The middle matters? An uncertainty-aware diagnostic for two-endpoint emissions screening

Two-endpoint interpolation is a convenient screening approximation for fuel-blend emissions, but its sufficiency depends on the fidelity of the resolved interior response. This study evaluates a signed, species-resolved endpoint-adequacy diagnostic for CO and NOₓ using a reduced aviation-fuel descriptor and an oxygenated composition-contrast probe across 144 operating states and 41 fractions (5904 reconstructed evaluations). A k-nearest-neighbour inverse-distance-weighted reconstruction and a cross-validated cubic response surface give substantially different local residuals and unqualified threshold-exceedance rates. An explicit non-negative C/H/O descriptor projection has an analytically determined active-set transition at w = 0.44744, explaining the shared flat response below this composition and the onset of a numerical step near w = 0.45. The transition arises from the descriptor mapping; the subsequent local emissions minimum is not established as a kinetic effect. At 5 ms, the cubic-model cross-validation errors correspond to approximate multiplicative response-error scales of 11.7% for CO and 17.0% for NOₓ, both larger than the illustrative 5% screening tolerance. These error scales are not calibrated bounds on the endpoint residual. A conservative three-way uncertainty gate accordingly gives an indeterminate result with the available evidence. No interior value has been independently verified by matched direct PSR kinetics or experiment. The contribution is a reproducible descriptor-switch diagnosis and an uncertainty-qualified decision framework for identifying when additional interior evidence is required, not a claim of verified aviation-fuel emission nonlinearity.