Scale-dependence of persistence in precipitation records

Long-term persistence or Hurst-Kolmogorov behaviour has been identified in many hydroclimatic records(1). Such time series are intriguing because they are the hallmark of multi-scale dynamical processes that govern the system from which they arise(2). They are also highly relevant for water resource managers because these systems exhibit persistent, for example, multi-decadal, mean shifts or extremes clustering that must be included into any long-term drought management strategy. During recent years the growing number of palaeoclimatic reconstructions has allowed further investigation of the long-term statistical properties of climate(3,4) and an understanding of their implications for the observed change(5). Recently, the consistency of the proxy data for precipitation was strongly doubted, when their persistence property was compared to the corresponding estimates of instrumental records and model results(6,7). The latter suggest that droughts or extremely wet periods occur less frequently than depicted in the palaeoclimatic reconstructions. Here, we show how this could be the outcome of a varying scaling law and present some evidence supporting that proxy records can be reliable descriptors of the long-term precipitation variability.