KGS OFR 2003-55C: Assessment of Mean Annual Streamflow and Precipitation, State of Kansas.

by Brownie Wilson and Geoff Bohling, Kansas Geological Survey.

This open file report is one of five assessment reports under KGS OFR 2003-55 conducted by the Kansas Geological Survey for the Kansas Water Office and the State Water Plan. This report series serves in part as the contract completion of KWO Contract No. 03-121 entitled, "Assessment: Water Management and Water Conservation." This contract was supported by the State Water Plan Fund.

By 2015, meet minimum desirable streamflow at a frequency no less than the historical achievement for the individual sites at time of enactment.

By 2015, achieve sustainable yield management of Kansas surface and ground water sources, outside of the Ogallala aquifer and areas specifically exempt by regulation. Sustainable yield management would be a goal that sets water management criteria to ensure long term trends in water use will move as close as possible to stable ground water levels and maintenance of sufficient stream flows.

INTRODUCTION

In the FY 2003 Kansas Water Plan (KWP), the Kansas Water Authority approved objectives for the year 2010. The objectives were developed to define targets to quantify achievements as part of the implementation of the KWP. The above objective is included in the Water Management Section of the FY 2004 Kansas Water Plan.

In support of this objective, this assessment report reviews trends in mean annual streamflow that account for the influence of precipitation for selected USGS gaging stations from 1948 to 2001. A central premise to this assessment is that it associates the total precipitation that falls within the drainage areas of USGS gaging stations with the mean annual flow rates for each gaging station. This association is based on the potential that the total amount of precipitation would run off and flow past a gaging station. Influences from land use patterns, soil-infiltration rates, surface diversions, and a host of other parameters that affect runoff are not accounted for. This assessment simply establishes trends in mean annual flow that account for a measure of the variations in precipitation over time.

The Kansas Water Office (KWO) used KWP funds to contract with the Kansas Geological Survey to complete and finalize this assessment as part of the contract "Assessment: Water Management and Water Conservation", KWO contract no. 03-121.

ASSESSMENT DATA SETS

Mean annual streamflow values were obtained from the USGS via the National Water Information System (NWIS), located at http://waterdata.usgs.gov/nwis. The NWIS provides streamflow statistics for all USGS gaging stations in Kansas, including annual mean streamflow in cubic feet per second. To establish long-term trends in streamflow, only gaging stations that had at least 20 years of recently recorded streamflow values were selected, which limits the 310 gaging stations in Kansas that contain calculated mean flow values to 126 stations.

To associate the influence of precipitation with each gaging station, the drainage area for each station was identified from 14-Digit USGS Hydrologic Unit Basins (HUC14). The Natural Resources Conservation Service and USGS created the HUC14 basin delineations from 1:24,000 scale topographic maps. Each USGS gaging station was coded as to which HUC14 basin made up its drainage area.

Total monthly precipitation data was obtained from the National Climate Data Center (NCDC) at http://lwf.ncdc.noaa.gov/oa/ncdc.html. The number of stations in Kansas that contain usable monthly precipitation data varies from year to year and ranges from 188 to 301 stations with an average of 274 for the 1948 to 2001 time period. NCDC precipitation data became widely available in 1948, both in terms of a greater number of obtainable stations and their spatial distribution across the state.

The monthly NCDC precipitation values were used to calculate the total annual precipitation for each station. Although it is not common, each year there are times when a total monthly precipitation value was not recorded for a particular weather station. In consultation with Mary Knapp, State Climatologist at the KSU Weather Library, missing monthly values were replaced with averages from surrounding weather stations if a station was missing four or fewer monthly values during a calendar year. If a weather station was missing more than 4 months of precipitation values during a single calendar year, then that year of data was removed from the data set.

For each year from 1948 to 2001, the annual precipitation values were used to create continuous 1x1 kilometer gridded surfaces across the state. An example of the 2001 precipitation surface is shown in Figure 1. Each grid cell contains an interpolated precipitation value based on the surrounding precipitation values from the weather stations for that year. These interpolated surfaces were then used to summarize the total amount of annual precipitation that fell within each of the HUC14 drainage areas assigned to the selected USGS gaging stations.

Associating total annual precipitation to each USGS gaging station's drainage area raises some additional analysis issues. Since this assessment is limited to Kansas, only those gaging stations that have drainage areas lying completely within the state boundaries could be properly associated with precipitation patterns. Many Kansas gaging stations have drainage areas that capture precipitation and runoff from adjacent states where the HUC14 data sets have not been completed. Including these gaging stations would not properly account for the total annual precipitation that falls within the total drainage area identified for each gage.

A second limiting factor with the selection of gages to associate mean annual streamflow and precipitation comes from gaging stations that are located below large lakes or reservoirs. In these cases, streamflow can be heavily regulated based on the operation of the surface water impoundment and this greatly influences the recorded streamflow at the downstream gages. As such, all gages that are located below a major surface water reservoir or lake were also not included in this assessment.

Basing the gage selection requirements on stations that have at least 20 years of recently recoded streamflow, have the majority of their drainage areas lying completely within Kansas, and are not located below major surface water impoundments limits the usable USGS gaging stations to 51 stations (Figure 2).

