International Special Edition 3/12EditorialFEFLOWs metamorphosis completedProf. Hans-Jörg G. DierschDirector of GMC FEFLOW 6.1This is a special FEFLOW issue supplemen- Its role and place in the FEFLOWting the 3rd International FEFLOW UserConference held at Berlin, Germany in developmentSeptember 2012. Let us remember: Threeyears ago at the preceding 2nd FEFLOW timelineuser conference 2009, the major FEFLOWrelease 6.0 introduced a completely newworking environment that began to replacethe classic motif-style GUI, which had beenpresent over a period of about 20 years. Continue see at page 2ContentFEFLOW 6.1Its role and place in the FEFLOWdevelopment timeline 1FEFLOW 6.1 for geothermalmodeling 3Advances in visualizationand user interface 4Powerful features for a newlevel of workflow efficiency 6Three years since the firstascent of FEFLOW 6– where are you? 8 Cover Fig.: GeothermalThe FEFLOW-PEST interface 9 modeling (Detail)Solving large problemsin FEFLOW 10 Rainer Gründler & Volker ClausnitzerThe FEFLOW book 11Case Study With the release of FEFLOW 6.0 in 2010, windows, FEFLOW 6.0 brought a newSpeeding up and improving theconceptual modeling process 12 the long-established Motif user interface visual quality in interactive groundwaterPrognoses of groundwater was joined by a modern Qt-based GUI modeling. The efficiency of many typicalinfluenced flooding of that reflects the technological advances in workflows was greatly enhanced by theabandoned mining areas both GUI toolkit and graphics technology. conceptual distinction between selectionin Eastern Germany 14 In particular, the new user interface inte- and assignment operations, combinedWater regulation measures grated the capabilities of the previously with generous Undo/Redo support.in the area of Schönebeck separate FEFLOW Explorer visualizationAnalysis and evaluation 17 software and employs OpenGL to make At the same time, the complete separationNews 18 full use of the broad availability of hard- of simulation kernel and user interface• Modeling software Leapfrog & FEFLOW to combine in dewatering and remediation ware-accelerated graphics processing. provided the foundation for the develop- solution ment of alternative, possibly purpose-tai-• Indian delegation learns about DHI’s flood forecast systems in Europe Offering several model-view types (3D, 2D lored graphical or command-line-based• New Staff slice, and 2D cross section) that can be FEFLOW user interfaces – such as the• FEFLOW Training in 2012 used simultaneously in multiple view FEFLOW console application that has
2 FEFLOW 6.1 already been made available with FEFLOW face pioneered the practical applicability simulators, which will be facilitated via an 6.0, or the FEFLOW-PEST interface appli- of graphically interactive groundwater OpenMI wrapper around the new Open cation presented with the 6.1 release. modeling. It will now be completely repla- FEFLOW API. ced by the Qt-GUI whose flexibility is an Replacing the Motif-GUI essential prerequisite for the future A key long-term goal (6.3 or 6.4) is the At the release of version 6.0 most FEFLOW FEFLOW development. The release of ver- handling of fully unstructured (tetrahe- functions were accessible via the new Qt- sion 6.1 thus concludes a transition period dral) FE meshes. It is expected that a typi- GUI. However, some essential features still that was marked by the twin GUI (Qt and cal tet-mesh workflow will involve the required using the traditional Motif-GUI. Motif) of FEFLOW 6.0. import of (possibly CAD-generated) mesh geometry and topology directly associa- ted with an elemental material index. The necessary introduction of a "material" con- cept to handle predefined parameter groups is conceivable with version 6.2. First steps beyond a prismatic mesh with strictly vertical edges are also planned for 6.2, specifically the ability to completely deactivate mesh elements (with respect to both computation and visibility) and a relaxation of the mesh geometry where slice-connecting element edges will no longer need to be strictly vertical while Fig. 1: FEFLOW development timeline maintaining the general prismatic mesh topology. Declared ambition for the next release was As a side benefit, FEFLOW no longer requi- thus the complete representation of all res any X-server software which simplifies Users strongly call for a user-friendly appli- FEFLOW features in the new interface. the installation, removes the potential for cation to control PEST automated calibra- X-server-related technical complications tion, technical optimization (well position, FEFLOW 6.1 delivers on this promise. The and reduces third-party dependency. etc.), and uncertainty analysis of FEFLOW remaining gaps have been closed and as a models. Medium-term, it is intended to result the 6.1 Qt-interface supports crea- There are significant improvements and develop the new-generation FEFLOW/ ting, editing, assigning, and visualizing of new features in FEFLOW 6.1 that distinct- PEST interface into a convenient GUI- ly go beyond the urgently awaited com- equipped scenario manager/batch proces- • Transient material parameters, pletion of the Qt-GUI – they are the sub- sor with remote-execution control via • Material parameters for unsaturated ject of separate articles in this issue. TCP/IP for platform independence and flow, potential cloud-computing capability. • Anisotropic conductivity distributions, Looking ahead • Discrete features (fractured, pipes, chan- Prompted by numerous requests from the As of version 6.2, the FEFLOW map inter- nels, ...), user community, a second, Python-scrip- face will