Automatic river planform identification by a logical-heuristic algorithm

Abstract

A solution is provided to automatically identify river reaches based on their planform typologies via a holistic synthesis of multi-dimensional, quali-quantitative information. Inspired by fuzzy logics, an algorithm is proposed which appraises the degree of fit of a range of archetypes based on formal criteria, applied over a discretized version of the river. The algorithm is fed by the values of a set of indicators relevant to the formal criteria. Indicators are determined from interpretation of remotely sensed data for each discrete “slice”. To ensure a spatially holistic view, the local indicators are first integrated over a variable significant length, thereby guaranteeing contextualization at the appropriate scale. For each indicator a specific Value Function (VF) translates its objective value into a partial fitness judgment; these judgments are then integrated within a multiattribute VF. The locally highest scoring VF determines the corresponding planform choice in each slice. As a by-product, the stretch (collection of consecutive slices) corresponding to that planform archetype is identified. Implementation of the algorithm for the Baker River in Chile shows promising results. The approach ensures self-consistency and repeatability of outputs. Beyond an initial subjective assessment of input criteria based on expert judgment, the method can be fully automated. As such it can support systematic, large scale river analyses and monitoring applications.