Pico presentations of research from the project

Transcription

1 Pico presentations of research from the project QUICS fellows

2 Optimal temporal resolution of merged radar gauge rainfall for urban applications F. Cecinati University of Bristol, Civil Engineering Amsterdam International Water Week, 2 nd November

3 Problem 1: Techniques to merge radar and rain gauge data are sensitive to data quality Solution 1: Accumulation to coarser temporal resolution Problem 2: Urban hydrology requires fine temporal resolutions

4 Solution 2: 1. All rain gauge and radar data are aggregated at the same coarse resolution T. 2. The data is merged with Kriging with External Drift (KED). 3. The results are disaggregated to resolution t using the radar pattern. 4. The optimal temporal resolution is identified trying products with different combinations of T and t in the InfoWorks model and comparing the output with water level observations.

5 Acknowledgements: This work was carried out in the framework of the Marie Skłodowska Curie Initial Training Network QUICS. The QUICS project has received funding from the European Union s Seventh Framework Programme for research, technological development and demonstration under grant agreement no The authors would like to thank the UK Met Office and the Environment Agency, which provided the radar rainfall data and the rain gauge data to develop this study, and the British Atmospheric Data Centre for providing access to the datasets.

6 How uncertainty of simulating water resources is affected by different input data information content Carla Camargos 1 Stefan Julich 2, Martin Bach 1, Lutz Breuer

7 What is the problem? Quantify Discharge

8 How to solve it? Regional maps Global maps D2 DEM D1 S2 SOIL MAP S1 L2 LAND USE MAP L1 DEM 1 + LU 1 + S 1 DEM 1 + LU 2 + S 1 DEM 1 + LU 1 + S 2 DEM 1 + LU 2 + S 2 8 MODEL SETUPS: DEM 2 + LU 1 + S 1 DEM 2 + LU 2 + S 1 DEM 2 + LU 1 + S 2 DEM 2 + LU 2 + S 2

9 (Statistical Parameter Optimization Tool for Python*) * Houska, et al. (2015) SPOTting Model Parameters Using a Ready-Made Python Package, PLoS ONE, 10(12), e

10 Thank you for your attention! Contact:

11 Surrogate Modelling for Simplification of Urban Drainage Simulators Mahmood Mahmoodian Amsterdam International Water Week 2017 Side event: QUICS 2 nd Nov 2017 Amsterdam, The Netherlands

12 Challenge: Complex/Detailed Urban Drainage Models Detailed Simulator Representing Physical Processes Detailed Network Structure Adapted from InfoWorks ICM help Detailed Tank and CSO Structure Zoom in Computationally expensive for: Real-time control (RTC) Calibration Optimisation Uncertainty propagation

13 Possible Solutions: 1. Use computational techniques 2. Develop your own fast simulator 3. Develop a surrogate of already existing simulators (e.g. InfoWorks, ) 1. Data-driven approaches Statistical model to captures the input-output mapping of the original simulator. 2. Projection-based approaches In which dimensionality of the parameter space is reduced by projecting the governing equations onto a basis of orthonormal vectors. 3. Hierarchical or multi-fidelity approaches E.g. ignoring some of the processes which are less relevant or by reducing the numerical resolution. 4. Hybrid approaches

14 Method 1: Develop your own fast simulator Applied in RTC Main advantages: Significantly fast Easy implementation Main disadvantages: Calibration challenge Case-specific Uncertainty quantification?

15 Method 2: A data-driven approach Gaussian Process Emulator (GPE) Olson & Chang (2013) Y: model output of interest µ β : a mean function which is considered linear in time ξy : a vector of covariance matrix parameters. Training data: Unseen initial conditions and rainfall scenarios Validation Main advantages: Non-intrusive and generic method Uncertainty quantification Main disadvantages: Limited number of parameters Short simulations

16 Method 3: A hybrid method (intrusive + data-driven) Storage Tank Volume Change: mass balance Validation Intrusive Data-driven Learning from synthetic data Main advantages: Super fast! Useful for longer simulations Main disadvantages: Partly-intrusive and case-specific Uncertainty quantification?

17 Conclusion: Take Away Message There is NO generic remedy of surrogate modelling for all applications Selection criteria: 1. Reduce the runtime significantly 2. Give a quantification of the uncertainty 3. Be easily applicable in practice Example: real-time control (RTC) Data-driven GPE approach! Limited number of parameters Short predictions Uncertainty bound

18 Thank you for your attention! Keep It as Simple as Possible! (Law of parsimony, Ockham's Razor)

19 Parameter estimation of hydrologic models using censored and binary observations Omar Wani

20 Parameter Estimation using Censored Observation O. Wani et al., Water Research, 2017

21 Likelihood Function O. Wani et al., Water Research, 2017

22 Prediction Phase O. Wani et al., Water Research, 2017

23 Omar Wani, M.Sc. Ph.D. Candidate: ETH - Swiss Federal Institute of Technology, Zurich Eawag - Swiss Federal Institute of Aquatic Science & Technology omar.wani@eawag.ch

24 Uncertainty analysis in Integrated catchment modelling Antonio M. Moreno-Rodenas, Franz Tscheikner-Gratl, Jeroen Langeveld and Francois Clemens

25 1- Integrated modelling for water quality systems.

