• Small structural asymmetries between PV areas can substantially bias PVGR yield data, highlighting the need for pre-retrofit data analysis or shading simulations in future studies. • Correction methods reduce the apparent yield gain during the vegetation period from ∼ 3 % to a net range of 0.95–1.14 %, reflecting the combined effect of increased module mounting height and vegetation cooling. • Lack of maintenance and vegetation overgrowth can temporarily negate or reverse yield gains. The integration of photovoltaics (PV) with green roofs is widely promoted as a synergistic building technology, yet reported impacts on PV electricity generation remain inconsistent. A major factor is the limited comparability bewteen PV systems in field studies, where structural differences are often confounded with vegetation effects. This study presents a long-term, large-scale assessment of PV yield changes following the retrofit of an extensive green roof, while explicitly accounting for such structural biases. Two adjacent PV systems on an operational office building in Austria were monitored, one remaining gravel-covered and the other retrofitted with an extensive green roof. Despite identical modules and layouts, pre-retrofit measurements revealed inherent performance differences between the two arrays, primarily caused by asymmetric shading. To account for this bias, different empirical normalization approaches based on pre-retrofit data and a simulation-based correction were applied. Performance in 2023 was variable, with temporary negative deviations caused by vegetation overgrowth and operational anomalies. For March–August 2024, when the green roof was well-maintained, the raw comparison indicates a yield advantage of 2.99 % for the greened system. After applying the correction methods, the net system-level increase ranges between 0.95 % and 1.14 %, depending on the method. This increase reflects both the effect of the green roof itself and the additional mounting height required for the retrofit. While the study indicates a small benefit for the specific configuration, it also underscores the importance of identifying and correcting structural biases when assessing PVGRs.
Möslinger et al. (2026) studied this question.