Velocity estimation using a Bayesian network in a critical-habitat reach of the Kootenai River, Idaho

Abstract

[1] Numerous numerical modeling studies have been completed in support of an extensive recovery program for the endangered white sturgeon (Acipenser transmontanus) on the Kootenai River near Bonner's Ferry, ID. A technical hurdle in the interpretation of these model results is the transfer of information from the specialist to nonspecialist such that practical decisions utilizing the numerical simulations can be made. To address this, we designed and trained a Bayesian network to provide probabilistic prediction of depthaveraged velocity. Prediction of this critical parameter governing suitable spawning habitat was obtained by exploiting the dynamic relationships between variables derived from model simulations with associated parameter uncertainties. Postdesign assessment indicates that the most influential environmental variables in order of importance are river discharge, depth, and width, and water surface slope. We demonstrate that the probabilistic network not only reproduces the training data with accuracy similar to the accuracy of a numerical model (root-mean-squared error of 0.10 m/s), but that it makes reliable predictions on the same river at times and locations other than where the network was trained (root mean squared error of 0.09 m/s). Additionally, the network showed similar skill (root mean square error of 0.04 m/s) when predicting velocity on the Apalachicola River, FL, a river of similar shape and size to the Kootenai River where a related sturgeon population is also threatened.

Coastal foredune evolution, Part 2 : modeling approaches for meso-scale morphologic evolution

“…Palmsten et al (2014) adapted the model of Plant and Stockdon (2012) for use on the Gold Coast of Queensland, AU, to hindcast dune retreat distance. Palmsten et al (2013) demonstrated that a Bayesian network developed at one geographic location can be applied at a second location, given similar morphology and similar hydrodynamic forcing.…”

Application of Bayesian Networks to hindcast barrier island morphodynamics

“…Additionally, quantitative and qualitative data, such as presence or absence of dune erosion or relative amount of infrastructure damage following a storm, can be incorporated to aid in decision making. Finally, Palmsten et al (2013) demonstrated that BNs are transferable to new settings, if there is adequate similarity in the observational data between study sites. Given the data limitations that many coastal managers face, development of generalized BNs that use extensive datasets from locations such as Fire Island, NY, Duck, NC, and Assateague, MD (Field Research Facility Coastal Observations and Analysis Branch, 2004;U.S.…”

Bayesian hindcast of acoustic transmission loss in the western Pacific Ocean

“…Examples of machine learning that can be trained to reproduce more complicated model results include clustering, artificial neural networks, extra tree regressors, or Bayesian networks. Examples of machine learning techniques applied to emulate numerical models in environmental applications and decision making include modeling wave transformation nearshore [ Plant and Holland , ], depth averaged riverine velocity [ Palmsten et al ., ], groundwater flow [ Fienen et al ., ], and atmospheric TL [ McCarron et al ., ]. Similar to the present case of acoustic propagation, real time computation of models in these examples can be prohibitive and initial and boundary conditions, as well as model coefficients, can be uncertain.…”

“…Similar to the present case of acoustic propagation, real time computation of models in these examples can be prohibitive and initial and boundary conditions, as well as model coefficients, can be uncertain. The advantage of the Bayesian approach is that once the simplified model is developed and trained over a variety of dynamic and uncertain situations, computation time is insignificant relative to execution of the full model, and produces results with skill similar to the original model [ Palmsten et al ., ]. The time savings is crucially important in real world applications, where there is a need to support rapid decisions based on dynamic environmental forecasts including uncertainty.…”