Extreme value analysis of significant wave height using data from a single location often incurs large uncertainty due to small sample size. Including wave data from nearby locations increases sample size at the risk of introducing dependency between extreme events and hence violating modelling assumptions. In this work, we consider extreme value analysis of spatial wave data from the 109-year GOMOS wave hindcast for the Gulf of Mexico, seeking to incorporate the effects of spatial dependence in a simple but effective manner. We demonstrate that, for estimation of return values at a given location, incorporation of data from a circular disk region with radius of approximately 5° (long.-lat.), centred at the location of interest, provides an appropriate basis for extreme value analysis using the STM-E approach of Wada et al. (2018).