Is Hydrologic Modeling Truly Advancing?
— Charles Dickens, A Tale of Two Cities
This conference, held in Las Vegas this spring, is proof that Dickens’ assessment of human affairs is timeless and that it applies even within our own unique specialties. Progress is made in hydrologic modeling because many talented people are at work, but inexplicable omissions are evident.
The First Federal conference brings together most of the individuals, agencies, and companies that build hydrologic modeling systems. Builder’s hydrologic modeling systems define the “view of the world” that a model uses, define data input requirements, and create and distribute software.
The goal of hydrologic engineering, as quoted from Elements of Applied Hydrology, Chapter 9, Don Johnstone and William P. Cross, is:
The problem is almost infinitely complex . . .
Hydrologic modeling, at its best, is an important tool to advance this dream. The information needed for hydrologic modeling has never been more readily available. Continuous improvements are being made in data collection, data management, computational speed, and in interactive/graphical access to model calculations and data.
Satellite connections from remote stations are giving us access to real-time hydrometeorological data on the Internet. Mapping of topography, soils, geology, and vegetation for watersheds in GIS systems gives us convenient access to physical information.
Given all of the new resources that could make hydrologic modeling more accurate, are hydrologic models becoming more accurate?
The inexplicable omission in hydrologic modeling is that almost no attention is being given to this question. The accuracy or reliability of current ‘commonly used models’ receives very little attention.
If you apply model X, will the result of a 100-yr flood be within 20 percent? 50 percent? Does it matter if model X was developed using data in a different hydrologic region?
Only one paper at the conference, by Phillip J. Zarriello of the USGS, examined modeling accuracy (Comparison of Nine Uncalibrated Runoff Models to Observed Flows in Two Small Urban Watersheds).
People who build and distribute the models must evaluate the suitability and accuracy of a modeling system. The end users of hydrologic models tend to accept results on faith–a faith that will be ephemeral if they apply two or more commonly used models to the same watershed.
Studies of model suitability/accuracy are quite easy to do (USGS streamflow data is widely available) when a model is applied to gaged streams. It represents the flood regimen or the hydrologic regimen of these streams. If accuracy on gaged streams is poor, accuracy on ungaged streams will be no better.
Studies that define modeling accuracy are the logical first step to improving accuracy. Several speakers at the conference, including Ralph Wurbs of Texas A&M, advocated education and training for the end users of hydrologic models.
Education needs to be defined:
- What would this education consist of?
- What is the relationship between education and modeling accuracy?
- Is there an inherent limit to the accuracy for some types of hydrologic models? If so additional education would not improve results for these models.
Does anyone remember ‘Carter’s Little Liver Pills?’ Truth-in-labeling laws have curtailed a “free-wheeling” era when almost any substance could be sold for almost any purpose.
The equivalent of truth-in-labeling for the builders of hydrologic modeling systems are statements about the suitability and accuracy of their modeling systems for their intended purposes.
The grand dream of Johnstone and Cross is not yet realized – the problem remains almost infinitely complex. Data on modeling system accuracy will tell us how close we are to the grand dream, and where we need to go from here.