Problem Statement

Hydrologic models are a critical tool for water resources planning and management, as well as for preparing for and responding to various natural disasters. There are a large number of hydrologic models out there (Figure 1), each with their own underlying assumptions, simplifications and implementations. In the past, modeling of large geographical domains has typically been done with a single model under the (typically implicit) assumption that this model will be a reasonable representation of hydrologic behaviour everywhere. More recently, there has been growing recognition that models have specific strengths and weakness and that the choice of model matters. One prime example is NOAA's NextGen Water Resources Modeling Framework. This framework is meant for flood forecasting throughout the entire USA, and is built around the capability to run different models in different places.

A key challenge with such approaches is to know which model to use in which place. This is referred to as "model structure uncertainty". There is increasing evidence that different models can produce (substantially!) different simulations, but understanding about which model to use where, as well as clear guidance for model selection based on modeling purpose, is still lacking. This module is intended to introduce students to these topics, to provide hands on experience with using different models in different places, and to start thinking critically about how the simplifications inherent in each model may make the model more or less appropriate for use in different locations.

Module Overview

This module uses a combination of open-source data and model code to show why model choice matters. Students take on the role of a new hire in an engineering consultancy firm. The student will first go through some on-the-job training to get familiar with commonly used data sources and tools. The student is then tasked to generate hydrologic model simulates for a basin in western Washington State. Finally, the consultant is asked to apply their models to a different basin, located near Houston in Texas. Students will contrast the performance of both models in both regions, and provide a summary of the usefulness and appropriateness of using either model in either scenario.