Premium
Selecting Between One‐Dimensional and Two‐Dimensional Hydrodynamic Models for Ecohydraulic Analysis
Author(s) -
Gibson S. A.,
Pasternack G. B.
Publication year - 2016
Publication title -
river research and applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.679
H-Index - 94
eISSN - 1535-1467
pISSN - 1535-1459
DOI - 10.1002/rra.2972
Subject(s) - residual , discretization , geology , divergence (linguistics) , hydrology (agriculture) , base flow , cobble , flow velocity , flow (mathematics) , soil science , mathematics , environmental science , habitat , drainage basin , geometry , geotechnical engineering , ecology , algorithm , geography , mathematical analysis , cartography , philosophy , linguistics , biology
Abstract Aquatic habitat assessment and river restoration design require geospatially explicit maps of hydraulic conditions. Diverse mechanistic ecohydraulic models compute spatially explicit depth and velocity results to evaluate habitat suitability spatially as a function of these abiotic conditions. This study compared depth and velocity results from two‐dimensional (2D) and one‐dimensional (1D) hydraulic models with algorithms that laterally discretize 1D velocity and interpolate depth and velocity spatially based on the Laplacian heat mapping approach. These ‘conveyance distributed’ methods constitute ‘best 1D modelling practice’ and were compared with 2D results for the first time. The 1D and 2D models were applied to three morphologically distinct reaches (leveed, meandering, and anastomosing) for three flows (base, bankfull, and flood flows) of the partially regulated, gravel/cobble lower Yuba River in north–central California. The test metrics were the coefficient of determination ( R 2 ) and the median absolute residual ( ε ˜ ). These metrics quantified the incremental uncertainty 1D approximation incurs, results which make explicit cost–benefit processes of model selection possible. Finally, velocity residual maps were analysed to identify regions and processes where residuals were high, indicating divergence from the 1D assumptions. Paired data (1D–2D) fell between 0.94 ≥ R 2 ≥ 1.00 ( R 2 mean = 0.98 and R 2 median = 0.99) for depth and median absolute residuals were all 3.8 ≤ ε ˜ ≤ 7.2% (i.e. 50% of residuals are approximately within ±1.7 to 3.6%). Higher flows and lower gradient reaches had lower residuals and higher R 2 . Velocity diverged more, particularly for base flow in anastomosing reaches (0.42 < R 2 < 0.58). One‐dimensional, conveyance distributed, assumptions performed better for other channel types, where 0.69 < R 2 < 0.81 ( R 2 mean = 0.75 and R 2 median = 0.77), with median absolute residuals between 9.6% > ε ˜ > 22.4% (i.e. ~ ± 4.6 to ±11.2%), whereε ˜mean = 14.2% andε ˜median = 13% (~ ±7.1 and 6.5%). The conveyance distributed 1D velocity model performed best, where the orthogonal flow assumptions obtained and where side channels did not transition from backwater to conveying area between flows. Copyright © 2015 John Wiley & Sons, Ltd.