Study of shrimp distributions by means of locally weighted regression

The distribution of shrimp is modeled through the use of locally weighted r egression based on survey data. Multivariate local regression was successfu lly employed to generate distributional surfaces and to construct approxima te F-tests for three shrimp species found in waters off Kuwait: Penaeus sem isulcatus, Metapenaeus affinis, and Parapenaeopsis stylifera. Distributiona l differences among the three species can be clearly visualized with local regression maps. Penaeus semisulcatus is distributed throughout the survey waters; M. affinis and P. stylifera are primarily concentrated in waters no rthward from 29.1 degreesN latitude. Because Kuwait Bay has a high density of small shrimp representing all three species, it may be regarded as a nur sery ground. A second nursery ground for P. semisulcatus was possibly ident ified in the southern coastal waters of Kuwait.