Laserfiche WebLink
70 <br />60 <br />50 <br />> 40 <br />U <br />a <br />30 <br />o <br />i " 20 <br />10 <br />Range of Error in Water Levels (meters) <br />Figure 6. Frequency distribution of errors in model - simulated wate <br />levels (simulated minus measured) for individual well locations i n <br />1972, 1982, and 1995. <br />Table 1 <br />Average Observed and Simulated Salinity <br />in the Alluvial Aquifer <br />Observed Observed Salinity Simulated Salinity <br />Salinity Spatially Spatially <br />Arithmetic WeightedU Weightedi Percent <br />Date Average (mg/L) Average (mg/L) Average (mg/L) Error <br />February 1972 2262 (25) 2177 (25) 2140 (554) -1.7 <br />February 1982 2269 (24) 2146 (24) 2370 (554) 10.4 <br />February 1995 1779(25) 1983(25) 1770(554) -10.7 <br />( Spatially weighted average computed by the Theissen method with all available data <br />for each event. <br />2 Number in parentheses indicates the number of samples /finite - difference cells used in <br />the averaging procedure. <br />simulated water levels. The frequency of errors in <br />the model - calculated water levels for the 12 individual wells is <br />shown in Figure 6. For 11 of the 12 wells, the errors were relatively <br />small (0 m to 1.2 m) and reasonably unbiased. Model simulations <br />for one of the 12 wells were consistently poor and under pre- <br />dicted, thus making the overall frequency distribution appear neg- <br />atively skewed (Figure 6). <br />The simulated water table at individual locations was in good <br />agreement with measured water levels at individual observation <br />wells through time. Figure 7 shows three snapshots of simulated ver- <br />sus measured water levels for the months of February 1972 (Figure <br />7a), March 1985 (Figure 7b), and March 1993 (Figure 7c). These <br />snapshots are representative of most months throughout the 24 <br />year simulation period. For most wells, there is no consistent bias <br />in the flow calibrations; simulated water levels fluctuate randomly <br />about measured elevations in Figure 7. Small errors in simulated ver- <br />sus measured water levels at individual wells should not have a sig- <br />nificant effect on the calculated regional hydraulic gradients and <br />velocities, which are factors that strongly control solute transport. <br />Dissolved solids concentration data were measured in water <br />samples collected from 25 to 26 wells in February of 1971, 1972, <br />1982 and 1995. An attempt was made to sample the same well net- <br />work that was used throughout the study period. In several instances, <br />we were uwrrvyeu or in a CGnulriuii irrat prevented Sarirpuug. in <br />these instances, new wells that were located near the old wells were <br />sampled. The February 1971 data were the basis for defining ini- <br />tial conditions for the simulation. The simulated salinity (represented <br />as a spatially weighted average computed by the Theissen method) <br />was within 11% of measured values for 1972, 1982, and 1995 <br />(Table 1). A comparison of the measured and simulated spatial dis- <br />tributions of salinity indicates that there is reasonably good agree- <br />ment between the general salinity patterns with some exceptions <br />(Figure 8). In 1972 and 1982, a ridge of high salinity ( >3000 <br />mg/L) existed in the middle of the area (Figure 8a and b). Although <br />this general pattern was well simulated, the model overestimated <br />salinity in some areas for 1982 (Figure 8b). The salinity was found <br />to vary substantially over a short distance in wells sampled during <br />the study period. This is due to differences in pumping rates across <br />the study area, and temporal differences in the salinity of the water <br />leaking into the aquifer from the irrigation canal. The variability in <br />salinity over a short distance and the relatively small number of wells <br />sampled (25 to 26) might account for some of the discrepancy <br />between simulated and measured salinity in areas where the model <br />tended to overestimate. In February 1995, measured salinity had <br />decreased substantially across the study area (Figure 8c), probably <br />in response to decreased ground water pumping and increased irri- <br />gation from surface water sources. The February 1995 salinity <br />pattern was reasonably well simulated by the model (Figure 8c), <br />although the model tended to underestimate salinity at the individual <br />well locations. The frequency distribution of the error in model -cal- <br />culated salinity for individual wells sampled in 1972, 1982, and 1995 <br />is shown in Figure 9. Although some of the errors are rattier large, <br />the errors are reasonably distributed about zero, which indicates rel- <br />atively unbiased estimates. Based on the results of the model cal- <br />ibration, the model satisfactorily reproduces historic spatial patterns <br />of ground water levels and salinity, although prediction errors for <br />some individual well locations are relatively large. Therefore, the <br />obvious strength of the model is to simulate general water level and <br />salinity trends or patterns, which is well suited to its intended use <br />in the study. <br />80 <br />N "o O a l+ -I In rn m cn C., In -I n � O "q N <br />vi i i I? CV CV r+ O O C C-" N m 'r 7 vi <br />Figure 7. Measured and simulated water levels in the alluvial aquifer, <br />February 1972, March 1985, and March 1993. Dark surface is simu- <br />lated water table. Spheres are measured water levels for individual <br />wells. Where the spheres intersect the dark surface, the calculated and <br />measured water levels are the same. <br />