A GIS-based open source pre-processor for georesources numerical modeling

https://doi.org/10.1016/j.envsoft.2014.08.011Get rights and content

Highlights

  • A novel approach to generate TOUGH2 input file through GIS is presented.

  • TOUGH2GIS automatically creates locally refined unstructured 3D grids.

  • Rock types attribution to grid blocks is automatically and efficiently performed.

  • Seamless integration into an available open-source GIS package.

  • An application to a sample dataset shows method feasibility.

Abstract

TOUGH2 is an integral finite differences numerical simulator for non-isothermal multiphase flow in fractured porous media, which can manage complex spatial discretizations. Numerical simulation accuracy is affected, among other things, by grid resolution. Increasing the grid resolution requires computational and operating costs depending on the number of nodes and variables processed. The complexity of the management of the model increases when unstructured grids and local refinement are used. In order to improve the management and optimize the activities to update the model, an open source pre-processor has been developed using the open source codes GRASS GIS, SQLite and AMESH. Operations such as domain discretization, rock type assignment and mesh file generation have been automatized. Graphical interfaces allow for a user-friendly utilization. Operating errors and time required by pre-processing activities to generate and update locally refined unstructured grids have been reduced. Productivity in numerical modeling has been substantially increased.

Introduction

Numerical simulation is a powerful decision-making support tool. It allows the study of physical processes and to evaluate the environmental consequences of resources exploitation and the sustainability of complex geosystems management choices. Concerning the environmental impact assessment of georesources exploitation, the approach with numerical models is, de facto, unavoidable.

In general, numerical simulation requires the development of a conceptual model, an outline that allows for a simplified analysis of the real system and provides its representation in terms of initial and boundary conditions, equations of state, fluids and rocks properties and thermodynamic variables. This structured set of information must be properly handled and turned into a numerical model in order to be used as input of a numerical simulator. This implies that the investigated system volume has to be spatially discretized into finite volume elements (blocks) to which petrophysical, thermodynamic and chemical parameters, initial and boundary conditions have to be associated (Pruess et al., 1999).

The investigated domain for a full field numerical model can cover an area of hundreds of square kilometers and can have a depth of some kilometers.

Tens up to hundreds of thousands of blocks can be required in spatial discretization, and several thermodynamic variables have to be processed for each grid element and each flow interface.

Discretizing the investigated domain with a large number of small blocks greatly improves the accuracy of the simulation outcomes, but can represent a limitation in all those activities concerning the numerical model creation or modification such as history matching and sensitivity analysis, due to the huge amount of data to be managed as well as the increase of typing mistakes and computational time.

Moreover, the numerical model is a dynamic representation of the physical system, continuously growing in details as the resource is exploited. As soon as new data are available these must be included in the model. Thus, the use of proper pre-processing tools is an essential condition for an efficient and effective management of such complex numerical models.

The lack of appropriate tools to be used for applied research purposes has led us to develop TOUGH2GIS, an open source pre-processor for the TOUGH2 (Pruess et al., 1999) simulator based on the Geographical Information System (GIS) Geographic Resources Analysis Support System (GRASS GIS; Neteler et al., 2012), a free open source software released under the GNU General Public License version 2 or later. GRASS GIS is used worldwide for environmental modeling purposes.

A brief introduction to TOUGH2, an overview of the spatial discretization methods and a summary of the existing pre-processors for the TOUGH family codes are provided in the following subsections. In Section 2 TOUGH2GIS is introduced, describing the functionalities implemented for the conceptual and numerical model creation. In Section 3 an application of the pre-processor to a sample dataset is presented to show its potential, allowing the reader to evaluate the developed tools.

Developed since the early eighties at Lawrence Berkeley National Laboratory (LBNL, Berkeley, CA), TOUGH2 is a widely used numerical simulator for geothermal reservoir engineering and other applications, such as underground carbon storage, nuclear waste and environmental management (O'Sullivan et al., 2001). It is very robust and flexible due to a modular structure which allows to study systems characterized by different thermo-fluid-dynamics conditions, by choosing one of the interchangeable modules generically called Equation of State (EOS) module. In particular, it is able to study non-isothermal multicomponent and multiphase fluid flows in continuous or fractured porous media in 1D, 2D, and 3D domains. Due to the adopted Integral Finite Difference Method (IFDM; Edwards, 1972, Narasimhan and Witherspoon, 1976), TOUGH2 is able to use locally refined unstructured grids (Aurenhammer, 1991) for spatial discretization.

