Abstract
A recent study published in Science reported that greater photovoltaic policy stringency (PSI) was associated with lower Shannon diversity, lower Simpson diversity, lower species richness, and slightly higher species evenness in China. We exactly reproduced their published model and then evaluated seven alternative survey-effort specifications across Richness, Shannon diversity, Simpson diversity, and Evenness. Existing reanalyses added total birdwatching duration and observer number but retained the original Duration variable, which is derived from these quantities. We found that once total observation time and observer participation were represented explicitly, PSI became statistically indistinguishable from zero for all four biodiversity measures. Retaining Duration contributed little additional information and produced coefficient patterns that were difficult to interpret ecologically. Nonlinear representations of total observation time and observer participation substantially improved model fit, whereas additional nonlinear flexibility in environmental controls and PSI provided little benefit. The semi-spline specification was consistently selected by the Bayesian Information Criterion (BIC) across all four biodiversity measures. Under this preferred specification, the reported PSI associations were no longer statistically detectable. These findings indicate that the reported biodiversity effect is highly sensitive to survey-effort specification and that nonlinear survey-effort adjustment is strongly supported by the data.