Diese Präsentation wurde erfolgreich gemeldet.
Wir verwenden Ihre LinkedIn Profilangaben und Informationen zu Ihren Aktivitäten, um Anzeigen zu personalisieren und Ihnen relevantere Inhalte anzuzeigen. Sie können Ihre Anzeigeneinstellungen jederzeit ändern.

Issues and Strategies for Integrated Model Calibration

312 Aufrufe

Veröffentlicht am

For more information about our work please feel free to visit our website at http://earthfx.com/

Veröffentlicht in: Ingenieurwesen
  • Als Erste(r) kommentieren

  • Gehören Sie zu den Ersten, denen das gefällt!

Issues and Strategies for Integrated Model Calibration

  1. 1. 1 Issues and Strategies for Integrated Model Calibration MODFLOW and More 2015 Dirk Kassenaar, E.J. Wexler P.J. Thompson, M.G.S. Takeda Earthfx Inc.
  2. 2. 2 Integrated Modeling ► Integrated modelling can provide significant insights into the overall system behavior and response to complex stresses ► Numerous technical and non- technical issues: ► Rainfall runoff models are plagued by numerical daemons  Mary Hill, June 1, 2015 ► Without the non-linear pressure saturation relationship of variably saturated flow the terrestrial system would simply not work  Stephan Kollet, June 1, 2015 After USGS
  3. 3. 3 Presentation Objectives ► Issues and Strategies for Integrated Modelling  Is integrated modelling different?  Technical Issues: ► Complex non-linear processes, compensating errors, long run times…  Non-Technical Issues: ► Knowledge limitations, different conceptual models, biases, terminology… ► Strategies for addressing these issues:  We present a general strategy and flow chart for model development, with some examples
  4. 4. 4 Background ► Integrated Stratigraphic/Groundwater modelling  Some GW modellers have only a limited background in geology ► Geology is a “knowledge boundary”  Re-conceptualization of the stratigraphic model is rarely undertaken once the GW model calibration process has begun. ► Geologic refinements and issues usually addressed with K zones or parameter estimation ► Integrated SW/GW modelling  Similar knowledge boundaries, limitations and modelling issues  “Compensating errors” (adjustment of GW model parameters to account for SW processes, and vice versa) is a bigger issue
  5. 5. 5 Presentation Outline ► Technical Issues and Challenges  Discussion of issues, with examples of soil zone response and dynamic GW feedback to illustrate challenges ► Strategies for integrated model calibration  Presentation of an integrated model development “flow chart”  Other guidelines and recommendations ► Non-technical issues  Data management, blind spots, “Renaissance Hydrogeology”
  6. 6. 6 Technical Issues ► Historic simplifications  GW: Baseflow separation, too many constant heads  SW: Lumped parameter catchment models, deep groundwater reservoirs, hydrology/hydraulics ► Calibration approaches  GW: Emphasis on matching heads and spatial patterns ► Less emphasis on regional flux calibration; recharge guesstimates  SW: Emphasis on matching streamflow peaks ► Limited emphasis on spatial and low-flow calibration ► Both surface water and groundwater modellers have “blind spots” and convenient simplifications that must be addressed early in the integrated model development process
  7. 7. 7 Technical Issues ► The shallow subsurface, where the integration happens, is highly transient and complex ► Significant fluctuation in system feedback  GW Feedback is highly variable – wet year/dry year, seasonal  Empirical baseflow separation is only a first guess ► Strong seasonality means the average conditions never exist  Steady state calibration can be very limited in the upper system ► In summary, dynamic feedback is reality – get on with it  Recognizing the dynamic nature is essential to the calibration process
  8. 8. 8 Integrated Model Development Flowchart: Step 1 ► Identify areas and scale of integration ► Pre-identify areas of strong transient interaction  Shallow depth to water – Dunnian rejected recharge ► Enhanced ET in areas with shallow depth to water table  Dynamic wetlands – storage  Riparian zones and “contributing areas”  Reaches with significant river pickup and loss ► Headwaters, springs, intermittent streams ► Seepage areas ► Identify, but avoid, these areas during initial model construction!
  9. 9. 9 GW Feedback Zones ► Dunnian rejected recharge may likely occurs in areas with:  Depth to water table less than 2 m  Areas with flowing wells, springs and headwater seeps
  10. 10. 10 Time-varying GW Feedback ► The “contributing area” that generates true runoff depends on the time-varying position of the water table ► Example: Dunnian process area varies seasonally between 5 and 25% of the study area ► Runoff occurs, but it is a groundwater dependent process!
  11. 11. 11 GW Discharge to the Soil Zone (Daily) Click for Animation Daily GW discharge to soil zone
