Exploring detrending techniques in detecting long-memory of ozone time series in Malaysia by simulation

Abstract

Air pollutants, and specifically ozone in the atmosphere, have received extensive attention since last few decades, mainly because of their adverse effect on people's health. Generally, the collected ozone data are often recorded as time series and long-memory behaviour in ozone levels usually exist. Long-memory or persistency is one of the statistical properties in time series which can be estimated by the Hurst coefficient, H determination. Currently, many methods to estimate H are available. Most of them, even if very effective, need prior information to be applied (in particular about the stationary nature of the series). In order to assess the long-term ozone behaviour in Malaysia, this study aimed to explore the role of detrending techniques of three existing methods used in detecting long-memory. Simulation series in the range of 0.1 ≤ H≤ 0.9 without any assumption on the stationary nature of the time series were used to detect long-memory. The quality of estimation was evaluated in terms of biases and variability. These methods are then applied to the daily mean hourly ozone concentration at 6 monitoring stations in Malaysia over 9 years. Our aim is to plan an optimal procedure to estimate the value of the Hurst coefficient and in addition to explain the degree of persistency in long-term ozone concentration in data Malaysia.