Comparison of Methods for Smoothing Environmental Data with an Application to Particulate Matter PM10

Data smoothing is often required within the environmental data analysis. A number of methods and algorithms that can be applied for data smoothing have been proposed. This paper gives an overview and compares the performance of different smoothing procedures that estimate the trend in the data, based on the surrounding noisy observations that can be applied on environmental data. The considered methods include kernel regression with both global and local bandwidth, moving average, exponential smoothing, robust repeated median regression, trend filtering and approach based on discrete Fourier and discrete wavelet transform. The methods are applied to real data obtained by measurement of PM10concentrations and compared in a simulation study.

Robust Trend Estimation for AR(1) Disturbances

We discuss the robust estimation of a linear trend if the noise follows an autoregressive process of first order. We find the ordinary repeated median to perform well except for negative correlations. In this case it can be improved by a Prais-Winsten transformation using a robust autocorrelation estimator.