INTRODUCTION
Macroecological models and theories that attempt to predict patterns in the abundance, distribution and energetics of species often implicitly or explicitly assume that the ecosystem under study is static, and thus exhibits patterns that are unchanged from year to year. If macroecological patterns are observed to vary in functional form over time, as for example might occur under disturbance, then patterns of deviation from the static predictions can potentially provide useful information for improving models, identifying the drivers of change, and forecasting the future state of the system (Newman 2019; Harte & Newman 2014).
There is accumulating evidence from multi-site comparisons that ecosystems undergoing change in response to either anthropogenic or natural disturbance regimes do exhibit macroecological patterns that differ from those observed in relatively undisturbed sites in similar types of ecosystems (Newman et al. 2020; Rominger et al. 2015; Supp et al. 2012; Harte 2011; Carey et al. 2006; Kempton & Taylor 1974). These studies also reveal a weakness of the Maximum Entropy Theory of Ecology (METE), which is based on an inference procedure derived from information theory (Harte 2011). METE predicts patterns of distribution, abundance, and energetics using instantaneous values of a community’s state variables: total area, total abundance, total species richness, and total system metabolism. Numerous examples exist of the success of METE for systems in which the state variables are relatively static, as well as its failures in disturbed ecosystems in which state variables are changing in response to disturbance (Newman et al. 2020; Harte et al. 2017; Xiao et al. 2015; Harte et al. 2013; White et al. 2012; Harte et al. 2009; Harte et al. 2008).
In contrast to the cross-site comparisons, an actual change in the performance of METE on a single system, from success to failure over time, has not been reported before. Here we describe an alpine plant community experiencing a period of dry conditions. In southwestern North America, a region that includes our study site, the period 2000-2018 was the second driest 19-year period in the past 1200 years (Williams et al. 2020). The increasing frequency of hotter, drier years with low winter snowpack and earlier snowmelt over the last decade, coupled with a 10% loss of species, increasing mortality, and declining recruitment rates (see Results) during the study period along with the site’s extremely poor soil development, strongly suggest that this community is stressed and experiencing demographic decline.
Our analysis showed that the species-area relationship and the species-abundance distribution increasingly deviated from the predictions of METE over a period of six years from 2014 to 2019, in conformity with previous speculation (Newman et al. 2020; Harte and Newman 2014; Harte 2011) that increasing theory failure over time will be observed in ecosystems undergoing change. Our results suggest that the goodness of fit of ecological models may contain useful information on the state of the system that can be applied to forecasting and generalizing ecological models to more dynamic systems.