Statement of Aldo Vecchia


Flood Risk for the Devils Lake Basin


Testimony submitted to the United States Senate Appropriations Subcommittee on Energy and Water Development, Byron L. Dorgan, Chairman


March 25, 2008

Memorial Building Auditorium

Devils Lake, North Dakota


Submitted by


Aldo Vecchia

U.S. Geological Survey

North Dakota Water Science Center

821 E. Interstate Avenue

Bismarck, North Dakota  58503



Opening Remarks

Thank you for the opportunity and privilege to summarize some of my recent research on flood risk analysis for Devils Lake.  This statement is based on U.S. Geological Survey Scientific Investigations Report 2008-5011, “Climate Simulation and Flood Risk Analysis for 2008-40 for Devils Lake, North Dakota,” which was authored by me and prepared in cooperation with the Federal Emergency Management Agency.  Detailed scientific justification for this statement is provided in the report and citations therein.  The report can be accessed online ( or a printed copy can be requested by email ( 




            Since 1992, Devils Lake has risen more than 25 feet, filling Stump Lake and reaching its highest level in more than 100 years (fig. 1).  Devils Lake and Stump Lake currently consist of one water body with an elevation of 1,447.1 feet, about 3 feet below the existing base flood elevation established by FEMA (1,450 feet) and about 12 feet below the outlet elevation to the Sheyenne River (1,459 feet).  Devils Lake could continue to rise, causing extensive additional flood damages in the basin and, in the event of an uncontrolled natural spill, downstream in the Red River of the North Basin.  The purpose of this testimony is to describe the cause of the recent flooding, present some findings regarding long-term climatic variability in the Devils Lake Basin, and evaluate the probability of continued lake level rises in future years.




Figure 1.  Recorded elevations of Devils Lake for 1868–2007.



Cause for Recent Flooding

            The Devils Lake Basin is currently in a wet cycle that began in about 1980.  Precipitation averaged about 4 more inches per year during 1980–2006 than during 1950–79 (fig. 2).  The increase occurred primarily in July-December when there tended to be a much higher frequency of summer and fall rainstorms during 1980-2006 than during 1950–79.  The increased precipitation resulted in a dramatic increase in inflows to Devils Lake beginning in 1993 (fig. 3).  The long lag between the onset of wetter conditions in about 1980 and the more than sevenfold increase in inflow during 1993–2006 can be attributed to the unusual hydrologic conditions of the Devils Lake Basin.  Much of the increase in precipitation during 1980–93 went toward filling soil moisture deficits, the upstream chain of lakes, and the thousands of smaller lakes and wetlands in the upper basin; thus little of the precipitation reached Devils Lake as runoff.  Following the summer flood of 1993, most of the lakes and wetlands in the upper basin were full and inflow to Devils Lake increased dramatically.  Because the Devils Lake Basin is so large (about 4,000 square miles), a small amount of precipitation runoff corresponds to a large volume of water reaching Devils Lake.  Total inflow to Devils Lake and Stump Lake during 1993–2006 was 3.7 million acre-feet, which when spread out over the entire basin averaged only about 1.4 inches of runoff per year.  Net lake evaporation (evaporation from the lake minus precipitation that fell on the lake) during 1993–2006 was only 1.4 million acre-feet.  Therefore, inflow exceeded net lake evaporation by 2.3 million acre-feet, causing Devils Lake to rise more than 25 feet and fill Stump Lake.


Figure 2.  Estimated annual precipitation for Devils Lake for 1950–2006.



Figure 3.  Estimated annual inflow for Devils Lake for 1950–2006.

Past and Probable Future Climatic Conditions in the Devils Lake Basin

            The recent wet period is not unusual from a long-term perspective.  Climate reconstructions based on tree rings and lake sediments indicate that wet periods similar to the current one occurred in the Devils Lake Basin many times during the past two thousand years.  In fact, our research indicates that climatic conditions in the Devils Lake Basin from the past 5,000 years may have consisted of two equilibrium climate states:  a dry state similar to 1950–79 and a wet state similar to 1980–2006.  Unless future rainfall patterns are altered significantly because of climate change, the occurrence of any intermediate states, or more extreme dry or wet states, is unlikely.  Transitions from wet to dry or dry to wet occur abruptly and precipitation during wet states is more variable from year to year than during dry states.

