MARKAL: Frequently Asked Questions (FAQ's)

How can I get MARKAL

In order to work with MARKAL, you need a number of software elements:

  • MARKAL proper;
  • a shell for MARKAL, either MUSS (the MARKAL User's Support System, an MS-DOS program) or ANSWER (the most recent Windows compatible shell);
  • GAMS;
  • an optimiser (MINOS, CPLEX or OSL).

Information regarding the acquisition of software can be obtained here.

What are the costs

MARKAL can be provided at no cost, but an ETSAP R&D contribution is expected from commercial and government institutions. In addition, you must buy a shell, an optimiser (e.g. OSL) and GAMS, the language in which the standard MARKAL model is written.
Prices vary significantly, mainly because the price of GAMS depends on: Number of platforms and CPUs; Number of potential simultaneous users; Solvers you are interested in; and wheter you are an academic user.
The total costs (MARKAL plus user-interface plus GAMS and solvers) range from about 3,300 USD for academic users up to 15,000 USD for unaffiliated commercial users.

Where can I get a database

The MARKAL databases are owned by the institutes which have developed the databases. Please contact the MARKAL user in your country if they are willing to let you share their database, for more details, see ETSAP members. MARKAL is distributed with a pair of demo databases.

What computer-equipment is required

The model runs on a personal computer with a Pentium processor (Windows95 and Office95 or higher for ANSWER).

How much time is required to build a model

The construction of a small MARKAL model for a country can be finished in 1 month, if you are a MARKAL expert and the statistics are available. However, the bulk of the work is in the collection of the development of the process database. For example, the building of the MATTER MARKAL model took about 1 man-year of modelling, but 10 man-years of data collection. Building such a model is a time-consuming exercise that should not be underestimated.
It should be noted that MATTER is a rather complex model compared to typical MARKAL models.

Is MARKAL a truth machine

A MARKAL model produces output that is determined by the input parameters and by the model algorithm. In this sense, the model works like a pocket calculator. The model is not an oracle, but a method to produce a techno-economic scenario for the future in a logical, traceable manner. Key paradigms are an ideal market (competitive partial equilibrium) and the predictability of technological development over a period of several decades.

How long does it take to run the model

Typical MARKAL models run in 3-10 minutes depending upon size and assuming a state-of-the-art LP (Linear Programming) optimizer. MARKAL-MACRO (NLP, Non-Linear Programming) runs in 0.25-4 hours depending on model size, although advancements in computer capability tend to decrease these running times. In addition, ETSAP is currently investigating the possibility the NLP formulation into a Mixed Complementarity Programming (MCP) formulation. For a certain class of MCP problems, the use of the PATH solver for MCP leads to significantly lower solution times compared to the analogous NLP problem.
The run of a base case of the (rather complex) Western European MATTER 4.2 model (including report writing) takes about 60 minutes on a computer with a 266 MHz pentium processor and the OSL optimiser. Each scenario variant with a new emission penalty takes another 20 minutes.
As a reference, the Netherlands MARKAL NL97 model needs about 4 minutes to solve.

What does the MARKAL model look like

The model consists of a database of processes, a demand vector, and import and export data. Processes are characterised by their inputs and outputs of energy and materials, by their costs and by their emissions. Examples of process data that serve as MARKAL input can be found in the list of publications for MATTER .

How large is a MARKAL model

The MATTER model consists of approximately 50 types of energy carriers, 150 materials, 100 demand categories and a database of several hundred processes. A full list of processes is available for the MATTER 1.0 model version (see the list of publications or the PDF report format(248 kbytes). ). The most recent MATTER 4.2 MARKAL model consists of a matrix of 25,000 rows and 50,000 columns. The .DD file (the matrix file) is approximately 5 Mb in size.
It should be noted that the size of a MARKAL model can vary a lot. The MATTER model is relative large and complex compared to a typical MARKAL model. E.g. the MARKAL model for the Dutch energy system consists of a matrix of 11,000 rows and 15,000 columns. The matrix file is about 1 Mb in size.

Who uses the results

The results are mainly used by governments, international bodies such as IEA, OECD, IPCC, and by organisations that fund R&D for energy technology. The results are used for development of environmental strategies (mainly greenhouse gas emission reduction and reduction of NOx and SO2 emissions), energy policy making, industrial policy making, and for evaluation of policy instruments. Fields where the model has not yet been applied extensively, but where further applications can be visualised are in the development of sustainability strategies, industry strategy development or the assessment of waste policies.

What are the advantages of MARKAL-MACRO

The MARKAL-MACRO model features in addition to the MARKAL model:

  • Useful energy demands endogenous and responsive to price
  • Estimates of GDP impact and feedback
Additional data needed includes: calibration data obtained from a MARKAL reference run, and the capital/labour ratio.
See MARKAL-MACRO for more information.

What about other flexible demand formulations (MARKAL-ED, MARKAL-MICRO)

Besides the MACRO module, MARKAL has also formulations for other flexible demand formulations, viz. the elastic MARKAL-ED and MARKAL-MICRO:

  • LP or NLP representation where demand levels are responsive to price
  • NLP cross-sector capability
These formulations need sector and/or cross-sector demand elasticities. See MARKAL-MICRO for more information, or GERAD's site on MARKAL-ED.

What about endogenous technology learning (MARKAL-ETL)

A novel feature of the MARKAL family of models is endogenous technology learning (ETL). In 1998, the Swiss ETSAP Partner, the Paul Scherrer Institute (PSI), and the Dutch ETSAP Partner, ECN, carried out the first MARKAL experiments with endogenous technology learning. PSI experimented with a small scale model for the global electricity demand. ECN used its large scale MARKAL application for the Western European energy system. The first PSI and ECN experiments were presented at the ETSAP Workshop in October 1998 .
Recently, ECN has expanded on the concept of technology clusters (preliminary results presented at the ETSAP Workshop in May 1999.
It is the intention to include the ETL feature in the official MARKAL version (distributed by ETSAP) in the near future.
More details and reports are now available.