Anthropogenic climate change is expected to cause an increase in fire danger over south-eastern Australia during the 21st century, primarily driven by increased surface temperature. Studies of future fire weather in Victoria, Australia, have so far mostly utilised direct output from general circulation models, which have inadequate resolution for resolving the dynamics of local fire danger and are prone to substantial biases that may affect the seasonality of dry fuels. In this paper, we assess the ability of the Multivariate Adaptive Constructed Analogs (MACA) method to downscale output from general circulation models over Victoria, and replicate statistical attributes of fire danger indices. We find that climatological descriptors of meteorological variables of wind, temperature and humidity are captured extremely well, and fields on extreme fire days are well captured. We find that the method works very well for statistically downscaling fire weather elements over Victoria and provides a vehicle to assess the regional variation of fire weather projections over Victoria.