Analysis of Climate Change’s E ﬀ ect on Flood Risk. Case Study of Reinosa in the Ebro River Basin

: Floods are one of the natural hazards that could be most a ﬀ ected by climate change, causing great economic damage and casualties in the world. On December 2019 in Reinosa (Cantabria, Spain), took place one of the worst ﬂoods in memory. Implementation of DIRECTIVE 2007 / 60 / EC for the assessment and management of ﬂood risks in Spain enabled the detection of this river basin with a potential signiﬁcant ﬂood risk via a preliminary ﬂood risk assessment, and ﬂood hazard and ﬂood risk maps were developed. The main objective of this paper is to present a methodology to estimate climate change’s e ﬀ ects on ﬂood hazard and ﬂood risk, with Reinosa as the case study. This river basin is a ﬀ ected by the snow phenomenon, even more sensitive to climate change. Using di ﬀ erent climate models, regarding a scenario of comparatively high greenhouse gas emissions (RCP8.5), with daily temperature and precipitation data from years 2007–2070, and comparing results in relative terms, ﬂow rate and ﬂood risk variation due to climate change are estimated. In the speciﬁc case of Reinosa, the MRI-CGCM3 model shows that climate change will cause a signiﬁcant increase of potential a ﬀ ected inhabitants and economic damage due to ﬂood risk. This evaluation enables us to deﬁne mitigation actions in terms of cost–beneﬁt analysis and prioritize the ones that should be included in ﬂood risk management plans.


