Future Climate Projections in Algeria Using Statistical DownScaling Model

: In this study, we perform a statistical downscaling to investigate projected future changes in minimum temperature (T-min), maximum temperature (T-max), and precipitation (PRCP) for the three periods the 2020s (2011 – 2040), the 2050s (2041 – 2070), and the 2080s (2071 – 2100), with respect to the reference period 1981 – 2010 over Algeria by applying the Statistical DownScaling Model (SDSM). The NCEP reanalysis data and CanESM2 predictors of three future scenarios, RCP2.6, RCP4.5 and RCP8.5 are used for model calibration and future projection, respectively. In order to get realistic results, bias correction was also applied to the climate variables. The evaluation of the SDSM performance indicated that model accuracy for simulating temperatures and precipitation was statistically acceptable. The predicted outcomes exhibit strong warming for both extreme temperatures under the worst-case scenario (RCP 8.5), it is more pronounced for the maximum temperature and over the Sahara region. The results indicate that the highest changes are expected to increase by 3.6 to 5.0°C for the minimum temperature and 5.0 to 8.0°C for the maximum temperature for the strong radiative forcing pathway (RCP8.5) by the end of the century as compared to the reference period. Under the optimistic scenario (RCP2.6), the strength of the warming is projected to increase up to 2.0°C for both extreme temperatures. For the precipitation, the projections indicate for all scenarios a significant decrease in rainfall by approximately 20% over the northwest region and central Sahara, while non-significant change is expected for the center and eastern coastal regions. Conversely, the projections of rainfall under different emission scenarios exhibit an increase (~10 – 40%) at the central and eastern high plateaus in the north and the extreme west and south of the Sahara. The study reveals several discrepancies among considered stations in the projections of seasonal rainfall under different emission scenarios where most of them exhibit a significant increase of precipitation in summer. Our findings corroborate previous studies by demonstrating that Algeria’s climate will warm further in the future. The results might be beneficial for policymakers for planning strategies and may help to mitigate the risks linked to climate change. by 2020s, 3.6 5.0°C by by of end


