Functional time series forecasting with dynamic updating: An application to intraday particulate matter concentration

TL;DR

This paper develops and compares statistical methods for forecasting intraday particulate matter concentrations using functional time series data, with dynamic updating to improve accuracy, validated on real environmental data from Graz, Austria.

Contribution

It introduces dynamic updating techniques for functional time series forecasting of environmental pollution, enhancing prediction accuracy with sequential data incorporation.

Findings

Improved forecast accuracy with dynamic updating methods.

Effective application to real-world air quality data.

Demonstrated benefits over static forecasting approaches.

Abstract

Environmental data often take the form of a collection of curves observed sequentially over time. An example of this includes daily pollution measurement curves describing the concentration of a particulate matter in ambient air. These curves can be viewed as a time series of functions observed at equally spaced intervals over a dense grid. The nature of high-dimensional data poses challenges from a statistical aspect, due to the so-called `curse of dimensionality', but it also poses opportunities to analyze a rich source of information to better understand dynamic changes at short time intervals. Statistical methods are introduced and compared for forecasting one-day-ahead intraday concentrations of particulate matter; as new data are sequentially observed, dynamic updating methods are proposed to update point and interval forecasts to achieve better accuracy. These forecasting methods are validated through an empirical study of half-hourly concentrations of airborne particulate matter in Graz, Austria.