Introduction
Floods are natural hazards that produce great material damage and human losses worldwide [1]. The increasing population density and infrastructure on river banks contribute to increased floodplain vulnerability, which can result in severe social, economic and environmental damage [2].
The fourth of the nine essential rules of flood risk management indicates that it should be taken into account that "The future will be different from the past. Climate and societal change as well as changes in the condition of structures can all profoundly influence flood risk" [3].
For all these reasons, it is essential to design methodologies that enable one to estimate flow regime variations and evaluate flood risk modification due to climate change. [4][5][6].
The technical document of the IPCC IV forecasts a probable increase in the frequency and intensity of precipitation episodes, as well as a decrease in average values in summer for mid-latitudes countries as Spain. The Fifth Assessment Report (AR5) of the IPCC (2013- 14), points out that it is probable that the frequency or intensity of intense rainfall has increased in Europe, and in relation to future changes, that extreme precipitation events over most of the mid-latitude lands will most likely be more intense and more frequent [7]. Additionally, climate change can specially affect those regions in which the snow phenomenon is relevant in hydrological behavior.
As a matter of these facts, the 2019 flash flood European notifications beat the last 7 years of notifications clearly [8], and Reinosa (Cantabria, Spain) suffered two important floods on January and Reinosa (Cantabria, Spain) is located in the north of the Iberian Peninsula, in the junction of Hijar, Ebro and Izarilla rivers, immediately upstream the Ebro reservoir. Hijar catchment, with 27.6 km length and 147.6 km 2 surface, drains the highest area of Cantabrian mountain range that belongs to the Ebro river basin. The catchment height ranges between 865 and 2023 m above sea level, with a mean height of 1351 m. A203 gauging station is located in Hijar River, immediately upstream of Reinosa, with a time of concentration of approximately 15 hours.
The average rainfall of the head of the Ebro for the period 1920-2002 was 1190 mm/year, concentrated in autumn and winter, characteristic of an Atlantic climate, with some continental patterns producing 33 L/s/km 2 . The average annual temperature is 8.9 °C in Reinosa. The main land uses in the catchment are forest and scrub. Figure 2 shows Hijar river basin overall and site map. Reinosa belongs to the Hijar river basin, and had potential significant flood risk (ES091_HIJ_01_02_04_05_06) during the implementation of DIRECTIVE 2007/60/EC, as shown in Figure 3. Reinosa (Cantabria, Spain) is located in the north of the Iberian Peninsula, in the junction of Hijar, Ebro and Izarilla rivers, immediately upstream the Ebro reservoir. Hijar catchment, with 27.6 km length and 147.6 km 2 surface, drains the highest area of Cantabrian mountain range that belongs to the Ebro river basin. The catchment height ranges between 865 and 2023 m above sea level, with a mean height of 1351 m. A203 gauging station is located in Hijar River, immediately upstream of Reinosa, with a time of concentration of approximately 15 hours.
The average rainfall of the head of the Ebro for the period 1920-2002 was 1190 mm/year, concentrated in autumn and winter, characteristic of an Atlantic climate, with some continental patterns producing 33 L/s/km 2 . The average annual temperature is 8.9 • C in Reinosa. The main land uses in the catchment are forest and scrub. Figure 2 shows Hijar river basin overall and site map.  (Figure 1). Reinosa (Cantabria, Spain) is located in the north of the Iberian Peninsula, in the junction of Hijar, Ebro and Izarilla rivers, immediately upstream the Ebro reservoir. Hijar catchment, with 27.6 km length and 147.6 km 2 surface, drains the highest area of Cantabrian mountain range that belongs to the Ebro river basin. The catchment height ranges between 865 and 2023 m above sea level, with a mean height of 1351 m. A203 gauging station is located in Hijar River, immediately upstream of Reinosa, with a time of concentration of approximately 15 hours.
The average rainfall of the head of the Ebro for the period 1920-2002 was 1190 mm/year, concentrated in autumn and winter, characteristic of an Atlantic climate, with some continental patterns producing 33 L/s/km 2 . The average annual temperature is 8.9 °C in Reinosa. The main land uses in the catchment are forest and scrub. Figure 2 shows Hijar river basin overall and site map. Reinosa belongs to the Hijar river basin, and had potential significant flood risk (ES091_HIJ_01_02_04_05_06) during the implementation of DIRECTIVE 2007/60/EC, as shown in Figure 3. Reinosa belongs to the Hijar river basin, and had potential significant flood risk (ES091_HIJ_01_02_04_05_06) during the implementation of DIRECTIVE 2007/60/EC, as shown in Figure 3.  DIRECTIVE 2007/60/EC implies an objective quantification of flood risk, for low (500 years return period), medium (100 years return period) and high probabilities (10 years return period), that will allow the efficient development of flood risk management plans by implementing a set of combined actions to reduce floods' consequences [9].
To evaluate different climate change scenarios, the Spanish Meteorological Agency (AEMET) regionalized a set of global climate models, by using two statistical downscaling methods. The usefulness of this regionalization was assessed by their fitting to the observed data in the control period . A comparison based on a set of statistics show that although the fit is good for annual mean values, annual maximum values for both regionalization methods are not adequately simulated, since they provide lower extremes with a smaller variability. However, different fitting was observed depending on the Spanish region [10].
Although results might show a decrease in the magnitude of extreme floods for climate model projections downscaled by AEMET [11], a different pattern might be concluded, when taking into account snowmelt, if the melting flow rate's increase due to higher temperatures is bigger than the decrease caused by less snow accumulation.
The main aim of the work is to estimate flow and flood risk variation due to climate change by using different climate models regarding a scenario of comparatively high greenhouse gas emissions (RCP8.5), with daily temperature and precipitation data.

Overall Methodology
According to DIRECTIVE 2007/60/EC, Spanish hazard and flood risk maps can be downloaded at https://sig.mapama.gob.es/snczi/, regarding the following scenarios: (a) floods with a low probability, or extreme event scenarios (return period ≥ 500 years); (b) floods with a medium probability (likely return period ≥ 100 years); (c) floods with a high probability (return period ≥ 10 years).
By relating flow rates in A203 gauging station with hazard and risk maps, correlations with flooding surface, economic damage and casualties can be achieved.
In order to evaluate climate change, projections of a series of climatic variables are entered as input data in a calibrated hydrological model in the study basin. The ASTER® model is a distributed hydrological model that calculates snowmelt and accumulation regarding energy balance [12,13]  DIRECTIVE 2007/60/EC implies an objective quantification of flood risk, for low (500 years return period), medium (100 years return period) and high probabilities (10 years return period), that will allow the efficient development of flood risk management plans by implementing a set of combined actions to reduce floods' consequences [9].
To evaluate different climate change scenarios, the Spanish Meteorological Agency (AEMET) regionalized a set of global climate models, by using two statistical downscaling methods. The usefulness of this regionalization was assessed by their fitting to the observed data in the control period . A comparison based on a set of statistics show that although the fit is good for annual mean values, annual maximum values for both regionalization methods are not adequately simulated, since they provide lower extremes with a smaller variability. However, different fitting was observed depending on the Spanish region [10].
Although results might show a decrease in the magnitude of extreme floods for climate model projections downscaled by AEMET [11], a different pattern might be concluded, when taking into account snowmelt, if the melting flow rate's increase due to higher temperatures is bigger than the decrease caused by less snow accumulation.
The main aim of the work is to estimate flow and flood risk variation due to climate change by using different climate models regarding a scenario of comparatively high greenhouse gas emissions (RCP8.5), with daily temperature and precipitation data.