extend beyond file-based data • Budget node groups, ting console application is currently under and also include compatibility with ESRI • Borehole-heat exchangers. development. ArcGIS as well as GIS map services such as Google Maps, OpenGIS, etc. This implies Familiar to many users for over two de- Currently in prototype stage is the coup- capability development for on-the-fly pro- cades, the classic FEFLOW Motif user inter- ling of the FEFLOW engine with other jection and data processing of maps. Continue from page 1 However, at that time the new inter face appears forever(!). The software archi- term change of the GUI and software did not fully cover the entire FEFLOW tecture has been modernized and gene- architecture: it was the ongoing support functionality and a resort to the classic ralized to attain a much higher extensibi- of our user community that allowed for GUI was still required in cases invol- lity and efficiency. Indeed, release 6.1 the necessary steps to be taken in a very ving, for example, unsaturated condi- marks a new era in the FEFLOW develop- careful and rigorous manner. In fact, it tions, fracture modeling, transient ma- ment and paves the way for scientific and was an open-heart surgery. But finally terial data, or certain specific boundary technological extensions which have been and fortunately, the patient lives (sur- conditions. deferred so far. vives) – even better, it got a drastic reju- venation to keep running over the next Now, with the release of FEFLOW 6.1 the We thank all users for their understan- decades, surely in the leading group. work is complete and the classic GUI dis- ding and patience related to the long-
Products 3FEFLOW 6.1 for geothermalmodelingAlexander RenzWith the rapid growth of the geothermal features were added and some historicalenergy market, simulation models for heat shortcomings could be overcome. As atransport have emerged from an exotic result, FEFLOW 6.1 provides more features,exercise to a standard application of higher accuracy and more productivity forFEFLOW during the last decade. With its geothermal projects than ever before.ability to consider all relevant processes(advective, conductive and dispersive heat Closed-loop systemstransport, temperature-related fluid densi- The incorporation of Borehole Heatty and viscosity) and providing the flexible Exchangers (BHE) including their internalmesh geometry needed for the implemen- heat transport is a unique feature in com-tation of complex structures for deep or mercial groundwater modeling systems.near-surface geothermal models, FEFLOWhas become an industry standard in many The BHE are now fully accessible in thecountries. new user interface, including the possi- bility to import large numbers of BHEMotivated by this development, DHI- and their properties directly from mapWASY has put a special focus on geo- files.thermal energy whenever extendingFEFLOWs functionality. Consequently also The new BHE boundary condition allows athe development of the new user interface direct prescription of power or tempera- Parallel or serial connections of BHE can be Fig. 2: Virtual radiuswas taken as an opportunity to incorporate ture differentials without the previously implemented to allow the correct descrip-the latest experiences in this sector; new necessary application of a plug-in. tion of BHE arrays. To assist the user in choosing the correct mesh geometry around the BHE, additio- nal visualization styles are available to show the real and virtual well radii as well as the ideal distance to neighboring nodes. This is of critical importance for the accurate calculation of BHE and ground temperatures. The temperature development of the indi- vidual BHE elements can be monitored in a new chart. Open-loop systems With its new Multilayer Well boundary condition, FEFLOW 6.1 uses a different and more physical approach to calculate heat transport in wells screened in mul- tiple model layers. More precise tempera- ture values can be obtained for extraction wells of open-loop systems by using thisFig. 1: BHE Array approach.
4 FEFLOW 6.1 Fig. 4: Geothermal gradient expression Fig. 3: Geothermal General improvements 6.1. Good examples are the possibilities to to assign temperature boundary condi- gradient applied Modelers of geothermal energy projects apply the geothermal gradient directly as tions from a calculated temperature field will significantly profit from a number of initial temperature distribution using the using the copy & paste functions. more general improvements in FEFLOW new expression-based assignment tool, or Advances in visualization and user interface Volker Clausnitzer & Julia Mayer The much anticipated completion of the Qt user interface goes distinctly beyond reproducing the features of the now-lega- cy Classic interface. FEFLOW 6.1 brings powerful new GUI tools for model editing as well as advanced new visualization capabilities. Model editing Full GUI control of transient material pro- perties: For many, this was perhaps the most critical feature still missing in FEFLOW 6.0. Version 6.1 provides con- venient options for manual assignment and the use of time series as well as powerful map import. Similarly, FEFLOW 6.1 not only closes the gap regarding the editing of multilayer wells, borehole heat exchangers, and variably-saturated flow models, but in each case provides more Fig. 1: Backward pathlines shown with period sections
