FISHERY BULLETIN: VOL. 80, NO. 3 



in the former instance, however, was there an 

 attempt to validate the regression using indepen- 

 dent data. 



Thus, the value of correlative or regression 

 models of fisheries is twofold: First, significant 

 correlations can serve to guide research into 

 identifying the causes of annual variation in 

 catch; secondly, if validated, such models may 

 forecast harvest in cases where more detailed de- 

 terministic models cannot be developed for lack 

 of information. 



The issue of model validation is especially im- 

 portant. Correlational models which best regress 

 to the available data are often not the best predic- 

 tors (Saila et al. 1980). In an effort to overcome 

 this difficulty we have employed an amended 

 version of stepwise regression analysis which 

 does not rely solely on overall good ness-of- fit, but 

 rather identifies those variables most likely to 

 provide good predictions of data points not used 

 in the actual regression. 



To our knowledge multiple correlational mod- 

 els of the relationships between environmental 

 variables and commercial landings in the Mary- 

 land portion of Chesapeake Bay have not been 

 attempted in an algorithmic fashion. Because 

 most of the dominant species reproduce in Mary- 

 land waters, the influence of environmental 

 variation on harvest of these species may be par- 

 ticularly strong. Thus, we have developed multi- 

 variate regressions of the landings of major 

 commercial species, using as predictors those 

 environmental variables considered to have bio- 

 logical significance for the species being exam- 

 ined. Although measures of catch per unit effort 

 (CPUE) would have been preferable as indica- 

 tors of stock size, adequate effort data were not 

 available. The results provide valuable insight 

 into factors which may contribute to determining 

 the population dynamics of these species and 

 may also prove to be of value in establishing man- 

 agement practices. 



SPECIES ADDRESSED AND 

 DATA AVAILABLE 



Seven dominant species in Maryland landings 

 were selected for analysis. American oyster, 

 Crassostrea virginica; blue crab, Callinectes 

 sapidus; soft shell clam, Mya arenaria; and 

 striped bass, Morone saxatilis, were chosen be- 

 cause they are the four species which yield the 

 greatest dollar value to the Maryland economy. 

 Menhaden, Brevoortia tyrannus, and alewife, 



Alosa pseudoharengus and A. aestivalis arbi- 

 trarily combined, were selected because they 

 have been dominant in number of pounds har- 

 vested. The bluefish, Pomatomus saltatrix, was 

 included because its harvest has increased dra- 

 matically in recent years, and there was interest 

 in determining if this increase might be related 

 to environmental variation. 



A 33-yr record of annual catch data for 24 com- 

 mercial species was available from records 

 maintained by the Chesapeake Biological Lab- 

 oratory and the NO AA Current Fisheries Statis- 

 tics series. These records report the total Mary- 

 land landings (Chesapeake Bay and Atlantic 

 Ocean) for each year. The Chesapeake Bay por- 

 tion of the harvest heavily dominates the catches 

 of the chosen species (85% or more of total). Be- 

 cause of the difficulty in obtaining sufficient 

 information to separate Bay catch from the State 

 total, the total was assumed to be representative 

 of Chesapeake Bay. 



ANNUAL CHARACTERIZATIONS OF 

 ENVIRONMENTAL DATA 



The environmental variables for which long- 

 term records exist are water temperature, air 

 temperature, salinity, and precipitation. All four 

 have potential relevance to the levels of commer- 

 cial harvest. Cross correlative relationships 

 among these variables would be accounted for in 

 the stepwise multivariate regression procedure 

 employed in the analysis (discussed below). Daily 

 recordings of these variables exist for a period 

 exceeding 40 yr as taken from the Chesapeake 

 Biological Laboratory pier at the mouth of the 

 Patuxent estuary in Solomons, Md. Because this 

 location is central to the Maryland portion of 

 Chesapeake Bay, these data were assumed to be 

 characteristics of conditions in the bay as a 

 whole. Gaps in air temperature and precipita- 

 tion from 1960 onwards were filled by data taken 

 at the nearby Patuxent River Naval Air Test 

 Center in Lexington Park, Md. 



While catch figures represented the total land- 

 ings for a season, environmental data existed 

 with much finer temporal resolution. Our goal 

 was to pair each annual catch figure with a value 

 of an environmental property which might be 

 representative of the effect that variable had on 

 the stock during the year the daily readings were 

 accumulated. One straightforward way of char- 

 acterizing a year is to calculate the annual 

 average of the variable in question. The stock, 



612 