Overall Methodology
According to DIRECTIVE 2007/60/EC, Spanish hazard and flood risk maps can be downloaded at https://sig.mapama.gob.es/snczi/, regarding the following scenarios: (a) floods with a low probability, or extreme event scenarios (return period ≥ 500 years); (b) floods with a medium probability (likely return period ≥ 100 years); (c) floods with a high probability (return period ≥ 10 years).
By relating flow rates in A203 gauging station with hazard and risk maps, correlations with flooding surface, economic damage and casualties can be achieved.
In order to evaluate climate change, projections of a series of climatic variables are entered as input data in a calibrated hydrological model in the study basin. The ASTER®model is a distributed hydrological model that calculates snowmelt and accumulation regarding energy balance [12,13] Water 2020, 12, 1114 4 of 14 Two global climate models (GCM), downscaled by the Spanish Meteorological Agency (AEMET), have been selected; we validated them with real flow rates from the control period (1961 to 2000) and used under the highest greenhouse gas emissions pathway (RCPs 8.5) from the Fifth Assessment Report of the Intergovernmental Panel on Climate Change.
The relationship between calculated daily flow rates and instant (15 minutes period) flow rates is not necessary for these downscaled climate models, as climate change's effect will be calculated in relative terms comparing climate models in the calculation period in two equal length stages (2007-2038) and (2039-2070).
Once the flow rate variation for the climate projection is estimated and applied on real flow, new impacts on hazard and risk maps can be established. Figure 4 shows the overall methodology scheme. Two global climate models (GCM), downscaled by the Spanish Meteorological Agency (AEMET), have been selected; we validated them with real flow rates from the control period (1961 to 2000) and used under the highest greenhouse gas emissions pathway (RCPs 8.5) from the Fifth Assessment Report of the Intergovernmental Panel on Climate Change.
The relationship between calculated daily flow rates and instant (15 minutes period) flow rates is not necessary for these downscaled climate models, as climate change's effect will be calculated in relative terms comparing climate models in the calculation period in two equal length stages (2007-2038) and (2039-2070).
Once the flow rate variation for the climate projection is estimated and applied on real flow, new impacts on hazard and risk maps can be established. Figure 4 shows the overall methodology scheme.

Hazard and Risk Maps
Hazard maps are the result of two-dimensional hydraulic models, indicating flooded area, flood depths and flood velocities. Return periods were given by Spanish Authority in the framework of DIRECTIVE 2007/60/EC implementation by "Maximum Flow Maps" software, based on a regionalized gauging station study [14].
Flood risk maps shall show the potential adverse consequences associated with flood scenarios, and express, among other things, terms of the indicative numbers of inhabitants potentially affected

Hazard and Risk Maps
Hazard maps are the result of two-dimensional hydraulic models, indicating flooded area, flood depths and flood velocities. Return periods were given by Spanish Authority in the framework of DIRECTIVE 2007/60/EC implementation by "Maximum Flow Maps" software, based on a regionalized gauging station study [14].
Flood risk maps shall show the potential adverse consequences associated with flood scenarios, and express, among other things, terms of the indicative numbers of inhabitants potentially affected and the type of economic activity of each area potentially affected. Additionally, economic damage can be estimated.
According to DIRECTIVE 2007/60/EC, these maps serve as a tool for establishing priorities and for making additional technical, economic and political decisions related to flood risk management. Measures to be implemented in the at-risk areas can be prioritized, depending on the results of the cost-benefit analysis. They also allow us to evaluate climate change's effects. The risk associated with flood events is established based on the vulnerability of the threatened element and the hazard to which it is exposed. The estimation of indicative number of inhabitants potentially affected depends on updated census data. The Spanish Water Authority has also established a methodology to estimate the economic value of flood damage partially based on the PREEMPT project [15,16].
Finally, a relationship, as shown in Table 1, between flow rates for different return periods in Hijar, A203, and hazard and risk maps can be estimated, and subsequently a trend line, as shown in Figure 5, can be fitted: Measures to be implemented in the at-risk areas can be prioritized, depending on the results of the cost-benefit analysis. They also allow us to evaluate climate change's effects. The risk associated with flood events is established based on the vulnerability of the threatened element and the hazard to which it is exposed. The estimation of indicative number of inhabitants potentially affected depends on updated census data. The Spanish Water Authority has also established a methodology to estimate the economic value of flood damage partially based on the PREEMPT project [15,16].
Finally, a relationship, as shown in Table 1, between flow rates for different return periods in Hijar, A203, and hazard and risk maps can be estimated, and subsequently a trend line, as shown in Figure 5, can be fitted:    ASTER® is a distributed hydrological model that calculates streamflow in a river using rainfall, temperature, windspeed and radiation as climate inputs. It has been applied in several countries and implemented by Spanish water authorities as an accurate model in river basins with strong snow phenomena, as in our case study (Hijar catchment) [17,18]. Figure 7 showsthe snow-water equivalent for two consecutive days in Hijar model.

