The Changing Precipitation Patterns at Devils Lake, ND

Transcription

1 The Changing Precipitation Patterns at Devils Lake, ND 1. Introduction The landlocked Devil's Lake in North Dakota is the largest natural water body in the state, having expanded from the size of a large pond in 1940 to an area of over 250 sq. miles in January The decades-long expansion of the lake accelerated in the early 1990s and has brought widespread flooding to the surrounding lands and communities. A continued rise of the lake level will cause ongoing hardship and economic damage and may eventually lead to an uncontrolled outflow into the Sheyenne River, with accompanying degradation of downstream water quality (Vecchia 2008). On a timescale of several years to decades, changes in the elevation of Devils Lake are controlled by variations in the quantity of precipitation that falls over the 3810 sq. mile drainage basin. Figure 1 shows that the lowest elevation of the lake occurred in 1940 after more than two decades of unusually dry conditions and that a sudden increase in precipitation beginning in the 1990s was accompanied by a dramatic rise in the lake level. Figure 1. Elevation of Devils Lake (black line) and annual (light blue line) and trailing decadal (dark blue line) basin-average precipitation. 1

2 The striking difference between the precipitation regimes of the early 20 th century and of recent years is shown in Table 1: Lake Level Number of Years Year Range 10-Year Annual Average Precipitation Below 1410 feet inches Above 1440 feet inches Table 1. Range of decadal average precipitation occurring in years for which the Devils Lake elevation was below 1410 feet or above 1440 feet. All of the years in which the lake level was below 1410 feet had preceding decadal average precipitation of less than 17.8 inches per year, but all of the past 13 years in which the level has exceeded 1440 feet have had preceding decadal averages greater than 18.7 inches per year. The two primary goals of this study are: first, to examine the characteristics of the precipitation changes that have profoundly affected the Devils Lake region; and second, to examine atmospheric processes and mechanisms that may explain the changing precipitation regimes at Devils Lake. An understanding of the changes that have already occurred may aid in assessing the likelihood of future scenarios for Devils Lake. 2. Analysis of Precipitation Changes Several high-quality gridded historical precipitation datasets make it possible to examine in detail the changes that have taken place in the characteristics of precipitation over and near Devils Lake. Temporal Distribution of Precipitation Changes Annual precipitation totals for the Devils Lake drainage basin are shown in Figure 2. For this analysis, basin-average precipitation totals were extracted from the high-resolution Parameter-elevation Regressions on Independent Slopes Model (PRISM) precipitation dataset (Daly et al. 2002). The recent wet regime appears to have begun in 1991, when the annual precipitation was 22.5 inches, greater than in all but two years in the previous 96 years of record. Since 1991, this amount of precipitation has been exceeded in four additional years. Figure 3 shows that the annual totals since 1991 have been concentrated in the upper half of the 115-year distribution; the difference in median annual totals before and after 1991 is more than 3 inches. The regime shift in 1991 is very distinct and statistically significant, as shown by a likelihood ratio test for gamma distributions fitted separately to the annual precipitation amounts between 1948 and 1990 and between 1991 and The year 1948 was chosen for the beginning of the first period both because the period was a time of relative stability in the decadally-averaged precipitation (Figure 1) and because high-quality gridded atmospheric data are first available from 1948 (see Section 4). The likelihood ratio test examines the probability that annual precipitation values in the two periods were drawn from the same gamma distribution. The outcome of this test shows that the 2

3 probability that the two distributions are drawn from different populations is approximately 0.995, and thus the change is highly statistically significant. The recent increase in precipitation at Devils Lake has not occurred evenly throughout the year but has been concentrated in certain months. Figure 4 shows the median precipitation by month for years between 1948 and 1990, and for years between 1991 and Significantly increased precipitation has been observed in May, June, and July, and also in October. Little change in precipitation has occurred in the winter months (November-April) since 1991 (Figure 2). Figure 2. Annual basin-average precipitation totals at Devils Lake, from the PRISM dataset. The blue columns show the annual totals summed over November through April, and the green columns show the May-October totals. 3

