Laserfiche WebLink
The absolute difference between simulated and observed diversions across all tinge <br />periods will be less than 10 percent of observed diversions. <br />T11e number of flooded and dry cells will be less than one percent of t11e model cells >IZ <br />t11e active model domain and will be randomly distributed. <br />The difference between simulated and mapped areas of groundwater <br />evapotranspiration will be less than 10 percent when examined at the end of the <br />transient simulations. <br />4.2 Objective Function <br />The automated calibration process will utilize a combination of optimization techniques that <br />niininuzes the residual between model-calculated values and observed values for a <br />corresponding set of model inputs (Doherty, 2004). Mininuzing the residual involves use of an <br />objective function that is comprised of a defined set of error terms for eac11 target along with a <br />weighting function for that target. There are three key parameters that constitute the <br />calibrations targets: head, streamflow and stream gain/loss. T11e Head is broken down further <br />into three categories (surveyed wells, non-surveyed wells with multiple measurements, and <br />non-surveyed wells with a single measurement) to allow different weigllt>11g factors to be <br />applied to each, as discussed in Section 4.3. T11is objective function takes the follow>11g form: <br />OF = wi(Hi) + wz(Hz) + wz(H,) + w3(S) + w~(G) <br />Where: <br />OF is fl1e Objective Fwlction <br />H~ is the sum of Head residuals of water level elevations at surveyed wells <br />H~ is t11e sum of head residuals of water level fluctuations atnon-surveyed wells that <br />have more than one measurement <br />Hs is the sum of head residuals of depth to water atnon-surveyed wells that have one <br />measurement <br />S is the sum of streamflow residuals at stream gaging stations <br />G is the sum of stream gain-loss residuals at reaches between gaging stations <br />wi, wz, war w~, and ws are the weight>lzg factors for each target <br />Each target may be subdivided into groups based on geographic location, data accuracy, <br />location along the mainstem or a tributary, or other factors. In these cases there will be a <br />separate target term and weighting factor i11 the objective function. The various measures must <br />also be scaled so that the different error terms are comparable. For example, streamflows are >11 <br />flow rates (cfs) that can vary by hundreds to thousands of units compared to water level <br />elevations that are in length units (feet) and typically vary by tenths of units. Tl1e various terms <br />i11 fl1e objective fwlction will be scaled to remove these unit dependencies. Further details on <br />objective functions are provide >IZ the PEST manual (Doherty 2004) and in Anderson and <br />Woessner (1992), among other references. <br />4.3 Weighting Factors <br />Weightuzg factors will be utilized to develop the goodness of fit statistics for the model <br />calibration by consideruzg the reliability of uzdividual measurements. The weighting factors <br />represent an estimate of the measurement error. These errors are an estimate of t11e underlying <br />accuracy of the measurement, not a measure of variation in fl1e data over time. For example, <br />P~ SPDSS T~8 2 Final TM 10-08-0t~.doc 17 <br />