Contaminant fate and transport models used for hydrogeological risk assessment commonly include an assumed constant source concentration to predict downgradient concentrations in groundwater. This assumption is unrealistic in many cases where nonaqueous phase liquids (NAPLs) are the contaminant source. In this study the effect of a constant and transient (declining) source term on contaminant plume chemistry and groundwater impacts predicted by a hydrogeological risk assessment model is investigated for the release of dissolved phase constituents from an unleaded petroleum fuel light nonaqueous phase liquid (LNAPL). Two transient source models were developed: a mass balance model and an exponential decay model. Both use Raoult's Law to describe the aqueous phase partitioning and depletion of organic compounds from a multicomponent LNAPL source to groundwater. The models were used to estimate the changing LNAPL composition and effective aqueous solubility of six common constituents (benzene, toluene, ethylbenzene, xylene, methyl-tert-butyl-ether, and tert-amyl-methyl-ether) in a representative unleaded petroleum fuel. The predicted source concentrations were then compared with groundwater quality data from a contaminated site. The source concentrations estimated from these models were propagated into a hydrogeological risk assessment model using the Remedial Targets Worksheet v3.1 developed by the Environment Agency of England and Wales. Different groundwater impacts and associated estimates of risk were predicted between the constant and transient source simulations. These are illustrated using simple calculations and established risk assessment methods. The constant-source model predicted higher contaminant concentrations at the compliance point, which also persisted over a much longer duration. Consequently, there was a greater requirement for remedial action estimated using the hydrogeological model in scenarios that assumed a constant-source. This could result in unnecessary remedial action in cases where no unacceptable risk is predicted when a more representative description of a declining source is used.