EDWARDS and KLEIBER: NONRANDOMNESS ON LINE TRANSECT ESTIMATES OF DOLPHINS 



TREND ESTIMATES (^Vs^,) 



03 



C 

 OJ 

 Q 



Q) 



3 



CO 



> 



03 

 LU 



LU 



o 



Z 

 LU 

 QC 

 LU 



LU 

 O 

 DC 

 LU 



a. 



-100 



INCONSISTENT BIASES 



Lo (2. 1) NS 

 Hi (4, 5) NS 



Lo(2. 1)S 

 Hi (2. 1) NS 



r 



Lo(2. 1)NS 

 Hi (4, 5) S 



-I 1 r 



B C A B C A 



DATA TREATMENT 

 A Not Stratified ; Not Smoothed 

 B Stratified ; Not Smoothed 



C Stratified ; Smoothed 



Figure 5. — Comparisons of estimated versus actual change ("trend") in 

 school abundance from one sampling period to another. Changes in abun- 

 dance were estimated as the ratio of school abundance (estimated or ac- 

 tual) under one set of model conditions (SDj) to school abundance (esti- 

 mated or actual) under some other set of conditions (SDo). Lo and Hi refer 

 to actual abundance of schools (Lo = 1,250 schools. Hi = 2,500 schools). S 

 and NS refer to topography dynamics (S = topography sliding sideways at 

 1 knot (dynamic). NS = static topogi-aphy). Number in parentheses refer 

 to parameters generating topogi'aphies. Two and 4 refer to number of 

 peaks along each axis. One and 5 refer to peak gi-adient (1 = gradual slope, 

 5 = precipitous slope). Three change estimates, resulting from three dif- 

 ferent types of data stratification, were generated for each comparison; 

 A) no stratification before estimating school abundance, B) stratification of 

 r squares based on observed (raw) encounter rates per square, and C) 

 stratification of 1° squares based on smoothed encounter rates per 

 square. Comparisons are expressed as (1 - (Estimated change/actual 

 change))*100, so that differences appear as percentages. Differences are 

 when estimated changes equal actual changes. 



static environment because this condition led to 

 very concentrated clumping of schools within a 

 few 1° squares. Vessels concentrated most of 

 their effort in these few squares with very high 

 density. Overestimates were less pronounced in 

 the cases of a simple environment because here 

 the areas of higher density were much more dif- 

 fused and not so different from areas of low den- 



sity. The gradient of increasing density toward 

 the topographic peaks built up much more 

 slowly, so that vessels sampled many more 

 squares with relatively low density than had 

 been the case for the complex, static envu'on- 

 ment. The overestimate of abundance was rela- 

 tively lower for the case of the complex, dynamic 

 environment for essentially the same reason; the 



869 



