Non-Stationary Lags and Legacies in Ecosystem Flux Response to Antecedent Rainfall

Ecosystem function can be affected directly by climate, including by meteorological extremes, and also by sustained lags and legacies on timescales that surpass those of the weather events themselves. However, important gaps remain in our understanding of the influence and timescale of persistence of antecedent climate, known as environmental memory, on terrestrial carbon and water fluxes. Identifying the interactions between the lagged response to climate and the legacies to climate extremes, and whether the influence of memory varies through time, has not been fully explored. Here, we used a novel k-means clustering plus regression approach to examine timeseries of the sensitivity of terrestrial fluxes to antecedent precipitation at 65 eddy-covariance sites across a range of ecosystems. Quantifying the sensitivity to past precipitation and temperature reveals that the role of memory in ecosystem fluxes varies across sites and in time. When memory was accounted for in the model, relative improvement in modeled site flux r 2 compared to an instantaneous model varied between 0% and 57%, with mean of 12%. Our results show that vegetation type was a stronger predictor of memory importance than site aridity, implying a need to understand vegetation resilience conferred by physiological traits and acclimation capacity. The influence of memory varied strongly through time at many sites, with the role of different timescales exhibiting consistent non-stationarity. Our results demonstrate the importance of accounting for time-varying vegetation response to antecedent rainfall in land surface models to accurately predict future terrestrial fluxes.Plain Language Summary To predict how changes in future climate and weather extremes might impact terrestrial ecosystems, we need to understand the timescales of vegetation response to antecedent climate. Prevailing methods of exploration assume such responses to be stationary, that is constant through time. We present a novel approach that shows how the memory of plants to climate conditions change through time. We show that the carbon and water fluxes of vegetation can be significantly sensitive to antecedent rainfall and importantly that this sensitivity can vary substantially through time. Plant functional type is a key indicator of the role of memory to precipitation, while the response to antecedent rainfall is not determined by site aridity. Predicting future changes in the global carbon sink requires understanding how vegetation responds to climate across timescales. Identifying these timescales at which plants respond to climate is critically important as the climate changes, especially if extremes (e.g., heatwaves) become more frequent due to compounding effects.