The gaging record from Vicksburg, Miss., extends back to 1815 and represents the longest available record of Mississippi River discharge into the Gulf of Mexico. In order to judge how well the Vicksburg discharge record represents large-scale events within the Mississippi Basin, we examined gaging records for the Illinois River, Missouri River, Ohio River, Arkansas River, and Red River to identify the longest representative record for each river (figs. 5, 6, 7, 8, 9). We then compared discharge records for the Illinois, Missouri, and Ohio Rivers with each other and with the long-term gaging records for the Mississippi River at St. Louis, Mo., and at Vicksburg, Miss. (figs. 10, 11).
Inspection of discharge records shows that annual mean discharge records from different locations on the main river within each basin show very similar patterns and that the pattern of annual mean discharge along the major rivers has not been significantly altered by manmade structures. For example, the variation in the three long-term records from the Illinois River shown in figure 5 is very similar during the 60 years of overlap between 1939 and 1999. The prominent maxima and minima in discharge match in all three records, and the similarity between records does not change for data collected after completion of dams in 1938 and 1939.
Comparison of long-term discharge records from the Ohio River (fig. 7) also shows a close correspondence in variation of flow from different stations along the river. During the 65 years that the records overlap, the variation in the three discharge records is very similar, indicating that the numerous dams and locks built along the Ohio River between 1929 and 1980 have not altered the overall variation in annual mean daily flow. The discharge records summarized in figures 5, 6, 7, 8, and 9 indicate that the longest available gaging record from each river can be used as a reliable proxy for annual mean discharge.
Comparison of the discharge record of the Mississippi River at St. Louis with records for the Illinois River at Marseilles and the Missouri River at Hermann shows that the St. Louis record closely matches the Hermann record (fig. 10). Although there are differences in details, the Hermann and St. Louis records are also similar to the Marseilles flow record. For example, all three records show prominent discharge maxima in 1927 and 1993 and prominent minima in 1940 and 1956 (fig. 10).
Discharge records from the Mississippi River at St. Louis, the Ohio River at Metropolis, the Arkansas River at Van Buren, and the Mississippi River at Vicksburg are compared in figure 11. The discharge records in figure 11 have many major features in common but differ in detail, reflecting the composite nature of the Vicksburg record. For example, the St. Louis, Metropolis, and Vicksburg records include prominent discharge minima in 1895 and a prominent discharge maximum in 1927. High discharge is shown in all four records for 1950-51 and 1974-75. All records show low discharge in the early 1960's and low to average discharge between 1930 and 1940. High discharge at Vicksburg in 1979 was largely due to high discharge in the Ohio River. Figure 11 indicates that the Vicksburg discharge primarily reflects conditions in the Central United States with modifications due to the contribution from the Ohio River.
The Vicksburg record shows substantial interannual variability about average mean discharge, but it also contains several extreme discharge events and several decadal-scale features representing extended intervals of overall below-average or above-average discharge (fig. 11). Low-discharge (drought) intervals are evident from 1894 to 1903, 1930 to 1942, and 1952 to the late 1960's. High-discharge or abnormally wet intervals are present from 1840 to 1852 and from 1981 to 1997. The interval from 1858 to 1870 could be considered a high-discharge interval although the departures from the long-term mean are not as consistent and extreme as the departures in the 1840-52 and 1981-97 intervals. Very high discharges occurred in 1847, 1927, and 1993.
The low-discharge intervals at Vicksburg correspond to several well-known severe droughts. The 1930-42 low-discharge interval corresponds to the Dust Bowl that resulted in farm failures in the lower Great Plains (Woodhouse and Overpeck, 1998). Note that the Dust Bowl event is more evident in the St. Louis and Van Buren records than it is in the Metropolis record, which reflects the regional nature of the drought conditions.
The low-discharge interval at Vicksburg from 1952 to 1969 includes the major drought of the 1950's, which is best represented in records from the southern Great Plains and the Southwestern United States (Woodhouse and Overpeck, 1998), and the 1960's drought. The 1950's drought is well represented in the discharge record at St. Louis, indicating that the drought affected much of the Great Basin.
Discharge at Vicksburg returned to near mean conditions around 1960 but then was reduced for several years in the early and middle 1960's. The 1960's drought is usually associated with the Northeastern United States (Cook and others, 1999), but the discharge records of the Mississippi at St. Louis and the Arkansas River show that dry conditions (low discharge) also occurred in the central and northern Great Plains during the 1960's. The low-discharge values match climate reconstructions for the continental United States based on tree-ring studies (Cook and others, 1999), which show that moderate to severe drought conditions prevailed across large areas of the Great Plains in 1963, 1964, 1966, and 1967 (fig. 12).
