Identifying distinct thermal components of a creek

Abstract

Statistical and heat budget methods for analyzing temperature dynamics of creeks are limited by the ability to resolve thermal processes and fine-grained thermal structures, respectively. Here we describe a hybrid method that identifies distinct thermal components in a stream's heat budget using only temperature data and an algorithm that employs mutual information to "unmix" signals in the temperature data. Spatial resolution is limited only by the number of temperature-logging sensors, which can be quite high for distributed-temperature sensors. Process resolution is at the level of thermal components, defined as distinct collections of heat flux elements sharing coordinated (nonindependent) dynamics. Inference can be used to relate thermal components to meteorological forcing and structural heterogeneity in the fluvial system and to suggest novel hypotheses for further testing with targeted heat budget studies. Applying the method to a small, arid-land creek produced two novel hypotheses: (1) lateral conduction of heat from adjacent dry land (bed, terraces) appeared to cause a substantial heating of the stream, augmented by off-channel flow paths, and (2) riparian vegetation was associated with a subtraction of heat from the stream at a rate proportionate to solar insolation, exceeding the maximum decoupling effect of shade by at least 2°C at midday, and suggesting upwelling heat flux from water to tree canopy proportional to sunlight. The method appears useful for generating new hypotheses, for selecting informative sites for detailed heat budgets, for determining the dimensionality of heat budgets in natural streams, and more broadly for associating thermal components to fluvial structure and processes.