Using Paleocurrents for Basin-Scale Hydrological Analysis: an Example from the Amazon Basin

Paleocurrent analysis is a long-established technique of great value to stratigraphy and paleoenvironmental reconstruction. In the case of the Amazon - the largest river system in the world - a continental-scale drainage reversal (from a Pacific-draining system to the extant Atlantic-draining system) has been suggested, typically hypothesized during or after the Miocene, with paleocurrent observations playing a significant role in attempts to constrain the timing and nature of such a reversal (Horbe et al., 2013). Yet, while a well-developed statistical framework exists for outcrop-level interpretation of paleocurrents, questions remain with regard to extending paleocurrent observations to draw continent-scale conclusions. Using a stream model of the Amazon River system derived from 1 arcsecond SRTM data, we present results of a Monte Carlo analysis to simulate the preservation, discovery and measurement of paleocurrent indicators in the rock record. We consider variable degrees of prior knowledge of the system and variable ability to distinguish relative stream orders to assess the effect of these factors on data confidence. We show that, where the location of the most major rivers can be well-constrained, at least 17 independent observations are required for the random sample to resemble the overall drainage pattern 95% of the time if determining between two potential directions; at least 23 observations are required to determine between four potential directions. If no prior knowledge of the basin's configuration is known, 82 and 71 observations are required, respectively, to reach a 95% confidence. Disregarding smaller streams does not consistently decrease the evidence required to reach 95% confidence. We conclude that paleocurrent observations are more appropriate to smaller-scale studies, and regions where external information such as topography, radioisotope analysis and biostratigraphy can provide useful context and constraints.