We develop a novel risk assessment approach that integrates complementary, yet distinct, spatial modeling approaches currently used in wildfire risk assessment. Motivation for this work stems largely from limitations of existing stochastic wildfire simulation systems, which can generate pixel-based outputs of fire behavior as well as polygon-based outputs of simulated final fire perimeters, but due to storage and processing limitations do not retain spatially resolved information on intensity within a given fire perimeter. Our approach surmounts this limitation by merging pixel- and polygon-based modeling results to portray a fuller picture of potential wildfire impacts to highly valued resources and assets (HVRAs). The approach is premised on using fire perimeters to calculate fire-level impacts while explicitly capturing spatial variation of wildfire intensity and HVRA susceptibility within the perimeter. Relative to earlier work that generated statistical expectations of risk, this new approach can better account for the range of possible fire-level or season-level outcomes, providing far more comprehensive information on wildfire risk. To illustrate the utility of this new approach, we focus on a municipal watershed on the Pike and San Isabel National Forests in Colorado, USA. We demonstrate a variety of useful modeling outputs, including exceedance probability charts, conditional distributions of watershed area burned and watershed impacts, and transmission of risk to the watershed based on ignition location. These types of results can provide more information than is otherwise available using existing assessment frameworks, with significant implications for decision support in pre-fire planning, fuel treatment design, and wildfire incident response.