Contributes of Modelling and Data Assimilation Techniques for
Contributes of Modelling and Data Assimilation Techniques for Water Quality Operational Modelling in Estuaries: The Case of the Tagus Estuary ngela Canas
Table of Contents l l l l 1. Introduction 2. Research Plan (objective, research question and methodology) 3. State of the Art 4. Improvement of Model Configuration 5. Implementation of Data Assimilation Module 6. Test of Data Assimilation Module in Improving TEPOMS 7. Preliminary Conclusions
1. Introduction l Relevance of estuaries: case of Tagus Diverse services: Estuary Tagus River Lisbon Guia wastewater underwater outfall Estoril coast beaches Natural Reserve of Tagus Estuary Economical (transport, resources, disposal of waste) Social (aesthetic and recreation services) Almada navigation industrial belt primary production Environmental (high primary production, nursing and habitat for ecological communities) Conflicts affect services sustainability
1. Introduction l Operational oceanography: “making, disseminating, and interpreting measurements of the seas and oceans in order to provide forecasts of future conditions” (Prandle, 2000) Operational forecast systems Adequate answer to coastal and estuarine areas management needs (Water Framework Directive) Prandle, 2000, Coastal Engineering, 41, 3 -12
1. Introduction l Tagus Estuary Pre-Operational Modelling System (TEPOMS) Fernandes (2005) at Maretec (IST) 3 nested domains: 72 h 1 - 2 D, hydrodynamics 162 x 162 (300 m) Objective: monitorization of Guia outfall pollution on estuary and beaches water quality Main end-user: SANEST S. A. (outfall operator) Tagus Estuary 168 x 223 (2 km) 3 - 3 D (11), hydrodynamics, outfall pollution dillution and dispersion Beaches outfall 100 x 60 301 x 105 (10 m) (35 m) + Meteo station (Guia) Hydrometic station (Ómnias) 2 - 2 D, 3 D (11), hydrodynamics, water quality Water sampling Sensor data + Validation and posting (http: //www. mohid. com/tejo-op/) + Forcing: Meteo (Meteo IST) Tide (FES 95) River flow (INAG)
1. Introduction l TEPOMS forecast strenghts: – Model using state of the art technology (MOHID Water): l l Accounts for relevant processes (tide, wind, river) Easily actualized (modular code) (Prandle, 2000) Easily runnable (MPI, several OS) (James, 2002) TEPOMS known forecast defficiencies: – – – Large scale circulation not adequate (e. g. slope current 2501500 m) Meteo forcing not accounting topography Poor grid resolution: l l – Horizontal: inside estuary Vertical: at outfall Lack of data assimilation
2. Research Plan l Objective: – l Research Question: – l Improve the TEPOMS numerical modelling tool (MOHID Water) to help meeting a Tagus Estuary operational water quality forecast system needs “Is it possible to improve the existing TEPOMS skill on forecasting hydrodynamic variables through an alternative model configuration or the use of data assimilation techniques? ” Methodology: – – – Assess TEPOMS face to face with state of the art operational systems for coastal and estuarine areas Test hypothesis of TEPOMS improvement through advanced modelling techniques Test hypothesis of TEPOMS improvement through data assimilation techniques
3. State of the Art: Operational Forecast Systems l l Aim: – Water level / storm surge: North Sea (DMI, DCSM), North America – Water quality (new): Mediterranean Sea (MFS) (Go. MMOOS, PORTS) Structure: Data assimilation Validation posting l Measureme nt network Modelling approaches: – 2 D: circulation and water level forecasts – 3 D: biological/water quality forecasts – Larger scale boundary conditions: nesting in other systems (e. g. MFS, MERCATOR, HYCOM) – Atmospheric pressure effect: level = - (Pmb – Prefmb). (0. 01 m/1000 mb) (Inverted barometer, e. g. Cañizares, 1999) – Coupling with meteorological models
