Which tributaries disrupt downstream fining along gravel-bed rivers?

Abstract

Tributaries and other lateral sediment sources can have a significant impact on river bed sediment texture and, in turn, on channel form. Sufficiently voluminous or distinct sediment inputs redefine the mainstem grain-size distribution, punctuating downstream maturation and isolating a sequence of discrete sedimentary links. Within these links fining processes usually dominate, such that models of sorting and abrasion, when applied to individual links, provide reasonable predictions of grain-size change. Links represent the fundamental natural unit within which fining models can be tested, developed and applied. Identification of significant lateral sources is therefore important, yet, beyond vague references to relative tributary size, sediment load, and sediment calibre, no criteria exist for the a priori discrimination of such sources. In this paper a procedure for identifying significant lateral (tributary) sources, without the benefit of grain-size information, is outlined. A high-resolution characterisation of bed material texture along two Canadian gravel-bed rivers facilitated classification of all their perennial tributaries as either significant or insignificant. Three absolute tributary basin parameters and their relative counterparts, chosen to reflect the likely controls on tributary significance, are then used to develop a discriminant function which isolates a large proportion of significant tributaries while minimising incorrect classifications. Examination of consistently misclassified (anomalous) tributaries reveals the importance of lateral source spacing and of inconsistencies in the geomorphic history of the contributing basins. In turn, a general tributary categorisation procedure is suggested which includes a logistic regression model for attaching probability statements to individual classifications. The generality of the discriminant and logistic functions cannot be assessed because of the lack of other suitable data sets.