Adaptive sampling of spatial point processes

Graduation date: 1982Adaptive sampling designs are those in which\ud observations made during a survey may be taken into\ud account when selecting the sites for further observations.\ud Increases in survey precision may be achieved through\ud adaptive procedures. The most efficient designs\ud incorporate flexible sample size along with sequential\ud selection of sites. In surveys of animal or plant\ud populations, adaptive sampling techniques may take\ud advantage of spatial tendencies associated with flocking,\ud schooling, or patchiness.\ud The framework of "sampling processes" is introduced\ud to deal conveniently with sampling theory and methods for\ud spatial point processes. A sampling process is a\ud stochastic process that depends on the detectability\ud functions associated with the survey methods as well as\ud on the structure of the spatial point process under study.\ud A realization of a sampling process consists of the\ud number of points of the point process that an observer\ud actually detects from different locations in the space.\ud Sampling theory and methods for spatial point processes\ud thus become largely a matter of studying the structure\ud of sampling processes. Many survey techniques for\ud the assessmemt of plant and animal abundance may be\ud analyzed within this framework.\ud Practical adaptive sampling strategies pursued\ud in this thesis include both design-unbiased and modelunbiased\ud adaptive sampling strategies The design-unbiased\ud strategy has the advantage that, even if assumptions\ud about the population being surveyed are incorrect,\ud the estimate is still unbiased and we still have an\ud unbiased estimate of mean square error. The modelunbiased\ud strategy relies for its unbiasedness on\ud certain assumptions made about the population model.\ud The key assumptions required may in fact be realistic\ud in many spatial survey situations with properly located\ud sampling sites. The two classes of adaptive strategies\ud are applied in modified form to bird survey and shrimp\ud survey methods.\ud The design-unbiased adaptive strategies used on Alaska shrimp surveys are estimated to give increases\ud in precision of 24% and 30% over the comparable\ud nonadaptive procedures of the same expected sampling\ud effort. Higher gains of 37% and 98% are computed\ud to be theoretically attainable by refining the adaptive\ud procedures