Urban heat island is becoming a serious problem that has been analyzed by many research fields. Statistical analysis is one of those fields of research. Statistical analysis uses meteorological data to grasp the features of urban heat island. However, in measurement data from climate observations, erroneous data might occur because of various problems. Outliers are often included in the actually observed data, necessitating detection of the outliers. We propose the method and criteria for detecting outliers included in spatio-temporal data such as meteorological data. We found six outlier patterns included in spatio-temporal data, and we proposed criteria for deleting these patterns. We could delete the six outlier patterns using two criteria. This method will be able to detect and delete outliers included in spatio-temporal data.