Premium
Multicriteria Optimization Model to Generate on‐DEM Optimal Channel Networks
Author(s) -
Bizzi Simone,
Cominola Andrea,
Mason Emanuele,
Castelletti Andrea,
Paik Kyungrock
Publication year - 2018
Publication title -
water resources research
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.863
H-Index - 217
eISSN - 1944-7973
pISSN - 0043-1397
DOI - 10.1029/2018wr022977
Subject(s) - mathematical optimization , channel (broadcasting) , minification , energy (signal processing) , scaling , energy minimization , computer science , elevation (ballistics) , digital elevation model , upstream (networking) , a priori and a posteriori , mathematics , geology , geometry , statistics , telecommunications , physics , philosophy , remote sensing , epistemology , quantum mechanics
The theory of optimal channel networks (OCNs) explains the existence of self‐similarities in river networks by multiple optimality principles, namely, (i) the minimum energy expenditure in any link, (ii) the equal energy expenditure per unit area of channel anywhere, and (iii) the minimum total energy expenditure (TEE). These principles have been used to generate OCNs from 2‐D networks. The existing notion of OCN considers the concavity of river longitudinal profiles as a priori condition. Attempts to generate OCNs starting from a random 3‐D digital elevation model (DEM) and minimizing solely TEE have failed to reproduce concave profiles. Yet alternative approaches can be devised from the three optimality principles, for instance, focusing on the local energy expenditure (LEE). In this paper, we propose a Multiobjective modeling framework for Riverscape Exploration (MoRE) via simultaneous optimization of multiple OCN criteria. MoRE adopts a multiobjective evolutionary algorithm and radial basis functions to efficiently guide DEM elevation variations required to shape 3‐D OCNs. By minimizing both TEE and the variance in LEE, MoRE successfully reproduces realistic on‐DEM, OCN‐based riverscapes, for the first time. Simulated networks possess scaling laws of upstream area and length and river longitudinal profile resembling those of real river networks. The profile concavity of generated on‐DEM OCNs emerges as a consequence of the minimization of TEE constrained to the equalization of LEE. Minimizing TEE under this condition generates networks that possess specific patterns of LEE, where the scaling of slope with basin area resembles the patterns observed in real river networks.