---

The World Gas Model (WGM) is a multi-period mixed complementarity model for the global natural gas market allowing for endogenous capacity expansions in the liquefied natural gas, pipeline and storage sectors. The model contains over 80 countries and regions and covers 98% of world wide gas production and - seasonally and over the years varying – consumption for three demand sectors as well as a detailed representation of border-crossing natural gas pipelines and contractual and spot trades in liquefied natural gas. The standard WGM is a deterministic model. All input parameters are certain and model agents can optimize their decisions for a specific known future. Since uncertainty is a major issue in markets – e.g. in demand, market prices and resource bases –we expand the World Gas Model to include stochasticity. We employ a scenario approach where in early periods the economic agents face different possible futures (i.e. scenarios). The optimal decisions include hedging for the different possible futures. To give an example of this scenario approach, consider the following scenario tree, in Figure 1. This scenario tree contains 31 nodes. Node 01 represents the first model year 2005. In the years 2010 through 2020 there are two scenario nodes. Each year after 2020 is represented by four different scenario nodes.

 ---

Each scenario node has its own set of input parameters that can differ in aspects such as market power assumptions, consumption levels and production capacities. In Figure 1, the scenario ending in node 28 could be a scenario where in 2010 the members of the Gas Exporting Countries Forum (GECF) would start to collaborate like a cartel and in 2025 a harsh decline in production levels of some western countries would occur. The scenario ending in node 31 would be a business-as-usual scenario where all baseline assumptions play out over the time horizon. Besides specific input parameters each scenario node also has its own set of decision variables. So, for example, the investment level in a certain pipeline in the year 2035 depends on the scenario and may take different levels between scenario nodes 16, 17, 18 and 19. All agents maximize their expected profits, having perfect information about all possible scenarios and their probability of realization. Decisions, notably investment levels, are optimal ‘on average’ among the different scenarios of which the relevant scenario node is a part. In early periods, before 2025, the optimal decisions are hedged against the outcomes of different futures. The decisions taken in 2005 have consequences for all future periods. On the contrary, a decision taken in scenario node 08 in year 2025 only has consequences for its succeeding nodes in a single scenario, nodes: 12, 16, 20, 24 and 28.
In this paper we address uncertainty of input parameters and its impact on model results. We present and discuss results of this stochastic version of the multi-period World Gas Model, with a focus on the impact on infrastructure investments, production and consumption levels, trade flows by pipeline and LNG; and market prices. We contrast stochastic results with the results from deterministic runs, in which each scenario is optimized separately. A main result of the stochastic case is that the timing of investments is affected by the hedging behavior of market players facing an uncertain future and deciding on capacity expansions.

---