Abstract Despite the current evidence on the thermal benefits of vegetation and water bodies, further research is needed to investigate how cooling capacities are influenced by particular types, amounts, and spatial arrangements of green infrastructure (GI). However, there are no commonly agreed typologies that can be confidently used to compare and report the existing climatological effects of GI. Two previous studies were conducted to respond to this gap, and a conceptual GI typology matrix was developed according to functional, structural and configurational attributes. The present research presents a streamlined version and tests its applicability for the automated mapping, classification and thermal evaluation of GI using remote sensing data. A combination of parameters is introduced, including surface cover fractions and FRAGSTATS metrics estimated from very high-resolution hyperspectral imagery, LiDAR and cadastral data. The proposed framework can be applied at different spatial scales to analyse large urban areas rapidly and with high spatial accuracy. This paper also proposes a replicable workflow that can be implemented by researchers and practitioners to map existing vegetation conditions, prioritise greening interventions and assess thermal conditions with greater confidence. In this study, this workflow was successfully applied at local scale to classify green open spaces, tree canopies and water bodies in the city of Sydney. Evidence presented here demonstrates the applicability of the proposed system to evaluate and compare the intra- and inter-typology variability of land surface temperatures (LSTs), which can be potentially applied for performance assessment across other ecosystem service categories. Despite satisfactory results, the proposed typology is contingent on further developments and tests on a larger spatial extent and with a greater number of observations. Further statistical analysis is also required to determine the cooling capacity of typologies and quantify the influence of measured parameters on LSTs.