With the mean annual flow and annual precipitation that fell within the drainage area for each gaging station that fit the selection requirements identified, a statistical regression model was fit to the data to establish a trend in mean annual streamflow that accounts for variations in precipitation over time. The regression analysis estimated one coefficient describing the influence of annual precipitation and another describing the remaining annual trend in mean annual streamflow apart from variations due to precipitation. Adjusted mean annual streamflow was then computed from the regression equation by replacing the actual annual precipitation values at each gaging station with the average annual precipitation (over the 1948-2001 interval) for that location. The R-squared value for the regression and a measure of the significance of the annual trend term were also recorded. Where possible, the streamflow trends were aggregated into KWO basin averages and the State of Kansas.

ASSESSMENT RESULTS

Of the 51 original gaging stations that had at least 20 years of mean annual flow calculated, the majority of their drainage areas located within the boundaries of Kansas, and were not located below a major surface water impoundment, 45 stations had a mean annual flow calculated every year from 1982 to 2001 and thus could be used to aggregate annual flow into a composite value for a 20-year period. The mean annual flow, average trends in flow statistically adjusted to account for annual precipitation based on the entire period of record for all the gages and based on only records from 1982 to 2001, and the average precipitation that fell within the drainage areas of the gages are shown in Figure 3.

Based on these 45 stations, the overall average trend in streamflow that has been statistically adjusted for precipitation shows a fractional decline in flow based on the entire period of record for all the gages. However, of the 45 gaging stations, only 15 or 33% had statistically significant annual trends, based on the all available flow record, after factoring out variations in precipitation. The average R-squared values for each gaging station was 0.59775, which means that about 60 percent of the variations in mean annual flow can be explained by a combination of variations in annual precipitation and a linear annual trend.

Calculating the trend in the adjusted streamflow rates using only records from 1982 to 2001 results in a slightly higher flow rate in terms of average annual cubic feet per second, but also has a fractional rise in flow over this time period. Only 5 stations, or roughly 10 percent of the 45 gaging stations, had statistically significant annual trends for this time period. The average R-squared value for 1982 to 2001 trend lines was 0.6468.

When reviewing the average trends in mean annual streamflow that have been statistically adjusted for variations in annual precipitation by KWO planning basin shows a general east-west division. In general, western and south-central basins in Kansas have declining trends in mean annual streamflow that have been statistically adjusted for precipitation, based on the all available flow records, while the eastern basins, with the exception of the Marais Des Cygnes, show rising trends in the statistically adjusted mean annual streamflow.

However, if the calculated flow trends are based on 1982 to 2001 records, the slope directions change in most cases. The western and south-central basins generally show rising trends in the statistically adjusted mean annual streamflow over the last 20 years with the Cimarron Basin being the exception. In comparison, several of the eastern Kansas basins have declining adjusted flow trends, based on 1982 to 2001 records, with the Missouri and Walnut basins being the exceptions.

However, regardless of either time period, the annual trend in the adjusted streamflow for most gages is not statistically significant. In addition, more stream gages are located in eastern Kansas, based on this assessment's selection criterion, which can skew regional averages.

Appendix A shows by KWO planning basin, the individual trends in mean annual streamflow that have been statistically adjusted for variations in annual precipitation, and when possible, the composite trend for stations where a mean annual streamflow value was established for every year from 1982 to 2001.

CONCLUSION

Using regression analyses, mean annual streamflow from 1948 to 2001 was statistically adjusted to account for variations in total annual precipitation for USGS gaging stations that had at least 20 years of recent flow data, had the majority of their drainage areas located within the boundaries of Kansas, and were not located below a major surface water impoundment. Based on gages that had a mean annual streamflow value for every year from 1982 to 2001, flow trends based on the entire period of record for all the gages indicate a fractional decline in the flow with the variations in annual precipitation accounted for. Of the 45 gaging stations used to establish this trend, only 15 or roughly 33 percent had a statistically significant trend. Basing the adjusted flow trends on flow records over the last 20 years, 1982 to 2001, the composite trend in adjusted flow shows a fractional rise. However, only 5 or roughly 10 percent of the gages used had statistically significant trends.

On a regional scale, most of the western and south-central KWO planning basins had declining trends in the statistically adjusted mean annual streamflow, based on the entire flow history, while most of the eastern Kansas basins had rising trends in the statistically adjusted mean annual streamflow. In most cases, the trends reversed directions when based on the regression equations over the last 20 years (1982 to 2001) of flow history. Again, in most cases, these trends could not be considered statistically significant.

Return to KGS OFR 2003-55.

Appendix A

This appendix provides plots of the mean annual streamflow, total accumulated annual precipitation, and the trends in mean annual streamflow statistically adjusted for precipitation by KWO planning basin. Plots are provided for each unique gaging station and when possible, the composite trend for stations that had a mean annual streamflow value every single year from 1982 to 2001. The R-squared value, which is a measure of the statistical relationship between streamflow and the combination of annual accumulated precipitation and time, is also provided. A measure of the statistical significance in the trend line can be seen with the "Probt" value where if Probt is less than 0.1 indicates a significant trend.

On the map, click on a gaging station in a basin to open a new browser window that shows the station plots and if available, basin composite plots.