Snow Accumulation Routine
Precipitation will be snow or water form at different temperatures depending on several atmospheric conditions. This range of temperatures can vary between −2 and 2 °C. The model sets a temperature, named rain/snow temperature, at which 50% of precipitation is snow, with a bilineal increase and decrease of snow form up to −2 and 2 °C.

Snowmelt Routine
As aforementioned, snowmelt is calculated using the energy balance equation. Melting takes place when air temperature is below a specific temperature ( − snowmelt temperature).
Snowmelt can be divided into two types: (a) rainfall snowmelt; (b) no rainfall snowmelt. ASTER®is a distributed hydrological model that calculates streamflow in a river using rainfall, temperature, windspeed and radiation as climate inputs. It has been applied in several countries and implemented by Spanish water authorities as an accurate model in river basins with strong snow phenomena, as in our case study (Hijar catchment) [17,18]. Figure 7 shows the snow-water equivalent for two consecutive days in Hijar model.  ASTER® is a distributed hydrological model that calculates streamflow in a river using rainfall, temperature, windspeed and radiation as climate inputs. It has been applied in several countries and implemented by Spanish water authorities as an accurate model in river basins with strong snow phenomena, as in our case study (Hijar catchment) [17,18]. Figure 7 showsthe snow-water equivalent for two consecutive days in Hijar model.

Snow Accumulation Routine
Precipitation will be snow or water form at different temperatures depending on several atmospheric conditions. This range of temperatures can vary between −2 and 2 °C. The model sets a temperature, named rain/snow temperature, at which 50% of precipitation is snow, with a bilineal increase and decrease of snow form up to −2 and 2 °C.

Snowmelt Routine
As aforementioned, snowmelt is calculated using the energy balance equation. Melting takes place when air temperature is below a specific temperature ( − snowmelt temperature).
Snowmelt can be divided into two types: (a) rainfall snowmelt; (b) no rainfall snowmelt.

Snow Accumulation Routine
Precipitation will be snow or water form at different temperatures depending on several atmospheric conditions. This range of temperatures can vary between −2 and 2 • C. The model sets a temperature, named rain/snow temperature, at which 50% of precipitation is snow, with a bilineal increase and decrease of snow form up to −2 and 2 • C.

Snowmelt Routine
As aforementioned, snowmelt is calculated using the energy balance equation. Melting takes place when air temperature is below a specific temperature (T mi -snowmelt temperature). Snowmelt can be divided into two types: (a) rainfall snowmelt; (b) no rainfall snowmelt.

Rainfall Snowmelt
It is the snowmelt produced by rainfall energy. Assuming that the snow surface temperature is equal to 0 • C (273 K), snowmelt caused by energy from rainfall is [19]: where: P = Precipitation (mm) f r = fraction of precipitation in the form of rain T r = temperature of rain ( • C) In addition to energy from rainfall, other secondary terms as radiation and condensation energy must be added to calculate snowmelt when rain takes place.

No Rainfall Snowmelt
It is the snowmelt on a dry day. Due to difficulties estimating wind speed, hydrological models usually use non-forced convection or other empirical relations depending on air temperature. The following equation is used to estimate no rainfall melt M c (mm) [20]: where: This melt factor (M f ) indicates the seasonal variation, and can be represented with a sine function [21,22].

