Future Precipitation Scenarios over Italy

To support the development of national adaptation policies and measures addressing climate change impacts over Italy, this work aims to analyze projected changes in mean temperatures and precipitations, and extreme events such as droughts and floods, highlighting some local trends in the different Italian regions that have been little considered to date. The investigations are made on the basis of a set of high-resolution Euro-CORDEX models (horizontal resolution 0.11°, about 12 km) to infer quantitative assessments about the danger of climate changes under three different Representative Concentration Pathways (RCPs): business as usual scenario, i.e., without a reduction in green-house gas emissions (RCP 8.5), medium stabilization scenario (RCP 4.5) and mitigation scenario (RCP 2.6). After filtering the models with limited performances in reconstructing the current climate, the multi-model climate change scenarios were characterized by comparing the ensemble mean values computed for the base-line period (1971–2000) with those elaborated for the short- (2021–2050), medium- (2041–2070) and long-term (2071–2100). Two WMO ETCCDI indices were considered to investigate climate extremes: Consecutive Dry Days and extreme precipitations. Despite some uncertainties (related to discrepancies among the models), drought conditions and extreme precipitations will likely exacerbate in the coming decades without mitigation (RCP 8.5). Such conditions will be less critical if partial mitigation actions will be undertaken (RCP 4.5) and are expected to be significantly reduced with decarbonization policies (RCP 2.6).


Introduction
Climate changes are currently affecting every country, destroying the basic elements of life for people and natural ecosystems around the world. Data from the Munich Reinsurance Company suggest an increase in meteorological and climatic catastrophic events (such as floods and droughts) since the mid-1980s and advise that climate-related hazards are at historically high levels at present [1].
The Sustainable Development Goals (SDGs) adopted by the United Nations in September 2015 state the need to take urgent action to fight climate change and its impacts [2], with the SDG 13 dealing only with "Climate Action" [3].
Because fresh water plays a central role in the functioning of the biosphere and supporting life, one of the greatest concerns is the response of the Earth's hydrologic cycle to global warming, as hundreds of millions of people could suffer from water shortages and coastal flooding as the world warms. Indeed, the natural cycle of water availability is now under threat, a situation known as 'water vulnerability' [4]. The first major driver of alterations in the hydrological system is land-use changes caused, for example, by the growth in urban areas. Water withdrawal is another cause of water vulnerability, especially in the agricultural sector: agriculture accounts for 33% of total water use in Europe and reaches up to 80% in parts of Southern Europe [4]. The third main cause of water vulnerability is climate change, even if its effects are difficult to discern given the natural variability in the hydrological cycle. There is growing evidence that the recent increase in floods and droughts in the world may be related to anthropogenic influences: some indications of the increase in precipitation extremes observed over different regions of the world have been highlighted in different IPCC reports [5,6] and other studies, such events in the last decades of the century, despite a reduction in the number of wet days, and a decrease in annual precipitation.
Focusing on Italy, Buccignani et al. [28] performed two simulations at a spatial resolution of 0.0715 • , driven by the global model CMCC-CM, over the period 1971-2100, by employing the Representative Concentration Pathways RCP 4.5 and RCP 8.5. They found a significant warming expected in Italy at the end of the 21st century, and a general reduction in precipitation, particularly in spring and summer.
In this study, some multi-model scenarios have been elaborated on the basis of several Euro-CORDEX simulations [29] at a 0.11 • spatial resolution under three hypotheses of radiative forcing (RCP 2.6, RCP 4.5 and RCP 8.5), in order to obtain some new measures about the uncertainties of the future projections.
As it is more interesting to analyze changes in the rainfall regime than the mean precipitation variation to investigate how precipitation change patterns affect different socioeconomic sectors [9], two WMO ETCCDI indices [30], specifically designed to describe climate extremes, have been computed to investigate the drought conditions and extreme precipitations expected over Italy in coming decades.
The study is presented as follows: the datasets and the adopted methodology are presented in Section 2; the results are drawn in Section 3; discussions and conclusions are reported in Section 4.

