How does the areal averaging influence the extremes? The context of gridded observation data sets

Abstract

Gridded data sets of observations are frequently used for the evaluation of climate extremes in climate models. However, it is necessary to understand how the behaviour of extremes is affected in such data sets. The aim of the paper is to analyse how the smoothing effect is related to the number of stations used for calculating the values in each gridcell. A very dense network of stations with highly correlated records of temperature and precipitation was used. The set of grid points was established and for each grid point all data from stations located in the circles with a radius of 75 km in the case of temperature and of 50 km in the case of precipitation were used. Two distributions were compared. The first one was built of the data averaged from all selected stations. The second one was built of all the data from the same stations. Apart from high correlation of averaged records, the effect of reducing extremes is strong. In the case of temperature in winter, the areal averaging causes the overestimation of temperature in the left tail of distribution for the percentiles below the tenth. In summer, the strongest effect is in the right tail and the temperature in underestimated. In the case of precipitation, the higher the number of stations averaged the lower the number of dry days. At the same time, the highest daily totals are underestimated in the highest 5 percentiles (equal and higher than the 95th percentile). All the effects increase with the increasing number of averaged records.

Keywords

• temperature extremes
• precipitation extremes
• station data sets
• gridded data set