Model Calibration and Validation
Calibration and validation have been carried out, comparing real daily flow rates, in A203 Hijar in Reinosa Gauging Station with the ASTER model daily flow rates being calculated with real meteorological data. The available continuous period for the observed daily flow rate ranges from 2014 to 2019, defining the calibration period as October 1st 2014 to September 30th 2016 and the validation period from October 1st 2017 to December 22nd 2019, including the aforementioned extraordinary event. Figure 8, shows results during the calibration period, with a correlation coefficient of 0.90 and a Nash index (NSE) of 0.78. The maximum observed daily flow rate was 23.6 m 3 /s, and the maximum calculated daily flow rate was 23.5 m 3 /s. Total observed runoff was 189 hm 3 and calculated runoff was 200.7 hm 3 , with 27 hm 3 of overall snow-water equivalent accumulation. validation period from October 1st 2017 to December 22nd 2019, including the aforementioned extraordinary event. Figure 8, shows results during the calibration period, with a correlation coefficient of 0.90 and a Nash index (NSE) of 0.78. The maximum observed daily flow rate was 23.6 m 3 /s, and the maximum calculated daily flow rate was 23.5 m 3 /s. Total observed runoff was 189 hm 3 and calculated runoff was 200.7 hm 3 , with 27 hm 3 of overall snow-water equivalent accumulation.  Altimetry temperature gradient = −3.1 (°C/1000 m). Figure 9 shows results during the validation period, with a correlation coefficient of 0.90 and a Nash index (NSE) of 0.80. The maximum observed daily flow rate was 83 m 3 /s, and the maximum calculated daily flow rate was 69 m 3 /s. Total observed runoff was 231 hm 3 and calculated runoff was 289 hm 3 , with 20 hm 3 of overall snow-water equivalent accumulation.

Climate Projection
Daily rainfall and temperature data from two regionalized climate models (analog method-ANA and statistical regionalization method-SDSM) can be obtained from AEMET website (http://www.aemet.es/es/serviciosclimaticos/cambio_climat) for RCP 8.5.
Two climate models have been selected from 24, to analyze the combination of rainfall and temperature: • ANA-BCC-CSM1-1-m (BCC); • ANA-MRI-CGCM3 (MRI). These two models have been pre-selected as generating the highest rainfalls. To evaluate their suitability, daily flow rates for the control period  have been calculated with ASTER for real meteorological data and both climate models. Figure 10a shows that annual maximum real daily flow rates are above climate models, and the trend line shows a higher increase as well. Figure 10b shows that annual mean real daily flow rates fit the MRI better at the beginning of the control period, while a fit better with BCC is shown at the end of the control period.

Climate Projection
Daily rainfall and temperature data from two regionalized climate models (analog method-ANA and statistical regionalization method-SDSM) can be obtained from AEMET website (http://www. aemet.es/es/serviciosclimaticos/cambio_climat) for RCP 8.5.
These two models have been pre-selected as generating the highest rainfalls. To evaluate their suitability, daily flow rates for the control period  have been calculated with ASTER for real Water 2020, 12, 1114 9 of 14 meteorological data and both climate models. Figure 10a shows that annual maximum real daily flow rates are above climate models, and the trend line shows a higher increase as well. Figure 10b shows that annual mean real daily flow rates fit the MRI better at the beginning of the control period, while a fit better with BCC is shown at the end of the control period. Supplementary Materials for rainfall stations (9008E, 9012) and temperature stations (2225, 2232) is given.
Two climate models have been selected from 24, to analyze the combination of rainfall and temperature: • ANA-BCC-CSM1-1-m (BCC); • ANA-MRI-CGCM3 (MRI). These two models have been pre-selected as generating the highest rainfalls. To evaluate their suitability, daily flow rates for the control period  have been calculated with ASTER for real meteorological data and both climate models. Figure 10a shows that annual maximum real daily flow rates are above climate models, and the trend line shows a higher increase as well. Figure 10b shows that annual mean real daily flow rates fit the MRI better at the beginning of the control period, while a fit better with BCC is shown at the end of the control period.

