Skip to content

Science questions

Manabendra Saharia edited this page Jun 1, 2018 · 1 revision

Problem Statement

Flood frequency estimates maybe more uncertain than appreciated.

Target

Drive a stochastic (ensemble) event to reveal true FF uncertainty and understanding sensitivities across case study basins in the Western US

Three modeling components to be explored

  1. Hydrologic Model Structure - Expanding from one model to multi model ensemble
  2. Model Parameters - Vary model parameters
  3. Initial Conditions - Develop many model Initial Conditions that are internally consistent for each model structure

Current Bureau status and problems

  • Stochastic simulations of historical events with randomly perturbed initial conditions (ICs) and thousands of precipitation input scenarios from defined precipitation frequency curves using one model and model parameter set.
  • Current FF estimates may be overconfident or, alternatively, require more calibration and parameter development than is necessary for some studies.

What will this project use?

Model

  • Framework for Understanding Structural Errors (FUSE) that allows users to systematically change hydrologic model components (structures) to create a rigorous multi-model ensemble.
  • FUSE uses a common numerical solver across all model structures, so that changes in hydrologic model simulations are due to only the changes in model structure.
  • FUSE will mimic operational models such as HEC-HMS

Dataset

  • Newman (2015) hydromet dataset with 100 realizations

Calibration techniques

  • Multi-Objective COMplex evolution, MOCOM method etc.

Sensitivity analysis framework

  • DELSA etc.

Project Trajectory

Year 1: End-to-end system in the prototype basin

  1. Multi-model ensemble and MOCOM calibrations for the pilot basin.
  2. ICs generated quickly using the Newman dataset to drive each model and parameter set in a 30+ year continuous simulation.
  3. Event based simulations in the historical simulation period
  4. Multiple calibrated model parameters for each model ensemble member via MOCOM
  5. DELSA

Year 2 and 3

  • Apply the end-to-end FF and DELSA systems to the remaining basins