  12. 12. 12 Step 2: Data and Model Tool Integration ► Integrated relational database  You need an integrated database to build an integrated model  Reduce barriers to integrated understanding and calibration  Need ability to assess cross-system response, trends, etc. ► Integrated modelling tools  Spatial visualization of SW processes – look beyond the gauge  Temporal visualization of shallow GW dynamics  Encourage both the SW and GW team to “visit the other domain”
  13. 13. 13 Step 3: Integration Conceptualization ► Address the shallow conceptual model  Discuss soil zone properties, thickness, storage, drainage, interflow  Develop compatible groundwater layer 1 geometry and properties ► Avoid the temptation to over-simplify the shallow system.  Resist “old habits” previously used to avoid dry GW cells ► MODFLOW NWT – stable representation of shallow complexity  Beware of SW “discharge to deep groundwater”
  14. 14. 14 SW vs GW Conceptualization ► SW Conceptual Model  Macropores  Preferential flow  Throughflow  Interflow  Subsurface stormflow  Infiltration/percolation/ drainage/recharge  Event mobilized GW  Soil/rock contact zone interface flow  Seepage faces ► GW Conceptual Model  1-D or 3-D Richard’s equation from Lin, 2010
  15. 15. 15 Storage and 3D movement of water in the Soil Zone ► Soil zone moisture content Beach Deposits Till Upland - Till uplands drain both vertically and downslope - Lateral drainage to the beach deposits from the till uplands enhances recharge - Soil zone storage helps supply rate limited GW recharge to the lower layers Click for Animation
  16. 16. 16 Soil Zone Drainage (GW Recharge) ► When moisture is available (winter months) there is a near constant, but rate limited, drainage from the soil zone ► Click for Animation Beach DepositsTill Upland
  17. 17. 17 Step 4: Sub-model Development ► Focus on:  SW and GW model construction and parameter preparation  Data review, assessment and pattern identification  Understanding of general sensitivity ► GW: Focus on the deeper GW flow system ► SW: Pre-calibrate to a gauged sub-catchment with relatively modest GW/SW interaction  Assume parsimony (consistency) when later extrapolating parameters to adjacent catchments.
  18. 18. 18 Step 5: First Integration Simulation ► Get the models and the team working together ► Re-conceptualize as necessary ► Write a draft report to formulate your understanding and impress your boss/client with your progress
  19. 19. 19 Time Step ► The timing of the SW and GW processes is very different, and a major source of contention ► Daily time step in GSFLOW:  Too fine for GW modelers  Too coarse for SW modelers Click for Animation
  20. 20. 20 Step 6: Sub-model Refinement ► Uncoupled model refinement  Update the conceptual model as necessary  Refine model parameters  Focus on the timing of the interaction ► GW: Focus on transient shallow system response  Ensure that surface discharge and groundwater discharge to streams matches observed wetland patterns and surface stream flows ► SW: Focus on the split between interflow and recharge ► In this final uncoupled simulation phase, the modellers must recognize that model response will not reflect interaction
  21. 21. 21 Step 7: Final Integrated Calibration ► Lots of re-thinking and even re-conceptualization  System response timing and lag is sensitive ► Two key benefits of the final integrated calibration process  Model Input: Measured total precipitation  Calibrate to: Measured total streamflow ► Baseflow separation is only good for the preliminary stages ► Focus on matching low flows, and not just the peaks  Balanced calibration to heads (GW) and flux (streamflow)
  22. 22. 2222 Aquifer Head vs. Stream Stage • GW/SW discharge reverses during each storm event • Baseflow separation does not account for reversals • GSFLOW Simulated Hydrograph at Oro-Hawkstone stream gauge Storm Event Reversal: Stream level higher than aquifer Dry period: Aquifer level higher than stream = GW discharge
  23. 23. 23 VL-GSFLOW GW Recharge ► GSFLOW provides ground water recharge estimates on a daily basis Click for Animation
  24. 24. 24 Non-Technical Issues and Strategies ► Expect to do a lot of education: clients and peer reviewers  Include a plenty of simplified details about model integration in your reports (no one wants to read the manuals) ► Don’t get too attached to preliminary results  Integrated conceptual models frequently require change,  Watch for “blind spots” ► Management: Identify a someone who knows a little about everything to oversee integration  A polymath or renaissance hydrogeologist is needed for mediation, and “compromise”
  25. 25. 25 Conclusions ► Integrated Modelling is different; It requires:  Integrated calibration strategies ► Don’t become attached to your initial uncoupled calibration estimates! ► Consider re-conceptualization, even late in the integrated process  Integrated data management ► Data silos and barriers will only hide the relationships and response lag between the systems ► Integrated modelling and calibration tools  An integrated and balanced modelling team ► The skill, multi-disciplinary knowledge, and ability of the SW and GW experts to address their “blind spots” is very important ► Our experience after building 9 fully-integrated GSFLOW models: It’s hard, but it’s worth it.