INTRODUCTION:
North Africa is considered as one of the most exposed regions to negative effects of climate change impacts in the world (IPCC 2014). This region is for several years experiencing extreme temperatures leading to severe droughts and water rarity, extending desertification, loss of marine ecosystems and biodiversity (IPCC 2018). The region is highly exposed to climate change and threatened by extreme heat and water shortages. According to the World Meteorological Organization's (WMO) REPORT of the 2020' State of the Global Climate (https://public.wmo.int/en/resources/library/state-of-globalclimate-2020), the average global temperature during 2020 is about 1.2 degrees Celsius above that of the pre-industrial period  baseline, classified as one of the three warmest years on record globally. Considering the climate change context, Working Group I of the Intergovernmental Panel on Climate Change (IPCC-WG I) revealed in its fifth assessment report (IPCC 2013), for the extreme radiative forcing scenario RCP 8. 5, an increase in atmospheric temperature expected for North Africa region ranging from 2 to 3 degrees Celsius for the period (2046)(2047)(2048)(2049)(2050)(2051)(2052)(2053)(2054)(2055)(2056)(2057)(2058)(2059)(2060)(2061)(2062)(2063)(2064)(2065), and from 3 to 6 degrees Celsius for the long-term period (2081-2100) compared to the 1986-2005 reference one. Therefore, the effects of global warming would be most severely felt by the increase in extreme air temperatures that would intensify the occurrence of extreme events and conditions of heat stress.
Moreover, the north Africa region is classified as a climate change hotspot (Diffenbaugh and Giorgi 2012). In their study,  demonstrate that climate observations and simulations show an upward trend of the heat extremes days and a decrease of the cool days and nights, particularly since the 1970s.
In early July 2018, the heat extended to the North Africa region, with records set at 5 cities in Algeria, the highest being +51.3 °C at Ouargla, classified as a national record (WMO 2019).
However, very few studies have investigated the projected climate simulations under different scenarios over Algeria. Most of the climate projections studies over the country were conducted in a regional context of the Middle East-North Africa (MENA) or Mediterranean regions (Almazroui 2016, Almazroui et al. 2016Driouech et al. 2020;Goodess et al. 2013). Zeroual et al. (2019) used CORDEX-Africa regional climate models simulations to assess future changes in the climate zones of Algeria as defined by Koppen-Geiger based on two Representative Concentration Pathway scenarios (RCP4.5 and RCP8. 5) for the period from 1951 to 2098. They found a gradual but significant expansion of the surface area of the desert zone while the rate of expansion of desert climate will increase in the future, particularly during the period from 2045 to 2098 according to projections for the pessimistic emission scenario (RCP8.5).
Although, various research studies applied global and regional model scenarios to simulate future temperature and precipitation changes and to assess impacts at a MENA scale (Bucchignani et al. 2018;Bigio et al. 2011;Sowers et al. 2011;Goodess et al. 2013;Giannakopoulos et al. 2009). Since the last decade, there is an increasing number of MENA studies based on the hub of the Coordinated Regional Downscaling Experiment (CORDEX: http://mena-cordex.cyi.ac. cy/), focusing on individual regional model simulation and validation (Zittis et al. 2021;Almazroui et al. 2016;Almazroui 2016;Bucchignani 2016 ), as well as several future projections and impact assessment studies for the region (Zittis et al. 2021;Driouech et al. 2020;Ozturk et al. 2018).
The most significant results of these studies confirm that precipitation in North Africa is likely to decrease while temperatures are likely to rise. Paeth et al. (2009) showed that according to the Regional Model REMO, the annual total precipitation is projected to decline between 10 and 20%, and the temperature to increase between 2 and 3°C by 2050 under SRES A1B scenario conditions, where more pronounced drying is expected.in north-western parts of North Africa.
A part of the Mediterranean region, the Integrated Research Project (CIRCE) models simulations over Algeria as well as most parts of the region were consistent in indicating warming ranging from 0.8 to 2.0°C in winter (DJF) and from 1.2 to 2.5°C in summer (JJA) for 2021-2050 compared with 1961-1990 under the A1B emission scenario (Goodess et al. 2013). In general, the six CIRCE regional and global models used show a tendency towards warmer and drier conditions over the Mediterranean, consistent with earlier studies (e.g. , Giorgi and Lionello 2008) accompanying by a general increase in the number of very hot days and nights, together with longer warm spells and heatwaves. Driouech et al. (2020) indicate that the projected changes over the CORDEX-MENA domain through contrasting ALADIN-climate regional model for the future period (2071-2100) against the present-day period  under both RCP4.5 and RCP8.5 are statistically significant (at 95% level). The mean temperature will increase around 2°C for RCP4.5 and at least 4°C for RCP8.5 in North Africa (Eastern Morocco and most Algeria). The future scenarios indicate the intensification of heatwaves occurrence over this region particularly in the inland, while the total annual precipitation amounts will decrease from 5 to 20% in the North and increase in nearly all the remaining parts notably in the Sahara exceeding +40%. Zittis et al. (2021) for a business-as-usual pathway (RCP 8.5), indicate that in the second half of this century, the MENA region including Algeria will experience unprecedented super-and ultra-extreme heatwave conditions. These events involve excessively high temperatures (up to 56 °C and higher) and will be of extended duration (several weeks).
Similarly, Varela et al. (2020) show that maximum temperature is expected to rise throughout the entire MENA region for the 21st century under the RCP 4.5 and 8.5 scenarios. The increment of maximum temperatures is expected to range between 4˚C and 7˚C, while the mean and maximum intensity of heatwaves is also expected to increase for almost the whole MENA region.
On the other hand, Ahmadalipour et al. (2018) investigate the mortality risk for people aged over 65 years caused by excessive heat stress across the MENA region for the historical period of  and two future scenarios of RCP4.5 and RCP8.5 during the 2006-2100 period. Their results show that without mitigation measures, the mortality risk will be 8 to 20 times higher by the end of the century, compared to the reference period.
The climate change studies aim to determine the changes in the future climate compared to a referenced past period or simulated one, under different radiative forcing pathway scenarios. The used models should be evaluated through their reproduction of the past climate (Collins et al. 2020).
To quantify the impacts of climate change, the outcomes of general circulation models (GCMs) are frequently used to simulate the impacts of increased greenhouse gases on climatic variables. However, these models are limited because of their coarse resolution at a subgrid-scale, and regional and local scale processes are occurring on spatial scales much smaller than those resolved in GCMs.
Scientists employ various techniques to bridge the gap between the resolution of climate models and regional and local scale processes, yielding localized information on future climate behaviours in order to fit the purpose of local-level analysis and planning and to assess the impact of climate change including the application of climate change scenarios to different sectors. These techniques are known as "downscaling". Furthermore, the benefit of the selected downscaling method could be evaluated once introduced in a practical impact study by assessing the climate risk and proposing adequate implementation measures (Hussain et al. 2015). Mainly, regional modelers use two categories of downscaling. One approach is dynamical downscaling where outputs from GCM's are used to drive higher-resolution regional climate models with a better representation of local conditions, but it is computationally expensive. The second approach is statistical which requires less computational effort than dynamical downscaling, where statistical links are established between large-scale climate phenomena and observed local-scale climate, and can therefore Nowadays, SDSM is broadly adopted due to its reliability, simplicity and great performance as well as the free accessibility of the dedicated software (Wilby et al. 2013;Saidi et al. 2020 The scientific community to downscale different climatic parameters like precipitation and temperature has also used SDSM as a user-friendly software package extensively. Most scientists and researchers used a different kind of statistical models and come to the point that SDSM is one of reasonable and reliable approach among all others (Tahir et al. 2018;Khazaei et al. 2020;Zehtabian et al. 2016;Khan et al. 2006). They indicate that the SDSM is the most capable of reproducing several statistical characteristics of observed data in its downscaled results with a 95% confidence level.