The low-discharge interval of 1894 to 1903 includes 3 years of very low discharge and several years of average flow. This 1894-1903 interval corresponds to severe drought recorded in the Great Plains and parts of the Southwestern United States (D'Arrigo and Jacoby, 1991; see summary in Woodhouse and Overpeck, 1998, fig. 3). Reconstructions based on tree ring studies indicate that the Great Plains drought was particularly severe in 1884, 1885, 1900, and 1901 (Cook and others, 1999). The most recent high-discharge interval (1981-97) contains the major flood of 1993. Note that the other definite decadal-scale high-discharge interval in the Vicksburg record (1840-52) also contains a major flood event comparable to the 1993 flood. The next highest discharge in 1927 is a more isolated event, but the 1927 flood was devastating (Bourne, 2000).
Comparison of a proxy for wet and dry conditions for the Great Plains based on tree-ring studies (Meko, 1992) and the flow record of the Mississippi River at Vicksburg is shown in figure 13. Inspection of figure 13 indicates that the tree-ring-derived proxy and the discharge record are similar. With minor exceptions, the two data sets covary between 1865 and 1965. The correlation between the records is very high (r2 = +0.82) between 1890 and 1930. Between 1840 and 1865, the correspondence between the tree-ring proxy and the discharge record degrades, largely due to a strong trend in the tree-ring proxy from wet conditions (positive values) to dry conditions (negative values) that is not present in the flow data. Between 1820 and 1840, the discharge record appears to lead the tree-ring proxy by a few years, creating an offset in the two time series. We cannot explain the offset for 1820-40 but suspect that errors in counting rings to develop the tree-ring chronology are the most likely causes. Overall, our data indicate that the discharge record of the Mississippi River at Vicksburg matches major features of the climate record in the Great Plains region derived and compiled from historical and tree-ring studies.
Assessing natural discharge variations for the Rio Grande is complicated. Construction of reservoirs, removal of water for irrigation, and rerouting and reshaping of the river channel have greatly reduced flow and altered the natural pattern of flow variation (Mueller, 1975). Thus, discharge near the mouth of the Rio Grande is highly perturbed by human activities. Another complication is that Rio Grande flow is not always continuous. Flow is greatly reduced or even eliminated in dry years downstream of El Paso, Tex. In these dry years, the flow then increases or resumes in the Rio Grande near Presidio due to the intersection of the Rio Conchos with the Rio Grande. The Rio Conchos contributes up to 75 percent of flow in the lower Rio Grande downstream of Presidio (Moring and Setser, 2000). However, in modern times, human activities have reduced the flow in the lower Rio Grande below the junction with the Rio Conchos by about 87 percent (Mueller, 1975).
In order to study discharge records showing the fewest effects of human activities, we used long-term gaging records from the upper part of the Rio Grande upstream from Albuquerque, N. Mex., and from just downstream of the intersection of the Rio Conchos with the Rio Grande to represent historical flow. Inspection of available records suggests that the record from Embudo, N. Mex. (fig. 14), is the best available record for the upper Rio Grande. The Embudo station is upstream of major dams and has a continuous record from 1912 to 1999 and a separate series from 1890 to 1903. For the lower Rio Grande downstream of the intersection of the Rio Conchos and the Rio Grande, we selected the record from Presidio because it was the longest, albeit discontinuous, record (fig. 14).
The most striking characteristic of the Embudo record (fig. 14) is the long interval of relatively low flow from 1950 to 1978. The only extended interval of high overall flow is from about 1915 to 1924. Maximum flow events occurred in 1891, 1920, 1941-42, and 1985-87.
The Presidio record shows many of the same general features as the Embudo record but differs in details. For example, the two records show very high flow maxima in 1941 and 1942 and an extended interval of low flow from 1950 through 1978. However, flow maxima in 1990 and 1991 in the Presidio record correspond to average or below-average flow in the Embudo record. The Presidio record indicates high overall flow between 1904 and 1912.
The long dry interval evident from 1950 to 1978 in the Embudo and Presidio records includes the major 1950's drought in the Southwestern United States and the 1960's drought (Cook and others, 1999). In many areas of the Southwest, and especially west Texas and New Mexico, the 1950's drought was more severe than the Dust Bowl event (for example, Stahle and Cleaveland, 1988; Swetnam and Betancourt, 1998). The Rio Grande record indicates that the 1950's drought was actually the beginning of a nearly 30-year drought in the Southwestern United States. The wet conditions of the early 1960's that ended drought in the central and northern Great Plains did little to reduce drought in the Southwest.
