C. and Wang P., 1999. MT3DMS: A Modular Three-Dimensional Multispecies Transport Model for Simulation of Advection, Dispersion, and Chemical Reactions of Contaminants in Groundwater Systems; Documentation and User’s Guide; Department of Geological Sciences, University of Alabama, 1999