2-1. Study area and data
Algeria with its geographical contrasts is located in Northwestern Africa, bordering the Mediterranean Sea between Morocco and Tunisia ( Precipitation in this region is scarce and unevenly distributed; the mean annual total varies between 15 and 120mm. To conduct this study, historical records of observed precipitation (PRCP) and temperature (minimum  (Table 1)  Meteorological Office, a part of the WMO regional basic synoptic network. The location of these stations on the map of Algeria is shown in Fig. 1. Furthermore, all meteorological data used in the study are quality controlled and homogenized.

Description of the SDSM
The SDSM is a hybrid tool combining stochastic weather generator (SWG) and multiple linear regression (MLR) based on downscaling methods (Wilby et al. 2002), requiring two types of daily data: Observed station data representing local climate (predictand) and large-scale atmospheric data (predictors) of a grid box closest to the station. Before model calibration, as pre-processing for the application of the SDSM model, data are quality controlled and checked for any outlier or missing data value. Then, the fourth root transformation has been conducted for precipitation, which is usually a skewed climatic variable to render it normal before using it in a regression equation (Wilby et al. 2002;Mahmood et al.2013;Huang et al. 2011). In SDSM, there are two methods to optimize the model: (1) ordinary least squares (OLS) and (2) dual simplex (DS). In this study, OLS is used because it is faster than DS and its results are comparable. To select the most potential future climate predictors, different correlation tests were completed by screening the variables in SDSM. Each predictor was selected based on a high correlation with the predictand and the magnitude of its probability (p-value) at a significant level of (0.05). Each of the selected predictors is further assessed for its accuracy using graphical methods such as a scatterplot.

Bias Correction:
To eliminate the biases from the daily time series of SDSM's downscaled data, a bias correction method is performed by applying the two following equations (Salzmann et al. 2007;Mahmood and Babel 2013;Saidi et al. 2020).
T deb = T sce − (T ̅ con − T ̅ obs ) P deb = P sce x ( P obs ̅̅̅̅̅̅̅ P con ̅̅̅̅̅̅̅ ) Where: Tdeb and Pdeb are the de-biased daily time series of temperature and precipitation, respectively, for future periods; TSCE and PSCE represent the daily time series of temperature and precipitation downscaled by SDSM for future periods respectively (e.g., 2011-2100). con and con represent in order the long-term mean monthly values of simulated temperature and precipitation by SDSM for the control period (e.g. 1961−2005). obs and obs represent the long-term monthly mean observed values for temperature and precipitation.

