<p>Flares in the upstream oil and gas (UOG) industry are an important and poorly quantified source of black carbon (BC) emissions and may be a dominant source of black carbon deposition in sensitive Arctic regions (Stohl et al. 2013). &#160;Accurate estimation of flare BC emissions to support informed policy decisions, accurate climate modeling, and new international reporting regulations under the Gothenburg protocol is a critical challenge.&#160; To date few studies have focussed on the primarily buoyancy-dominated turbulent non-premixed flames typical of upstream oil and gas flares, such that existing emission factor models are highly uncertain (see (McEwen and Johnson 2012)).&#160; Although recent progress has been made in measuring black carbon from flares in the field (e.g. (Conrad and Johnson 2017; Johnson et al. 2013), data have also shown that emissions of individual flares may vary by more than 4 orders of magnitude.&#160;</p><p>The objective of the current study is to develop a robust data-backed model to predict black carbon emissions from flares considering variations in flare gas composition, flow rates, and stack diameters. &#160;Laboratory measurements of black carbon (soot) for a range of turbulent non-premixed jet diffusion flames of up to 3&#160;m in length were performed at the Carleton University Flare Facility in Ottawa, Canada.&#160; Two hundred and thirty cases spanning five flare stack diameters (25.4 to 76.2 mm), exit velocities from 0.16 to 15.15 m/s, and a broad range of industrially-relevant multicomponent (C1-C7 hydrocarbons, CO<sub>2</sub>, N<sub>2</sub>) flare gas compositions were studied. &#160;Emissions were captured in a large (~3.1 m diameter) sampling hood and forwarded to gas- and particulate phase analyzers.&#160;</p><p>Black carbon concentrations were measured via a Sunset Labs thermal-optical instrument using the OCECgo software tool (Conrad and Johnson 2019) to quantify uncertainties via Monte Carlo analysis.&#160; BC yields were subsequently calculated using a mass-balance methodology (Corbin and Johnson 2014).&#160; Variability in BC yield was well-predicted by an empirical model incorporating both the aerodynamic and chemistry effects.&#160; For this range of conditions, it was observed that primary independent variables (such as exit velocity and higher heating value) act as reasonable surrogates for sooting propensity.&#160; Further experiments are underway to test the proposed model over a broader range of conditions.&#160; However, results to date represent a significant advance in our ability to predict black carbon emissions from flares.</p>