Fig 1.
Maps of average wet-day mean precipitation with units of mm/day (top) and relative linear trend in μ with units of %/decade (bottom) for the period 1950–2020, based on the ERA5 reanalysis.
Only statistically significant trends are shown in the bottom figure (α = 0.05). The wet-day mean precipitation μ is a statistic calculated for each year by selecting ‘wet days’ (days with more than 1 mm recorded precipitation) and estimating their average value, and is also referred to as ‘mean precipitation intensity’. Both mean and trends were estimated from these annual statistics (The coastline data ETOPO1 from https://www.ngdc.noaa.gov/mgg/global/, doi:10.7289/V5C8276M).
Fig 2.
Map of average wet-day frequency in terms of fractions [0, 1] (top) and relative linear trend in fw in terms of %/decade (bottom) for the period 1950–2020 based on ERA5 reanalysis.
Only statistically significant trends are shown in the bottom figure (α = 0.05). The wet-day frequency fw is a statistic calculated for each year by dividing the number of ‘wet days’ (days with more than 1 mm recorded precipitation) with the total number of days per year. Both mean and trends were estimated from these annual statistics (The coastline data ETOPO1 from https://www.ngdc.noaa.gov/mgg/global/, doi:10.7289/V5C8276M).
Fig 3.
Daily and annual total global rainfall in Gt/day estimated from the ERA5 reanalysis.
The blue curve represents the rainfall area between 50°S and 50°N. The thick curves show the annual mean Pt estimated from daily estimates. Dashed lines are best-fit linear trend fits from an ordinary linear regression on the annual mean estimates. Brief ‘negative’ spikes can be seen for January 1st in 1950 and 1979 when the two sets of integrations started and did not calculate a full 24-hr cycle of precipitation. These results indicate a long-term increase in the total global precipitation amount from 1440 Gt/day in 1950 to 1511 Gt/day in 2020.
Fig 4.
Daily and annual fractional global area with rainfall from the ERA5 reanalysis.
The blue curve represents the rainfall area between 50°S and 50°N. The thick curves show the annual mean Ap estimated from daily estimates. Dashed lines are best-fit linear trend fits from an ordinary linear regression on the annual mean estimates. Brief ‘negative’ spikes can be seen for January 1st in 1950 and 1979 when the two sets of integrations started and did not calculate a full 24-hr cycle of precipitation. These results indicate that the global surface area with 24-hr precipitation has decreased between 1950 and 2020 from 0.43 to 0.41.
Fig 5.
Daily and annual spatial mean precipitation intensity in mm/day from the ERA5 reanalysis estimated as the ration Pt/Ap and the results presented in Figs 3 and 4.
The annual spatial mean precipitation intensity has increased between 1950 and 2020 from 6.17 mm/day to 6.81 mm/day.
Fig 6.
Results from the multi-resolution decomposition of daily precipitation fields based on 2-D Haar wavelet transform (blue 1961–1990, pink 1991–2020).
The boxplots show the distributions of the wavelet energy for each spatial scale. The boxes show the ranges between the 25th and the 75th percentiles and the medians are shown as black lines in the boxes. The whiskers extend from the 1st to the 99th percentiles. The outliers are shown as points. The inset on the top-right corner emphasises the shift in the distribution of values between the two normal periods (blue 1961–1990, red 1991–2020). The envelopes show the range from the 1st to the 99th percentiles and the tick lines are the means of the distribution at each scale.
Table 1.
Summary of the comparison of the multi-resolution decomposition wavelet energies over the entire globe between 1961–1990 and 1991–2020.
q50(En2l) is the median of the energies at spatial scale l. The relative change is the ratio between the energy medians in 1991–2020 and in 1961–1990, multiplied by 100. The linear trend is the angular coefficient of the best-fitting line of the daily energies in the period 1991–2020. Note that only spatial scales greater than or equal to 0.25 degrees are shown.
Table 2.
Summary of the comparison of the multi-resolution decomposition wavelet energy percentages over the entire globe between 1961–1990 and 1991–2020.
q50(En2l) is the median of the energy percentages %En2(Pl) at spatial scale l. The other columns are the same as in Table 1, but for the energy percentages.
Fig 7.
Annual mean energy for the individual wavelet components (thick coloured curves) and a linear model best-fit based on the annual global mean temperature (dashed lines), also estimated from ERA5.
The different colours represent different wavelet components as shown in the legend. The best-fit was estimated for each individual wavelet component through an ordinary linear regression against the global mean temperature estimated from the ERA5 reanalysis. These results show that the wavelet evolution and trend, characterising the changes in precipitation patterns, is approximately matched by that of the global mean temperature.
Fig 8.
Hurricane Katrina in the Gulf of Mexico.
The ERA5 daily precipitation field on 2005–08-28 is shown in Panel b), units are mm/day and latitude and longitude are displayed. Panel a) shows the simulated precipitation for the same day adapted to the 1961–1990 climate based on multi-resolution analysis. Panel c) shows the simulated precipitation for the same day projected onto the 2021–2050 climate assuming a constant growth rate in the wavelet energies similar to those in the past. The procedure used to compute the inflation or deflation coefficients is included in the R-markdown scrip at the end of the S1 Appendix. The coastline data from http://thematicmapping.org/.