WSP11764 - Laserfiche WebLink

Home Browse Search

<::> c.o CD CD DATA BASE DEVELOPMENT A series of computer procedures was developed to generate representative monthly values from existing data. These procedures were also used to check for errors in the data base and to estimate missing values. A complete guide tothe use ofthe component programs is given in appendix A. WATSTORE Retrieval All data used in this project were retrieved directly from the USGS WATSTORE data base by an automatic communication link between the USBR CYBER and USGS IBM computers. These data include mean daily streamflows, daily specific conductances, and periodic chemical analyses of the following: calcium, magnesium, sodium, chlo- ride. sulfate, bicarbonate, alkalinity, TDS (residue on evaporation at 180 oc). and specific conduct- ance. Chemical analyses are reported for time- weighted composite samples until September 1969 and for monthly discrete samples thereafter. Initial Analysis Potential errors in the periodic chemical data were identified by testing the differences between the observed and predicted values of: (1) total solute concentration, for which the predicted value was the sum of constituents and the observed value was the residue on evaporation (increased to account for lost bicarbonate); (2) specific conduc- tance, for which the predicted value was computed from equivalent ionic conductances; and (3) net charge, for which the predicted value was assumed to be zero. Obvious errors were checked and corrected, as described in appendix A. Possible errors in daily values were identified from outliers in the log-log regression of conductance on streamflow. Typographical errors in the WATSTORE data were confirmed and corrected by comparing them with values in USGS Water Supply Papers. Other obviously erroneous values were set to zero and treated as missing values in subsequent procedures. Changes made to WATSTORE data are documented in table 2 of appendixes B through P. Single missing solute concentrations in individual chemical analyses were estimated assuming a net charge of zero. When two or more values were missing from an analysis, no estimates were attempted; but the sum of six ions, necessary to compute mass fraction, was estimated from regression on evaporation residue TDS. The periodic chemical analyses, including missing value estimates, were then combined with daily values of streamflow and conductance to estimate solute concentrations for days samples were not taken, Generation of Monthly Mean Values Composite sample analyses were assigned to each day of the composite period; discrete sample analyses were assigned to the day of collection. Substitutions for missing daily values were gener- ated by interpolation in a hierarchical process in which streamflow was estimated first, followed by conductance, and finally solute concentrations. This sequence was employed because streamflow was used to weight the interpolation of conduc- tance, and conductance, in turn, was used to weight the interpolation of concentration. The general formula used was: (. k - ko ~. _, Xk = rkoXko + kn - ko (WknXkn - WkoXko~ Wk (1) where: X = Streamflow, conductance, or solute concentration W = A weighting factor k = Julian date within a period of missing values ko = Last Julian date with observed data before the period of missing values kn = First Julian date with observed data after the period of missing values Weighting factors in equation 1 are defined as follows: 1. Streamflow: Wk = 1, for all k 2. Conductance: Wk = l/Q/. where B is the slope of the log-log regression of daily con- ductance on daily streamflow (Q) for the entire period of unregulated flow. Wk= 1, ifB is not Significantly different from 0 (because of major flow regulation). 3. Solute concentrations: Wk = 1 ILk, where L is specific conductance on day k. 6