Experimental investigations were conducted to characterise the impacts of crosswind and burner aspect ratio on the flame evolution characteristics and flame base drag length of gas diffusion flames on rectangular burners. The burners have the same surface area of approximately 100 cm(2). The tests to capture the flame base drag length were conducted three times for each condition with the differences between the original and repeated tests being less than 6%. The thermocouple readings were corrected for the effect of radiative and convective heat exchange with the surroundings. Overall, 84 independent test conditions were conducted on 4 different burner aspect ratios, 3 fuel supply rates and 7 crosswind conditions. The changing behaviour of the flame with different burner aspect ratios, heat release rates and crosswind speeds were carefully analysed. The appearance of "blue flames" in the upstream edge of the main diffusion flames just above the burner in relatively strong winds was analysed. Unlike the flame tilt angle and flame height which either increase (the former) or decrease (the later) monotonically with the increase of wind speed, the flame base drag length was found to increase with the wind speed firstly until a critical point and then decrease with further increase of the crosswind for a given heat release rate. This is thought to be due to the competing influence of thermal buoyancy and wind induced inertial forces. The transition point for the maximum flame base drag length with regard to crosswind was found to decrease with the increasing aspect ratio of the burner for a given heat release rate. A new physics-based correlation considering decay phase with the crosswinds was proposed for the flame base drag length incorporating all important physical factors including inertia force, fire induced thermal buoyancy, Froude number, dimensionless heat release rate and fuel/air density ratio. The proposed formulations were found to correlate well with the current measurements of gas burner fires as well as some published data in the literature for pool fires on the ground which were not used in their derivation. (C) 2018 The Combustion Institute. Published by Elsevier Inc. All rights reserved.