A Comparison of Moderate and Extreme ERA‐5 Daily Precipitation With Two Observational Data Sets

Abstract

Natural hazards related to extreme precipitation (river floods, flash floods, landslides, debris flows, and avalanches) cause casualties, damages to infrastructures and buildings, and have direct and indirect economic impacts (MunichRE, 2018). For infrastructure planning and prevention measures, information about rare events, that is, events that occur on average only once in a hundred years, is important. Such information can be obtained from precipitation data with statistical tools. Assessing the accuracy in high quantiles depends on spatial domain sizes and temporal availability. Different types of global precipitation data sets are available (Sun et al., 2018): global precipitation data sets are based on ground observations, satellite observations, combinations of ground observations and satellite observations, and on short-term weather model forecasts in reanalyzes data sets. Reanalyzes combine past observations with weather forecast models to reconstruct past weather. The main advantage of this type of precipitation data set is its regular spatial and temporal coverage. Reanalyzes ensure consistency of the precipitation data with the atmospheric conditions, which are important for weather and climate process studies. Here, we focus on ERA-5 precipitation (C3S, 2017). ERA-5 is the latest reanalysis product from the European Centre for Medium-Range Weather Forecasts (ECMWF). ERA-5 precipitation is computed in short-term forecast started from reanalysis initial conditions (Hennermann, 2020). The ERA-5 precipitation production process does not include precipitation observation inputs. Hence comparison with observational data makes sense, keeping in mind that observation data have (partly substantial) uncertainties as well (Kulie et al., 2010;Prein & Gobiet, 2017;Sun et al., 2018). ERA-5 precipitation has already been widely used since its release in 2018, but very few assessments of this data set have been conducted over large regions. Only precipitation over restricted areas and precipitation associated with specific type of events have been assessed (

A novel method to identify sub-seasonal clustering episodes of extreme precipitation events and their contributions to large accumulation periods

“…For example, taking the average over the N ep episodes (as discussed in Appendix C) is the same as setting all weights equal to 1 N ep . Sitarz (2013) discusses a mathematical approach for defining a scoring system in sports, with two intuitively appealing properties. First, the first place should be rewarded more points than the second, and the second more than the third, and so on.…”

“…The second property means that someone gaining a place (or a rank) should be rewarded more if the initial rank is higher, as improving at upper ranks is more challenging than improving at lower ranks. We then follow the method of the incentre of a convex cone (Sitarz, 2013) to construct our weighting scheme (see Appendix B for a detailed description). The same weight q i is assigned to the ith episode of each classification (Cl n and Cl acc ).…”

“…Appendix B: Calculation of the weights Sitarz (2013) assume two intuitive conditions for a scoring system. First, more points are assigned to the first place than to the second place, more to the second than to the third and so on.…”

“…Any choice of points in K would satisfy the two conditions for a scoring system; however we would like to have a unique and representative value. The option chosen by Sitarz (2013) is to look for the equivalent of a mean value: the incentre of K. Formally, the incentre is defined as an optimal solution of the following optimisation problem by Henrion and Seeger (2010):…”

“…where S x denotes the unit sphere, ∂K denotes the boundary of set K and dist denotes the distance in the Euclidean space. By using the calculation presented in the Appendix of Sitarz (2013), and dividing the points of the first place (x 1 ) to get the weights (q i ), we obtain Table D1. Descriptive statistics of the S cl distributions for all parameter combinations.…”