The downstream as well as the upstream oil and gas industry has for a number of years been aware of the potential for flame acceleration and overpressure generation due to obstacles in gas clouds caused by leaks of flammable substances. To a large extent the obstacles were mainly considered to be equipment, piping, structure etc. typically found in many installations. For landbased installations there may however also be a potential for flame acceleration in regions of vegetation, like trees and bushes. This is likely to have been the case for the Buncefield explosion that occurred in 2005 (Buncefield Major Incident Investigation Board, 2008), which led to the work described in the present paper. The study contains both a numerical and an experimental part and was performed in the period 2006-2008 (Bakke & Brewerton, 2008; Van Wingerden & Wilkins, 2008). The numerical analysis consisted of modelling the Buncefield tank farm and the surrounding area with FLACS. The site itself was not significantly congested and it was not expected to give rise to high over-pressures in case of a hydrocarbon leak. However, alongside the roads surrounding the site (Buncefield Lane and Cherry Tree Lane), dense vegetation in the form of trees and bushes was included in the model. This was based on a site survey (which was documented by video) performed in the summer of 2006. A large, shallow, heavier-than-air gas cloud was defined to cover part of the site and surroundings. Upon ignition a flame was established in the gas cloud. This flame accelerated through the trees along the surrounding roads, and resulted in high overpressures of several barg being generated by FLACS. This is to the authors' knowledge the first time a possible effect of vegetation on explosions has been demonstrated by 3D analyses. As a consequence of these results, and since the software had been validated against typical industrial congestion rather than dense vegetation, a set of experiments to try to demonstrate if these effects were physical was carried out as well. The test volume consisted of a plastic tunnel, 20 m long with a semicircular cross-section 3.2 m in diameter allowing for representing lanes of vegetation. The total volume of the tent was approximately 80.4 m(3). The experimental programme involved different degrees of vegetation size, vegetation density (blocking ratio) and number of vegetation lanes (over the full length of the tunnel). The experiments were performed with stoichiometric propane-air mixtures resulting in continuously accelerating flames over the full length of the tunnel for some of the scenarios investigated. The main conclusions of the study are that trees can have an influence on flame acceleration in gas-air clouds, and that advanced models such as FLACS can be used to study such influence. More research is needed, however, because even if FLACS predicts flame acceleration in dense vegetation, no evidence exists that applying the code to trees rather than rigid obstacles provides results of acceptable accuracy. (C) 2010 Elsevier Ltd. All rights reserved.