Enhancing the accuracy of climate zoning for buildings through precise climate variables selection and novel misclassification index

Climate zoning for buildings (CZB) is crucial for ensuring that architectural design principles are appropriately tailored to the specific environmental conditions of each region. Inadequate consideration of climate factors leads to the suboptimal application of architectural design principles, compromising the effectiveness of energy-efficient building strategies. In response, this work proposed integrating multivariate clustering of climate variables with spatial analysis and novel climate zoning misclassification index (CZMI) to produce a more accurate CZB development framework. For the first time in CZB research, machine-learning techniques were employed to identify the key climate variables. To explore the impact of spatial proximity on CZB results, k-means, and hierarchical clustering methods were utilized for classification in both spatially and non-spatially constrained variants. To assess the quality of the resulting CZB classifications, and to perform a comparison between the proposed and official climate maps, novel CZMI was used as a primary and clear indicator, measuring the degree of building energy needs overlap across climate zones. The proposed method demonstrates markedly improved separation and minimal overlaps between climate zones. Conversely, the comparison showed that while the official CZB map of Kazakhstan and the ASHRAE map both exhibit limitations in accurately distinguishing climate zones, the ASHRAE map performs better but still not ideally. The proposed framework not only enhances local CZB accuracy but also has the potential to be adopted globally. The presented framework through accurate climate maps, might serve as a pivotal tool to drive progress in building, construction, energy, and associated industries. •First machine-learning key variable search in climate zoning for buildings.•Integrated multivariate clustering and spatial analysis for precise climate zoning.•Introduced Novel Misclassification Index to assess accuracy in zone separation.•Hierarchical Clustering is preferable over K-means in climate zoning for buildings.•Increasing the number of climate variables does not enhance classification efficacy.