Products 5comfortable and intuitive editing than hasbeen available in the Classic interface.The concept of selections for parameterassignment introduced in version 6.0 hasbeen employed to extend discrete-featuremodeling capability: Arbitrary mesh nodeson different slices can now be connectedto one-dimensional discrete features. Thisprovides, for example, an easy way tomodel inclined wells or boreholes.Data plottingFEFLOW 6.1 significantly extends the plot-ting capabilities of the previous version.Notably, forward and backward pathlinesand/ or streamlines can now be displayedsimultaneously while colored period sec-tions facilitate quantitative travel-time View windows direct three-dimensional manipulation. Fig. 3: Labeled cross- sectional lines and asso-analysis (Figure 1). FEFLOW 6.1 complements the existing FEFLOW 6.1 takes a first concrete step in ciated chart views cross-section, slice and 3D view-window this direction by offering a variety of visua-A new isoline implementation includes fle- types by a new chart-like view that allows lization options that, together with suit-xible, editable in-line labels and also pro- the monitoring of data plots along user- able hardware, can give a "real" sense ofvides more convenience: Based on the defined 2D polylines on a mesh slice, or 3D. Specifically, two video signals (one forcurrent (user-specified or automatic) value along an arbitrary 3D polyline (Figure 3). each eye) are provided simultaneouslyrange, the interface automatically sug- and separated via special glasses thatgests appropriate "round" isoline levels In any slice or 3D window, the default make use of either shutter technology orand allows the definition of major and navigation reference (full object view) can polarized light. With FEFLOW 6.1 it is pos-minor levels (Figure 2). now be replaced by a mesh-item selection sible to export model snapshots and (such as an element or node group) to movies in various 3D stereoscopic formats. obtain, for example, a user-defined center Given a 3D-capable monitor or projector, of rotation. Also, multiple view windows FEFLOW 6.1 also allows the interactive ste- may be temporarily defined as slaves of a reoscopic display in the familiar FEFLOW positioning-master view and will then 3D view windows. automatically follow the navigation of the master. Compared to conventional 3D visualiza- tion, the difference in perception is parti- Floating view windows for the positioning cularly evident in models with complex of views outside the FEFLOW application structures. While traditional 3D displays window and a full-screen mode offerFig. 2: Major and minor isolines with in-line labels more flexibility for presentation purposes as well as for the interactive work with multiple screens. The full-screen modeTo facilitate a quick display of the water can be easily activated for any view wind-table, FEFLOW 6.1 provides predefined ow and allows the display of the modelzero-pressure isoline and isosurface plot view at the maximum possible resolutionstyles. Appearance settings can be easily for the given screen. All tools (navigation,copied not only between entire views but selection, etc.) remain accessible and fullyalso for individual parameters or plot functional.styles. Stereoscopic visualizationThe visualization extensions also include Lastly, a technological trend merits men-an option for displaying calendar time ins- tioning that could become more prevalent typically require a rotation of the object to Fig. 4: Stereoscopic pro- jection can improve thetead of elapsed simulation time, and new in the near future: It appears increasingly obtain an adequate depth perception for perception of complexannotations such as embedded polyline likely that soon set-up and editing of such structures, a stereoscopic display pro- spatial structureslabels or title, header and footer text for three-dimensional spatial models may be vides natural depth much more directlyeach view window. assisted by stereoscopic visualization and and easily.
6 FEFLOW 6.1 Powerful features for a new level of workflow efficiency Olaf Arndt & Rainer Gründler Map usage FEFLOW 6.1 introduces embedded super- mesh and POW files as new map types. Several features were added to the map handling: • Filtering map records via SQL (Struc- tured Query Language) statements • Joining of multiple maps using common key attributes • Integrated map table data viewer Fig. 2: Expression based selections Data assignment FEFLOW 6.1 provides new tools to simplify the assignment of parameter data. In par- ticular, two new features are remarkable: • Assignment of discrete values to multi- ple parameters in a specific area at once • Assignment of parameter values by ex- pression (see Expressions section below) Expressions The FEFLOW Expression Editor is a power-Fig. 1: Joining maps with In addition, expression-based manipula- Fig. 3: Editing group of parameter values ful, generalized approach that replaces SQL filtering tion can be performed on map attributes during data assignment and 3D map selection was added for 3D maps. Selections In FEFLOW 6.1 the capability of selecting mesh items has been extended: New topology types (slice edges, join edges, slice faces, and join faces) are now suppor- ted and the selections were separated from the current model data. Each topolo- gy type has its own selection state which makes it possible to switch between nodal and elemental selection without losing the corresponding current selection. Additio- nal features of selections are: Fig. 4: Expression based value assignment • Undo/redo operations are aware of drop in during such operations and far surpasses the Debug function for mesh operations (refinement, de-refine- • Expression-based selection (see Expres- conditional editing of parameter values ment, etc.), hence, there is no need to sions section below) and process variables in the Classic user
