Cluster Analysis of Surface Water and Groundwater Levels

Conventional methods of identifying clusters of hydrologic time series are based mostly on using a Euclidean distance measure. Such approaches agglomerate time series with similar means, and the temporal fluctuations in the clustered series are not necessarily highly correlated. The approach described in this paper clusters time series that meet prespecified Pearson correlation and p-value criteria. The approach is demonstrated by the combined analysis of 397 time series of both groundwater levels and surface-water levels using a threshold correlation of 0.90 and maximum p-value of 0.05. Clusters with unique characteristics in both the time and frequency domains were extracted. Four clusters included both surface water and groundwater stations with close synchronous connections over distances on the order of 5–10 km, three clusters included groundwater stations that were synchronously influenced by daily wellhead operations, one cluster included widely separated stations with common diurnal tidal influences resulting from direct connection to a tidal bay, and one cluster included stations that had direct hydraulic connections. These results will aid in the development of management strategies as well as conjunctive surface water and groundwater models.