Ground-based direct imaging surveys like the Gemini Planet Imager Exoplanet Survey (GPIES) rely on Adaptive Optics (AO) systems to image and characterize exoplanets that are up to a million times fainter than their host stars. One factor that can reduce AO performance is turbulence induced by temperature differences in the instrument's immediate surroundings (e.g.: "dome seeing" or "mirror seeing"). In this analysis we use science observations, AO telemetry, and environmental data from September 2014 to February 2017 of the GPIES campaign to quantify the effects of "mirror seeing" on the performance of the GPI instrument. We show that GPI performance is optimal when the primary mirror (M1) is in equilibrium with the outside air temperature. We then examine the characteristics of mirror seeing by calculating the power spectral densities (PSD) of spatial and temporal Fourier modes. Inside the inertial range of the PSDs, we find that the spatial PSD amplitude increases when M1 is out of equilibrium and that the integrated turbulence may exhibit deviations from Kolmogorov atmospheric turbulence models and from the 1-layer frozen flow model. We conclude with an assessment of the current temperature control and ventilation strategy at Gemini South.