The code dedicated to model calibration, iTOUGH2 (Finsterle, 2007), solves the inverse problem by automatically calibrating a TOUGH2 model and allows to perform inverse simulations providing an efficient tool for sensitivity and uncertainty analysis. The forward simulator integrated in iTOUGH2 is consistent with TOUGH2.

TOUGH2 is written in Fortran and therefore can be used under different operating systems.

Its input is a set of human readable fixed format ASCII files, thoroughly described in the user's manual (Pruess et al., 1999). All numerical model information (such as grid characteristics, initial and boundary conditions, etc.) can be stored in the main input file (for the sake of simplicity thereinafter referred as *.inp) or split into separate files linked to *.inp through appropriate keywords. In particular, MESH is the file containing geometrical information about grid nodes and interfaces between blocks, GENER is the file containing information about nature, strength, and time-dependence of sinks and sources, and INCON is the file containing the complete specification of thermodynamic initial conditions.

With regard to the terminology used to define the different spatial discretization techniques, the literature is lacking uniformity. Therefore, the scheme in Fig. 1 summarizes the terms used in the present paper.

As already mentioned, IFDM allows for domain spatial discretization using structured or unstructured grids. Structured grids allow for the position of all the grid blocks/nodes and their connections to be implicitly identified (i.e., there is no need to provide coordinates and geometrical information for each element).

Structured grids can be distinguished into: i) regular, when all the blocks have the same size and shape; ii) irregular, when the spacing between the blocks varies. This means that structured grid elements are usually rectangular (Cartesian grids). A particular case of a structured regular grid is the regular hexagonal grid, which can be implicitly defined.

In an unstructured grid (such as a Voronoi grid) the positions of the grid blocks and their connections have to be explicitly defined, providing coordinates and geometrical information for each of them.

Grids can be refined in two ways: i) a global refinement is used when the entire domain is subjected to the refinement process, and can be applied to either a structured regular grid or an unstructured grid; ii) a local refinement is applied to increase grid resolution in the regions of interest, and is obtained either generating a structured irregular (telescopic) grid (Townley and Wilson, 1980) or an unstructured grid.

Telescopic refinement is relatively easy to perform and manage (see Fig. 2a), but it necessarily involves the refinement of areas of little or no interest, hence generating grids with many superfluous blocks and therefore increasing computational time.

On the other hand, an unstructured grid allows for a more efficient local refinement, with smaller blocks in the areas of interest and greater elsewhere, in many cases reducing the total amount of grid blocks (see Fig. 2b).

TOUGH2 doesn't have a native Graphical User Interface (GUI) tool to easily manage input files (that are tedious to generate and change by hand), thus several third-party pre-processors have been developed, both by software houses and by scientific research groups. Petrasim (Alcott et al., 2006), WinGridder (Pan, 2003), mView (Avis et al., 2012) and Leapfrog (Newson et al., 2012) are worth mentioning among commercial software programs. They provide efficient and user-friendly pre-processing environments but are closed source codes. This represents a substantial restriction in research activities, which typically require a customization of the software used in order to better follow the constraints of the simulation process. Among the software developed by research groups, the following are worth to be mentioned: MulGeom (O'Sullivan and Bullivant, 1995); GeoCad (Burnell et al., 2003); G*Base (Sato et al., 2003); Simple Geothermal Modelling Environment (Tanaka and Itoi, 2010); TOUGHER (Li et al., 2011); PyTOUGH (Croucher, 2011, Wellmann et al., 2012).

One of the major tasks in the pre-processing of 3D models is the automatic generation of the grid nodes, in particular for locally refined unstructured grids. But many of the above mentioned software require the grid nodes to be provided as an input, previously created by a different software.

