Few multiyear precipitation–reduction experiments find a shift in the productivity–precipitation relationship

Well-defined productivity-precipitation relationships of ecosystems are needed as benchmarks for the validation of land models used for future projections. The productivity-precipitation relationship may be studied in two ways: the spatial approach relates differences in productivity to those in precipitation among sites along a precipitation gradient (the spatial fit, with a steeper slope); the temporal approach relates interannual productivity changes to variation in precipitation within sites (the temporal fits, with flatter slopes). Precipitation-reduction experiments in natural ecosystems represent a complement to the fits, because they can reduce precipitation below the natural range and are thus well suited to study potential effects of climate drying. Here, we analyse the effects of dry treatments in eleven multiyear precipitation-manipulation experiments, focusing on changes in the temporal fit. We expected that structural changes in the dry treatments would occur in some experiments, thereby reducing the intercept of the temporal fit and displacing the productivity-precipitation relationship downward the spatial fit. The majority of experiments (72%) showed that dry treatments did not alter the temporal fit. This implies that current temporal fits are to be preferred over the spatial fit to benchmark land-model projections of productivity under future climate within the precipitation ranges covered by the experiments. Moreover, in two experiments, the intercept of the temporal fit unexpectedly increased due to mechanisms that reduced either water loss or nutrient loss. The expected decrease of the intercept was observed in only one experiment, and only when distinguishing between the late and the early phases of the experiment. This implies that we currently do not know at which precipitation-reduction level or at which experimental duration structural changes will start to alter ecosystem productivity. Our study highlights the need for experiments with multiple, including more extreme, dry treatments, to identify the precipitation boundaries within which the current temporal fits remain valid.