Datasets and Analysis Method
The whole Italian territory was included in the analysis domain ( Figure 1). of both droughts and floods in Europe in the coming decades [6], with increasingly serious detrimental socio-economic impacts. Some significant signals regarding precipitation changes over Italy have been pointed out by Brunetti et al. [27], who found an increase in the precipitation intensity and extreme events in the last decades of the century, despite a reduction in the number of wet days, and a decrease in annual precipitation.
Focusing on Italy, Buccignani et al. [28] performed two simulations at a spatial resolution of 0.0715°, driven by the global model CMCC-CM, over the period 1971-2100, by employing the Representative Concentration Pathways RCP 4.5 and RCP 8.5. They found a significant warming expected in Italy at the end of the 21st century, and a general reduction in precipitation, particularly in spring and summer.
In this study, some multi-model scenarios have been elaborated on the basis of several Euro-CORDEX simulations [29] at a 0.11° spatial resolution under three hypotheses of radiative forcing (RCP 2.6, RCP 4.5 and RCP 8.5), in order to obtain some new measures about the uncertainties of the future projections.
As it is more interesting to analyze changes in the rainfall regime than the mean precipitation variation to investigate how precipitation change patterns affect different socioeconomic sectors [9], two WMO ETCCDI indices [30], specifically designed to describe climate extremes, have been computed to investigate the drought conditions and extreme precipitations expected over Italy in coming decades.
The study is presented as follows: the datasets and the adopted methodology are presented in Section 2; the results are drawn in Section 3; discussions and conclusions are reported in Section 4.

Datasets and Analysis Method
The whole Italian territory was included in the analysis domain ( Figure 1). A set of 14 high-resolution Euro-CORDEX models (daily meteorological data at a horizontal resolution of 0.11°, about 12 km, ref [31] was considered (Table 1) to elaborate climate change scenarios from 1971 to 2100. A set of 14 high-resolution Euro-CORDEX models (daily meteorological data at a horizontal resolution of 0.11 • , about 12 km, ref [31] was considered (Table 1) to elaborate climate change scenarios from 1971 to 2100. Because of the uncertainty regarding future anthropogenic emissions, as well as future land-use changes, future climate projections were elaborated using model simulations realized under three different hypothesis of radiative forcing; each provides a plausible description of how the future may evolve with respect to a number of variables, including socio-economic change, technological change, energy and land use, greenhouse gas emissions and air pollutants [32]: RCP 8.5 should be seen as very high baseline emission scenario (business-as-usual scenario), RCP 4.5 can be interpreted as an intermediate mitigation scenario, while RCP 2.6 is representative of a mitigation scenario leading to a very low forcing level.
In addition, the observational dataset E-OBS [33] was used as reference to characterize the current climate. E-OBS [34] provides gridded daily data at a spatial resolution of 0.25 × 0.25 degrees from 1961 over Europe. The delivered variables are mean, minimum, and maximum temperature, and total precipitation.
Following the recommendations stated in the Program for Climate Model Diagnosis and Intercomparison [35,36] the multi-model ensemble mean approach was adopted to elaborate future climate scenarios, in order to improve the accuracy and reliability of probabilistic projections compared to single-model realizations.
Moreover, the Wilcoxon rank-sum test (a non-parametric test) was used to identify the areas that will likely be affected by significant climate change (at 95%).
First of all, the models' performances in reconstructing the current climate (1971-2000) were checked [37]. For this purpose, a MESAN reanalysis dataset [38] was used in addition to E-OBS. Mod 1 and Mod 14, using, respectively, the regional models WRF331F and AL-ADIN53 (Table 1), showed some shortcomings in representing the observed climate conditions, especially in reproducing intra-annual variability in precipitations ( Figure 2). Therefore, they were discarded in the elaboration of future multi-model mean (MMM) scenarios.
Dry spell length and heavy precipitation intensity are affected by various characteristics of the climatic system, most notably precipitation (intensity, duration, amount, timing, phase-rain or snow). However, temperature patterns are also responsible for such phenomena. Therefore, gradual changes in surface temperature and total precipitation (pr) were considered. Changes in extreme events were analyzed using two WMO ETCCDI indices: Consecutive Dry Days (CDD) to investigate drought conditions, and R99PTOT to characterize extreme precipitation. Dry spell length and heavy precipitation intensity are affected by various characteristics of the climatic system, most notably precipitation (intensity, duration, amount, timing, phase-rain or snow). However, temperature patterns are also responsible for such phenomena. Therefore, gradual changes in surface temperature and total precipitation (pr) were considered. Changes in extreme events were analyzed using two WMO ETCCDI indices: Consecutive Dry Days (CDD) to investigate drought conditions, and R99PTOT to characterize extreme precipitation.