26 2- Uncertainty analysis Objectives: 1- Quantify uncertainties in the model outcomes. 2- Evaluate uncertainty propagation. Adapted from: Franz Tscheikner-Gratl, 2017 (QUICS deliverable)

27 Context Model Input Parameter Calibration Model structure Determinism Statistical Scenario Deep Uncertainty Epistemic Stochastic Ambiguity 3- Implementation example Step 1. Identification of uncertainties id Uncertainty sources Source Type Nature Step 2. Prioritization according to magnitude and contribution River 1 Temperature River (measurement) 2 Luminosity River (measurement) 3 River upstream Pollution 4 Baseflow hydrology 5 Pollution load rural catchment 6 River diversion/retention structures levels 7 River geometry 8 River energy losses/roughness Sediment evolution (unaccounted dredging 9 and transport) 10 Errors at measured water quality data Urban 11 Rainfall Data measurement errors Rainfall data input characteristics (timespace 12 resolution) 13 Soil characteristics for infiltration 14 Water infiltration in the sewer 15 Evaporation potential 16 Daily/seasonal pattern urban pollution load 17 Population density Urban drainage CSO pollution mean 18 concentration 19 Pumping capacity-activation levels 20 Georeference of main CSO structures 21 CSO weir geometry 22 Layout of connected draining areas 23 Transport line to WWTP capacity WWTP 24 WWTP reactor conditions 25 Temperature WWTP 26 Control WWTP 27 Water treatment chemical addition Integrated model Model extrapolation to simulate corrective 28 measures 29 Climatological scenarios

28 3- Implementation example Step 3. Propagate and evaluate uncertainties at each relevant step of the model design phase. Example 1. Effect of rainfall input characteristics in DO at the river Example 2. Model structural uncertainty for river routing processes: Testing 2 model structures: SM and VPMM. Global Sensitivity analysis and model bias comparison Example 3. Impact of uncertainties at the CSO water quality characteristics. Outcome: Design of optimal rainfall input estimation. Outcome: Improvement on river routing description. Outcome: Quantification of effect for data-based water quality description of CSO

29 Thank you!

30 Sub-kilometre spatial variability of rainfall and its representation in urban hydrology Manoranjan Muthusamy 1, Alma Schellart 1, Simon Tait 1, Gerard B.M. Heuvelink 2 1 Department of Civil and Structural Engineering, University of Sheffield, UK 2 Soil Geography and Landscape group, Wageningen University, The Netherlands

31 Introduction Analysing rainfall variability of catchments > 1 km 2 : Operational rain gauge data, radar data General hydrology Vertical lines Urban hydrology Horizontal lines Analysing sub-kilometre rainfall variability: Requires high resolution (spatial and temporal) data. More challenging! Schilling, W., Rainfall data for urban hydrology: what do we need? Atmos. Res..

32 Research Questions 1. Does sub-kilometre spatial variability have a significant effect on urban run-off prediction? Poster 1 1. How to represent this variability in lumped urban hydrological models? Poster 2

33 Thank you!

34 Numerical and experimental flow investigation in a surcharge manhole Md Nazmul Azim Beg 1 Supervised by: Rita F. Carvalho 1 and Jorge Leandro 2 1 Department of Civil Engineering, University of Coimbra, Portugal 2 Hydrology and River Basin Management, Technical University of Munich, Germany

35 Introduction Flows in manholes may include swirling and recirculation flow with significant turbulence and vorticity. However, how these complex 3D flow patterns could generate different energy losses and so affect flow quantity and also dispersion levels in the wider sewer network is unknown. Flow inside a surcharged manhole is investigated using laboratory based scaled and prototype manholes Experimental data is used to validate CFD model CFD model is utilized to explain flow phenomena

36 PIV at a scaled manhole PIV was done at a scaled manhole Data was compared with different RANS models: i) RNG k-ε, ii) Realizable k-ε, iii) k-ω SST model The RNG k-ε model showed the closest approximation

37 Investigation of head loss RNG k-ε model was used in three real scale prototype manholes with a diameter of 1.0 m, connected with a 300 mm inlet-outlet pipe Type A with a sump of 0.1m Type B is without sump Type C has a hydraulically shaped bottom You are welcome to my poster for detail