4 Figure 3. Histogram of annual basin-average precipitation totals at Devils Lake, from the PRISM dataset. The blue columns show the annual totals observered between 1895 and 1990, and the red columns show the totals. Figure 4. Median basin-average precipitation by month, from the PRISM dataset. The blue columns show the median values from the period, and the red columns show the median values from

5 Spatial Distribution of Precipitation Changes The spatial extent of the increase in precipitation that has affected Devils Lake is revealed by maps of the difference in average precipitation between years before and after 1991 (Figure 5 and Figure 6). The Global Precipitation Climatology Center (GPCC) dataset (Beck et al. 2005) was used in order to have data for southern Canada as well as the United States. Figure 5 (top left) shows that average precipitation has increased across almost the entire Great Plains region, from Texas to central Canada. Larger precipitation amounts have also occurred broadly over the northeastern U.S. In percentage terms, the precipitation increase is most dramatic (greater than 10 percent) in the central part of the continent, especially in North Dakota and South Dakota, and in north-central Canada (Figure 5, top right). Figure 5. Differences in average precipitation between the period and the period, from the GPCC dataset. The four panels show annual differences (top left), annual percentage differences (top right), May-July differences (bottom left), and May-July percentage differences (bottom right). 5

6 Precipitation amounts in May through July the period showing the most significant increase at Devils Lake have also increased broadly, with the greatest percentage changes occurring in parts of the U.S. Midwest, along the U.S. - Canada border from British Columbia to the Dakotas, and also in California and in north-central Canada (Figure 5, bottom right). The changes near Devils Lake are examined more closely in Figure 6, which is obtained from the PRISM dataset. Annual totals have increased by more than 50 mm (10 percent) in the eastern Dakotas and northwestern Minnesota. For the May-July totals, the most dramatic percentage increases have occurred in the region surrounding Devils Lake in northeastern North Dakota. Figure 6. Differences in average precipitation between the period and the period, from the PRISM dataset. The four panels show annual differences (top left), annual percentage differences (top right), May-July differences (bottom left), and May-July percentage differences (bottom right). The Devils lake drainage basin is outlined in red. 6

7 It is clear from the spatial analysis of precipitation changes since 1991 that wetter conditions have occurred very broadly in central and eastern North America. However, a zone of particularly marked summertime increases in precipitation is found in the Great Plains near the U.S.-Canada border. Thus Devils Lake appears to have experienced some of the most dramatic changes in average precipitation to have occurred across the entire continent. Frequency Distribution of Precipitation Changes In view of the widespread, significant increase in precipitation near Devils Lake since 1991, it is of interest to examine whether precipitation events have become more frequent, or more intense, or both. This analysis was based on the daily unified gridded precipitation analysis of the Climate Prediction Center (Chen et al. 2008). The daily basin-average precipitation was extracted for the Devils Lake drainage basin; the resulting frequency distributions (Figure 7) show that the frequency has increased in the period for every category of daily precipitation amount. The overall frequency of measurable (at least 0.01 inches of) precipitation increased from 31.5 percent of days to 35.7 percent of days. Figure 7. Frequency of occurrence of daily basin-average precipitation amounts at Devils Lake, from the CPC Unified Precipitation Analysis. The blue columns show the frequencies from 1948 to 1990, and the red columns show the frequencies. In the highest category of most intense precipitation events (1 inch or more), the frequency increased from 0.36 percent (1.3 events per year) to 0.59 percent (2.2 events per year), a proportional increase of 66 percent. This increase is much larger than the proportional increases at smaller precipitation 7