Taken together, the Embudo and Presidio records suggest high flow from about 1904 to 1924. The extended interval of high flow reflects the 1905-28 wet interval known from reconstructions of the Southwestern U.S. climate based on tree-ring studies. The flow maxima in 1941 and 1942 seen in the Embudo and Presidio records reflect the extremely wet year in 1941 (D'Arrigo and Jacoby, 1991). The climate reconstructions for the Southwestern United States indicate that 1987 was even wetter than 1941 (D'Arrigo and Jacoby, 1991). The flow records from the Rio Grande indicate that 1987 was a wet year but not as extreme as the 1941 event.
Thus, like those for the Mississippi River, Rio Grande discharge records contain decadal-scale features that correspond to known droughts and extreme high-flow years that match known major floods. Differences between discharge records for the Mississippi River at Vicksburg and the Rio Grande reflect the regional nature of droughts and wet events. For example, the Dust Bowl interval of the 1930's is not well represented in the Rio Grande discharge records because the 1930's drought was most severe and persistent in the central and northern Great Plains; see Woodhouse and Overpeck (1998) for a summary of regional impacts of major historical and prehistorical drought events.
Interannual climate variability in the Southwestern United States and many areas of the Gulf Coast and Great Plains is highly influenced by strong El Niño/Southern Oscillation events (Stahle and Cleaveland, 1993), which are briefly described below. Therefore, we studied flow records for the Mississippi River and Rio Grande to test for an ENSO influence in the flow records.
El Niño is the name given to the quasi-periodic occurrence of anomalously warm surface waters off the coast of South America and in the eastern equatorial Pacific. The surface water warming usually begins in December, thus the association with Christmas and reference to "the child." The warming of eastern equatorial surface waters corresponds to a quasi-periodic change in the difference in sea-level pressure between the eastern and western tropical Pacific, which is the Southern Oscillation. These coupled changes in sea-surface temperature and sea-level pressure are referred to as ENSO. ENSO influences atmospheric circulation outside of the tropical Pacific region, which, in turn, causes regional climate variations linked to ENSO through the atmospheric "teleconnections" (Diaz and Kiladis, 1992).
In the "normal" condition, sea-level pressure in the eastern equatorial Pacific is high relative to sea-level pressure in the western equatorial Pacific. Surface winds blow from the eastern equatorial Pacific to the west. The winds move surface water west and toward the poles (due to the Coriolis effect), resulting in upwelling of cool, subsurface nutrient-rich water. The persistent upwelling results in cool sea-surface temperatures along the coast of South America and in the eastern equatorial Pacific. Warm surface waters are moved to the west and pile up in the western Pacific (see fig. 15).
During an El Niño, or the warm extreme of ENSO, sea-level pressure in the eastern equatorial Pacific is reduced, and surface winds decline in intensity or even reverse direction and blow to the east. Upwelling of cool waters along the South American Coast declines or stops, and the warm surface waters of the western Pacific expand eastward toward the Americas.
A La Niña event is the opposite of an El Niño event and represents an extreme case of the normal situation. During La Niña, sea-surface pressure in the eastern equatorial Pacific increases, wind intensity increases, and surface waters become cooler than normal due to increased upwelling. More extensive discussions of ENSO are provided by Diaz and Kiladis (1992) and Zebiak (1999).
Establishing a record of significant individual El Niño and La Niña events is complicated because different indicators and boundary conditions have been used to define specific events. Some workers use an index based on the sea-surface-pressure difference between Darwin, Australia, and Tahiti, which is the Southern Oscillation Index or SOI. Other workers use sea-surface-temperature anomalies based on particular areas of the tropical Pacific Ocean, and the National Oceanic and Atmospheric Administration (NOAA) has developed a composite index that combines several ENSO indicators. In addition, records of the presumed proxies for El Niño and La Niña in areas within and outside of the equatorial Pacific are often used to identify ENSO events to augment or extend instrumental records (for example, see Quinn, 1992).
We compiled records of El Niño and La Niña events from several sources in order to develop a composite record for our study (fig. 4). Records from 1950 to the present are well constrained and show a great deal of agreement. Records before 1950 are based on sparse and less reliable data.
Flow records for the Ohio River at Metropolis, the Mississippi River at St. Louis and at Vicksburg, and the Rio Grande at Embudo and at Presidio are shown in figures 16 and 17 with major ENSO events from figure 4. We first used the Mann-Whitney U-test (Mann and Whitney, 1947) to determine if ENSO extremes (El Niño versus La Niña) were associated with significantly increased or decreased river flow. For each flow record, average daily flow values were obtained for each of the years designated as either El Niño or La Niña. A Mann-Whitney U test was performed on each data set using 
= 0.05.
= 0.05.
The results of the analysis show that there is no statistically significant difference (either positive or negative) in flow rate due to El Niño or La Niña in the St. Louis, Vicksburg, or Metropolis data sets (table 2). An interesting point is that the effects of El Niño on the Mississippi River flow at St. Louis and on Ohio River flow at Metropolis, while not statistically significant, are opposite. In contrast to the Mississippi River results, our analysis shows that El Niño years have statistically higher average daily flow rates than do La Niña years in the Rio Grande data sets (table 2).
