Eulerian-Lagrangian-Agent Method (ELAM)

Criterion	Explanation
General Description	An ELAM first represents the environment as a computational mesh (Eulerian component), such as a grid or mesh from geographic information system (GIS) or hydrodynamic modeling. Numerical particles are then simulated (Lagrangian component) in the domain with behaviors (agent component) representing responses to variables stored in the mesh. A key element in the ELAM approach is describing how the animal's perception of its surroundings varies with context and time.
Model Domain	Model domain is flexible; concepts can be applied to aquatic and terrestrial systems. Typically involves engineered structures that may affect fish behavior.
Developer	US Army Corps of Engineers
Hardware computing requirements	Not specified
Code language	Fortran90, Matlab
Original application	Initial publication was proof of concept where model was calibrated to conditions in Lower Granite Dam forebay and validated by comparing model predictions to measured passage proportions at 19 other configurations: 2 at Ice Harbor Dam, 5 at Wanapum Dam, and 12 at Lower Granite Dam.
Public/proprietary and cost	Cost varies. Use of the model is available within the U.S. government through interagency agreement. For entities outside the U.S. government, use of the model is available through a cooperative research and development agreement.
Physically or empirically based	Physical
Mathematical methods used	Flow field hydraulics modeled with Reynolds averaged Navier-Stokes (RANS) computational fluid dynamics (CFD) models. Some applications of ELAM use Adaptive Hydraulics (ADH) 2-D model for flow field hydraulics. Exponentially weighted moving average to acclimatize perceived stimulus intensity over time. "Just-noticeable difference" of changes in water acceleration. Downstream movement modeled as biased correlated random walk.
Input data requirements	Calibration requires horizontal and vertical distributions of migrants entering the dam forebay, detailed 3-D tracks of individual acoustically-tagged migrants, and proportions of fish entering the spillway, turbines, bypass, etc.
Outputs	Detailed simulated fish trajectories; typically aggregated into proportion of simulated fish passing or avoiding a structure.
Pre-processing and post-processing tools	Tecplot used to visualize fish trajectories and evaluate results within the context of our 3-D CFD model data sets.
Representation of uncertainty	Root-mean-square error (RMSE) values are based on applying the fitted parameter values to 5,000 individually modeled fish over ten simulation replicates using different random number seeds. Passage of neutrally buoyant passive particles is based on 5,000 release locations sampled through 10 bootstrap replicates of 500 individuals each. The passive particles share all simulation attributes except the contribution of volitional fish swim orientation and speed. "Pass Randomly" values are based on 10,000 replicates drawn from a uniform distribution where passage through all open routes sums to 100%. Water flow distributions through the dams are the steady-state averages from CFD modeling of field monitoring periods. Fit to fictitious fish passage data is based on ten separate simulated annealing fits to ten corresponding sets of random values drawn from a uniform distribution with passage through all open routes summing to 100%.
Prevalence	The model is well represented in the peer-reviewed literature. Used in Sacramento River Bank Protection Project and Yolo Bypass Salmonid Habitat Restoration and Fish Passage Project.
Ease of use for public entities	Use of the model requires an agreement with USACE. However, descriptions of the model in the peer-reviewed literature allow the techniques to be implemented without an agreement (see Gao et al. 2016).
Ease of obtaining information and availability of technical support	No formal user group and no commercial help desk
Source code availability	Not available
Status of model development	ELAM is a robust model framework that has been subjected to numerous calibration and validation efforts. Future work might involve using time-varying hydraulics, using in situ Lagrangian drifters, of approximate size and mass as the target species, released into the real flow field such that they occupy the same range of depths as target species, and working on how to best to characterize flow fields from the fish's perspective for behavior simulation.
Challenges for integration	In this model characterization document, ELAM was not considered an integrated model even though it includes flow field hydraulic modeling and fish behavior modeling. ELAM outputs can be readily incorporated into larger-scale movement or life cycle models, but ELAM is arguably overkill for use in those models except for the most critical junctions. ELAM requires detailed fish tracking data and is computationally intensive.

Browser not supported