8 categories. Figure 8 shows that the changes in the larger precipitation categories have accounted for the majority of the observed increase in average annual precipitation since 1991; specifically, the total precipitation falling in daily precipitation events of 0.5 inches or more has accounted for 2.4 inches (64 percent) of the 3.6 inch increase in average annual precipitation. Figure 8. Average annual precipitation totals occurring in categories of daily amount at Devils Lake, from the CPC Unified Precipitation Analysis. The blue columns show the annual averages from 1948 to 1990, and the red columns show the averages. Figure 9 illustrates the spatial distribution of proportional changes in the average frequency of daily precipitation amounts greater than 1 inch, confirming that events of 1 inch or greater have become about 60 percent more frequent near Devils Lake since Very large proportional increases have also occurred in other regions of eastern North Dakota and northeastern South Dakota. In view of the dramatic changes in the precipitation distribution in these areas, it is perhaps not surprising that major water management issues have arisen. 8

9 Figure 9. Percent difference in average frequency of daily precipitation events of more than 1 inch, between and Grid points having less than 1 event per year, on average, are masked. The Devils Lake drainage basin is outlined in red. Source: CPC Unified Precipitation Analysis. 3. Analysis of Upper-level Flow Changes The broad spatial scale of the precipitation increases across central North America in the most recent two decades indicates that the atmospheric circulation has undergone large-scale, though perhaps subtle, changes and that the new average pattern is more favorably disposed to wet conditions in many locations. It is beyond the scope of this study to determine exhaustively the nature of the circulation changes across the entire continent, but it is possible to identify a significant summertime pattern shift that has occurred in the northern Great Plains. Figure 10 shows the difference in the May-July average relative vorticity at 500 hpa between the period and the period; this analysis uses the NCEP-NCAR Global Reanalysis dataset (Kalnay et al. 1996). A zone of markedly increased average vorticity is observed near the U.S.-Canada border, extending from western Montana to the western Great Lakes. Larger relative vorticity reflects the presence of lower pressure and more storminess, which generate increased precipitation. The recent shift to a regime of higher average vorticity is illustrated in Figure 11, which shows the sequence of May-July average relative vorticity values within the green box outlined in Figure 10. The shift in the early 1990s stands out sharply, as 11 of 19 years since 1991 show May-July vorticity higher than the highest value in the period. 9

10 Figure 10. Difference in average 500 hpa relative vorticity in May, June, and July, between the period and the period The green rectangle shows the area in which the vorticity values were averaged to obtain Figure 11; the north-south black dotted line through the Great Plains shows the baseline for the cross-section in Figure 13. Figure 11. Average 500 hpa relative vorticity in May, June, and July within the region outlined in green in Figure

11 In addition to larger average relative vorticity in recent years, the frequency of positive vorticity events has also increased in the vicinity of Devils Lake. A positive vorticity event is sometimes referred to as a short wave and its passage aloft is often accompanied by a distinct precipitation event. Figure 12 shows the difference in average frequency of short waves exceeding a relatively low threshold vorticity of 2 x 10-5 s -1, confirming a concentration of increased frequency near the U.S.-Canada border of North Dakota. Figure 12. Difference in average frequency of occurrence of 500 hpa relative vorticity values of at least 2x10-5 s -1 in May, June, and July, between and The vertical structure of the average vorticity change is illustrated in a north-south cross-section along 100 W (Figure 13); the baseline of the cross-section is marked by the black dotted line in Figure 10. The 100 W meridian passes just west of Devils Lake, and the latitude of Devils Lake is approximately 48 N. Figure 13 reveals that the post-1991 increase in vorticity is vertically coherent and is centered almost directly above the Devils Lake region in this north-south vertical plane. A wide variety of additional analyses have been constructed to examine the post-1991 changes in circulation and airflow patterns near Devils Lake; some of these have been made available on the Prescient Weather Devils Lake website. At this juncture, however, it is sufficient to emphasize that historical reanalysis data have revealed a clear shift in the early summer months to a circulation that favors storminess and wet conditions over and near Devils Lake. It is also of interest to note that the same is true of October (Figure 14), which has seen the highest percentage increase in precipitation of any single month at Devils Lake (Figure 4). In October, the increase in average relative vorticity is centered farther west, along the Montana-Canada border but also extends over all of North Dakota. 11