Products 7 Fig. 6: Scatter plotFig. 5: Error bars and confidence intervalsinterface. Simple or complex user-defined • Confidence intervals tion of the FEM mesh, superelement Fig. 7: Observation point property editormathematical expressions can now be • Error bars with qualitative indication mesh, observation points and additionalused to create element or node selections • Error labels displaying the absolute location sets (2D/3D points, polylines, andthat are based on current parameter val- deviation loops). Besides standard affine and Hel-ues, process variables or mesh data such mert transformations, FEFLOW 6.1 sup-as coordinates or node/ element numbers. Mesh geometry ports all WGEO transformations as well. ByThe Expression Editor can further be used The classic version of FEFLOW provided means of a Microsoft-COM-based API it isto evaluate parameter values based on various tools to check the meshing quality possible to extend the set of availableuser-defined expressions and to assign (mesh geometry submenu). In FEFLOW transformations by custom transformations.them to an existing selection. Finally, 6.1 most of these tools are represented asFEFLOW 6.1 provides the means to create auxiliary parameters. They include:new parameter distributions that are de-fined by mathematical expressions. These • Slice distance (distance between nodesare updated dynamically during the com- at slice and the slice below)putation and can be visualized analogous- • Layer thickness (distance between ele-ly to other internal FEFLOW parameters. A ment center points of neighbored layers)prominent application example is the • Maximum interior angle of triangulardynamic display of groundwater draw- elementsdown during a simulation run or while • Elements violating the Delaunay criterionreplaying a simulation record. These parameters can be used like regularData calibration tools parameters for visualization, export, plot-FEFLOW 6.1 offers an extended set of tools ting, and in expressions. Scalar features File input and output Fig. 8: Geographic trans- formationsfor the calibration of FEM models. Refe- such as mesh area and volume have their As has been customary throughoutrence values can now be used in addition to place now in the Summary page of the FEFLOW history, the current version 6.1the regular observation points to visualize problem-settings editor. is able to read and write the file formatthe deviation between measured and com- of any previous version. New is the abi-puted parameters. New visualization op- Geotransformation lity to write into temporary files rathertions are available for process variables such The re-implemented geotransformation than overwriting the target files directly.as hydraulic head or concentration: dialog facilitates the complete transforma- This keeps existing files unaffected until
Products 9The FEFLOW-PEST interfaceErik McCurdyWhat is PEST? here: http://www.feflow.com/uploads/media/ Like the IFM module it will provide a userPEST (by J. Doherty, Watermark Numerical pest_feflow.pdf. interface for configuring and running PESTComputing) is a software tool which can for a FEFLOW problem but it will use abe extremely helpful when calibrating PEST in previous versions new approach: the new tool will be a se-models. It achieves this goal by using of FEFLOW parate executable which will start themodeling software in an inverse way: It An IFM module is available which provides standard PEST-executable. It will not beattempts to find the combination of input a user interface to configure and run PEST reliant on any particular version of PEST,values, which result in a set of known out- with FEFLOW 6.0 in Classic mode. A spe- which was a major disadvantage of theput values. This requires that the model be cial recoded version based on PEST version IFM module. Users can therefore now benefit from im- provements and newer features as they become avail- able with new PEST versions. The PEST files created by the tool can be edited manually prior to running PEST and thus PEST-functiona- lity not available through the user in- terface may be acti- vated. Parameter zones are easily defi- ned by creating ele- ment sets within FEFLOW and obser- vations by setting re- ference parameter values at observation points. Future plans Aside from perfor- mance improve-Fig. 1: A screen shot of the new FEFLOW-PEST interface ments, new features currently under de-run many times to test how the output 2.0 was used for this module which en- velopment include support for pilotvalues depend on the input values. abled it to run from within the FEFLOW points, Parallel PEST, and map services. process as it is loaded as dynamic linkPEST is model-independent, requiring library. Starting with version 6.1 the PEST DHI-WASY has committed substantialonly that the modeling software is able to IFM module will no longer be supported resources for advancing the FEFLOW-PESTread input and write output to text files. due to the discontinuation of FEFLOW interface and suggestions from the http://www.feflow.com/upUsing the text versions of the FEFLOW Classic mode. FEFLOW community will be very impor- loads/media/pest_feflow.fem- and dar-file formats it is possible to tant in deciding what to include in future pdfuse PEST with FEFLOW by constructing PEST and FEFLOW 6.1 releases.PEST configuration files manually. An ar- A new tool for FEFLOW has been deve-ticle describing this procedure is available loped to replace the legacy IFM module.
