Influence of temporal data aggregation on trend estimation for intense rainfall

Due to their influence on design guidelines, there is great interest in quantifying the possible impacts of climate change on extreme rainfall return levels. These studies can be affected by existing distortions in available rainfall historical series. The effect of the temporal aggregation (or time resolution), ta, of rainfall observations on the estimation of trends in annual maxima over the last 100 years is examined here. We have used long-term historical rainfall observations with various temporal aggregations, due to the progress of recording systems through time, at 10 representative meteorological stations located in an inland region of Central Italy. Series of annual maximum rainfall depths, Hd, for given durations, d, have been then derived. It is well known that Hd values derived from rainfall data characterized by every ta may involve underestimation errors, that for ta> ≈10 min can become important. Considering that all selected stations were installed in the first half of the twentieth century, each Hd series can be assumed as inhomogeneous since it contains values obtained by rainfall data with ta values ranging from fine (e.g. 1 min) to coarse (e.g. 24 h),thus with different levels of underestimation. By using a recently developed mathematical relation between average underestimation error and the ratio ta/d, we then correct each Hd value has been corrected through two different approaches, obtaining quasi-homogeneous series. Successively, commonly used climatic trend tests, including least-squares linear trend analysis, Mann-Kendall, Spearman's rank correlation, and Sen's method, have been applied to the inhomogeneous and quasi-homogeneous Hd series. The results show that the underestimation of Hd values with coarse ta plays a significant role in the analysis of the effects of climatic change on extreme rainfalls. Specifically, the correction of the Hd values can change the sign of the trend from positive to negative, mainly for series characterized by high probability to include Hd values with ta/d = 1.