12 Figure 13. Top panel: cross-section of average May-July relative vorticity along 100 W, for (black lines) and (red lines). Contours for negative values are dashed. Bottom panel: crosssection of the difference in average May-July relative vorticity along 100 W, between and Devils Lake lies just to the east of the cross-section at approximately 48 N. 12

13 Figure 14. Difference in average 500 hpa relative vorticity in October, between the period and the period Possible Role of Large-Scale Climate Oscillations Numerous studies in the past 15 years have discussed the importance of large-scale modes of decadal to interdecadal climate variability in affecting regional climate, including rainfall distributions, across the globe (e.g. Parker et al. 2007, Meinke et al. 2005). Among the phenomena that show decadal variability are the El Niño - Southern Oscillation (ENSO, Rodgers et al. 2004), the Pacific Decadal Oscillation (PDO, Mantua and Hare 2002), the Atlantic Multidecadal Oscillation (AMO, Enfield et al. 2001), and the North Atlantic Oscillation (NAO, Hurrell 1995). The annual and decadal averages of the indices that represent the phase of these oscillations are shown in Figure

14 Figure 15. Annual (shaded) and decadal (black lines) averages of ENSO, PDO, AMO, and NAO. An early hypothesis concerning the recent changes at Devils Lake was that phase shifts in one or more of the decadally-varying climate indices would be identified as causally related to the precipitation shift in North Dakota. The annually- and decadally-averaged climate indices do not appear to exhibit any marked and lasting changes on or around 1991, and a preliminary data mining analysis has not revealed any clear connection between the annual climate indices and the Devils Lake precipitation. However, a more sophisticated analysis that compares the smoothed climate indices with rotated empirical orthogonal functions of extreme precipitation frequency is showing promise and will be explored further in future research. Closer examination of the months in which precipitation has increased most significantly (May through July, and October) reveals significant climate index changes that coincided with the 1991 onset of wetter conditions. First, ENSO in the months of May through July has shifted to a more positive state since Figure 16 shows the sequence of May-July average ENSO index values and 14

15 the running decadal average. Beginning in 1991, five consecutive years occurred in which the May- July ENSO index was positive, and most years since 1991 have continued in the positive phase. The decadal average has reflected this trend by remaining higher in recent years than at most times in the record prior to Second, the NAO has shifted to a more negative state since 1991 in October, as shown in Figure 17. Negative values of the October NAO began in 1991 and have continued in most years since. Figure 16. Yearly values of average ENSO index values in May, June, and July (blue columns), and centered decadal averages of the same (red line). Source: Bivariate ENSO time series data from the NOAA Earth System Research Laboratory. 15

16 Figure 17. NAO index values in October (blue columns), and centered decadal averages of the same (red line). Source: Climate Analysis Section, Climate and Global Dynamics Division, NCAR Earth Systems Laboratory. The simultaneous appearance in 1991 of the apparent phase shifts in May-July ENSO, October NAO, and Devils Lake precipitation suggests quite strongly that the ENSO and NAO changes may be causally related to the wetter conditions at Devils Lake. This possibility is the subject of ongoing investigation, with a view to determining the mechanisms by which such effects are mediated. In the case of ENSO variability, the influence upon Devils Lake precipitation is likely transmitted via the circulation anomalies over the north Pacific. As shown in Figure 18, El Niño conditions in the May-July period are usually associated with lower than usual 500 hpa heights over the northwest Pacific Ocean near Kamchatka. The zone of lower than average heights extends across the central Pacific north of Hawaii and also includes the southwestern US. In contrast, in May-July seasons having La Niña conditions, heights are below-normal across the far North Pacific from the Bering Sea to the Gulf of Alaska, and including western Canada (Figure 19). Therefore in La Niña seasons, the upper-level trough that is a semi-permanent feature of the north Pacific is displaced eastwards and tends to reside closer to Canada than in El Niño seasons. These circulation differences over the north Pacific are very likely partially responsible for the altered flow patterns that create significant differences in the amount of precipitation falling in central North America in this season (Figure 20 and Figure 21). However, a more detailed exposition of the nature of the connection awaits further investigation. 16

