Estimates of local biodiversity change over time stand up to scrutiny

Two recent meta-analyses of local-scale biodiversity change over time, by the authors of the present paper, have been subject to a harsh critique. Here we use new data and analyses to respond to the main points of this critique. First, a central argument of the critique was that short-term time series lead to biased estimates of long-term biodiversity change. However, we show here that this conclusion was based entirely on two fundamental mistakes in the simulations used to support it. Second, we show that the critic's conclusion that there are negative relationships between temporal biodiversity change and study duration is entirely dependent on: (i) the unrealistic assumption that biodiversity change can be positive when study duration = 0; (ii) the use of only a subset of the available data; (iii) inclusion of a single outlier data point in a single study (out of 100 in this case); and/or (iv) a choice to use log ratios rather than slopes (when available) as the metric of temporal biodiversity change. In short, the evidence does not support the conclusion that studies of longer duration tend to find local diversity decline. Finally, the critique highlighted the obviously true fact that studies in the ecological literature represent a geographically biased sample of locations on Earth; this issue was noted in both original papers, and is relevant for all ecological data syntheses. This fact was used by the critics to cast doubt on our conclusion that, outside of areas converted to croplands or asphalt, the distribution of temporal biodiversity trends is centered on zero. As a scientific rule, future studies based on more or different data may cause us to modify our conclusion, but at present, alternative conclusions based on the geographic-bias argument rely entirely on speculation. In sum, the critique raises points of uncertainty typical of all ecological studies, but it falls far short of providing an evidence-based alternative interpretation for our results, which are now supported by syntheses of hundreds of individual data sets of temporal biodiversity change.