Front Cover

Convocation 2000

News

In Brief

Notable

Research Feature

Research News and Notes

Out and About

Papers and Presentations

Student View

Classified

Obituaries

Search This Issue

Division of University Relations Homepage

E-mail us

 

Floods taken into account

(June 8, 2000, Gazette)

Dr. Leonard Lye

By Andris Petersons
SPARK student

Floods in Mozambique, China, United States and Canada. There is no place in the world protected from the risk of flooding.

The research in flood risk analysis and estimating magnitude in rivers is there to alleviate these catastrophes. Dr. Leonard Lye, Engineering, is in the process of developing a new method of flood prediction. His project is titled Generalized Linear Models, Physical Factors, and Long Range Dependence in Flood Risk Analysis.

One common method of flood risk estimation uses a regional approach. Even if there is not enough data from the river it is still possible to estimate flood risk using the concept of homogeneous regions where flooding occurs in similar way.

“It is like trying to predict human behaviour,” said Dr. Lye. “You are not going to take a bunch of Chinese people to predict the behaviour of a bunch of Europeans. Sometimes you have a river that is near the boundary. You can use the equation for the region A, but you might get a total as if you would use it for the region B.

“Select someone near the border and you don’t know whether he behaves like a Scandinavian or Russian.”

To avoid the problems of defining homogeneous regions, Dr. Lye uses a technique called generalized linear models. With this method, a model is formulated so that the effect of the gauges and the effect of the year are taken into account. The generalized linear model automatically shows the regional differences.

“During a dry spell you expect rivers to behave in a similar fashion,” Dr. Lye explained. “You don’t usually expect two rivers side by side – one to be dry and the other one to be very high.”

When there is no data for a river it is also possible to predict floods with the help of local physical factors such as the size of the river basin, vegetation, and the amount of rainfall.

“We use all kinds of data from the map. People measure everything. However, you cannot just measure things from the map and say that’s it. There are things you don’t see just by measuring things on the map.”

Researchers from other provinces and other countries like to use Newfoundland data in their work. But they just take it as numbers and they may have no knowledge of what is going on. Sometimes there may be a river affected by tidal effects or by prevailing storm direction. For them if the number fits, even for the wrong reason, they will leave it in.

Another aspect of Dr. Lye’s research is dealing with the issue of long range dependence. A short term dependence analysis does not show long range dependence. There are small changes occurring from year to year that short term dependence cannot indicate. The most prominent case would be the major flood of the Red River Valley in Winnipeg in 1997. Dr. Lye has used this data in his research.

One of his findings was the relation between the flooding and the soil moisture stored in the heavy soils of the Red River basin. Heavy soil holds water for years and can cause potential flood risk.

“Though the Red River flood does not prove that our method is more correct,” Dr. Lye said. “It only shows that high flooding tends to come together in bunches. It is very clear in the Red River flood series that there is a bunch of low floods in 1930s and 1940s. Then a bunch of very high values – 1950s and 1970s. There is no cyclical effect. It is that they just tend to bunch together. Long range dependence does not necessarily mean that a flood 100 years ago will have an influence on flooding today.”

Still many researchers do not believe in long range dependence. However, this whole area of long range dependence is actually blossoming now despite some controversy. Lately long range dependence has become somewhat important in financial data analysis. Research on long range dependence is published in many financial type journals, monetary economics papers, and statistical papers. Earlier this type of research could be found only in hydrology type papers, or papers on water resources.

“In general my kind of research is statistical. Statistics is something people are really afraid of. It is complicated. I am not a statistician myself, although all my work is in statistics. You have to collaborate with people who know the statistics.

“At the end you want a simple model that captures what you want. The problem is how to translate this highly complicated stuff and make it usable for a simple engineer. They do the designing of flood control schemes. People have a tendency not to use anything that is complicated – I cannot understand it, so it can’t be right. You have to simplify this and explain it. It is a very good engineering practice to take whatever evidence you see and use it in your design and not ignore it.”

Dr. Lye communicates with researchers from all around the world – Malaysia, United States, Germany, to exchange data and information.

“We don’t have enough records. If I have data from all around the world I’ll use it. There is only a limited amount of easily accessible data,” he added.

“We can’t predict floods, only estimate its risk. Nobody knows. Only God knows.”

Dr. Lye’s project is funded by NSERC.

SPARK, Students Promoting Awareness about Research Knowledge, is a NSERC funded program designed to encourage writing about research.