| Summary: | In this project, various approaches were taken to analyse precipitable water vapor (PWV) in order to find its general trend and characteristics, such as trend during different period of a day. Weather station data, especially rain event, was considered throughout the analysis. “Before rain” and “after rain” PWV data was studied in time series. The observations obtained were inconclusive on whether rain has effect on PWV. Full PWV dataset and “without rain” PWV dataset were organized into 5 time period for every data analysis method. The 5 time periods are “daily basis”, “transition from night-time to day-time”, “day-time”, “transition from day-time to night-time” and “night-time”. Pre-analysis data manipulation and missing data detection was done minimize error in PWV analyses and to avoid error in software. Scatter plots (PWV vs Day) and boxplots were plotted. In general, observations gained from both full PWV dataset and “without rain” PWV dataset showed that range, median and mean of PWV values varies daily, regardless of time period. Discontinuity and outliers were found on certain days of both major set of data, regardless of time period as well. Besides, median of most days was found to fall in the range (50, 60) mm, regardless of dataset and time period. Cyclic trend seemed to be observed from scatter plots as well. Basic statistics, such as mean and median, were calculated. Mean were noted to be quite similar to median of PWV. Normalized surface plots were used to observe the trend in full PWV dataset and “without rain” PWV dataset. Min-Max Normalization equation was used with several variations to observe the trend in PWV data. However, no evident trend was observed from the plots although different color map was used to help in visualizing the trend. Other than that, correlation of PWV with other parameters, such as relative humidity and dew point temperature, was considered to find out PWV trend in relation to the parameters. However, correlation was not further explored after misconception about correlation has been cleared. In the end, Time-Series Analysis, ARIMA model with rain intervention, was done on PWV data. ARIMA (2, 1, 3) (2, 1, 3)288 model was successfully derived to fit July to Oct 2013 PWV data and rain was found to have no significant effect on PWV from the analysis.
|
|---|