Nationally Consistent, Locally Incomplete: A Bayesian Remote-Sensing Audit of Rooftop Photovoltaic Registries

Tracking the energy transition requires accurate data on distributed energy resources, yet rooftop photovoltaic (PV) systems are notoriously difficult to account for due to their decentralized nature. Official registries often suffer from unknown levels of incompleteness. This paper introduces a Bayesian framework to audit these registries using remote sensing. By treating deep learning-based image detections as a noisy measurement instrument rather than a ground-truth target, the authors apply capture-recapture and detection-probability modeling—techniques borrowed from ecology—to estimate the true installed capacity with quantified uncertainty.

The researchers applied this framework to France, comparing their corrected remote-sensing estimates against the national transmission system operator's (TSO) grid connection data. Nationally, the two datasets converge within 3.3%, confirming the validity of the Bayesian correction. However, this national agreement masks significant local discrepancies. The audit identified 25 geographical units (departments) where the registry significantly under-reports capacity, with coverage falling as low as 39% in some regions. The authors demonstrate that 45% of this missing capacity is attributable to administrative fragmentation, where data is lost as it passes from local distribution companies to the national grid operator. Additionally, the study quantifies a truncation bias in public data, where privacy rules censor municipalities with fewer than ten installations, obscuring the fine-grained territorial dynamics of solar deployment.

This study provides a scalable, independent tool for grid operators and policymakers to audit energy registries worldwide. By moving beyond the assumption that administrative records are inherently accurate, the framework allows for the identification of systemic reporting failures. The findings highlight that administrative fragmentation is a major, quantifiable source of data loss in decentralized energy systems. This methodology is particularly relevant for emerging markets where official registries may be unreliable or non-existent, offering a way to establish a baseline for renewable energy deployment.