|
<br />1134
<br />
<br />T, C, GRAND ET AL.
<br />
<br />daily mean, with mean daily variation over the entire base flow period equal to 6.5%, We ran simulations for the
<br />periods of July 25-30, August 25-30, October 25-30 and December 25-30, analysing invertebrate availability
<br />results from the last day of each of these simulation periods, These simulation dates were chosen to reflect the range
<br />of temperatures observed in Green River backwaters and to bracket the temperature at which invertebrate
<br />production is maximized in our model (Equation (3). Because the level of within-day flow fluctuation affects
<br />backwater temperature, temperature treatment values differed slightly among flow fluctuation scenarios and
<br />between backwaters. Hence, we treated mean daily temperature as a random effect in our statistical analysis of
<br />invertebrate availability,
<br />As suggested by the forms of Equation (3) through 8, I5 and M are expected to have strong but predictable effects
<br />on invertebrate availability in backwaters (d; is linearly related to I5 and exponentiaUy related to M regardless of the
<br />value of Ie, mean daily backwater temperature or the magnitude of within-day flow fluctuations), After confirming
<br />the predictability of these effects using single-parameter sensitivity analyses, we set I5 to 1000 invertebrates per m2
<br />and M to 0.00008 g,
<br />A mixed model analysis of vmiance was run on the simulation results to examine the significance of each of the
<br />three main effects (and any interactions) on predicted invertebrate availability, All p-values are those associated
<br />with Type III sums of squares as generated by the subset model, indicating that the magnitude of each effect is
<br />independent of its order of entry in the fun model (Wilkinson, 1990). Because results were similar for the two
<br />backwaters (each had the same tel11lsin the best-fit model, all p-values < 0.001, both R2 values -?: 0.96), we plot
<br />only results from the backwater five analysis (Figure 5).
<br />
<br />Flow,fluctuation simulations
<br />
<br />Our final set of simulations addressed the backwater habitat model's primary objectives: understanding the
<br />effects of within-day flow fluctuations on backwater temperature, wetted area, water exchange with the mainstem
<br />and availability of invertebrate food for fish. Five flow scenarios were created as described above (see Invertebrate
<br />Production Sensitivity Analysis methods), except that they included all days from July through 31 December 1994
<br />(see Figure 4), We ran the model for all six backwaters at each of the five flow scenarios, Thus, the experiment
<br />produced 30 output sets for comparison, Replication was not required as the habitat model is completely
<br />deterministic. Whenever backwater variables responded similarly to flow fluctuations, we present results from only
<br />backwaters 5 and 6 (representing the range of backwater sizes typically observed in the Green River study area),
<br />When results are noticeably variable among sites, results from aU six backwaters are provided.
<br />
<br />RESULTS
<br />
<br />Backwater tempera.ture model calibration
<br />
<br />Calibration results from the two backwaters analysed were in good agreement, indicating that the model captured
<br />the mechanisms driving temperatures fairly well. Although the two analyses did not rank any paranleter
<br />combinations identically, the same eight of 49 combinations appeared within the top ten rankings for both
<br />backwaters. Three of these eight combinations were ranked highly by both analyses (0.4 and 1,0, 0.5 and 1.5, 0.4
<br />and 1.25 for Ie and We, respectively), and, hence, were identified as the best fit parameter values. Visual inspection
<br />of the relationship between calibration parameter values and the calculated sums of squared differences revealed
<br />that none of these combinations was really any better than the others at predicting backwater temperature.
<br />Furthermore, all three pairs of parameter values resulted in a reasonable approximation of the true backwater
<br />temperature, reproducing the observed time lag relative to mainstem temperatures and the known range of diurnal
<br />temperature variation quite well. Hence, we used the highest mutually ranked set of parameter values (Ie::::: 0.4 and
<br />We = 1.0) for all subsequent analyses. The regressions of observed versus predicted temperatures had high R2
<br />values (0,86 and 0.92 for backwaters 5 and 6, respectively), slopes near 1.0 (0.84 and 0.95, respectively) and
<br />average errors of 20C or less (1.88 and 1.27, respectively), aU of which indicated a close fit of model predictions to
<br />observed data. An analysis of the residuals of these regressions indicated that, even though there was no overall
<br />correlation of error with mainst.em temperature, elTor in backwater temperature varied with mainstem temperature
<br />
<br />Copyright ((; 2006 John Wiley & Sons, Ltd,
<br />
<br />River Res, Applic, 22: 1125-1142 (2006)
<br />DOl: 10,L002/rra
<br />
|