10 FEFLOW 6.1 Solving large problems in FEFLOW Vladimir Myrnyy The performance limit of a single-proces- Institute for Algorithms and Scientific SAMG threads is completely independent sor core essentially having been reached Computing which implements a hierar- of the separately entered number of several years ago, any increase in the com- chic multi-grid method using OpenMP. threads to be used for matrix assembly putational power of modern computers is parallelization. The selected options per- now mainly related to multicore archi- SAMG Solver sist for future FEFLOW sessions on the tectures with recent advances including For several years now SAMG has been a same workstation. many cores of graphical processors stable and efficient alternative to the (GPUs), specific processors with many default PCG for symmetric and BiCGStab When SAMG cannot achieve the targeted integrated cores (MIC), and the processor for non-symmetric matrices for large accuracy, a warning message will appear vector extensions SSE and AVX. models. In FEFLOW 6.1 the user can select in the log window reporting both the FEFLOW 6.1 provides basic parallelization Table 1: Benchmark models description capabilities to use multicore shared- Model Flow Mass Simulation Mesh Number of Number of memory architectures and accelerate the transport period elements nodes simulation process. The performance 1 √ √ 100 d Triangular 28940 17712 benefit is strongly related to model size 2 √ 100 d Triangular 459155 277158 and can easily reach several days of com- 3 √ steady Quadrilateral 640000 667521 putational time for models with millions of 4 √ √ 10 d Triangular 1852160 1114038 nodes. Each time step of a simulation process can between three versions: 2.3a5, available actual and target residual values. If that be roughly split into two steps: matrix since FEFLOW 5.4; 2.5a1, the default; message appears frequently during the assembly and solution of the linear equa- 2.6a1, the latest, recently released version. simulation, a critical evaluation of the The major improvement of version 2.5a1 overall accuracy is indicated. was a new automatic solver control for the benefit transient simulations. It attempts Benchmarks to minimize set-up cost by reusing set-ups The performance benefit of parallelization from previous calls, and also dynamically was tested on four models covering diffe- switches between multi-grid and an alter- rent problem classes – one relatively small native (one-level) solver – whichever ap- one for comparative reasons, and three pears to be more efficient. Version 2.6a1 larger ones with hundreds of thousands of provides a better OpenMP parallelization mesh elements each (Table 1). All tests as well as additional one-level solvers, were executed on a server with two Xeon including ICCG for symmetric matrices X5690@3.47Ghz running the Windows and the direct solver PARDISO. Server 2008 operating system. The FEFLOW 6.1 Table 2: CPU time in minutes for different solvers user interface of- in single-threaded mode fers controls for Model PCG/ SAMG SAMG SAMG selecting the BiCGStab 2.3 2.5 2.6 Fig. 1: FEFLOW 6.1 tion system based on the assembled SAMG version as parallel-computing 1 1.0 3.2 1.3 0.8 dialog window matrix. The assembly step is parallelized well as the number 2 5.4 12.3 5.7 4.6 via an efficient mesh partitioning. A paral- of SAMG threads 3 0.9 1.3 0.6 0.9 lelization of the solution step is achieved (Figure 1). Notab- 4 144.4 190.3 138.5 125.3 by the SAMG solver from Fraunhofer ly, the number of
Products 11 Table 3: CPU time in minutes for 1, 2, 4, 8 and 12 threaded perfor- threads mance relative to PCG/ BiCGStab. Model 1 2 4 8 12 1 1.3 1.1 0.8 0.6 0.6 SAMG parallel 2 5.7 3.8 2.6 1.9 1.8 performance 3 0.6 0.4 0.4 0.3 0.3 Parallel perfor- 4 138.5 77.1 66.3 52.2 42.6 mance has been tested on theSingle-thread performance of iterative default SAMG version 2.5. The same foursolvers simulations were executed using 1, 2, 4, 8Table 2 compares single-threaded perfor- and 12 threads (Table 3). In every test themance, expressed in CPU-time over all same number of threads was used fortime steps of the simulation. The results both matrix assembly and SAMG. Theclearly show that in terms of single-threa- general trend shows better speed up (Fi- The speedup coefficient is significantly less Fig. 2: Speedup with respect to the numberded speed SAMG 2.3 is inferior to the gure 2) for larger number of mesh nodes than the number of used threads, because of threadsPCG/ BiCGStab solvers in all cases. with an additional dependency on the of inevitable sequential parts in the codeMoreover, the results illustrate the benefit problem class. For instance, model 2 has and thread-handling overhead. Still theof adaptive switching to a simple one-level fewer mesh nodes than model 3, but sol- speedup can increase linearly for large mo-solver introduced with version 2.5, which ves flow and transport at every time step dels containing more than one million meshsaves a multi-grid set up: versions 2.5 and and thus derives more benefit from multi- nodes as seen for model 4 in Figure 2.2.6 deliver comparable or better single- threading.The FEFLOW bookHans-Jörg G. DierschA new textbook titled FEFLOW – Finite The book consists of 15 chapters, which saturated media, fractured media, andElement Modeling of Flow, Mass and Heat can be formally grouped into two major unconfined and confined aquifers. Appro-Transport in Porous and Fractured Media parts: (I) fundamentals and (II) finite ele- priate boundary, constraint and initial con-will be published by Springer next year. ment method (FEM), supplemented by a ditions complete the model formulations.This comprehensive book covers on near- number of appendices. Part I covers thely 1000 pages the full spectrum of state- fundamentals of modeling flow, mass and Chapters of part II explain and emphasizeof-the-art modeling from fundamental heat transport processes in porous and the FEM in a number of areas: (1) funda-concepts to problem solutions, extensive- fractured media. It starts with prelimina- mental concepts (also in relation to otherly and systematically describing FEFLOWs ries, where all important notations, defini- numerical methods such as finite dif-underlying basics and computational tions and fundamental algebra used ference, finite volume and spectralcapabilities. It