Results
Climate change scenarios are presented in terms of seasonal anomalies, i.e., by considering the differences between the climate future values and the reference ones in the four seasons: winter (DJF), spring (MAM), summer (JJA) and autumn (SON).
Such variations are described by means of some figures, each composed of 12 panels (four columns by three rows) to characterize the expected changes for each season-in the order DJF, MAM, JJA and SON from the first to the fourth column-and for each emission scenario-RCP 2.6, RCP 4.5 and RCP 8.5, from the the first to the third row. Moreover, the grid cells of each panels are stippled if they are characterized by a significant change according to the Wilcoxon rank-sum test applied to compare the two related samples and determine whether they have the same distribution.

Gradual Changes of Climate
The seasonal mean temperature anomalies at the short-, medium-, and long-term with respect to the baseline period 1971-2000 are described, respectively, by  In these figures, each grid point is stippled. This means that a significant temperature increase is expected everywhere, according to Wilcoxon rank-sum test.

Results
Climate change scenarios are presented in terms of seasonal anomalies, i.e., by considering the differences between the climate future values and the reference ones in the four seasons: winter (DJF), spring (MAM), summer (JJA) and autumn (SON).
Such variations are described by means of some figures, each composed of 12 panels (four columns by three rows) to characterize the expected changes for each season-in the order DJF, MAM, JJA and SON from the first to the fourth column-and for each emission scenario-RCP 2.6, RCP 4.5 and RCP 8.5, from the the first to the third row. Moreover, the grid cells of each panels are stippled if they are characterized by a significant change according to the Wilcoxon rank-sum test applied to compare the two related samples and determine whether they have the same distribution.

