Abstract
Abstract
All data, including air temperature data, must be verified by conducting quality control using the step check method. Step check quality control is carried out by looking at the difference of a parameter in a certain period compared to the threshold value that was already determined. Therefore before carrying out step check quality control, it is necessary to determine the ceiling and floor boundaries of the difference in air temperature data every hour. The data used in this study are hourly air temperature data and hourly present weather data from weather observations at the South Tangerang Climatological Station during 2016 - 2020. In determining the threshold for air temperature step check quality control, the air temperature data is paired with weather condition data to obtain a threshold value according to rain and no rain conditions. The threshold conducted in this study is based on a check for unusual climatological values, where the limits for an unusual and impossible jump in hourly air temperature changes are determined based on a certain percentage of the data distribution. This study uses percentile analysis to determine the threshold, where 5% in the lower and upper part of the data distribution are used as the threshold. The results show various thresholds every hour. The increase in temperature dominates the changes of hourly air temperature in no-rain conditions. The highest threshold for temperature increase occurs at 00.00 – 01.00 UTC at 3.2°C and continues to decrease over time. The highest threshold for temperature decrease occurs at 09.00 UTC - 10.00 UTC at 2.2°C. In rain conditions, the increase in temperature can still occur. However, the decrease in temperature mainly occurs. The highest threshold for temperature increase during rainy conditions is 1.8°C at 01.00 - 02.00 UTC, while the highest threshold for the temperature decrease is 5.8°C at 06.00 UTC – 07.00 UTC. With these results, observers can first carry out quality control with the Step Check method before filling in the data into the system database. Thus, any suspect data either from reading errors or tool errors can be minimized and finally produce a valid dataset.