RESULTS AND DISCUSSION
The results indicate that the most relevant atmospheric predictors for both maximum and minimum temperature were: Mean sea level pressure, 850 and 500 hPa Geopotential, 850 hPa Specific humidity and Air temperature at 2 m. Besides these predictors and for the Saharan region, we selected the predictors 500 hPa Relative vorticity of the true wind and 500 hPa Wind direction. However, for the precipitation, the dominant predictor variables were: surface and 500 hPa Specific humidity, surface Meridional wind component, surface Divergence of true wind, 850 hPa Zonal wind component, 500 hPa Meridional wind component and Total precipitation. The highest correlation scores were obtained for the maximum and minimum temperatures while the lowest ones were for precipitation.

Calibration and validation of SDSM
Using the selected predictors for each predictand, the model is calibrated under unconditional (temperature) and conditional (precipitation) processes on a monthly scale except for the Saharan stations and for precipitation, the model is calibrated on seasonal time scales to increase the number of wet days. For our case, R varies between 77 and 98% for both calibration and validation of minimum and maximum temperatures, while the R2 is ranging between 60 and 96% and the mean of NSE between 0.77 and 0.90 (Table 3.). However, for the precipitation, weaker correlations were found, R and R2 were globally equal to lower than 60% notably for the Saharan stations, where they were less than 30%, this suggests that the model performance is less effective for the precipitation parameter in the Saharan zone.
These results corroborate previous studies on modelling daily precipitation (Saidi et al. 2020;Wilby et al. 2002;Gagnon et al. 2005). This is because of the complex characteristics of daily events and to the fact, modeling precipitation is one of the most challenging climate variables and in conditional models, there is an intermediate process between regional forcing and local station weather which in turn depend on regional-scale predictors such as humidity and atmospheric pressure (Gebrechorkos et al. 2019).
Data of both observed and simulated precipitation and minimum and maximum temperatures during the period 1991-2005 are mapped using the universal Kriging interpolation method with SAGA-GIS software by converting point data to raster data to carry out and to compare the spatial distribution of these parameters over the country (Figs. 2-4). The observed and simulated patterns of minimum (Fig. 2) and maximum temperatures (Fig. 3) illustrate the observed North-South temperature gradient as well as the effect of elevation. The temperatures distribution over the high plateaus is well reproduced. The Saharan region remains the hottest whereas the highlands in Tamanrasset enjoy a temperate climate and fairly mild weather due to its high altitude. The spatial distribution of simulated precipitation produced is significantly consistent with observations (Fig. 4), thus, higher values of precipitation appeared in the far-North-Eastern cities and the coastal central zone, and lower values mainly observed in the South.
The simulated downscaled data during the validation period resulted globally in similar spatial patterns compared to observations as they all show obvious variations across the country. Consequently, SDSM provides reasonable downscaling data when using NCEP large-scale predictors representing the observed current climate and can be therefore used to project future climate.

Future Temperature and Precipitation Change Scenarios
The future climate parameters have been extracted based on each model-scenarios and compared with respect to the baseline period 1981-2010 considered as the present climate, using the largescale atmospheric predictor variables derived by CanESM2 for the three emission scenarios: RCP2.6, RCP4.5 and RCP8.5. Twenty ensembles of synthetic time series at the daily scale were generated for the daily temperatures and precipitation and for each emission scenario of future decades; the average value of these 20 ensembles was used for the appropriate period.

Temperature
Future temperature projections from all downscaled models under different scenarios highlight an increase for both minimum and maximum temperatures in intensity in the country up to the end of the 21st century. Overall, strong warming is projected for the "business-as-usual" Representative 1.9-2.9 °C (4.3-6.1 °C) for SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios respectively. During the summer, the projected temperature increases under SSP5-8.5 for more than 6 °C over the North-Africa region and more than 5°C over the Sahara.