38 Thank you!

39 stupscales: an R package for spatio-temporal Uncertainty Propagation across scales J.A. Torres-Matallana 1,2, U. Leopold 1, G.B.M. Heuvelink 2 1 Luxembourg Institute of Science and Technology, 2 Wageningen University and Research Centre Final dissemination nd event Amsterdam International Water Week 2 November, 2017, Amsterdam, The Netherlands

40 Objectives 1. Development of an open software to perform uncertainty analysis of environmental models 2. Definition of classes and methods for identification of model input uncertainty 3. Definition of methods for propagation of input uncertainty through the model by Monte Carlo simulation 4. Definition of functions for analysis of uncertainty propagation by visualisation, and global and spatiotemporal statistics

41 stupscales Conceptual workflow

42 Contribution 1. We demonstrate that the conceptual workflow proposed for stupscales can be applied to solve real problems about model input uncertainty. Usefulness in urban drainage modelling: 1. Evaluation of green infrastructure developments in the urban catchment 2. Dimensioning of storage structures in combined sewer systems 3. Control of pollution in receiving water bodies

43 Thank you! 1. Contact:

44 Risk-averse model based decisions for water quality failures caused by sewer overflows Ambuj Sriwastava PhD Student, University of Sheffield

45 Motivation Combined sewer overflow (CSO) spills managed by the water utility companies need to comply with the local regulations. Utility companies usually face the risk of paying penalties and/or suffering reputational damage if they fail to comply. Urban drainage models are used to simulate CSO overflow quality so as to make appropriate decisions. Understanding the potential uncertainty in such models may lead to a better informed decision making.

46 Problem definition Evaluate the performance of proposed solutions against the risk of non-compliance with the emission standards for Ammonia concentration in the CSO discharges Retrieved October 22,2017, from

47 Objectives Quantify uncertainty in the simulation of ammonia concentration in the CSO spills Evaluate the impact of the modelling uncertainty on the performance evaluation of the proposed solutions Explore risk-averse decision making criteria to determine optimal solutions satisfying CSO emission standards set by regulators.

48 Thank you!

49 Sampling design optimisation for rainfall prediction using a nonstationary geostatistical model Alexandre Wadoux Wageningen University

50 Objective 1: Account for the non-stationarity of the radar error variance Objective 2: Rain-gauge network optimisation

51

52 Thank you!

53 Uncertainty Quantification of Water Quality Models in Rivers Vivian Camacho James Shucksmith Alma Schellart

54 Parameter Uncertainty in ADE Water Quality Model - Study datasets used in the development of the longitudinal dispersion equations - Evaluate equations based on: model standard deviation, bias, accuracy, coefficient of determination and Nash Sutcliffe Coefficient - Determine 95% confidence intervals of predicted concentrations

55 Sensitivity Analysis of River Dommel WQ Model Modelling of slow-river processes in low-lying areas with urban and agricultural pollution Rural runoff quantification using Rainfall-Runoff WALRUS model Urban runoff hydrodynamic simulation using SOBEK River ecological services and food web processes representation using PCDitch Parameter Sensitivity Analysis using distributions of observed data

56 Thank you!

57 DEALING WITH THE PROPAGATION OF UNCERTAINTIES IN INTEGRATED CATCHMENT STUDIES Franz Tscheikner-Gratl, Manfred Kleidorfer and Jeroen Langeveld

58 Framework Broadening of existing frameworks Implementation in to integrated modelling best practice Stepwise modelling approach Uncertainty as integral part of the modelling workflow

59 Further Information Deliverables of the QUICS Project Special Issue in Open Access Journal water Acknowledgements This work was carried out in the framework of the Marie Skłodowska Curie Initial Training Network QUICS. The QUICS project has received funding from the European Union s Seventh Framework Programme for research, technological development and demonstration under grant agreement no

60 Partners and Acknowledgements This project has received funding from the European Union s Seventh Framework Programme for research, technological development and demonstration under grant agreement no

61 Workshops after Lunch 13:45 15:15: Workshop 1 (G102): Improve rainfall estimations using Kriging with External Drift with non-stationary variance Workshop 2 (G107): Surrogate Modelling for Simplification of urban Drainage Simulators Workshop 3 (G108): CEUPUB (Chaos Expansion Uncertainty Propagation in Ungauged Basins) Workshop 4 (G109): Software tools for uncertain propagation of input variables in urban drainage modelling Workshop 5 (G110): Modelling urban sediment transport: From surfaces to sewer network Workshop 6 (G111): Calibration and uncertainty analysis of a SWAT model using SPOTPY