Comparing Artemis AI Access Values to Drone-Acquired Access Values

Accurate estimation of solar access values (SAVs) is critical for rooftop solar design, as SAVs quantify the extent of unobstructed sky exposure and resulting shading impacts. This study evaluates the statistical equivalence of SAV estimates generated by the Artemis platform - an artificial intelligence-driven, physics-informed modeling system - against SAV measurements obtained through Exactus drone-based surveys. Using a matched dataset of 116 rooftop locations across eight residential buildings in California and Texas, with monthly, seasonal, and annual observations, we compare the two approaches under a dependence-aware statistical framework. Equivalence was assessed using the Two One-Sided Tests (TOST) procedure with predefined margins of +/-1, +/-3, and +/-5 SAV points, incorporating spatiotemporal correlation to account for repeated measurements across locations and time. Results indicate that the mean difference between Artemis and Exactus SAVs is small and that the associated confidence intervals fall entirely within even the strictest equivalence margin (+/-1 SAV). These findings demonstrate that Artemis provides statistically equivalent SAV estimates to drone-based measurements, supporting its use as a scalable, low-touch alternative for accurate rooftop solar assessment.