BOTSFORD ET AL.: CYCLIC COVARIATION IN CALIFORNIA FISHERIES 



peak at +1 and +2 yr that remains significant at 

 +1 yr after first-differencing. The same char- 

 acteristics are present, though less strong, when 

 the predecline and postdecline periods are con- 

 sidered individually. This lag of 1 or 2 yr between 

 northern California and central California salm- 

 on catch is commensurate with the shift in the 

 point of negative correlation for the predecline 

 situation (Fig. 7). The cross-correlation between 

 crab catches at the two locations does not show 

 significant results nor do higher correlations 

 persist after first-differencing. 



We can consider the implications of observed 

 correlations for three classes of possible mecha- 

 nisms. The first class of mechanisms involves 

 cyclic environmental factors which indepen- 

 dently drive the cycles in each species. Differ- 

 ences in life history between the two species 

 could be responsible for the phase difference be- 

 tween the two cyclic processes. In the second 

 class of mechanisms, one species is cyclic because 

 of environmental factors or an endogenous mech- 

 anism within the population and the second spe- 

 cies is cyclic because of some linkage to the first 

 species. The third possibility requires neither 

 species to be inherently cyclic. Rather, a biologi- 

 cal interaction between the two species results in 

 cyclic behavior in both (e.g., as in a classical 

 predator-prey system). The computed correla- 

 tion functions place constraints on specific tim- 

 ing of the mechanisms in each of these classes. 

 These can be compared with known life history 

 characteristics and suspected interactions to 

 eliminate some possibilities. 



The life histories of the two species follow simi- 

 lar temporal patterns. The eggs of Dungeness 

 crab and fall-run king salmon hatch in midwin- 

 ter. The pelagic crab larvae drift for several 

 months then settle as first crabs during April 

 and May. The salmon fry remain in streams for 

 several months then enter the ocean in late 

 spring through summer. Some adult crabs enter 

 the fishery 3 yr after hatching but most are 

 caught at age 4. King salmon first enter the fish- 

 ery about 2% yr after hatching, and most are 

 caught at 3% yr and some are caught 4% yr after 

 hatching. 



That there was no significant positive cross- 

 correlation between the two catch records at 

 lag indicates that a cyclic environmental factor 

 which drives the cycle of one species through an 

 effect on one age-class cannot also affect the same 

 age-class of the other species. This implies, for 

 example, no direct interaction between the age 



classes of the two species. This is to be expected 

 since most crab larvae have settled before salm- 

 on smolts begin entering the nearshore pelagic 

 environment. 



The positive cross-correlation, which indicates 

 that good (bad) salmon catches are followed 3 to 5 

 yr later by good (bad) crab catches, may be a re- 

 sult of a cyclic environmental factor which affects 

 early salmon survival in 1 yr and similarly af- 

 fects larval crab survival 3 to 5 yr later. This 

 environmental factor need not affect exactly the 

 same age class in both species. For example, a 

 positive effect on growth and survival of matur- 

 ing salmon in their penultimate year at sea and a 

 simultaneous positive effect on ovary develop- 

 ment in female crab could increase salmon catch 

 in the following year and crab catch 4 to 6 yr later 

 through increased egg production in the follow- 

 ing year. 



Salmon have been observed to prey heavily on 

 pelagic crab megalopae (Orcutt 1978). If this 

 mechanism is considered as increased larval 

 crab mortality when salmon are abundant, it 

 does not fit the conditions implied by the correla- 

 tions. However, if abundant crab megalopae lead 

 to a good crab year class while increasing the 

 growth and survival of adult salmon, then the 

 observed cross-correlation would result. A mech- 

 anism by which salmon were more available to 

 the fishery during years of high crab larval 

 abundance could also cause the observed covari- 

 ation. 



The negative cross-correlation indicates that 

 good (bad) crab catches are followed 1 to 2 yr 

 later by bad (good) salmon catches. That this 

 does not persist following first-differencing is 

 commensurate with it being a result of fluctua- 

 tion in an auto-correlated series (e.g., abundance) 

 rather than a less auto-correlated series (e.g., re- 

 cruitment). The mechanism which is a priori the 

 most likely cause of this observation was the one 

 investigated in detail in this paper: the cycle in 

 salmon catch is actually a cycle in fishing effort 

 for salmon and that this cycle is driven by the 

 highly cyclic crab catch. 



The conclusion resulting from analyses of the 

 hypothesis of behavioral switching by fishermen 

 is that an immediate response to crab abundance 

 is not a likely cause. The strongest evidence for 

 this was the comparison of the cyclic nature of 

 early with late season salmon catch. The other 

 two analyses are less powerful because both the 

 availability of salmon and the variation in wea- 

 ther conducive to salmon fishing introduce vari- 



799 



