Abstract: Objectives: Under intensifying climate change and increasingly frequent extreme heat events, urban thermal environment assessment, climate adaptation planning, and risk identification require more refined spatial classification frameworks. Local climate zone (LCZ) classification links urban spatial morphology with local climatic differences and provides an important basis for urban environmental research. However, LCZ classification in complex urban environments still faces bottlenecks such as class confusion and insufficient accuracy. The objective was to develop a refined LCZ classification method suitable for complex urban spaces and evaluate its local applicability. Methods: With Wuhan as the study area, multi-source geospatial data for 2020 were integrated to construct features related to three-dimensional building morphology, surface cover, vegetation conditions, spectral responses, and functional constraints. A two-stage LCZ classification method was formed by combining knowledge-rule-based initial classification with random forest supplementary classification. Classification accuracy, multi-scale performance, comparison with the global LCZ product, and land-surface-temperature-based external consistency were evaluated. Results: The 120 m LCZ classification result achieved an overall accuracy (OA) of 0.899, with an overall accuracy of 0.827 for built-up LCZ classes. Compared with the global LCZ product evaluated using the same validation samples, the OA and Kappa coefficient of the proposed result were higher by 0.130 and 0.186, respectively. Multiscale comparison showed that the 120 m classification scale achieved a relatively balanced performance between accuracy and spatial detail representation. The land surface temperature consistency test showed that the classification result effectively reflected thermal differences among urban morphology and surface-cover units, supporting the climatic interpretability and applicability of the LCZ classification method. Conclusions: The results can provide technical support for high-accuracy LCZ classification, refined urban thermal environment assessment, and climate adaptation planning in complex urban spaces.