Nonrecursive Digital Filters to Process 30- and 10-minute-sampled Groundwater Observation Time Series with Tidal Components

Abstract

When observations of groundwater levels or groundwater qualities are made in coastal hydraulically highly permeable strata using stationary auto-logging devices, the collected observation time-series data often contain semidiurnal and diurnal oscillations caused by predominant tidal components as well as longer-period components due to atmospheric-pressure changes and other agents. Because the periods or frequencies of the predominant tides are accurately known, the tidal and longer-period components in observation time-series data can be separated from each other using a digital filter that screens fluctuation components by frequency. This paper presents and evaluates low-pass digital filters for 30- and 10-minute-sampled time series that reduce or eliminate the predominant tidal components with diurnal and shorter periods while preserving longer-period components. The presented filters are the simplest running-mean filters, cosine filters using windows, and, for 30-minute-sampled time series, a newly produced tide-killer filter designed to completely eliminate the specific predominant tidal components. These nonrecursive filters are easy to apply to time-series data using built-in functions of spreadsheet software. They would be useful for processing groundwater observation time-series data collected from coastal areas.