We’ve already discussed what the term ‘100 year flood’ actually means. So now the question has become, how do we figure out what the 100 year flood is for any particular location? It’s an important thing to know for any number of reasons ranging from determining if that dream house you want to buy is too close to the creek to the farmer deciding how much of his land he can plant on to the millionaire land developer deciding if he can build a Walmart on his property. This is the ultimate goal of most hydrology studies and it’s often referred to as flood frequency analysis.

You mostly hear ‘100 year flood’ in the media, but lots of different floods are important. In my day job we use everything from the 2 year to the 500 year flood. You can refer back to my previous post to review what that number means, but it’s referred to as the recurrence interval. The recurrence interval for any particular design varies depending on the importance of the project and the amount of risk you’re willing to put up with for that project. In most design situations the engineer will consider multiple recurrence intervals. For purposes of this discussion I’m mostly going to use the 100 year recurrence interval.

So, how do we decide how much flow the creek has to have before it’s a 100 year flood? Or how deep the ditch out back will get during a 100 year flood? In an ideal world you would do a detailed hydrology study with lots of modeling and calibrating. If you have a good hydrology model all you have to do is input a rainfall amount and it will spit out a flow. (There’s a whole different branch of civil engineering that turns that flow into a depth in the creek.) Some of the most advanced models can even pull the rainfall data out of those Doppler radar images the tv weather folks like to show. The problem… hydrology modeling requires a lot of detail and is time consuming and expensive. Hydrology models are extremely sensitive to minute changes and aren’t always reliable because they can be thrown off by something as simple as two rainstorms in two days. So hydrology modeling is generally only done in larger watershed wide projects.

The next best solution involves someone getting their feet wet, because if you’re really lucky there might be a stream gage near your project. A stream gage is a location where specialized equipment has been set up to keep track of the water depth in a creek, river, etc… If you’ve collected data from your stream gage for a long enough time period you can use the magic of statistics to define what constitutes a 2 year storm, 100 year storm, and so on. The log-Pearson Type III methodology is generally accepted as the best.

Sadly, there are very few stream gages in Tennessee. That number grows significantly if you include old stations that are no longer in operation, but it shrinks again if you start excluded stations that don’t have enough data to try and figure out the larger recurrence interval storms. Just how many years of data you need in order to realistically determine a 100 year storm is a matter of some scientific debate, but it’s not an insignificant time period. The most recent study to tackle this included gage stations with at least 10 years of data collected which amounted to 453 stations. You can draw your own conclusions on the validity of using 10 years of data to determine a 100 or 500 year storm but I suspect the 10 year limit was a compromise between rigorous science and real world needs.

If you are unlucky enough to have a project that isn’t near a gage station then you have to employ a little more ~~voodoo~~ statistical analysis. I’m not going to try and give a detailed description of how this works, mostly because I don’t know much about it myself. In broad terms, The gage data is analyzed to develop a relationship between the measured flow rate and certain site specific variables. Once a statistically relevant relationship is determined, you have an algebraic equation that you can apply that to other, non-gaged, locations. Due to differences in climate, geology, etc… each equation is only valid for a particular area. Equations developed in the Nevada desert can’t be used in rainy Seattle, and the same can be said for equations developed in flat, sandy Memphis versus the mountains around Johnson City, TN.

It’s like deciding that if your dog ate three cups of food a day when she weighed 20 pounds and six cups a day when she grew to 40 pounds, then your new dog will eat the same amount when he weighs that much. There’s some variability from one dog to the next, but if the new dog is the same breed as the old then you’re generally going to be right. But if your new dog is a different breed, then you have to start from scratch. The vet already knows how much each type of dog eats because she has all the historic data, just like the US Geological Survey knows how much flow a 100 year storm will generate for just about anywhere in the country.

If you’re interested in too much detail, read this USGS study.

## Leave a Reply