from the different fisheries or recovery sites. In general, CWT data do not 

 provide reliable estimates of the total recovery fraction (adtrecv) because 

 of low, uncertain sampling rates. They likely do provide, however, the best 

 available indication of the allocation of recovered fish among fisheries and 

 subbasin returns. 



Reliable estimates of survival parameters are very difficult to obtain. In 

 an ideal situation, modelers would have independent data on spawner den- 

 sities and smolt production, juvenile passage survival, and survival from 

 the estuary through adulthood. Such data do not exist. For most stocks, 

 the best available data will be in the form of a continuous record or time 

 series of counts of individuals at a given life stage, such as estimates of the 

 number of adults returning to the confluence of a major tributary. The 

 trick to calibrating the model is to use these data in combination with all 

 other sources of information to derive a set of parameter estimates that 

 provides the best fit to the data. 



A General Process Calibration depends in part on the version of juvenile production that is 

 for Calibration used. When using the more conventional model of juvenile survival (Ver- 

 sion 1) is used, we found the following process to be successful. First, we 

 obtained estimates of as many parameters as possible from the literature, 

 opinions of experts, other models, and CWT data. From this information, 

 we estimated all but the juvenile survival parameters and total adult re- 

 covery fraction. 



Second, we estimated the general shape and magnitude of the egg-to- 

 presmolt survival function. Direct estimates of smolt production resulting 

 from different spawning levels generally are not available. Even if stock- 

 recruit data are available, they usually are too scattered to allow precise 

 estimation of a survival function. Ultimately, considerable professional 

 judgment is needed to construct an approximate juvenile production func- 

 tion. The term "production function" refers to S*f(S), where S is the num- 

 ber of eggs, and f(S) is the egg-to-presmolt survival function. Since the pa- 

 rameters of the production function and survival function are the same, the 

 terms are interchangeable at times. 



The principal factors to consider in constructing a juvenile production 

 function are: (1) Does the function have the proper shape? (2) If the pro- 

 duction function has a maximum, does it correspond to the level of spawn- 

 ing adults that would be expected to produce the maximum number of 

 presmolts? (3) Do the relative differences in egg-to-presmolt survival at 

 low density and at high density seem reasonable? 



It is critical that the shape of the production function for juveniles be cor- 

 rect, since this function directly determines the compensatory capacity of 

 the population. Accurate estimates of the absolute number of presmolts 

 produced are less important because our principal interest is in the number 

 of adults produced. In the final step (described below), we scale the adtrecv 

 parameter so that the combined parameter set provides the best fit to the 

 available data, and compensates for some of the error we might have intro- 

 duced through estimates of other parameters. 



The final step in calibration is to estimate adtrecv, cvegsv, and cvadtrv. 

 For this purpose, a second ancillary model has been constructed. The cali- 

 bration model uses the parameters of the SLCM and deterministic analogs 

 of its relationships to predict elements of a historic time series using previ- 

 ous observations in the time series. The parameter adtrecv is introduced 



21 



