Magnus is a high productivity oil field in the northern North Sea.First oil was produced in 1983 and the plateau of 150 mstbod ended in 1995.During the 10 years post-plateau a variety of reservoir management techniques have been employed to arrest decline rate. Now, through exploitation of a gas injection EOR opportunity the oil rate is again rising.And looking to the future a major project to increase reservoir access by adding drilling slots is being executed that will maintain significant oil production beyond the next decade. The field remains opportunity-rich, prioritisation of drilling targets is complex: EOR wells vie with infill waterflood targets and extended reach wells to the field periphery.The non-uniqueness inherent in any reservoir simulation history match means that a conventional full field model cannot sufficiently reduce uncertainty on future drilling and facilities decisions.A particular challenge emerges when the reservoir process changes and future performance may be sensitive to aspects of reservoir description that have little influence on the history match.The gas injection project presents such a challenge. A solution is found through the emergence of genetic algorithm techniques applied to reservoir simulation.A semi-automatic workflow is developed where input parameters are varied, each combination of parameters yielding a quantitative measure of history match quality.The methodology is notable for generating multiple, significantly different, reservoir descriptions, each with an acceptable history match but with potential for quite different performance during prediction.This approach better determines those parameters that control performance; it increases the exploration of uncertainties and allows development of robust mitigation strategies. We present results of genetic algorithm techniques applied to dynamic and static full field model descriptions and to the development of pseudo relative permeability curves to provide effective flow functions for all phases in the full field model.We show how the development of a reservoir simulator that is able to address diverse development options against the uncertainty of reservoir description has impacted reservoir management. Introduction Magnus was discovered in 1974 and came on production in 1983; it is the most northerly of the presently producing fields in the UKCS (Fig. 1).There are over 90 well penetrations in the field comprising exploration, appraisal and development wells and sidetracks.The current owners of the Magnus field are BP (Operator) 85%, Nippon Oil Exploration Ltd. (7.5%), ENI UK (5%) and Marubeni (2.5%) The reservoir is formed by stacked turbidite sandstones of Late Jurassic age with the Magnus Sandstone Member (MSM) overlying the Lower Kimmeridge Clay Formation (LKCF), see MacGregor et al[1].A type log is shown in Fig. 2.The MSM is a high NTG (0.8 – 0.9) series of stacked high-density turbidites.Individual lobes are typically 2–7 m thick fine-coarse sandstones with occasional interleaved shales.Units are named A-G from base to top and many units are fining-upwards cycles, or show fine tops.Two shales are ubiquitous and act as regional seals, the B and F shales.Other shales are locally present and act as vertical flow barriers. The underlying LKCF is a mud dominated low-density turbidite system with NTG nearer to 0.3; the thickest sand unit is c.5m thick.