With　more　record-breaking　skyscrapers　built　in　big　cities　around　the　world,　horizontal　urban　sprawl　no　longer　dominates　the　research　of　urbanization　rather　than　the　vertical　growth　of　cities.　In　such　a　context,　the　urban　heat　island　problem　cannot　be　understood　by　solely　studying　the　impact　of　the　horizontal　urban　expansion　because　the　3D　structure　of　the　urban　landscape　could　severely　alter　the　natural　heat　flux　transport　over　the　land　surface　and　thus　lead　to　bigger　heat　island　problems.　In　addition　to　our　current　knowledge　of　impact　of　2D　landscape　changes　on　urban　thermal　dynamics,　it　is　crucial　to　understand　the　effects　of　3D　landscape　pattern　on　the　thermal　environment,　in　order　to　maintain　a　sustainable　and　eco-friendly　urban　development.　This　study　investigated　the　2D/3D　landscape　pattern　metrics　and　their　association　with　the　land　surface　temperature　(LST)　changes　in　a　case　study　area　of　Shanghai　City　using　the　extreme　gradient　boosting　(XGBoost)　regression　model　and　Sharpley　Additive　exPlanations　(SHAP)　interpretation　method　based　on　datasets　of　land　cover　and　digital　surface　model　(DSM).　Major　findings　include,　1)　3D　landscape　pattern　metrics　could　better　describe　the　undulation　and　heterogeneity　of　urban　surface　and　were　essential　when　explaining　the　variation　of　LST　compared　with　conventional　2D　landscape　pattern　metrics,　2)　Low-rise　and　high-rise　buildings　tend　to　alleviate　LSTwhile　buildings　with　medium　height　heating　the　surroundings;　3)　the　cooling　effect　of　vegetation　was　significantly　strong;　4)　different　urban　functional　types　impact　the　surface　temperature　in　the　way　determined　by　their　3D　urban　landscape　pattern.　These　findings　may　help　urban　planners　and　landscape　designers　achieve　the　goal　ofminimizing　urban　heat　island　using　computer　models　of　3D　urban　structure.　(C)　2020　Elsevier　B.V.　All　rights　reserved.