To further explore the relation between ENSO events and flow in the Rio Grande, we did simple time-series experiments with the flow data between 1915 and 1997 for the Embudo record and between 1930 and 1997 for the Presidio record. We performed a cross-spectral analysis on flow data and the SOI. To make the SOI comparable to our data, we converted the monthly SOI index (see
http://daac.gsfc.nasa.gov/COMPAIGN_DOCS/FTP_SITE/INT_DIS/readmes/soi.html) to an annual mean (fig. 18). The power spectra for SOI (figs. 19, 20) show peaks at frequencies of approximately 0.150, 0.175 (Embudo only), and 0.237, which translate to cycles having periods of 4.5 to 6.5 years. Note that the difference in the SOI power spectra for comparison with Embudo and Presidio flow data is due to the different length of the records. In other words, the 0.175 frequency is not well represented in the shorter SOI data set.
http://daac.gsfc.nasa.gov/COMPAIGN_DOCS/FTP_SITE/INT_DIS/readmes/soi.html) to an annual mean (fig. 18). The power spectra for SOI (figs. 19, 20) show peaks at frequencies of approximately 0.150, 0.175 (Embudo only), and 0.237, which translate to cycles having periods of 4.5 to 6.5 years. Note that the difference in the SOI power spectra for comparison with Embudo and Presidio flow data is due to the different length of the records. In other words, the 0.175 frequency is not well represented in the shorter SOI data set.
Analysis of the Embudo flow record results in cycles having periods similar to the SOI. The variation in the SOI and the Embudo flow records is coherent at periods of approximately 4.5, 5.4, and 6.5 years (fig. 19). The analyses suggest a clear relation between ENSO events and flow in the upper part of the Rio Grande. Analyses of the Presidio flow record show the presence of the 4.5-year cycle and demonstrate that variation in the SOI and flow are coherent at the 4.5-year period. The 6.5-year period observed in the 1930-95 SOI record is not present in the Presidio flow record. Our analyses indicate that flow in the Rio Grande downstream from the Rio Conchos is related to ENSO events but less strongly than flow variations in the upper part of the basin as monitored by the Embudo record.
The link between Rio Grande discharge and ENSO events is consistent with previous studies based on instrumental records (Diaz and Kiladis, 1992) and tree-ring records (Stahle and Cleaveland, 1993), which demonstrate correlation between ENSO and climate in the Southwestern United States. Tree-ring data (D'Arrigo and Jacoby, 1992) demonstrate that the El Niño results in increased winter precipitation in northern New Mexico, which would have a direct effect on the Embudo record due to melting of increased winter snow pack. It seems likely that the ENSO signal in the lower Rio Grande is complicated by variations in the summer monsoon. Most of the flow in the lower Rio Grande derives from the Rio Conchos, and the Rio Conchos is influenced by summer monsoon rains (Moring and Setser, 2000). However, the Southwest summer monsoon also shows some correlation with ENSO (Harrington and others, 1992).
Another complication in the ENSO signal is due to decadal-scale variations in sea-surface temperature of the North Pacific Ocean known as the Pacific Decadal Oscillation (PDO). Historical trends indicate that when the PDO is positive (indicating that central North Pacific sea-surface temperatures are cooler than normal), ENSO influence on the Western United States diminishes (McCabe and Dettinger, 1999). Despite these complications, our data indicate that variation in flow of the Rio Grande is related to ENSO and that the correlation is strongest in the upper part of the drainage basin.
In contrast to correlations with individual ENSO events or short-term variability, a long-term reconstruction of Mississippi River flow may provide important information on changes in the long-term variation of ENSO and changes in mean climate state. For example, instrumental and coral-based studies show that the tropical Pacific climate shifted to warmer and wetter conditions in 1976. This shift coincided with a change in variability of ENSO from predominantly interannual to a period of about 4 years (Urban and others, 2000). The flow records from the Mississippi River and the Rio Grande show a shift to wetter conditions in the late 1970's. The change is most striking in the Vicksburg record, which shows that the last 25 years has the highest overall flow of any 25-year period since 1815. A smoothed Vicksburg flow record (fig. 21) shows the change to wetter conditions in the 1970's.
U.S. Geological Survey, U.S. Department of the Interior
This page is https://pubs.usgs.gov/bulletin/b2187/results.html
Contact: Harry Dowsett (hdowsett@usgs.gov)
Last modified 09.26.01 (krw)
This page is https://pubs.usgs.gov/bulletin/b2187/results.html
Contact: Harry Dowsett (hdowsett@usgs.gov)
Last modified 09.26.01 (krw)