only focused on aggregate variability, leaving a conceptual gap. Here, we address this gap with a novel framework for quantifying the aggregate and compositional variability of communities and ecosystems through space and time. We demonstrate that the compositional variability of a metacommunity depends on the degree of spatial synchrony in compositional trajectories among local communities. We then provide a conceptual framework in which compositional variability of 1) the metacommunity through time and 2) among local communities combine into four archetype scenarios: spatial stasis (low/low), spatial synchrony (high/low), spatial asynchrony (high/high) and spatial compensation (low/high). We illustrate this framework based on numerical examples and a case study of a macroalgal metacommunity in which low spatial synchrony reduced variability in aggregate biomass at the metacommunity scale, while masking high spatial synchrony in compositional trajectories among local communities. Finally, we discuss the role of dispersal, environmental heterogeneity, species interactions and suggest future avenues. We believe this framework will be helpful for considering both aspects of variability simultaneously, which is important to better understand ecological stability in natural and complex landscapes in response to environmental changes.

No abstract

Nine of the 12 sampled census tracts were previously surveyed in 2011 to build a longitudinal database on human and environment interactions within the Phoenix (AZ) region.Part of the sampling frame (n=188) included households in census tracts that responded to the 2011 survey. An additional 1,212 addresses were obtained from the Marketing Systems Group, which provides an extensive list of addresses from the U.S. Postal Service. This amounted to 101 new addresses randomly selected from each census tract. An additional 14 were drawn as backup addresses (i.e., to swap with bad addresses or any duplicates from the previous survey sample).The Survey Center at the University of Wisconsin administered the survey from June through August of 2017. A total of 1,400 surveys were mailed. Surveys and reminders were delivered through the postal service in a series of four mailings. The first wave included the printed questionnaire and a self-addressed, postage-paid envelope, along with a $5 incentive and a prepaid postcard to request a Spanish version. The second mailing was a reminder postcard sent a week later. The third and fourth mailings included the questionnaire and return envelope, which were sent 2 to 3 weeks after the previous mailings. The response rate was 39%, which yielded a sample of 496. This rate is based on the American Association for Public Opinion Research's Response Rate 2, which is calculated as the number of completed and partial questionnaires divided by the total N minus both vacant and undeliverable addresses. Only 80 respondents who completed the 2011 survey also completed the 2017 survey (response rate of 43% based on the American Association for Public Opinion Research's Response Rate 2). Modeling ApproachOrdinary least squares (OLS) regression was an appropriate specification for our analysis, given the relatively normal distribution of the life satisfaction variable and the relatively consistent linear relationships between life satisfaction and our continuous explanatory and control variables. The basic OLS model is specified as follows:where Y is the life satisfaction of an individual i in neighborhood n; β 0 is the intercept; β 1n is the effect of the objective neighborhood built environment (NBE) factors; β 2in is the effect of the perceived NBE factors, which are unique to individual i; β 3in is the effect of the linking factors (captured at the neighborhood level), including perceptions of neighborhood social capital, identification, disorder, and opportunities for nature engagement and exercise, which also vary by individual i; β 4i are the effects of the controlled individual inherited and acquired characteristics; β 5n are the effects of the controlled neighborhood demographic characteristics; and e in is the error term. Accounting for Neighborhood ClusteringRespondents were sparsely spread out among the 57 census block groups (average of 7 persons per neighborhood; 12 neighborhoods had only 1 person and 1 neighborhood had 56 persons). The intraclass correlation, or proportion of varia...