fills the gap we recognize in throughout the text are explained. methods) including upwind schemes withthe existing reference materials and va- Chapters deal with a more general theory their pros and cons, numerical stabilityrious white papers, where the matter has for all relevant phenomena on the basis of and accuracy, sparse matrix solution tech-been presented incoherently and parts the continuum approach, develop syste- niques, treatment of nonlinearities, suit-appeared incomplete. We believe that matically the basic framework for impor- able time-integration procedures, properadvanced software such as FEFLOW needs tant classes of problems and derive all rele- approaches for flux computation, budgetan appropriate fundamental reference vant constitutive relations based on uni- evaluation and property of local conserva-work that provides a high transparency of fied and rigorous principles. The general tivity; (2) flow in saturated porous mediathe theoretical and numerical back- set of model equations for the multiphase with free-surface simulation on rigid andground, makes the available solution and multispecies, chemically reactive, moving meshes including multilayer wells;methods accessible, reliable and better variable-density, non-Darcy and non-iso- (3) flow in variably saturated porousunderstandable, and helps users to extend thermal processes are reduced, in stepwise media with the different forms of Richardstheir insight in modeling and to open new manner, to important model subclasses in equation, preferred solution techniquesfields of application. practical modeling applied to variably and hysteresis; (4) variable-density flow in
12 FEFLOW 6.1 porous media including sharp interface features; and (8) special topics including sive theoretical fundament and modeling approximation, convection phenomena, mesh generation, pathline tracking and reference; however, due to its general double-diffusive fingering and useful others. In all areas a number of prototypi- approach it will also be useful for students approximation methods; (5) mass trans- cal examples and benchmarks are presen- and practitioners in engineering, geo- port in porous media with and without ted to illustrate the capability, usefulness, sciences, and other branches where chemical reactions; (6) heat transport in accuracy and efficiency of FEM. porous media flow dynamics and compu- porous media involving application of tational methods are of specific concern. borehole heat exchangers; (7) flow, mass The book is primarily written for current and heat transport occurring in discrete and future FEFLOW users as a comprehen- Consulting Case Study Speeding up and improving the conceptual modeling process Thomas Krom, ARANZ Geo Limited Leapfrog Hydro is a tool for knowledge proach which results in a rapid work in the FEFLOW flow and transport simu- discovery and numerical analysis within flow for the development of geological lator. To quote a user: "Having worked geological systems. The software is models. These models are grid free and with geological modeling tools since based on the implicit modeling ap- can then be transferred to grids for use 1990, I find Leapfrog Hydro to be the most innovative, cost effective, and flexi- ble geological modeling tool I have yet seen. Leapfrog Hydro helped us to quick- ly solve the most complex geological modeling challenge weve faced in years." Key tools Leapfrog Hydro is designed for inte- gration in your business processes and uses data from a wide variety of sourcesFig. 1: Data at the site. including: Borehole logs are blackdots while DC-resistivity data is shown as • GIS data in vector and raster format colored lines. • Images (georeferencing tools) Fig. 2: An E-W same cross-section through the data set, on the left the grouped borehole descriptions and on the right the same borehole data after the main aquifer (light blue) is separated out using the graphical selection tools in Hydro.
Consulting 13 The user can model geological units using the most appropriate data. Figure 4 shows a large sand lens modeled directly from DC resistivity data. The geological model can be used to guide the development of the FEFLOW model; and the FEFLOW model including applying geological zonation can be develo- ped literally in minutes once the geological model is developed. Figures 5 and 6 compare the FEFLOW model grid against the geological model, while Figure 7 shows the model in FEFLOW.Fig. 3: Cross-section from Figure 2, but with the geological model shown, note that projection width is less than in Figure 2.• Borehole logs: lithology, wire-line, well screens et al.• Well screen data (e.g. water levels and concentrations)• Data from electrical geophysical surveys (resistivity and TEM)• Point data (e.g. picks from old paper logs)• Scanned cross-sections Fig. 4: The main aquifer• Other geological modeling packages. and gravel lenses, where the cross-sections show the background till, theCase study large sand lens in the middle is modeled directlyThe following example is motivated by the from DC-resistivity data.need to maintain water quality across apeninsula where the water quality is threa-tened from a number of different conta-minant sources. The base data is approxi-mately 350 borehole logs that cover thesite (Figure 1) as well as DC resistivity data.Geologically the model is dominated byglacial deposits and lenses which partiallycover the modeling area. Fig. 5: Cross-section view showing the FEFLOWThe detailed geology can be regrouped mesh and the outside of the lower aquifer.using tools in Leapfrog Hydro to provide Note how Hydro hasa more appropriate classification for assigned the correct material property tohydrogeological purposes, for example each element/node.coarse and fine sands can be groupedtogether. Tools are also available to splitlithologies that are in fact part of multiplestructures and need to be modeled sepa-rately (Figure 2).Several geological models can be produ-ced for a site (Figure 3) to explain differentgeological interpretations and these canthen be easily converted to a finite ele-ment mesh. It is also possible to explorethe effects of alternative meshing solutions Fig. 6: Along the cross- section line we see theon the flow modeling simulations inde- FEFLOW model sliced andpendently of the geological modeling. the borehole data.
