RESULTS
SAR. The observed SAR in each of the census years is shown in a graph of local slope, z , versus log(N /S ) (Fig. 1), where number of individuals, N , and number of species, S , are measured at the same spatial scale at which z is calculated (Eq. 1). On each graph the METE scale-collapse prediction is shown as well. Note that because N scales linearly with area in a complete nested design, while S increases less rapidly than linearly with area, larger values of log(N/S ) correspond to larger areas. Fig. S1 shows the observed species-area curves used to calculate the observed slopes.
In 2014 there is good agreement between theory and observation. In the subsequent years the slope, z , declines faster than predicted asN /S increases. The deviation between theory and observation is larger at larger area.
The deviation of the observed slope from the METE prediction (Eq. 3) for each year is shown in Fig. 2a. From 2014 to 2018 there is a monotonic increase in that deviation, with a small decrease appearing in 2019.
SAD. The SAD in each of the census years is shown in a graph of log(n ) versus the rank order of the species (Fig. 3). The rank ordered observed abundances are shown as dots and the predicted rank abundance curve is shown as a thick line. The rank ordered results of 1,000 samples from the predicted SAD are shown as thin lines.
As with the SAR findings, agreement between the observed and predicted SADs is best in 2014 (Fig. 2b). In this case, however, we see constant, rather than increasing, disagreement in years 2015-2019.
Stress factor trends. We examined the correlation between our SAR and SAD errors with recruitment, mortality, net loss, number of individuals, and species richness (Fig. 4) over time and report the correlation coefficients of linear fits in Table 1. Fig. S2 shows the graphs of each of the error stress metric pairs along with their regression lines. The SAR error and net loss appeared to have the strongest relationship (Fig. 5). The SAR error and mortality had a moderately strong relationship, but no other even moderate correlations were found for the SAR error and none were found for the SAD error.