Missing Data in Environmental Time Series - a Problem Analysis

Abstract

Missing data in environmental time series lead to some general problems in many fields of environmental research and simulation. They cause not only difficulties in process identification and parameter estimation but also misinterpretations of spatial and temporal variations of environmental indicators. Mostly, time series represent samples of data at discrete time events based on various sampling intervals. For modelling and simulation of environmental processes time series must be mapped on a regular time grid. This procedure is known as re-sampling of time series and consists on data interpolation or, in the case of disturbed signals, on data estimation. Some well-known linear and nonlinear interpolation methods exist while data estimation can be done by static and dynamic approximation procedures. Regression type functions or in the case of cycling time series Fourier approximation are used. In opposite of that, digital filtering procedures deliver consistent equidistant data estimates based on major signal frequencies. In the paper different algorithms of data interpolation and approximation are applied on irregularly sampled water quality time of rivers with different hydraulic conditions. Additionally, low pass filters are checked to find out the best filter function for each environmental indicator.