14 FEFLOW 6.1 Training requirements are lower than for competing products. • Quality: Since time allotted to routine tasks is reduced, staff can carry out more analysis resulting in better results. Leapfrog Software is designed to aid in QA/QC. These will give you a market advantage over the competition A user survey shows that a single seat of Fig. 7: The exported Leapfrog Hydro on average is used by 3model in FEFLOW colored staff members. If each staff member canby hydraulic conductivity in the X direction. save 25 hours of work a year, the software Connection to FEFLOW The gridding starts by constructing the 2D super mesh in FEFLOW. Once the geologi- cal model is built the 3D FEFLOW mesh is constructed in Hydro using that 2D mesh as a guide, and material properties are assigned. The final 3D grid is then expor- ted to FEFLOW as a FEM file. Once the model is set up and calibrated in FEFLOW results can be shown in Hydro. Figure 8 Fig. 8: Results from a shows the water table and capillary suc- FEFLOW simulation in Leapfrog Hydro. tion for this case study. Benefits Thomas Krom Leapfrog Hydro software provides an ARANZ Geo Limited intuitive environment for geological • Productivity: The software aids staff in is paid for. Additionally, product quality PO Box 3894 Christchurch 8140 modeling, meaning a more comprehen- carrying out routine tasks more rapidly can be improved for the same resource New Zealand sive model of the geology can be built in than in traditional work flows as well as allocation for project work. www.aranzgeo.com less time. The user benefits from gains in: speeding up data cleaning and analysis. Prognoses of groundwater in- fluenced flooding of abandoned mining areas in Eastern Germany Bertram Monninkhoff & Junfeng Luo Introduction characterized by large industrial zones, produce electricity, some of the industri- Since the end of the 19th century, large producing lignite and electricity. Al- al activities are planned to be phased parts of Southern Brandenburg and though mining carries on to date and out by 2015. A number of old pits left Saxony in Eastern Germany have been three large power stations continue to by the mining process have already
Consulting 15 Fig. 1: Location of Cottbus-Nord and the surrounding mining pitsbeen transformed into a new landscape • a MIKE11 surface water model to de- water level development in the lake andformerly unknown in this region, crea- scribe different options for the outflow for detailed description of the surroundingting a huge new lake district. One of the to the river Spree. groundwater levels.lakes to be developed is the "CottbuserSee" to the East of the city of Cottbus The first two components are completely To describe the water level development(mining pit Cottbus-Nord). integrated and are used to simulate the of the lake, detailed information about theDHI-WASY was contracted by VattenfallEurope Mining AG to analyze the floodingprocess of this lake approximately 19 km²in size.Methodology and model setupTo identify the optimal flooding strategyto flood and control the lake CottbuserSee the following model components areused:• a 3D FEFLOW groundwater model• a FEFLOW plug-in termed IfmLAKE enabling detailed description of the flooding• a WBalMo water and allocation model to identify long term water needs and Fig. 2: Model compo- optimal allocation strategies nents system
16 FEFLOW 6.1 directly towards the Spree. For all alterna- tives, maximum discharges and related costs were analyzed. Results Results of the WBalMo simulations were used in FEFLOW (with IfmLAKE) to calcu- late the water levels in each single sub-lake and to analyze the influence of the pro- posed control strategy on groundwater levels in the region. An exemplary result of these simulations is shown in Figure 4. The strategy shown in this figure involves an additional surface water inflow from the river Spree under average flow conditions. The maximum withdrawal of surface water is defined according to the manage- ment rules currently established by the water authorities in the Spree river basin. The figure shows that in this case the floo- ding of the lake will take approximately Fig. 3: FEFLOW model inflow into the lake is needed. This in- The water management and allocation five years (up to a level of 63.5 m). The with applied conducti- vities and boundary flow consists of three components; (1) software WBalMo has already been used flooding would then be finished around conditions groundwater inflow, (2) surface water in previous analyses for the utilization con- 2023, which is approximately seven years inflow and (3) rainfall and evaporation cept of the mining pit Cottbus-Nord. The earlier compared with the flooding period at the lake surface. The second compo- model WBalMo Spree/Schwarze Elster was applying a strategy without additional sur- nent is calculated by the model WBalMo since extended and updated and builds a face water inflow. and includes detailed information on good basis for long-term management the long term water needs and availabi- analyses of the area in focus, and the lake Within the project new features of lity of the Spree. It also includes infor- Cottbuser See in particular. FEFLOW 6.1 could be successfully applied, mation about the maximum outflow rate of the planned diversion system, which is provided by the fourth model component (MIKE11). This model com- ponent system is shown in Figure 2. The FEFLOW plug-in IfmLAKE has already been described in DHI-WASY Aktuell 3/ 2011. For this project it has been exten- ded to offer the possibility of linking lake series. Within the FEFLOW model linking points can be defined including corres- ponding weir levels and widths. In case the lakes levels at both sides of a single weir are above the weir crest, both lakes are regarded as a single lake and the cor- responding water level – volume curves are integrated. This feature was used to Fig 4: Water level development within single lake basins with additional surface water inlet divide the lake Cottbuser See into eight separate parts. This was necessary to de- Optimization analyses of the WBalMo especially to derive exchange rates be- scribe the flooding of the different basins, model had to be restricted with respect to tween each sub-lake and the different especially the basin which was used to fill the maximum possible outflow discharge geological horizons. Here, expression- the lake with additional surface water towards the Spree. For this reason, first based selections could be used to define from the Spree river. hydrodynamic analyses using the 1D observation point groups for each single model MIKE11 were made for routes exchange area. These results could then An overview on the 3D FEFLOW ground- along the existing branch Schwarzer be used by Vattenfall Europe Mining AG to water model is shown in Figure 3. Graben, as well as for alternative routes analyze qualitative aspects of the flooding.
