Welcome to the Stream Network Stats WG

Spatial autocorrelation quantitatively represents the degree of statistical dependency between random variables using spatial relationships (Cressie 1993). It is an intrinsic characteristic of freshwater stream environments, where watersheds are nested within one another and sites are connected by stream flow through directed networks. Analyzing spatially correlated data requires the use of spatial statistical methodologies because the assumption of independence is violated, making many conventional statistical methods inappropriate (Cressie 1993). Spatial statistical methods have only recently been developed that represent the unique spatial configuration, longitudinal connectivity, flow volume, and flow direction found in freshwater ecosystems (Cressie et al. 2006, Ver Hoef et al. 2006, Peterson and Ver Hoef 2010, Ver Hoef and Peterson 2010). These methods provide significant potential advancements for ecological research and aquatic monitoring because spatial statistical models can be used to quantify patterns of spatial autocorrelation across multiple scales, to make predictions at unobserved sites with estimates of prediction uncertainty, and yield unbiased regression parameter estimates relating ecological variables to the environment (Cressie 1993). Our proposed working group will extend the capabilities of spatial statistical models for stream networks to include additional functionality available in traditional spatial statistical methods, so that a wider range of ecological and management questions can be fully addressed. To accomplish this goal, we will: 1) identify the most pressing needs in terms of analytical capabilities (i.e., what would be most useful for informing science and management) and begin developing these models, with possible extensions to include space-time models, generalized linear mixed models, computing for massive datasets, and others as identified by the working group, 2) assess the current state of software and functionality and determine whether it is sufficient to meet those needs, and develop new ones in conjunction with the previous objective, and 3) intensely analyze a single, nationally important, large, multivariate, stream dataset collected across the Northwestern (NW) United States (US) to gain ecological insights, evaluate methods, and demonstrate the new spatial statistical modeling capabilities.