            We developed a simulation model, called a two-state climate transition model, to simulate long-term precipitation in the Devils Lake Basin and provide information for determining what precipitation might be like in coming decades.  These simulations do not take into account significant departures from historical trends due to climate change or other factors. Some of the simulated data are shown in fig. 4.  There are 10 distinct wet periods and 10 distinct dry periods during the 1,500-year simulation period.  The average duration of the wet periods is 30 years and the average duration of the dry periods is 120 years. However, the actual lengths of the individual periods are highly variable—wet periods ranged from 3 to 80 years in duration and the dry periods from 15 to 367 years in duration. 

The remaining length of the current wet period will have a profound effect on future lake levels.  To illustrate this, we used a Devils Lake statistical simulation model developed by the USGS to generate potential future realizations, or traces, of lake levels for Devils Lake, given existing conditions on October 1, 2007.  Two sets of simulation runs were generated by assuming a fixed duration for the wet period of 2 more years for the first set and 30 more years for the second set.  Each trace was based on randomly generated possible future precipitation, evaporation, and inflow data that were consistent with the assumed duration of the wet period. The recorded annual maximum lake levels for 1980–2007, along with examples of 5 future lake-level traces for 2008–40, are shown in figs. 5 and 6.  The 5th and 95th percentiles of the generated lake levels for each year, computed from 1,000 simulated traces, also are shown.   For the simulations with the wet period lasting 2 more years (fig. 5), simulated lake levels generally decline after the wet period ends.  For the simulations with the wet period lasting 30 more years (fig. 6), the simulated lake levels are highly variable and in 2040, most of the traces are between about 1,434 and 1,456 feet.  Thus, if wet conditions continue, both lake-level increases and decreases of 10 feet or more could easily occur in the coming decades. 










Figure 4.  Five-year moving average of annual precipitation for 1,500 years generated from the climate transition model with 30-year average duration of wet periods (similar to 1980–2006) and 120-year average duration of the dry periods (similar to 1950–79).




Figure 5.  Historical and generated annual maximum lake levels for Devils Lake for 1980–2040, assuming the current wet period lasts until 2010.



Figure 6.  Historical and generated annual maximum lake levels for Devils Lake for 1980–2040, assuming the current wet period lasts until 2038.


Future Flood Risk for Devils Lake

It is impossible to predict exactly how much longer the current wet conditions will last.  However, we can use the climate transition model to estimate the probability that wet conditions will continue for any given length of time.  The model indicates that it is not likely the current wet cycle will end any time soon.  For example, there is a 72-percent chance the current wet cycle will last at least another 10 years, a 37-percent chance it will last at least another 30 years, and a 14-percent chance it will last at least another 60 years. 

Because it was impossible to predict exactly how long the current wet cycle will continue, a total of 10,000 simulated traces were generated from the statistical model as described previously, but for each trace the duration for the current wet period was generated at random using the climate transition model. Probabilities of future lake-level increases for Devils Lake were computed using the set of 10,000 simulated traces (Table 1 and fig. 7).  Each column in Table 1 shows the lake level that has a certain chance of being exceeded sometime between now and a specified future year.

            As indicated by Table 1, there is a relatively high risk of further lake level increases occurring in future years. For example, there is a 1-percent chance that Devils Lake will exceed the existing spill elevation to the Sheyenne River (1,459 feet) by 2013 and a 5 percent chance Devils Lake will exceed the spill elevation by 2034.  An uncontrolled spill could have serious water quality and flooding consequences downstream in the Sheyenne River and the Red River of the North.  There is a 10 percent chance of exceeding 1,454.1 feet (7 feet above the current elevation) by 2016 and a 20 percent chance of exceeding 1,454.1 feet by 2032.  If that happens, the existing levee protecting Devils Lake could be threatened and many roads and buildings in the basin would be flooded.

            Although there is a relatively high risk of future lake level increases, the lake is by no means certain to rise more than a foot or two above the historical record level of 1,449.2 feet set in 2006.  For example, judging by the last column of Table 1, there is about a 50-percent chance the lake will not exceed 1,450 feet any time between 2008 and 2040.  








Figure 7.  Exceedance elevations for Devils Lake for 2008–40, computed by using 10,000 traces from the Devils Lake stochastic simulation model (exceedance elevations are for calm conditions and open water).


Table 1.  Cumulative flood elevations for Devils Lake for 2008–40.

[Flood elevations are for calm conditions and open water]