Effects of measurement frequency on water-level summary statistics

Abstract

Wetland scientists and managers recognize the need to characterize hydrology for understanding wetland ecosystems. Hydrologic data, however, are not routinely collected in wetlands, in part because of a lack of knowledge about how to effectively measure hydrologic attributes and how frequently to measure water levels. To determine how measurement interval affects interpretation of water-level data, we analyzed data from seven wetlands in Oregon and Pennsylvania. We created subsets of daily data for each wetland, with measurement intervals of 2 to 28 days, then compared those subsets to the daily data for annual water-level summary statistics, monthly mean water levels, and occurrence/duration of threshold conditions (e.g., water in the root zone). Our primary goal was to determine if sampling at low frequencies can provide representative water-level data and accurate perceptions of the occurrence of water levels above thresholds. For annual water-level distributions, small data sets from 28-day measurement intervals provided summary data (e.g., median, quartiles, range) comparable to the 1-day reference data. For measurement intervals of seven days or less, average errors in estimates of stage (minimum, 25th, 50th, and 75th percentiles) were ≤0.03 m; for a 28-day interval, average errors were <0.05 m. Errors in estimates of maximum stage were considerably larger (0.11 m and 0.21 m for 7-and 28-day intervals, respectively) but can be circumvented using crest gauges. Errors in estimates of monthly mean stage varied greatly with measurement frequency (1–4% error for 7-day intervals, 5–15% error using one measurement per month), among wetlands and from month to month. Water-level durations above threshold values were problematic; for measurement intervals of 2 days and longer, 14-day exceedances of water in the root zone were frequently missed or spurious exceedance periods were identified. Overall, results show that sampling at monthly intervals, supplemented with crest gauges, provides a representative description of annual water-level distributions for use in classifying and comparing wetlands. More frequent sampling is required to characterize water levels for shorter (e.g., monthly) time periods and to reliably identify exceedance periods for water above threshold levels. More generally, the results remind us that the frequency and duration of sampling in hydrologic studies must be designed to ensure that data will support planned analyses.