Investigating daily categorized rainfall at global scale

An increasing number of regional studies suggest that the intensity and frequency of rainfall has been growing over time. Here, we investigate this assumption by performing a global scale analysis of trends in daily rainfall measurements. For this aim, we use the daily observations of the Global Historical Climate Network (GHCN) and reanalysis from ERA-Interim, available globally from 1979 to 2018 at 1°, 3-h resolution, which was aggregated at daily scale for consistency. To guarantee a minimum of reliability for the trend analysis, we identify a subsample of 2509 gauges from the GHCN dataset with more than 50 years of observations and perform a quality control of the records to remove outliers.

For each GHCN gauge and ERA grid cell, we define a set of rainfall thresholds based on percentiles of the empirical cumulative distribution function, defined as P_j with j = 0, 90, 92.5, 95, 97.5 and 100. From these thresholds, we define five classes of rainfall depth as [P_j, P_j+1]. Subsequently, we count the number of occurrences falling in each class for every year of observation, creating a time series of occurrences. We then investigate the statistical significance and sign of the trend for the time series through linear regression and non-parametric tests.

We find that, for both databases, the number of statistically significant positive trends for the highest classes exceeds the number of negative trends. This indicates that, in most locations, the heavy rainfall events have been significantly growing during time. Moreover, results show that the number of locations with positive trends is linked with the rainfall intensity. In other words, as the rainfall depth class increases, the number of stations (or grid cells) with statistically significant positive trends increases too.