Gradual Changes of Climate
The seasonal mean temperature anomalies at the short-, medium-, and long-term with respect to the baseline period 1971-2000 are described, respectively, by  In these figures, each grid point is stippled. This means that a significant temperature increase is expected everywhere, according to Wilcoxon rank-sum test.
Some differences in the temperature anomalies are clearly visible in the coming decades ( Figure 3): according to the RCP 2.6 configuration, the temperature increase will be limited within 1 • C (first row of Figure 3); in case of partial mitigation actions (RPC 4.5), the warming is projected to be less than 1 degree on average, and between 1.5 and 2 • C in summer (second row of Figure 3); without mitigation (RCP 8.5), temperature is expected to increase between 1.5 and 2 • C in autumn and winter over the Alps (third row).
In the middle of the century (Figure 4), warming is likely to increase, without exceeding 2 • C, under the RCP 2.6 hypothesis. The warming will be more serious in the other two emission scenarios, especially in summer, for which an increase of about 2.5 • C is projected in RCP 4.5 and 3 • C in RCP 8.5.
In the last thirty-year scenario ( Figure 5), climate changes are projected to be compatible with a warming of 1.5 • C, assuming the RCP 2.6 mitigation pathway (first row of Figure 5), whereas the temperature increase may reach 3 • C for the emission scenario RCP 4.5 (second row), and be even more than 5 • C in RCP 8.5 (third row).
It is worth noting that the projected warming in winter over Alpine region will be 1.5-2 • C, 3 • C,~5 • C, respectively, in the short-, medium-, and long-term, without Some differences in the temperature anomalies are clearly visible in the coming decades ( Figure 3): according to the RCP 2.6 configuration, the temperature increase will be limited within 1 °C (first row of Figure 3); in case of partial mitigation actions (RPC 4.5), the warming is projected to be less than 1 degree on average, and between 1.5 and 2 °C in summer (second row of Figure 3); without mitigation (RCP 8.5), temperature is expected to increase between 1.5 and 2 °C in autumn and winter over the Alps (third row).
In the middle of the century (Figure 4), warming is likely to increase, without exceeding 2 °C, under the RCP 2.6 hypothesis. The warming will be more serious in the other two emission scenarios, especially in summer, for which an increase of about 2.5 °C is projected in RCP 4.5 and 3 °C in RCP 8.5.
In the last thirty-year scenario ( Figure 5), climate changes are projected to be compatible with a warming of 1.5 °C, assuming the RCP 2.6 mitigation pathway (first row of Figure 5), whereas the temperature increase may reach 3 °C for the emission scenario RCP 4.5 (second row), and be even more than 5 °C in RCP 8.5 (third row).
It is worth noting that the projected warming in winter over Alpine region will be 1.5-2 °C, 3 °C, ~5 °C, respectively, in the short-, medium-, and long-term, without undertaking mitigation actions. Such a temperature increase may seriously affect the hydrological cycle and, therefore, the availability of water resources over the year.      As was observed in previous studies [39][40][41], precipitation changes are much more spatially heterogeneous compared with temperature changes, i.e., there are some areas in which a model projects an increase in precipitations, while another model highlights a      As was observed in previous studies [39][40][41], precipitation changes are much more spatially heterogeneous compared with temperature changes, i.e., there are some areas in which a model projects an increase in precipitations, while another model highlights a decrease of them. Therefore, confidence in the precipitation change is lower than confidence in the temperature change. This is due, in large part, to the fact that temperature changes directly depend on radiative forcing, whereas precipitation variations are affected by a number of complex interactions, including changes in the water-holding capacity of a warming atmosphere, atmospheric circulation, evaporation, and other factors (e.g., [42]). High-resolution RCM experiments show that local forcing associated with complex topography and coastlines can substantially modulate large-scale signals, often to the point of being of the opposite sign. This spread causes some uncertainty in the multi-model scenarios, an uncertainty has to be taken into account in the assessment of the final results.
In the coming decades ( Figure 6), a significant precipitation reduction is expected in summer and spring (in this case, only Southern Italy is affected) with values ranging between 5 and 10% depending on the emission scenario. In the other seasons, no significant signals are evident.
In the middle of the century (Figure 7), precipitations will likely decrease even more in summer and spring (in this case, only in Southern Italy) ranging between 10 and 20% in RCP 4.5 and RCP 8.5.
At the end of the century (Figure 8), a significant precipitation reduction is expected in spring and summer over most of Italy under the configurations RCP 4.5 (up to 30%) and RCP 8.5 (up to 50%).

Changes of Extreme Events
The model performances in reconstructing climate extremes in the base-line climate period  are checked first. Then, the results regarding future MMM scenarios are discussed.
Drought conditions were characterized by the WMO ETCCDI index CDD (Consecutive Dry Days), calculated by considering the maximum length of a dry spell (pr < 1 mm/d).
The multi-model reconstruction underestimates the CDD index of a few days in spring and in autumn, and of about 10 days in summer in comparison with E-OBS data (Figure 9). However multi-model performances may be considered satisfactory in reconstructing the pattern of seasonal CDD fields for the reference period, pointing towards more frequent events in summer in Southern Italy.
Water 2021, 13, x FOR PEER REVIEW 10 of 18

