ATLAS: Software for analysing the relationship between urban microclimate and urban morphology in a tropical city

This research applied clustering to unsupervised learning of meteorological data and evaluated the impact of urban morphology on temperature in different weather. Air Temperature Learning Algorithms (ATLAS) was developed for the cluster analysis of urban meteorological data and regression analysis of urban morphology and outdoor air temperature. The case study experiment was conducted on the university campus. Based on principal component analysis (PCA) reduced K-means clustering, this study evaluated the characteristics of reference meteorological data, and local solar irradiance and air temperature. Three weather conditions were identified for the given reference meteorological data. Revised selection criteria for clear, hot, and calm weather in tropical climate were proposed. The mean Rain(tot) shall be less than 0.01 mm, with 25th, 50th and 75th quartile to be 0 mm, respectively. The mean W-avg and its quartiles shall be less than 3 m/s. The mean, 50th and 75th quartile of S-tot, S-max, and S-avg shall be higher than 5000 Wh/m(2), 800 W/m(2) and 200 W/m(2), respectively. The mean, 50th and 75th quartile of T-avg, T-max and T-min shall be higher than 29 degrees C, 32 degrees C and 26 degrees C, respectively. The hourly ground level air temperature models were established in different weather. The results indicate the ground level air temperature in rainy and cool weather conditions are more predictable than hot and dry weather conditions. The sky view factor, exterior wall area, and green plot ratio have a greater impact on air temperature in sunny days than rainy days.