152 



Fishery Bulletin 91 11). 1993 



> 76' 00' 



Figure 1 



Chesapeake Bay geography. The four fishery locations are indicated by x's 



peared that year. In 1990, contacts were primarily made 

 in the first half of the fishing season; we used that 

 data (Fig. 3) to describe when Spanish mackerel ap- 

 peared and became abundant that year. Catch-size in- 

 formation for 1989 and 1990 was related to daily sur- 

 face-water temperature records at Kiptopeke and at 

 the Chesapeake Bay Bridge Tunnel, provided by the 

 National Ocean Survey (Fig. 1). 



The accuracy of our estimates 

 of catch size varied from data con- 

 tact to contact, because the esti- 

 mates were not always simple 

 catch data. Most estimates were 

 quite accurate, particularly when 

 catches were zero or very small 

 (e.g., records being "none caught," 

 "few caught," etc.), when, as 

 usual, the fishermen were will- 

 ing to give a specific estimate ("n 

 boxes caught"), or when the catch 

 was stacked in boxes on pallets 

 for shipment and could be counted 

 by us. In some cases, our esti- 

 mates were verbal (e.g., records 

 being "larger than last week," 

 etc.). To compare estimates, the 

 size of the catch from each data 

 contact in 1989 and 1990 was 

 scored in the following catego- 

 ries: (0) no Spanish mackerel 

 caught, (1) <1 22.6 kg box offish 

 caught, (2) 1-5 boxes, (3) >5-10 

 boxes, (4) >10-20 boxes, and 

 (5) >20 boxes. Adjacent categories 

 may show some overlap, because 

 the original records are inexact. 

 However, these categories permit- 

 ted a distinct separation of zero 

 or small catches (categories 0,1) 

 from large catches (categories 

 3.4,5), but a less-distinct separa- 

 tion of intermediate-sized catches. 

 We feel the error of these esti- 

 mates is small and does not af- 

 fect the broad spatial and tempo- 

 ral patterns described. 



To evaluate temporal distribu- 

 tions, differences in monthly 

 catches in 1989 were tested for 

 each location using a Kruskal- 

 Wallis one-way nonparametric 

 analysis of variance (Table 1) af- 

 ter ranking the scores (SAS 

 1988). This was supported by Tukey's multiple com- 

 parisons tests (Table 2), applied to the ranked scores 

 to evaluate specific monthly differences. Similar pro- 

 cedures were followed to evaluate spatial distributions. 

 We interpret significance tests on spatial differences 

 with caution, because the number of nets varied among 

 locations and information does not exist to standard- 

 ize nets and nominal effort. We feel this has little 



