Gas flaring, the pervasive oil and gas industry process of using turbulent buoyant non-premixed flames to destroy unwanted flammable gases, is an important global source of carbonaceous soot or black carbon (BC). However, experimental data and reliable models with which to predict these emissions over a range of flare gas compositions and operating conditions relevant to upstream oil and gas production sites (which account for 90 % of global flaring) do not exist. In the absence of alternatives, most official reporting and inventory estimates are based on single-valued emission factors that are known to be insufficient and inaccurate. This work addresses this gap through parametric experiments to measure BC emissions from vertical lab-scale flares at Reynolds and Froude number conditions directly relevant to flares at upstream oil and gas production sites, burning multicomponent C1-C7 alkane, CO2, and N2 flare gas mixtures typical of flares in North Dakota, USA and Alberta, Canada. Analysis reveals that BC emission rates fall under two different regimes corresponding to the transition buoyant and transition shear regimes of turbulent buoyant non-premixed flares previously proposed by Delichatsios based on visual flame observations and distinguished by the product of Reyolds number and the square of the fire Froude number. Within the transition buoyant regime, BC emissions are simply proportional to total volumetric fuel flowrate whereas within the transition shear regime BC emission are inversely proportional to the exit strain rate. For a narrow range of methane-dominated hydrocarbon mixtures relevant to upstream oil and gas flaring, empirical models are presented for each regime that reliably predict BC emissions scaled by the mean carbon number of the fuel. Though ultimately limited to the specific conditions considered in this study, these empirical models nevertheless are a significant improvement over existing, widely used single valued emission factors.