Quantitative Skills > Teaching Methods > Teaching Quantitative Literacy > Recurrence Interval

Recurrence Interval


Quantitative concepts: probability

Teaching recurrence intervals

by Dr. Eric M. Baer, Geology Program, Highline Community College
Jump down to: randomness | forecasting vs. recurrence | How to calculate | Teaching Examples and Resources

Essential Concepts

Generally, there are four concepts that our students struggle with when studying recurrence intervals:
Hydrologic technician for USGS. Details
  1. the concept of a recurrence interval
  2. the idea that random events that are independent of previous events
  3. the difference between recurrence intervals and forecasting
  4. how to calculate recurrence intervals when data have variable magnitudes and when they don't.

Recurrence interval

When geologic events are random or quasi-random, it is helpful to represent their frequency as an average time between past events, a "recurrence interval" also known as a "return time." For instance, there have been 7 subduction zone earthquakes in the Pacific Northwest in the past 3500 years, giving a recurrence interval of about 500 years.

How do geologists use this concept?

Recurrence intervals occur in a variety of geologic contexts including:
sign of flood
High flood level reaches street sign. Details
  • flood frequency
  • earthquake hazards
  • volcanic eruption frequency
  • severe weather and storms
  • meteor impacts
These many contexts occur throughout an introductory geoscience course and give opportunities to revisit and reinforce this concept.

What is meant by "random"?

Random events have a probability of occurring that do not depend on the past. While this may not always be true, for many geologic events this is a robust model. For more information on teaching and using probability, please look at our probability page.

Students express their misunderstanding of random events in a variety of ways:

rolling dice To reinforce the idea that geologic events are random, I often relate them to events that they recognize as random such as flipping a coin or rolling a 6 on a die. When we talk about streaks, I use the question "how does the die/coin know what it rolled the last time?" as a way to dispel misperceptions arising out of superstitions. Students can also see this by conducting an experiment flipping coins. I have each student flip a coin until they get three heads or tails in a row. Then I have the entire class flip one more time. They see that half of the class that continues their "streak" and half breaks their "streak."

Recurrence intervals vs. forecasting


Recurrence intervals refer to the past occurrence of random events. Forecasting refers to the future likelihood of random events. These are often confused because the recurrence interval (calculated from past events) is used to gauge the future probability of an event. However, the mathematics used with these two concepts are very different. The confusion between the past-determined recurrence interval and the forecasted probability is reinforced by the widespread use of "a 100-year flood" to mean a "flood with a 1% probability of occurring in any given year."

Teaching recurrence intervals

Determination of the recurrence interval is straight-forward when looking at past events. Where there is no associated magnitude or a limited magnitude (such as pumice-producing volcanic eruptions) the recurrence interval (T) is number of years in the record (N) divided by the the number of events (n). An example of an activity using these calculations is Determining Earthquake Probability and Recurrence.
T = N/n

When there is a magnitude associated with the data (such as discharge with a flood or seismic moment with an earthquake) the recurrence interval (T) is

T = (n+1)/m
where n is the number of years of the record and m is the magnitude ranking. Student activities using these calculations are Two streams, two stories... How Humans Alter Floods and Streams and Flood Frequency and Risk Assessment.
The Los Angeles River at Sepulveda Blvd had the following peak discharges between 1970 and 1979 (data from the U.S.G.S.):

Los angeles river The L.A. River at Sepulveda Blvd
Year Discharge (cfs)
1970 81,806
1971 123,006
1972 75,806
1973 112,006
1974 99,706
1975 114,006
1976 57,406
1977 95,106
1978 147,006
1979 112,006

The data can be reorganized, and a recurrence interval computed for each discharge. In this case, n is 10, because we are using 10 years of data.
The LA river showing damage after 1930 flood Damage on the L.A. River from flooding in 1930
Year Discharge rank (m) recurrence interval (n+1)/m
1976 57,406 10 1.1
1972 75,806 9 1.2
1970 81,806 8 1.4
1977 95,106 7 1.6
1974 99,706 6 1.83
1973 112,006 5 2.2
1979 112,006 4 2.8
1975 114,006 3 3.7
1971 123,006 2 5.5
1978 147,006 1 11
Note that these are not the true recurrence values for the L.A. River since only a selection of available data have been used.

Teaching forecasting

Once we start looking to the future, we are looking at forecasting which is governed by the mathematics of probability.
The probability (P) of an event with recurrence interval T is
P = 1/T

The probability PT that a given event will be equaled or exceeded at least once in the next r-years is:
PT = 1—Pr

Illustrating the difference between forecasts and intervals

A flood is a 100-year flood if the discharge has exceeded that value on average once every 100 years in the past. In this case the probability of such a flood occurring in the next year is 1/100 or 1%. Many students would assume that the chance of a 100-year flood occurring in the next 100 years is 100%, but that is not true.
It may be counter-intuitive to students that that a 100-year flood has a less than two thirds chance of occurring in 100 years. I explain that the chance of a 100-year flood not happening in the next 100 years is 99%100, or 36.6%. If the probability of an event NOT happening is 36.6%, the probability of it happening is 63.4%. I also explain that that there is a chance that 2 or even 3 100-year events will occur within a given 100 year period. As a result, the average recurrence will drop from these more closely spaced events. Indeed, there is only about a 35% chance that a single 100-year flood will occur in 100 years, an 18% chance that 2 100-year floods will occur, and even a nearly 2% chance that we will get four 100-year events within a given 100 year time period.

Teaching Examples

Resources


« Previous Page      Next Page »