3. State of the Art: Data Assimilation time = t 1. . . Model state forecast f l (x , Nx 1) time = t 2 Data assimilation scheme Contrast(y-Hxf) Model Correction(K) Measurements (y) Sequential assimilation: + error (P) Kalman filter (Kalman, 1960) SEEK (Pham et al. , 1998) SEIK (Pham et al. , 1998) P=LULt EOFs + error (R) Analysis (xa) Model state forecast . . . + error Optimal: Linear M Large cost (N) Sub-optimal: Non linear M: x Small cost (r+1) Sub-optimal: Non linear M: x, L Small cost (r+1)
4. Improvement of Model Configuration l Research hypothesis 1: “Yes, through the spatial expansion of large scale domain. ” Bathymetry (324 x 218): Old Meteo. Galicia ETOPO 2’ 0. 04ºx 0. 04º Better currents description: N N MOHID Water (tide only) Velocity modulus (m/s) Similar contrast with tide prediction: 0. 02ºx 0. 02º Basis for other Portuguese systems and for a 3 D version (account slope current) Peniche Cascais Sines Correl V 1 0. 99932 0. 99933 0. 98673 Correl V 2 0. 99933 0. 99934 0. 98690 RMSE V 1 0. 345 m 0. 361 m 0. 1308 m RMSE V 2 0. 427 m 0. 426 m 0. 1314 m
4. Improvement of Model Configuration l Tide gauges measurements 1 – Peniche 2 – Cascais 3 – Paço d’Arcos 4 – Trafaria 5 – Cacilhas 6 – Lisboa 7 – Seixal 8 – Montijo 9 – Alfeite 10 – Cabo Ruivo 11 – Sesimbra 12 – Alcochete 13 – Ponta da Erva 14 – Póvoa Santa Iria 15 – Vila Franca Research hypothesis 2: “Yes, through the incorporation of the inverted barometer effect. ” Inverted barometer mean sea level time series (Level 1 boundary, ERA 40 data) 2 D forced only with tide Contrast with measured detided water level:
4. Improvement of Model Configuration l Research hypothesis 3: “Yes, through the use of a boundary condition for stratification. ” Stratification Methodology: Relaxation to Levitus (1982) T/S climatology in boundary 33 layers resolution Other improvements (advection, bathymetry) Validation data: SST images (MODIS) CTD near outfall (01/02/05) Spatial distribution
4. Improvement of Model Configuration l Research hypothesis 4: “Yes, through the use of 3 D nested domains with different vertical resolution when and where needed. ” Study of sediment transport in Nazare canyon cartesian geometry Methodology: Implementation in MOHID Water code one-way connection with different 3 D vertical resolution sigma geometry surface temperature
5. Implementation of Data Assimilation Module l Research hypothesis 5: “Yes, through data assimilation, with SFEK/SEEK method, of water level measurements. ” Model historical data: Hydrodynamic_#. hdf 5. . . Water. Properties_#. hdf 5 Methodology: Measurements (time series) - Assimilation Pre. Processor Covariance calculation and EOF analysis: Assimilation Pre. Processor EOF set Eigenvalues - Sequential Assimilation Module - Test in schematic 1 D channel MOHID Water - Test in TEPOMS twin test Modules - Test in TEPOMS assimilating tide gauges water level Sequential Assimilation (SFEK, SEEK) Model Hydrodynamic Water. Properties
6. Test Data Assimilation Module in Improving TEPOMS l Twin test: – – True model: improved TEPOMS (without inverted barometer) Wrong model: improved TEPOMS with 0. 1 standard deviation error at mean sea level EOF analysis:
6. Test Data Assimilation Module in Improving TEPOMS Measurement location Error statistics Cascais Paço de Arcos RMSE Wrong (m) 0. 23 0. 19 RMSE SFEK (m) 0. 51 0. 59 RMSE SEEK (m) 0. 17 0. 21 EOFs become smaller and smaller: Validation location (Hoteit, 2001)
7. Preliminary Conclusions l l l Operating TEPOMS approaches state of the art modelling technology but defficiencies remain; Changes to TEPOMS model configuration provide improvements and affect MARETEC work (http: //data. mohid. com/data. xml, support tools) Water level forecast inside estuary remains an important problem
7. Preliminary Conclusions l An advanced data assimilation scheme was implemented in MOHID Water: – – Assimilation. Pre. Processor tool can be used for validation or data assimilation Sequential Assimilation Module usable for data assimilation aimed at: l l l initial modelling conditions (Objective Analysis) on-line correction of model hydrodynamic and water properties forecasts (time series of observations) Data assimilation in TEPOMS still does not provide acceptable results
- Slides: 18