Consulting 17Water regulation measures inthe area of SchönebeckAnalysis and evaluationSven Seifert, Thomas Koch & Bertram MonninkhoffThe town of Schönebeck is located in the adaption of the existing trench-system culates and provides the groundwaterGerman federal state of Saxony-Anhalt should be analyzed. For this the following recharge in area and time dependent dis-adjacent to the Elbe River between the model system has been set up. tribution as input for the groundwaterneighboring communities of Barby and model. FEFLOW uses these data to calcu-Magdeburg. Previously, the city was main- Hydrologic model system late the groundwater conditions andly supplied with water by local freshwater For the analysis of the actual state and the transfers the groundwater depth back towells located in the south of the town. effects of measures, a coupled hydro- ArcEGMO as an essential boundary con-Starting around 2004, these wells were logical model system was built incorpora- dition for the groundwater influencedconsecutively shut down and the water ting two models. The first one is a soil parts of the soil water model (Figure 1).supply was changed to long-distance water budget model [ArcEGMO, BAH Details to the coupling mechanism cansupply. Consequently an increase of water (www.arcegmo.de)] to include the in- also be found in DHI-WASY Aktuell 3/2011.logging in several parts of the city could fluence of inflow from the catchment area The coupling module IfmArcEGMO is now http://www.arcegmo.debe observed in recent years. During winter2010/11, the highest groundwater levelsin recording history were witnessed,accompanied by water damage especiallyin several basements in the town.Next to the above-mentioned changes inwater supply, the constantly rising waterlevel of the Elbe River, particularly in springdue to the snowmelt, can be seen as oneof the main reasons causing the problemsmentioned. Additional surface water fromwithin the catchment of Schönebeckcould not be dissipated into the Elbe Riverin time by the existing trenches. This situa-tion led to infiltration of surface waterfrom the trenches into the groundwater,aggravating the matter at hand.Finally, the sewage system in the city is sus- Fig. 1: Coupling betweenpected to interfere with the groundwater. It ArcEGMO and FEFLOWis assumed that ongoing renovation mea-sures are interrupting this process causingan additional rise of groundwater levels. of Schönebeck as well as the time depen- also available for FEFLOW 6.1 and can be Client: dent groundwater recharge. The second conveniently applied to any 3D FEFLOW Prof. Dr. habil. Frido ReinstorfThe aim of modeling was to investigate to one is a groundwater flow model model. University of Appliedwhich extent the groundwater level can (FEFLOW) to represent the regional sub- Sciences Magdeburg-be affected by passive and active measures surface flow of the groundwater. After transient calibration of the model Stendal, Department of Water and Wastein the area. Various measures are conceiv- system a total of seven possible measures Managementable for that case. Besides additional The two models are coupled with the were simulated. Some of them includedewatering wells, other measures such as FEFLOW plug-in IfmArcEGMO in both several modifications that culminate inadvanced water retention in rural areas or directions during runtime. ArcEGMO cal- additional sub-scenarios.
18 FEFLOW 6.1 Results The impact of the measures were calcula- ted and evaluated using the new nodal expression feature available in FEFLOW 6.1. Figure 2 exemplary shows the effect of three groundwater dewatering wells that were placed at strategic points to achieve the maximum result in ground- water reduction in Schönebeck’s problem zones. Each well was set with a pumping rate of 5000 m³/d. Using the created models and the perfor- med calculations, passive measures such as modifying trenches result only in relati- vely small improvements in the area of Schönebeck, even in case that these mea- sures are combined. Active measures such as dewatering wells however could attain the best results by far. Additionally, these types of measures can be regulated accor- ding to the occurring event. But due to ongoing maintenance and operation costs these measures would be rather expensi- ve. Currently the modeling results are being analyzed by the client and the city of Schönebeck, after which decisions will be made in implementing necessary improvements to the city water system in order to minimize groundwater driven Fig. 2: Achievable drop of groundwater table with dewatering wells flooding in future times. News Modeling software Leapfrog & FEFLOW to com- bine in dewatering and remediation solution New Zealand’s ARANZ Geo Ltd and bridged. The resulting solution, which will German DHI-WASY GmbH plan to com- incorporate invaluable knowledge gained bine their flagship Products to create a from current user experiences, is planned comprehensive groundwater modeling Both programs feature multiple functions to cover the entire groundwater modeling solution. Leapfrog is a fast and dynamic for visualization and editing and are flex- workflow and provide significant benefits 3D geological modeling software capable ible, supporting users in all parts of a pro- for users focusing on mine dewatering of modeling even complicated geologies ject cycle. By extending the current link- and remediation. The novel technology and was developed specifically for the ages between Leapfrogs hydrogeological will be available and supported world- minerals mining industry, while FEFLOW solution Leapfrog Hydro and the FEFLOW wide through DHI offices. is known as a professional software pack- software up to the seamless integration of age for modeling fluid flow and transport both products, the previously existing gap If you would like more information abouthttp://www.feflow.com/ of dissolved constituents and/ or heat between geological modeling packages FEFLOW or Leapfrog software, please visit leapfroghydro.html transport processes in the subsurface. and hydraulic modeling software will be http://www.feflow.com/leapfroghydro.html.