Changes of Extreme Events
The model performances in reconstructing climate extremes in the base-line climate period  are checked first. Then, the results regarding future MMM scenarios are discussed.
Drought conditions were characterized by the WMO ETCCDI index CDD (Consecutive Dry Days), calculated by considering the maximum length of a dry spell (pr < 1 mm/d).
The multi-model reconstruction underestimates the CDD index of a few days in spring and in autumn, and of about 10 days in summer in comparison with E-OBS data ( Figure 9). However multi-model performances may be considered satisfactory in reconstructing the pattern of seasonal CDD fields for the reference period, pointing towards more frequent events in summer in Southern Italy. The following three figures (Figures 10-12) illustrate the CDD anomalies expected, respectively, in the short-, medium-, and long-term.
In general, no signal highlights changes in drought conditions under the RCP 2.6 configuration. Instead, the maximum length of dry spell is expected to grow significantly, above all in summer in Southern Italy in the RCP 4.5 and RCP 8.5 hypothesis. The stippled region highlights a significant increase in CDD, ranging from 2 to 10 days, over much of Italy during the summer. The following three figures (Figures 10-12) illustrate the CDD anomalies expected, respectively, in the short-, medium-, and long-term. Water 2021, 13, x FOR PEER REVIEW 11 of 18     Changes in extreme precipitations were assessed by computing the WMO ETCCDI index R99PTOT (precipitation due to extremely wet days, when pr > 99th percentile of the reference period).
For this purpose, first, the 99th percentile (RR99) fields have been computed at seasonal scale by using both E-OBS and MMM data. Figure 13 allows for comparison between the corresponding seasonal RR99 patterns. It is worth noting that the MMM seasonal reconstructions are wetter than the corresponding E-OBS scenarios over Southern Italy, with overestimations that reach up to 100%. However, the multi-model pattern is able to satisfactorily describe the main characteristics of the precipitation regime by identifying the northern regions as the wettest in Italy, as well as the season cycle, with summer as the driest season and autumn as the wettest.
As for the other indices, the anomalies were estimated by computing the differences between the future and reference R99PTOT values in the short-, medium-, and long-term (Figures 14-16, respectively). In general, no signal highlights changes in drought conditions under the RCP 2.6 configuration. Instead, the maximum length of dry spell is expected to grow significantly, above all in summer in Southern Italy in the RCP 4.5 and RCP 8.5 hypothesis. The stippled region highlights a significant increase in CDD, ranging from 2 to 10 days, over much of Italy during the summer.
Under RCP 4.5 and RCP 8.5 configurations, a lengthening of CDD is expected in spring and autumn, with values ranging from 2 (RCP 4.5) to 5 (RCP 8.5) days after the middle of the century. This may be particularly worrying in spring because increasing drought conditions may seriously impact the agricultural sector.
Changes in extreme precipitations were assessed by computing the WMO ETCCDI index R99PTOT (precipitation due to extremely wet days, when pr > 99th percentile of the reference period).
For this purpose, first, the 99th percentile (RR99) fields have been computed at seasonal scale by using both E-OBS and MMM data. Figure 13 allows for comparison between the corresponding seasonal RR99 patterns. It is worth noting that the MMM seasonal reconstructions are wetter than the corresponding E-OBS scenarios over Southern Italy, with overestimations that reach up to 100%. However, the multi-model pattern is able to satisfactorily describe the main characteristics of the precipitation regime by identifying the northern regions as the wettest in Italy, as well as the season cycle, with summer as the driest season and autumn as the wettest.
As for the other indices, the anomalies were estimated by computing the differences between the future and reference R99PTOT values in the short-, medium-, and long-term (Figures 14-16      In general, no significant change is expected under RCP 2.6 configuration. Instead, extreme precipitations are likely to increase, with different intensities depending on both region and season in RCP 4.5 and RCP 8.5 hypotheses. There is some uncertainty in  In general, no significant change is expected under RCP 2.6 configuration. Instead, extreme precipitations are likely to increase, with different intensities depending on both region and season in RCP 4.5 and RCP 8.5 hypotheses. There is some uncertainty in In general, no significant change is expected under RCP 2.6 configuration. Instead, extreme precipitations are likely to increase, with different intensities depending on both region and season in RCP 4.5 and RCP 8.5 hypotheses. There is some uncertainty in localizing the most affected regions in the future, but a clear trend toward exacerbation of the weather conditions is evident during the century, locally, over northern regions and the Tyrrhenian coasts.

