Air quality data-base implementation by using time series statistic filling

In this work we present a set of procedures to achieve an extended air quality Data Base starting from sets of raw data which can be affected by wide gaps due to malfunction of sensors and/or weaknesses in collection software. The focus is on the statistic filling of incomplete time series, after evaluating a short range of methods. In the first step we analyse statistics and substituted undoubtedly abnormal isolated values by means of specific algorithms such as Wilcoxon rank sum test, Box-Jenkins model, trend analysis, regression etc. Then we proceed with missing data filling along time series choosing among linear interpolation, nearest neighbor and spatial averaging algorithms. Standard filling protocols such as those proposed by EPA (Environmental Protection Agency) and by NWS (National Weather Service) have also been considered. In order to enhance data set handling, we also developed an "ad hoc" Data Base and a software tool. This tool aims to manage data porting among different environments (Microsoft Access™ and Mathematica™ EDA), different data forms (lists and matrixes), diverse data fillings and data exportation formats.