8195 - Laserfiche WebLink

Home Browse Search

1170 Issvsmli "' llwlroG ;ir:lllr rrrliuu scribed by Searcy (1960) and Alley and Burns (1983). Hydrologic simulation modeling or water budgeting techniques can also be used to synthesize hydrologic records for comparison by means of the IRA method (Linsley et al. 1982). Climatic differences between the pre- and post-impact time periods obviously have the potential to substan- tially influence the outcome of the IHA analysis. Various statistical techniques can be used to test for climatic dif- ferences in the hydrologic data to be compared. When the IRA analysis is based upon actual hydrologic mea- surements rather than estimates produced from models, a reference site or set of sites uninfluenced by the hu- man alterations being examined can be used as climatic controls (Alley & Burns 1983). For example, a stream gauge may exist upstream of a reservoir that is thought to have affected a study site. Analyses can establish a sta- tistical relationship between stream flows at the study site and at the upstream reference site using synchro- nous pre-dam data sets for the two sites. This relation- ship can then be used to estimate the stream flow condi- tions that would have occurred at the study site during the post-impact time period in the absence of the reser- voir. The IRA method can then be used to compare the _ measured post-impact conditions with estimated unaf- fected conditions for the same time period. Alterna- tively, a time series of observed impact versus control differences that spans the time of perturbation at the im- pact site can be used to assess hydrologic impacts (Palter et al. 1992); this is the basis for the before-after-control- impact-pairs design suggested by Stewart-Oaten et al. (1986). In the absence of an appropriate control site, process-based hydrologic models that simulate climatic and runoff processes or other climate analysis tech- niques can be used to create model data sets for compar- ison by means of the IRA method (Maheshwari et al. 1995). Case Study Application We selected the dam-altered Roanoke River in North Carolina to illustrate the application of the IRA method for assessing hydrologic alteration. Although we chose a surface water system for this case study, we emphasize the applicability of the method to analyses of ground wa- ter alterations as well. In choosing appropriate estimators of the central ten- dency (e.g., mean, median) and dispersion (e.g., vari- ance, coefficient of variation) of the hydrologic parame- ters, careful consideration needs to be given to the efficiency of the estimator and to the efficiency and as- sumptions of the statistical tests used to evaluate the dif- ference between time periods. The mean is the most efficient estimator of central tendency when the under- lying distribution is normal, and various Mike tests based on the mean are applicable even when assumptions of /?'rr'blr?r et rrl. the standard t test (e.g., normal distribution, equal vari- ances) are violated (Stewart-Oaten et al. 1992). We used the mean as an estimate of central tendency and the co- efficient of variation (CV) as an estimate of dispersion. However, we programmed the IRA software to enable nonparametric analysis as well. For each of the 32 hydrologic parameters the differ- ences between the pre- and post-impact time periods in both the mean and coefficient of variation are pre- sented, expressed as both a magnitude of difference and a deviation percentage (Table 2). These comparisons of means and coefficients of variation for each of the 32 pa- rameters comprise the 64 different Indicators of Hydro- logic Alteration. Approximate confidence limits are also estimated for the.difference between means and CV, re- spectively (Table 2), using standard formulae that are ap- proximately valid when distributions are not normal or changed (e.g., have unequal variances) between time periods (Snedecor & Cochran 1967; Stewart-Oaten et al. 1992). ..,,m Since 1913, The U.S. Geological Survey (USGS) has collected daily streamflow measurements at Roanoke Rapids on the Roanoke River. Flow values are recorded as cubic feet per second (cfs), but all results are con- verted here to cubic meters per second (cros). Dam im- pacts on the Roanoke River system began with the com- pletion of Philpott Lake on the Smith River (in the upper watershed) in August 1950 and were followed by con- struction of Ken Reservoir in 1950 for flood-control pur- poses. In 1955 Roanoke Rapids Lake was built down- stream of Kerr Reservoir for "run-of--the-river" hydropower generation purposes. Another reservoir, Gaston Lakc. was subsequently built between the locations of Roanoke Rapids Lake and Kerr Reservoir, but its influ- ence on flow regimes in the lower Roanoke below Roanoke Rapids Lake are believed to be inconsequential. The pre-impact data set has therefore been defined as 1913-1949, and the post-impact data set covers 1956- 1991. Typical pre- and post-dam annual hydrographs are presented in Fig. 1. The IHA results for the Roanoke River are given in Ta- ble 2 and illustrated in Figs. 4-7. The relative differences between means ranged from -73% (annual 1-day maxi- mum flow) to +232% (low pulse counts) for the individ- ual attributes, whereas the average absolute difference for the five groups of hydrologic characteristics ranged from 15% (Group 1 monthly means) to 88% (Group 4: frequency and duration of pulses). For individual at- tributes the relative difference in CV ranged from -60% (mean August flow) to +72% (mean April flow); the range for the five groups was 26% (Group. 4: frequency and duration of pulses) to 41% (Group 3: timing of ex- treme, events). The results of the IRA analysis for the Roanoke. River reflect the effects of Kerr Reservoir operations for the purposes of flood control and operations for generation Conservation Biology Volume Ill, No. 4. August 1996