The response of precipitation characteristics to global warming from global 1 and regional climate projections 2

Abstract. We revisit the issue of the response of the precipitation characteristics to global warming based on analyses of global and regional climate model projections for the 21st century. The prevailing response we identify can be summarized as follows: increase in the intensity of precipitation events and extremes, with the occurrence of events of unprecedented magnitude, i.e. magnitude not found in present day climate; decrease in the number of light precipitation events and in wet spell lengths; increase in the number of dry days and dry spell lengths. This response, which is mostly consistent across the models we analized, is tied to the difference between precipitation intensity responding to increases in local humidity conditions, especially for heavy and extreme events, and mean precipitation responding to slower increases in global evaporation. These changes in hydroclimatic characteristics have multiple and important impacts on the Earth's hydrologic cycle and on a variety of sectors, and as examples we investigate effects on the potential stress due to increases in dry and wet extremes, changes in precipitation interannual variability and changes in potential predictability of precipitation events. We also stress how the understanding of the hydroclimatic response to global warming can shed important insights into the fundamental behavior of precipitation processes, most noticeably tropical convection.



Mean precipitation changes 97
In general, as a result of the warming of the oceans and land, global surface 98 evaporation increases with increasing GHG forcing.This increase mostly lies in the range of 99 1-2 % per degree of surface global warming (%/DGW; Trenberth et al. 2007).As a 100 consequence, global mean precipitation also tends to increase roughly by the same amount.

212
iii) An increase (ratio > 1) in the frequency of events for intensities higher than the 213 threshold mentioned above.The relative increase in frequency grows with the intensity of the 214 events, and it is thus maximum for the highest intensity events, an indication of a non linear 215 response of the precipitation intensity to warmer conditions.Note that, because of the 216 logarithmic frequency scale, the absolute increase in the number of high intensity events is 217 relatively low.

218
iv) The occurrence in the future time slices of events with intensity well beyond the 219 maximum found in the reference period.These are illustrated by the prescribed value of 10 220 when events occurred for a given bin in the future time slice, but not in the reference one.One 221 could thus interpret these as occurrences of "unprecedented" events.222 v) All the features i)-iv) tend to amplify as the time slice is further into the future, i.e.

223
as the level of warming increases, and are generally more pronounced over tropical than 224 extratropical areas (and over land than ocean regions, which we did not show for brevity).

225
Although the results in Figures 3 and 4 are obtained from one model, they are 226 qualitatively consistent with those we found for other CMIP5 GCMs (not shown for brevity).precipitation change only include wet days, i.e. days with precipitation greater than 1 mm/day.320 321 Also in these calculations, the increase in global mean precipitation is in the range of 322 1-2 %/DGW except for the GFDL experiment, which shows a very small increase (indicating 323 that in this model most of the precipitation increase occurs in the polar regions).In all cases 324 except for MIROC the increase in global SDII is greater than the increase in mean  precipitation, resulting in a decrease of the number of rainy days.The changes in the 95th, 326 99th and 99.9th percentile are maximum for the most extreme percentiles, showing that the 327 main contribution to the response of Figure 7 is due to the highest intensity events, i.e. above 328 the 99th and 99.9th percentiles, whose response becomes increasingly closer to the Cl-Cl one.

329
In fact, the increase in 95th percentile for the ensemble model average is lower than the 330 increase in SDII, and this is because in some models the threshold intensity in Figures 3-6, where the sign of the change turns from negative to positive, lies beyond the 95th percentile.

332
When only land areas between 60 o S and 60 o N are taken into account (bottom panel in Table 333 1), the changes are generally in line with the global ones, except for the CNRM model.Over 334 land areas we also find changes in the highest percentiles of magnitude mostly greater than 335 over the globe (and thus over oceans).

336
We can thus conclude that the shift to a regime of more intense but less frequent 337 events in warmer conditions is due to the fact that precipitation intensity, especially for 338 intense events (beyond the 95th percentile), responds at the local level primarily to the Cl-Cl-339 driven increase of water vapor amounts, while mean precipitation responds to a slower 340 evaporation process, driving a decrease in precipitation frequency.Noticeably, the MIROC 341 experiment does not appear to follow this response, i.e. in this model the increase in mean 342 precipitation appears to be driven by an increase in the number of light precipitation events.

343
While the data of Table 1 provide a diagnostic explanation of the hydroclimatic 344 response of Figure 7, it has also been suggested by very high resolution convection-permitting

33
One of the greatest concerns regarding the effects of climate change on human 34 societies and natural ecosystems is the response of the Earth's hydrologic cycle to global 35 warming.In fact, by affecting the surface energy budget, greenhouse gas (GHG) induced 36 warming, along with related feedback processes (e.g. the water vapor, ice albedo and cloud Earth Syst.Dynam.Discuss., https://doi.org/10.5194/esd-2018-64Manuscript under review for journal Earth Syst.Dynam.Discussion started: 12 September 2018 c Author(s) 2018.CC BY 4.0 License.Earth Syst.Dynam.Discuss., https://doi.org/10.5194/esd-2018-64Manuscript under review for journal Earth Syst.Dynam.Discussion started: 12 September 2018 c Author(s) 2018.CC BY 4.0 License.data.In this regard, we focus on the high end RCP8.5 scenario, in which the ensemble mean 85 global temperature increase by 2100 is about 4 o C (+/-1 o C) compared to late 20th century 86 temperatures (IPCC 2013), stressing that results for lower GHG scenarios are qualitatively 87 similar to those found here but of smaller magnitude (not shown for brevity).88 In the next sections we first summarize the changes in mean precipitation fields in our 89 ensemble of model projections, and then explore the response of different precipitation 90 characteristics, trying specifically to identify robust responses.After having identified the 91 dominant hydroclimatic responses, we discuss examples of their impact on different quantities 92 of relevance for socio-economic impacts, and specifically the potential stress associated with 93 changes in dry and wet extreme events, precipitation interannual variability and predictability 94 of precipitation events.

101Figure 1 .
Figure 1.Normalized mean global precipitation from 1981 to 2100 in the 10 CMIP5 GCMs simulation for the 104 118

Figure 3 .
Figure 3. Small right panel: Probability density function (PDF) defined as the normalized frequency of

Figure 4 .
Figure 4. Same as Figure 3 but for extra-tropical land areas.

227
We also carried out the same type of analysis for a high resolution RCM projection (12 km 228 grid spacing, RCP8.5 scenario) conducted with the RegCM4 model (Giorgi et al. 2012) over 229 the Mediterranean domain defined for the MED-CORDEX program (Ruti et al. 2016).

Figures 5
Figures5 and 6show PDFs and PDF ratios for three 30-year future time slices calculated over 231

Figure 6 .247
Figure 6.Same as Figure 5 but for the Alpine region.

Figure 7 .
Figure 7. Schematic depiction of the hydroclimatic response to climate warming emerging from the

345
smulations that ocean temperatures might affect the self-organization and aggregation of 346 convective systems (e.g.Mueller and Held 2012; Becker et al. 2017), which would also affect 347 the precipitation response to warming.Therefore, the study of this response might lead to a 348 greater understanding of the fundamental behavior of the precipitation phenomenon, and in 349 particular of tropical convection processes.

Figure 8 .
Figure 8.Total number of additional stress years due to increases in wet (R99.9) and dry (D25) events

Figure 9 .
Figure 9. Same as Figure 8, but with the inclusion of the SSP5 population scenario (see text for more

Figure 8
Figure8shows that, when only climate is accounted for, dry and wet extremes add