Uncertainty Assessment of Head Measurements Using Time Series Analysis

Head measurements are the most abundant data available for groundwater modelers. Application of automatic water level monitors has lead to a significant increase in both measurement frequency and measurement locations in the past decade. Head fluctuations are caused by variations in the stresses on the groundwater system, including variations in rainfall, evaporation, pumping, and surface water levels. For steady models, the question arises how to extract a representative steady head from a measured head series and what the confidence interval of this value is. For transient models, the question arises how accurate the measured head fluctuations can be explained by the measured stresses. We propose to answer these questions through application of time series analysis of the measured head series. Time series analysis gives the best fit of the head measurements using the measured stresses, and provides a confidence interval. Hence, a time series model quantifies the uncertainty in the head measurements, which may be used in the calibration of transient groundwater models. In addition, results of time series analysis may be used to compute a representative value for the steady head based on the entire measurement period. Analysis of the residuals of the time series model may provide further insight, including the effects of the measured stresses, whether stresses are missing in the model, or whether processes are missing (e.g., flow through the unsaturated zone).