Predicting the occurrence of persistent hotspots in ecosystem variables

Ecological resources and services (e.g. organisms, nutrient cycling) are distributed heterogeneously across landscapes. While spatial variation has been studied extensively, the pattern of hotspots and coolspots persisting over time – called persistent spatial variation (PSV) – has not. Yet this pattern imparts key information to managers about whether resources will be found consistently in certain locations or vary unpredictably. Anticipating whether an ecosystem variable will display PSV is thus a valuable prospect. We tested the ability of attributes of variables (e.g. niche breadth, abundance, temporal scale) to predict the occurrence of PSV. Using a new measure of PSV based on the F‐value of analysis of variance, we were able to 1) decompose the pattern of persistent hotspots into spatial and temporal components – ‘spatial variation’ of site mean values and ‘stability’ of time series at each site – and 2) identify predictors of these patterns in temperate lakes and tropical coastal rock pools. We found PSV to be highly predictable ( R2 = up to 0.80) from an estimate of stability taken at a single site, as well as from other factors related to stability. These factors included whether the variable was environmental (stable, slow) or was an aggregate of other variables (stabilized by statistical averaging). Species properties like niche position and abundance were modest predictors because they correlated with PSV components of site occupancy, spatial variation and stability. We conclude that PSV and the distribution of resources in space and time might be predicted from simple temporal indicators (e.g. stability at a single location) when data are scarce.