A Global Analysis of Streamflow in a Changing World

Abstract:

Streamflow regime influences human societies, aquatic ecosystems, and biogeochemical cycles. Despite its importance, a unified method for describing streamflow has not emerged; several hundred metrics of flow have been proposed, and their usefulness is still debated by the community. Here we show that a single frequency decomposition can capture most of the information present in currently used flow metrics. We also show that PCA dimensionality reduction can be used to further simplify the continuous output from a frequency decomposition into a few axes that describe fundamental combinations of frequencies commonly observed throughout the world. This same technique can be applied to raw hydrographs to condense non-frequency characteristics of streamflow into a few additional axes describing magnitude and seasonality. We found little evidence of clustering, instead observing continuous variation in flow regime associated with several ecological gradients, especially stream order (i.e. watershed size) and human influence. We also applied these descriptors to quantify global alterations in streamflow during the last ~30 years. Finally, we found that a transposed PCA dimensionality reduction effectively described the autocorrelation between days in the hydrological year, demonstrating that global day-to-day alterations in magnitude can be described by a handful of variables. We conclude that these continuous descriptions of streamflow regime provide a unified, global framework for characterizing flow regime, complementing and potentially subsuming the information described by previously proposed metrics.