By J. R. Slack and Jurate Maciunas Landwehr
These records of river and stream flow are available as a dataset, including an interactive data viewer and downloadable data files.It has long been known that surface-water conditions are generally correlated with fluctuations in meteorologic variables such as precipitation and temperature. (For example, see discussion by Newell, 1891.) However, the dynamics of streamflow are not just a simple first-order response to existing atmospheric conditions. "Streamflow", the surface-water discharge measured in a natural channel, differs from "runoff", the fraction of precipitation that appears in surface streams. (Definitions taken from Langbein and Iseri: 1960.) There is a complex, interactive relationship between hydrologic conditions and meteorologic fluxes. Inputs from diverse specific local precipitation events are collected over the surface of the watershed, so that the meteorologic fluxes are spatially integrated by the watershed. Meteorologic events and conditions are also temporally integrated because the watershed retains moisture both on and below its surface. Storage on the land surface occurs primarily in lakes, ponds and wetlands and in the seasonal or even multiyear accumulations of snow, as well as in the biomass covering the surface. Storage in the subsurface occurs in both the unsaturated and the saturated zones. Furthermore, water can enter or leave the watershed through the subsurface. Losses of water to the atmosphere (through evaporative and transpirative processes) occur throughout the watershed. This moisture may return again to the watershed as precipitation or it may leave the watershed entirely by advection. Thus, the watershed acts to dampen the noisy signal of specific instantaneous and local meteorologic events. Records of streamflow can provide a filtered account of meteorologic fluctuations over the watershed i.e., of prevailing climatic conditions.
The ability of streamflow records to reflect variations in the prevailing climate is conditioned on the absence of any other major causes that would radically alter streamflow patterns during the period of record. Such confounding processes would generally be anthropogenic in origin, that is, induced by human activity, either intentionally or unintentionally. Some human actions, such as the removal of water from a stream for consumptive use or the regulation of extremes by a control structure, have a direct effect on streamflow patterns. For other activities, the consequences may be indirect but can be equally significant, such as the effects of a change in the watershed storage capacity due to major land-use changes during the period of record. In either case, the pattern of past climate variation to be discerned in the streamflow record would be confounded by changes induced by anthropogenic activity. An exception might be made, however, if the non-climatic forcing factor was consistent in its nature and degree over the period of record. In this case, the effects are present as a constant background level within the record. In this case, the record of streamflow measured at a surface-water gaging station in the watershed could still be useful for the study of past climate variation.
A significant potential effect of a change in prevailing climatic conditions would be a shift in the seasonal patterns of precipitation and temperature that could induce a seasonal shift in the timing of hydrologic events as well. Such a change in the hydrologic patterns could only be discerned in a record whose values represent the average streamflow conditions over a sufficiently small time span, one that is shorter than the interval of the induced effect. In the case of a seasonal shift, the time step within the hydrologic record should be at least a month. Thus, the criteria for the suitability of a streamflow record for the study of potential effects of climate variation is that the discharge values be representative of at most monthly mean conditions and that the effects of anthropogenic controls, either intentional or unintentional, should be negligible for the average discharge over the time step of the record.
In order to characterize the variation in streamflow over the land area of the United States during the last past century, a period of great temporal and spatial climatic variation on the North American continent (for example, see discussion by Diaz, 1986), and to contrast this variation with that of concurrent meteorological conditions, it is necessary to assemble a large data set of suitable and long streamflow records. The primary source of such information for the United States is the U.S. Geological Survey (USGS). Currently, streamflow data are collected by nationally standardized procedures through the activities of the USGS network of District offices. (Except for three multi-State Districts, each District office corresponds administratively to a single State.) A discussion of these national procedures for the measurement and computation of streamflow can be found in the report of Rantz and others (1982).
The USGS National Water Storage and Retrieval System (WATSTORE) data base contains streamflow records for the United States and its trust territories going back into the 19th century. (See Hutchinson and others, 1975, for a description of the structure of WATSTORE.) The data base even contains data that predate the establishment of the initial USGS gaging station on the Colorado River at Embudo, New Mexico, in 1888 (Frazier and Heckler, 1972) that were deemed to be of appropriate quality and to satisfy appropriate standard practices. Streamflow data are available in published form, such as in the annual State Water Data Reports, and in electronic form from WATSTORE through the USGS National Water Data Exchange (NAWDEX) Office. (For contact information, see Blackwell, 1991, or Dodd and others, 1989.) However, given the development thathas occurred throughout the United States over the last century, many of the records in WATSTORE are affected, entirely or after some date, by confounding anthropogenic activities, such as discussed above. Thus, many records are not appropriate for the study of climate variation Consequently, an effort was undertaken to identify and assemble the long records of daily mean discharge held by the USGS that are relatively free from confounding anthropogenic effects, as judged according to a single set of consistent criteria.