In order to achieve full code customization capabilities, we have designed and implemented software tools dedicated for TOUGH2 geometrical input data file creation and management. In the implementation of these new tools it has been taken into account the need to both minimize operational errors and numerical model preparation time. The result is TOUGH2GIS, a suite of GUI-provided BASH scripts running in the GRASS GIS environment that allows a more effective and efficient creation and maintenance of 3D numerical models.

Regardless of their source format, the information required to create a numerical model can be used with a GIS in order to exploit its powerful tools and functionalities. In fact, in any GIS data can be created, stored and managed using database attribute tables and raster and vector maps.

It is therefore possible to easily generate and manage the conceptual model and to create the numerical model using a unique GIS environment (see Fig. 3).

In order to avoid the complexity related to 3D Voronoi grid generation and to preserve the IFDM orthonormality requirement, TOUGH2GIS generates 3D models as vertically extruded 2D Voronoi grids, that is a layered mesh with uniform layer thickness.

Moreover, the task of properties assignment to grid blocks according to the conceptual model may be a source of errors and is time consuming. With TOUGH2GIS elevation raster maps (representing surfaces between relevant rock types) are used to automatically associate each grid node with the proper material.

Section snippets

TOUGH2GIS

As previously mentioned, GRASS GIS has been adopted as the development environment in order to: i) create a homogeneous geodatabase with all the available information; ii) create a complete conceptual model of the investigated domain; iii) generate a numerical model discretized through either 3D locally refined structured or unstructured grids complying with the constraints imposed by the IFDM; iv) automatically assign rock types (i.e., rock properties) to grid nodes; and v) generate the

TOUGH2GIS application

This section deals with the application of TOUGH2GIS to create a numerical model of a geothermal reservoir, specifically created for testing purposes using data from the most up to date literature. All data files used in this section are available for download and usage at http://software.dicam.unibo.it/tough2gis.

The simulation domain (see Fig. 10) has an extent of 25 km due east and 27 km due north. The upper limit is at 2000 m.a.s.l., while the lower one is at −4850 m.a.s.l.; thus the

Conclusions

With TOUGH2GIS the pre-processing activities related to the numerical modeling carried out using TOUGH2 have been significantly improved. The time required for the creation and update of complex 3D numerical models, as well as the risk of operational errors occurring during the handling of huge amount of data, have been reduced. All pre-processing tasks, from the creation of the conceptual model to its conversion into a numerical model, have been integrated under the same homogenous GIS

Acknowledgments

The authors would like to thanks all reviewers (their comments have been vital to improve the paper) and PhD Marcella Livi for kindly providing support in language and grammar review.

Research for this paper was partly supported by funding from the National Interuniversity Consortium for Georesources Engineering, Rome, Italy (CINIGeo).

Author contributions: P.B. and V.B. concepted the research; P.B., S.B., V.B. and C.C. performed the research; C.C. coded TOUGH2GIS; V.B. coded the AMESH modified

References (28)

  • J. Carrera-Hernandez

    University of Alberta, Edmonton Canada

    (2008)
  • L.P. Chew

    Guaranteed-quality mesh generation for curved surfaces

  • A. Croucher

    PyTOUGH: a Python scripting library for automating TOUGH2 simulations

  • A.L. Edwards

    TRUMP: a Computer Program for Transient and Steady State Temperature Distributions in Multidimensional Systems

    (1972)
  • Cited by (23)

    • An integrated, mesh-independent geothermal modelling framework

      2023, Environmental Modelling and Software
    • TATOO – Python Topographic Analysis Tool Library for semi-automated setup of high-resolution integrated hydrologic models

      2022, Environmental Modelling and Software
      Citation Excerpt :

      Overall, the preprocessing of IHMs is “very arduous and time-consuming” (Gardner et al., 2018), and such tasks “can take a significant portion of the hydrological modeller's time and effort” (Gichamo et al., 2020). Berry et al. (2014), Gichamo et al. (2020) and Viviroli et al. (2007) showed that preprocessing tools reduce the time for model setup and execution considerably, while Krysanova et al. (1998) highlight the improved transferability. However, with limited resources, available models and tools are prone to be fastly outdated or slow and inefficient for new model applications (Chen et al., 2020).

    View all citing articles on Scopus
    View full text