Discussion and Conclusions
The availability of climate data has never been greater. However, availability of climate data does not automatically translate into usability, as the uncertainty of regional climate changes projections is still large and has to be accounted for, especially in impact and adaptation studies.
There is great concern regarding the effects of climate change on human societies and natural ecosystems. As the response of the Earth's hydrologic cycle to global warming is of the greatest interest, this study aimed to characterize the precipitation regime over Italy on the basis of 11 high-resolution Euro-CORDEX models by highlighting some local trends in the different Italian regions the have been less considered to date.
A range of future scenarios was elaborated by considering average temperature and total precipitation changes. Moreover, an analysis of changes in the rainfall regime was carried out by using the two climate extreme indices, CDD and R99PTOT, to describe drought conditions and extreme precipitation, respectively. Such scenarios were inferred under three Representative Concentration Pathways (RCP 8.5, RCP 4.5, and RCP 2.6) to investigate how socioeconomic factors may affect climate change in the future.
The signals inferred from the Euro-CORDEX simulations confirm and strengthen the obtained results by analyzing the ENSEMBLES and Med-CORDEX models in previous studies [39,41,43,44] and, thanks to the high-resolution models, updated knowledge about climate change in Italy was achieved from the greater orographic details.
Confidence in climate change projections varies by region and by climate variable. The most important result is that a warming is projected, with a marked seasonality and an intensity depending on the orography (high-altitude areas are the most vulnerable) and, especially, the greenhouse gas emission scenarios. In fact, in the near-term (to approximately 2040), projected warming will be similar under all emission pathways. However, by the late 21st century, the differences between possible emission pathways will have a considerable effect: the estimates indicate that the temperatures are likely to increase from an additional 1 • C (for a low-emission scenario, RCP 2.6) to about 4-5 • C (for a high-emission scenario, RCP 8.5) respect to the base-line period .
Although precipitation increases globally [6,9], relatively complex patterns of change have been discovered at the local level, with areas of increasing and decreasing precipitation. These variations are a result of climate processes and feedbacks that depend on local conditions, as local forcings (e.g., topography, land use), and energy and water fluxes affect convective activity. The projections point towards an increase in total precipitation in winter over North Italy from 5% to 30% over the century, and a likely decrease in summer of about 5-10% up to 50%, respectively, in the short-, medium-and long-term in RCP 8.5 and RCP 4.5.
Changes in climate extremes are a consequence of changes in both temperature and precipitation: in case of partial or no mitigation, coastal areas will be affected by an increase in extreme precipitation with values depending on region and season. In particular, coastlines will be affected by an increase of one or two more extreme precipitation events during the winter and autumn. However, no significant signals were seen if drastic green-house gas reduction is undertaken (RCP 2.6). Regarding drought conditions, no signal highlights significant changes under RCP 2.6 configuration. Instead, in RCP 4.5 and RCP 8.5, the maximum length of dry spell is expected to grow until 10-15 days during the summer in Southern Italy, indicating an increasing risk of droughts.
In general, climate extremes will be much less critical if mitigation actions are undertaken. However, it is worth noting that some climate impacts can no longer be avoided, even in the case of low anthropogenic radiative forcing.