With respect to the National Water Conditions data set, the USGS publishes monthly a report that geographically summarizes patterns of monthly mean discharge conditions at the nationally distributed sites. Currently entitled "National Water Conditions", the publication was begun inOctober 1944, under the title "Water Resources Review." (See discussion in Holmes, 1987.) The stations comprising the data set are referred to as "National Water Conditions Streamflow Index Stations" or just "Index Stations." The composition of the collection has changed over time. For example, the set current in 1988 as used in development of HCDN was slightly amended in 1989; see discussion of changes in "National Water Conditions October 1989." This collection of stations, in its entirety or as subsets of the whole, has served as the source of data for many studies of continental and regional streamflow patterns beginning with those of Harbeck and Langbein (1949). (Also see Busby, 1963; Bartlein, 1982; and Cayan and Peterson, 1989, as examples of uses of this data set.) It should be noted that in some cases, the data used in the monthly report does not reside in WATSTORE. Rather, the data may be furnished by another agency, such as the U. S. Corps of Engineers, or the discharge values are adjusted to account for impoundments, using additional information provided by another agency or one of the USGS District offices. In such cases, the data values in the monthly report reside only in the files of the USGS office that prepared it, namely the Hydrologic Information Unit of the USGS Water Resources Division.
Station lists from all of the above collections were compiled. It was not expected that the data sets would be congruent; however, it was expected that there would be a large overlap among them. That this was not found to be the case suggested that still other appropriate streamflow records for yet other stations might exist within the USGS files. H. C. Riggs (of the Office of Surface-Water of the USGS Water Resources Division) provided an initial review of the various annual State Water Data Reports, and verified that such data did exist. (H. C. Riggs, USGS, oral communication, 1988.)
The effort was then enlarged with the intent of identifying as many as possible of the discharge records held by the USGS whose monthly mean values would be appropriate for the study of the relationship of hydrologic conditions with meso-scale climate variation, as determined by a single set of criteria. Given that the USGS has spent considerable effort in automating the streamflow data files, searching WATSTORE is, for practical purposes, synonymous with choosing from all available daily mean discharge data held by the USGS. Given that daily mean discharge records are available for more than 12,000 surface-water gaging stations, both active and inactive, with in the WATSTORE data base, and given that the criteria required specific information about watershed conditions frequently available only from the originating District office, the help of a surface-water specialist in each District office was enlisted. The criteria by which station records were designated as acceptable for inclusion in the HCDN are as follows.
The rating assigned to a record of daily mean discharge reflects the professional judgment of the office that obtains and prepares the records. The rating of overall accuracy is conditioned on the accuracy of the stage measurements, the stability of the stage-discharge relationship, the accuracy and the frequency with which discharge measurements are made to establish the stage-discharge relationship, and the interpretation of the records. An accuracy of "good" implies that 95 per cent of the daily mean discharge values are assessed to be at least within 10 per cent of the true value. A discussion of the accuracy of the records and the assignment of ratings can be found in the USGS Water-Data Reports for each State, published annually, as well as in the report by Rantz and others (1982).
Although discharge records that were subject to diversion or streamflow transfers of any kind were generally not acceptable for inclusion in the HCDN, an exception was made if the diversion had existed virtually unchanged for the entire period of record -- that is, a diversion may exist and affect the actual value of streamflow, but its effect on the daily mean discharge record is unchanging and should not significantly mask the influence of climate variation in the record. Similarly, stations where streamflow was subject to regulation by dams were generally not acceptable. However, if the regulation was due to a low-head hydropower dam with a reservoir of small storage capacity having only a transient effect on high or low streamflow, with no effect on the monthly mean discharge values, then the discharge record for the station could be acceptable for inclusion in the HCDN. On the other hand, a station in a basin that has undergone a substantial land-cover change, for example from forest to agriculture to urban, is probably not suitable even if the changes were gradual.
It is noted that a streamflow record was considered suitable for inclusion in the HCDN if the monthly mean discharge values met the criterion for nonimpairment of "natural" streamflow conditions. However, for the majority of records, even the daily mean discharges satisfy this criterion. Consequently, if a record should be considered unimpaired only at a monthly or longer average time step, this qualification is given as a comment which accompanies the identifying information for the station.
