ABSTRACT
Environmental and ecological data occur in space and time, and precisely where and when they occur may affect what can be inferred from the data. For example, if precipitation is measured at gauges widely separated in space, average precipitation amounts over large regions may be estimated quite well but little may be inferred about the local or small-scale spatial variability of those amounts. This chapter describes sampling design for environmental monitoring within the following framework for spatio-temporal data. Model-based sampling design, as its name suggests, applies to situations in which a statistical model is assumed for Z(s). Sometimes understanding the second-order spatial dependence, or covariance, structure of a model is the primary inferential objective. Even when it is not, and estimation of mean parameters or spatial prediction of unobserved values of Z is the primary inferential goal, the success of these other goals may depend, in part, on how well the covariance parameters are estimated.
