This paper reports on research conducted in the Imperata project at CRES. Land infested with Imperata cylindrica has traditionally been regarded as being of low productivity. Recent research has shown tree growing to be an option for the Imperata areas of Southeast Asia. In this context, fire is seen as a major constraint. Rubber is one tree that is successfully grown in Imperata areas by many smallholders. In this paper, the physical impacts and private economic costs of fire associated with rubber planting are examined. Also, an aspect of the social costs of fire control are examined - in particular, the costs associated with the risk of fire spreading from one farm to another. These social benefits are not usually factored into private decisions regarding fire control, and thus may result in fire control, and tree plantings, being below the desirable level. A simple conceptual model of fire in Imperata areas is presented. This `fire model' is then incorporated within an existing bioeconomic model of a smallholder rubber agroforestry system involving Imperata. In the latter model, rice is initially the understorey to rubber followed after two years, by Imperata. The geographical focus of the study is South Sumatra, Indonesia. Much of the empirical data which was used to calibrate the model is from the Palembang region.