Introduction

Developing a monitoring program requires a chain of decisions, many complex, about a variety of issues. These decisions deal with questions ranging from policy and management (“What information is required to decide among the feasible actions?”) to the science of the system under study (“What is the best indicator of the ecological process of interest, and what is the best measurement protocol for that attribute?”) to purely statistical (“How much effort will it take to estimate a parameter with a precision of ±10% of its true value?”). This variety of challenging questions implicitly encourages compartmentalization and isolation in decision making. However, this compartmentalized approach can lead to suboptimal and even counterproductive decisions. Instead, a “whole-problem” perspective helps us see that all of the necessary decisions interact to influence the quality and cost of the monitoring program's ultimate product, information (Box 2.1).

A whole-problem perspective reveals that statistical considerations permeate the monitoring process. Information quality, a key measure of a monitoring program’s effectiveness, inherently entails consideration of the statistical concepts of bias and uncertainty (Box 2.2). Bias and uncertainty are controlled and accounted for through statistical methods of survey design, measurement, and analysis. Most importantly, effectively incorporating these statistical methods into the broader monitoring program requires statistical thinking to see how each source of bias and variability ultimately impacts information quality and cost and to identify where to allocate resources among the many program components to most efficiently improve information quality (Hooke 1980, Snee 1990, Chance 2002, Hoerl and Snee 2010).