Model calibration

Model calibration involved comparing the simulated water levels to observed historic water levels for both steady-state and transient model runs. Figure 12 shows the calibration graph for observation well 217, which is the only provincial observation well in the valley. The graph displays the observed long-term monthly mean water elevation and modelled groundwater elevation after model calibration (1961-1999). Also shown are observed and simulated discharge hydrographs for the nearby Kettle River for the corresponding time period. There is a regular seasonal pattern, similar to the stage hydrograph of the Kettle River. The ground-water level in the well varied between 1 and 1.8 m over the period of record, whereas the river experienced stage fluctuation of 3 m. The mean

□ Observed groundwater elevation (longterm monthly means) Simulated groundwater elevation, 1961-1999

-Observed Kettle R. discharge 1971-2000

-Simulated Kettle R. discharge 1971-2000

511.00

510.50

in 510.00

509.25

□ Observed groundwater elevation (longterm monthly means) Simulated groundwater elevation, 1961-1999

-Observed Kettle R. discharge 1971-2000

-Simulated Kettle R. discharge 1971-2000

A?i\

/r\\ D ■•

/ t\\ \

/Mi D '

- V\ ■

□ 1 L

Model Time (Julian Day)

450 400

100 "I

50 0

Model Time (Julian Day)

Fig. 12. Mean hydrograph of water table elevation (total head) in observation well 217 in Grand Forks aquifer and water surface elevation of Kettle River approximately 400 m from well 217 (Scibek et al. 2007).

monthly water table elevation varied only by about 1 m, with a standard deviation of 0.2 m. The shape of the well hydrograph is similar to the Kettle River hydrograph, but the amplitudes of seasonal fluctuations are damped, which would be expected with increasing distance away from the river channel.

The groundwater model is reasonably well calibrated for transient flow at the location of observation well 217. Calibration residuals for static (steady-state) water levels across the entire aquifer had an acceptable error distribution, but residuals tended to be high near the model boundaries, which might be anticipated due to lack of physical data in these areas with which to constrain the conceptual model (Fig. 13a). Observation wells where residuals were very large (>5 m) were examined in detail, and compared to nearby observation wells, the possible range of river water levels if the well was adjacent to the river, ground surface elevation, and the expected water table surface in that area. The root mean square (RMS) error for model was 8% (Fig. 13b).

0 0

Post a comment