Calculation of Flow Rates for Different Return Periods
Daily flow rates for the calculation period (2007 to 2070) have been estimated with ASTER model for both climate projections (BCC and MRI). Calculation period has been divided in two equal length stages (2007 to 2038) and (2039-2070) and a probability distribution of extreme values has been estimated using the Gumbel distribution, presenting the best fit goodness: where: u = location parameter; β = scale parameter; Tr = Return period.
Finally, daily flow rates and instant (15 minute period) flow rates observed in A203 can be fitted in a linear relationship. As climate projections flow rates are being studied in relative terms with observed flow rates, transformation from daily flow rates into instant flow rates is not necessary. Table 2 and Figure 11 show results and trend lines from percentile analysis for both calculated climate models. For BCC-CSM1-1-m a small decrease in flow rate projection due to climate change is estimated, with small variability for different probabilities, while for MRI-CGCM3 bigger increase is predicted, showing a smooth variability for different probabilities, describing a more extreme climate.

Hazard and Risk Maps
After calculating the projected flow rate by entering for each return period (2nd Stage (2039-2070)/1st Stage (2007-2038)) the flow rate ratio in the trend line fitted for flood surface, economic damage and inhabitants potentially affected ( Figure 5), new climate change estimations show similar results as for flow rate (Table 3).

Hazard and Risk Maps
After calculating the projected flow rate by entering for each return period (2nd Stage  (Table 3).
As for flow rate, the MRI-CGCM3 climate model shows an increase in flood risk effects due to climate change, while BCC-CSM1-1-m shows a decrease. Results are enlarged for economic value and inhabitants potentially affected, while flooded surface decreases.
A logarithmic trend line is the one that best fits, and represents how the atmosphere behaves that shows in Figures 12-14.     Flooded surface, inhabitants potentially affected and flood economic damage value can be calculated in absolute terms and are shown in Table 4 for ES091_HIJ_01_02_04_05_06. Flooded surface, inhabitants potentially affected and flood economic damage value can be calculated in absolute terms and are shown in Table 4 for ES091_HIJ_01_02_04_05_06.

Climate Change Effect
Comparison between observed meteorological data and projections of climate scenarios, for a common time interval, has revealed a significant difference between the series. This means that the results must in any case be interpreted in relative terms, evaluating trends and not absolute values.
Results for both selected climate models show important differences between each other in absolute and relative terms, indicating that there is significant uncertainty in the projections, especially due to rainfall; and they suggest that for local impact studies, further analysis to regionalize outputs from global climate models still needs to be done. For the results to be reliable, climate models must adequately represent extreme phenomena, as most of them are suitable for evaluating water resource evolution.
There is a large uncertainty in climate projections that influences the estimation of future flood risk. The crucial action in the case of adaptation to floods is to carry out a detailed analysis to choose the best climate model and use results in relative terms between the short term (1st Stage) and long term (2nd Stage) of the projection period [4].
The proposed methodology allows one to determine in terms of hazard and flood risk, the effect that climate change has on a given territory, considering both variable rainfall and temperature, and enables a fast evaluation of various climate models and a comparison of their differences.
This evaluation enables one to define actions in terms of a cost-benefit analysis and prioritize the ones that should be included in flood risk management plans due to climate change, as these management plans must contain effective actions in the short, medium and long term and cannot ignore the effect of climate change.
In the specific case of Híjar River Basin, and Reinosa, MRI-CGCM3 model trend analysis shows that climate change will increase damage to society, both economic and personal, while BCC-CSM1-1-m shows a small decrease. As the MRI-CGCM3 model fits better with annual maximum real daily flow rates than BCC-CSM1-1-m, it can be concluded that climate change will cause a significant increase of potential affected inhabitants and economic damage due to flood risk that will require mitigation actions.
This assumption might match with previous years' observations, especially given the high number of low probability events recorded in Spain. This is the specific case of the Hijar River that in 2019 registered two floods, the one on December 2019 being the largest ever registered in Reinosa Gauging Station and related with a low probability event (300 years return period). This conclusion is consistent with the Fifth Assessment Report (AR5) of the IPCC (2013-14), which points out that it is likely that the frequency or intensity of intense rainfall has increased in Europe.

Future Research
As already mentioned, this methodology could be applied in every potential significant flood risk area through a water authority flood risk management plan in order to prioritize actions.
Firstly, all available regional climate models could be selected and evaluated. Finally, a sensitive RCP analysis with RCP 4.5 scenario could be carried out.