Multimodal IVR framework for pre-intervention assessment of tree-abundance manipulations in historic garden heritage sites

Immersive virtual reality (IVR) enables the prototyping and comparison of design or management scenarios before irreversible on-site interventions. However, evidence on whether synchronized multimodal measures converge under subtle, ecologically plausible manipulations remains limited. We present an image-based IVR multimodal evaluation framework that integrates panoramic scene variants, semantic AOI segmentation, and synchronized eye tracking, EEG, peripheral physiology, and subjective ratings. In a historic-garden testbed, we manipulated tree abundance as three graded strategies (Conservative/Moderate/Radical) across two typologies (Grove/Courtyard). Lower tree abundance consistently shifted attention from vegetation to heritage architecture, but did not result in a monotonic increase in perceived sense of history or visiting intention, even when scenes approached archival planting baselines. The effects were typology-sensitive: groves showed more consistent shifts, whereas courtyards showed descriptive intermediate tendencies in selected attention and rating measures. Subgroup differences were also observed. The framework supports pre-intervention screening and communication where field A/B testing is impractical.