Special attention was paid to stations that are included in either the Hydrologic Benchmark Network or in the National Conditions Streamflow Index Stations set used in the 1988 monthly National Water Conditions reports. The objectives for defining both of these data collections are allied to the purpose for which the HCDN was constructed, and both have been of particular interest to meso-scale climate studies in the past. If any station in either of these two data sets was deemed not to be acceptable for inclusion in the HCDN by the criteria specified above, then that station was specifically identified as "rejected" for inclusion in the HCDN, with an explanation given for its unacceptability. For example, Crater Lake, Oregon, is a Hydrologic Benchmark station but the station records lake levels, not streamflow. Similarly, the St. Lawrence River at Cornwall, Ontario, near Massena, New York, is a National Water Conditions station even though the flow has been regulated in accordance with an international joint agreement with Canada since 1958. Indeed, there is no actual gaging station at the site; rather, the reported discharge is constructed on the basis of discharge at several points of regulation upstream from the site. Thus, records from these two sites are not included in the HCDN.
The geographic distribution of the HCDN stations throughout the United States, in relation to State and hydrologic boundaries, is shown in figure 1a-1e. The network density indicates a natural constraint: there are more streamflow gaging stations in the humid regions of the United States than in the arid zones. The length of acceptable record available at each HCDN station is shown in figure 2. The geographic distribution of the length of acceptable records reflects both the history of the streamflow gaging program of the USGS, which began in the arid regions west of the 100th meridian, and the history of the settlement of the United States, beginning in the humid East. The number of stations with a suitable record of given length, for the range of record lengths, is plotted in figure 3. Snapshots of the data set at four different times in the development of the USGS streamflow gaging program are shown in figure 4. In contrast, the total number of stations for which suitable data is available in any given water year is plotted in figure 5. It can be seen that after 1970 there is a national decline in the number of stations for which data is available. This can be attributed to several causes, including (1) a decrease in cooperative funding of streamflow gaging stations by several States, leading to the termination of data collection a sufficient number of sites in those States to have national consequences cumulatively; (2) increased streamflow regulation; and (3) increased urbanization. The geographic distribution of one of the basin characteristics available for the HCDN stations -- namely, the mean annual precipitation within the watershed -- is shown in Figure 6. The figure further emphasizes the occurrence of streamflow gaging stations in humid areas. The geographic distribution of the size of the drainage areas for the HCDN stations is shown in figure 7. This information is also summarized in figure 8, where the number of stations and the total number of suitable water years of daily mean discharge available are plotted over the range of sizes of the drainage areas of the HCDN stations.
A complete compilation of all of the station identification and descriptive information as well as any qualifying comments for each station are given in table 1, which is to be found at the end of the report. The HCDN station entries are given in order by USGS station number. The identifying information was taken primarily from the USGS WATSTORE Station Header file. It includes the USGS station number, the name of the station (reproduced exactly as officially given in the Station Header file), the latitude and longitude of the station, and the State and a code for the County in which the station is located. (The County codes are found in Appendix C, Chapter 3, Volume I, of the WATSTORE User's Guide by Hutchinson and others, 1975, and correspond to the Federal Information Processing Standards; i.e., FIPS, codes.) If the USGS District office responsible for the maintenance of the station is in a State other than that in which the station is located, the responsible District office is also identified. The 8-digit Hydrologic Unit Code (HUC), which uniquely identifies the geographic area representing part or all of the drainage basin is given for the station. (See Seaber and others, 1987, for a description of the delineation of hydrologic units.) The first two digits of the HUC denote in which of the 21 water- resources regions within the United States and its territories the station is located. The gage datum, which serves as a reference point in determining stage and may be assigned to any arbitrary reference point, is given here because it usually corresponds to the approximate height above sea level (NGVD of 1929) at the gage. The drainage area of the stream at the specific station location is provided, and the percentage of noncontributing area is given only if there is any such in the drainage basin. In addition, if the station is a National Water Conditions Stream Index or Hydrologic Benchmark Network station, it is so identified.
Any comments about the record available for the station in the HCDN are given in table 1. These include any qualifications about watershed conditions with respect to the suitability of the record as defined by the HCDN criteria as well as explanations for short records, as needed. If only monthly mean discharge values or if only a part of the entire period of record are accepted for inclusion in the HCDN, this is noted in the comments for the station, as well. Finally, an explanation is provided for rejecting any Hydrologic Benchmark Network or National Water Conditions station, if it was deemed unacceptable for inclusion in the HCDN. (Note that a list of just the comments for each station can be found in the file COMMENTS.DAT on the enclosed disk.)
