Auditing visible and structural vegetation signals for urban forest monitoring using street-view imagery and airborne LiDAR

Abstract

Street-view imagery captures visible greenness, while airborne LiDAR measures vegetation structure. We assessed their agreement and mismatch in Kazimierz, Krakow. We paired semantically segmented Google Street View images with classified airborne LiDAR metrics for 2,376 valid paired records at 30 m spatial support. We examined correlations, audit classes, temporal lag, return filtering, and image quality. Visible and structural vegetation were correlated (Pearson r = 0.694; Spearman rho = 0.750). The upper-image vegetation/high-LiDAR vegetation pair had r = 0.717. The audit identified 281 legible-green-structure, 27 latent-structural-greenness, 59 image-visible-green, 711 low-green-hardscape, and 1,298 mixed/moderate records. Associations remained positive across temporal lags and after excluding noise and overlap returns. Major vegetation errors occurred in 3.3%–8.3% of independently reviewed images. The two sensing approaches provide complementary evidence. Mismatches identify locations requiring site, tree-condition, or ownership checks before management action; the audit does not establish ecological condition.