17 Figure 18. Average 500 hpa height anomaly in May, June, and July, for the 10 years since 1948 having the most positive average ENSO index value in those months. Figure 19. Average 500 hpa height anomaly in May, June, and July, for the 10 years since 1948 having the most negative average ENSO index value in those months. 17

18 Figure 20. Average percent of normal precipitation in May, June, and July, for the 10 years since 1901 having the most positive ENSO index in those months. Figure 21. Average percent of normal precipitation in May, June, and July, for the 10 years since 1901 having the most negative ENSO index in those months. 18

19 5. Conclusions The recent dramatic expansion of Devils Lake, North Dakota, has occurred in response to a significant increase in precipitation that began in 1991 and has been concentrated mainly in the months of May, June, July, and October. Precipitation events in the Devils Lake basin have become both more frequent and more intense, on average. These changes have occurred as part of a widespread shift to increased precipitation across most of central North America from Texas to north-central Canada. Atmospheric conditions over the northern Great Plains have featured increased upper-level vorticity since 1991, reflecting lower pressure and more storminess. The cause of this shift is not yet clear, but it may be related to simultaneous changes that occurred in the phases of ENSO and the NAO. In May to July, the phase of ENSO has become more positive, while in October, the NAO has become predominantly negative. Increased precipitation over the past two decades has forced Devils Lake to rise and expand, with disastrous consequences for the surrounding properties. The cause of the increased precipitation appears to be subtle changes in the large-scale flow patterns over North America. Understanding the origin and evolution of those changes, and perhaps capturing their essence in dynamical computer models of the flow, is essential in attempting to foresee the behavior of Devils Lake in the decades to come. References Beck, C., J. Gieser, and B. Rudolf, 2005: A New Monthly Precipitation Climatology for the Global Land Areas for the Period 1951 to DWD, Klimastatusbericht 2004, ISSN , ISBN , Chen, M., W. Shi, P. Xie, V. B. S. Silva, V. E. Kousky, R. W. Higgins, and J. E. Janowiak, 2008: Assessing Objective Techniques for Gauge-Based Analyses of Global Daily Precipitation. J. Geophys. Res., 113, D04110, doi: /2007JD Daly, C., W. P. Gibson, G. H. Taylor, G. L. Johnson, and P. Pasteris, 2002: A Knowledge-Based Approach to the Statistical Mapping of Climate. Climate Research, 22, Enfield, D. B., A. M. Mestas-Nuñez, and P. J. Trimble, 2001: The Atlantic Multidecadal Oscillation and Its Relation to Rainfall and River Flows in the Continental U.S. Geophys. Res. Lett., 28, Hurrell, J. W., 1995: Decadal Trends in the North Atlantic Oscillation: Regional Temperatures and Precipitation. Science, 269, Kalnay, E., and co-authors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, Mantua, N. J., and S. R. Hare, 2002: The Pacific Decadal Oscillation. J. Oceanography, 58,

20 Meinke, H., P. DeVoil, G. L. Hammer, S. Power, R. Allan, R. C. Stone, C. Folland, and A. Potgieter, 2005: Rainfall Variability at Decadal and Longer Time Scales: Signal or Noise? J. Climate, 18, Parker, D., C. Folland, A. Scaife, J. Knight, A. Colman, P. Baines, and B. Dong, 2008: Decadal to Multidecadal Variability and the Climate Change Background. J. Geophys. Res., 112, D18115, doi: /2007JD Rodgers, K. B., P. Friederichs, and M. Latif, 2004: Tropical Pacific Decadal Variability and Its Relation to Decadal Modulation of ENSO. J. Climate, 17, Vecchia, A. V., 2008: Climate Simulation and Flood Risk Analysis for for Devils Lake, North Dakota. U.S. Geological Survey Scientific Investigations Report , 28p. 20