Finally, table 1 provides a snapshot of the length of acceptable record available for the station. The number of water years for which data is acceptable by the HCDN criteria is given, along with a graphical depiction of the acceptable water years in the station's record. Water years for which the data is acceptable are denoted by an "*" on the dashed time line, which begins in water year 1874 and runs through water year 1988.
For ease in searching for stations by location or by length of record, the contents of table 1 are restated in tables 2 and 3, respectively, following table 1 in the back of the report. Stations are listed in order of station number. Table 2 provides a concise summary of the station identification information given in table 1 for each HCDN station, as well as for any Hydrologic Benchmark Network or National Water Conditions station even if rejected, that is, deemed unacceptable for inclusion in the HCDN as defined by the criteria discussed above. (Table 2 corresponds to the file STATIONS.DAT on the enclosed disk.) Table 3 provides a summary of the record available at each station, giving both the number of water years of suitable data and the time line that graphically depicts the acceptable water years in the record available for the station. (Table 3 corresponds to the file WYEARS.DAT found on the enclosed disk.)
Table 4 (at the back of the report) contains the values of 11 watershed characteristics for each station in the HCDN, as taken from the WATSTORE Basin Characteristics File. These include main channel slope, stream length from gage to basin divide, mean basin elevation (measured from topographic maps by transparent grid sampling method), surface storage area of lakes, ponds and swamps in percent of contributing drainage area, area of lakes and ponds in percent of contributing drainage area, forested area in percent of contributing drainage area, area of glaciers in percent of contributing drainage area, soil (infiltration) index from the Soil Conservation Service, mean annual precipitation, precipitation intensity in 24 hours expected on the average of once each 2 years, and the mean minimum January temperatures. (For a further discussion of these characteristics, see Thomas and Benson, 1970, or Dempster, 1983.) Unlike the discharge data and the station identifying information in the previous tables, the values given in table 4 were taken with minimal review from WATSTORE.
The WATSTORE Basin Characteristics file was created for use in regional streamflow regression studies in 1970 (U.S. Geological Survey, written communication, 1970; also, discussed by Benson and Carter, 1973). It was last updated nationally in 1978 (U.S. Geological Survey, written communication, 1977) for a minimum set of required characteristics. These included all but three of those given in table 4, namely, area of lakes and ponds in percent of contributing drainage area, area of glaciers in percent of contributing drainage area, and soil (infiltration) index. Since that time, the file has been maintained and updated on a State-specific and generally study-specific basis by many different people. Consistent edit procedures were implemented in 1981. Although much of the information in the Basin Characteristics File warrants an "excellent" or "good" accuracy rating, it should be noted that the contents of the file has been subject to variable quality control measures over time. However, the data are of sufficient accuracy for qualitative comparisons and the information is valuable for its aggregate availability. It is provided in table 4 as auxiliary information for the HCDN stations. (Table 4 corresponds to the file BASINS.DAT found on the enclosed disk.)
Any stations found in either the National Water Conditions Streamflow Index set or in the Hydrologic Benchmark Network for which the available discharge records were considered to be unacceptable for inclusion in the HCDN are given in table 5 (at the back of the report). The stations are specifically noted here as exceptions because the discharge records available for these stations have frequently been used to study hydrologic variation in relation to climatic conditions, an objective of the HCDN. The list of stations not included in the HCDN although they are found in either the National Water Conditions Streamflow Index set or in the Hydrologic Benchmark Network totals 44, including 39 that have been used in the monthly National Water Conditions Reports and 5 that have been part of the Hydrologic Benchmark Network.
The distribution of the HCDN station information among the 21 water-resource regions is summarized in table 6 (at the back of the report). The table gives not only the number of stations per region but also the total number of water years for which the data are acceptable by the HCDN criteria. The stations are further differentiated into those for which the records of daily mean discharge are acceptable by the HCDN criteria and those for which only monthly mean discharge values are acceptable, as well as into those stations for which the entire period of record is acceptable in contrast to only part of the available record. For most stations, the entire period of record is acceptable according to the HCDN criteria, even with regard to daily mean discharge values.
This data and documentation are from the USGS HCDN CD-ROM. The original text and tables are available on the CD-ROM.