Precipitation
To investigate the future projected precipitation, relative future changes were calculated by normalizing the future change values by the historical values of the reference period to obtain a relative sense of how much the change compares to the recent climatology. Overall, we project mean annual precipitation to increase in the south-western, extreme south-eastern part and the eastern region of North highlands and a decrease in the remaining regions particularly in the north-western of coastal and plateaus regions. However, this latter area is characterized by low annual total precipitation values, leading to high percentage variations in any future period even if absolute changes are weak.
Expected changes in precipitation for the 2050s will be globally similar to that in the 2020s. Yet for the eastern high plateaus region and extreme south of Sahara, there is going to be a greater spatial expanse of strong values in relative precipitation variations (up +35% from historical values), particularly for RCP4.5 and RCP8.5.
Under the "business-as-usual" pathway, projected precipitations for the period 2071 to 2100 are from normal over the North of the Sahara to a substantial increase expected for the remaining parts of this zone (up to 50%), while precipitation reductions (~20%) are projected in the coastal region notably in the west.
These results are consistent with those revealed by Massoud (2020)

Comparison of projected seasonal precipitation change in selected stations:
To assess the precipitation changes at a seasonal scale, five meteorological stations established in five important Algerian cities are selected due to their importance economically and demographically and representing different climate regimes: Two coastal cities, Algiers in the center, and Oran in the west; Constantine which is in the northeast highland; Bechar in the northwest of the Sahara and Tamanrasset in the extreme south of the country. Fig.8 a,b show that for all scenarios, Bechar, Algiers and Constantine will experience a strong positive change of the precipitation in summer during the 2020s, the 2050s and the 2080s, while, there will be a decline in precipitation in Autumn for Bechar and in Winter for Constantine. Furthermore, the highest seasonal predicted change is for Constantine in summer with +140% under the RCP8.5 scenario by the 2080s as compared to the reference period 1981-2010. For the Constantine and Bechar stations, a nonsignificant trend is projected at the annual scale.
In Oran, scenarios exhibit a decreasing precipitation trend for all seasons and in the 2020s, the 2050s and the 2080s. This decrease is more stressed during summer with -30 to -70% under respectively RCP2.6 and RCP4.5 and -40 to -90% for RCP8.5 and that during the 3 selected time slices. However, specifically for the rainy season from autumn through winter, the precipitation is expected to fall to approximately −30 to -80% in this city.
Drying conditions are projected with all scenarios in Tamanrasset in winter and summer. The reduction of seasonal precipitation amounts in this region varies for all scenarios respectively, from -25% in summer to -50% in winter from now to the end of the century. In contrast, precipitation will increase by up to 20 and 30% during spring and autumn for the entire projected period. At the annual scale, we expect rainfall in Tamanrasset to be normal to slightly below normal (-5%) as compared to the baseline period. The study shows that the magnitude of the warming at the end of the century relative to the baseline period (1981−2010) highly depends on the emission scenario considered. Overall, for the future climate of the domain, the warming trend is observable for the maximum temperatures than the minimum temperatures especially for the worst-case scenario (RCP 8.5). It is more pronounced over the Sahara than in high plateaus and coastal regions.
Under the RCP8.5 scenario, the minimum temperature over Algeria will increase up to 2.4°C in near future (2011-2040), 2.4 to 4.0°C in the mid-future (2041-2070) and 3.6 to 5°C for the far future (2071-2100) periods. For the maximum temperature, this increase will be stronger, varying between 3 to 4°C in the 2020s, 3.6 to 5.0°C in the 2050s and from 5 °C up to 8 °C in the 2080s. Under the "business as usual" scenario, the warming is expected to be therefore 2 to 3 times higher in the late twenty-first century.
The future pattern of rainfall under all scenarios is constantly decreasing over the western coastal and western high plateau regions. The expected decrease is varying between 10 to down 20% relatively to 1981-2010 period. Inversely, normal conditions to an increase by up to 50% are respectively projected for the Northeastern and southern regions in Algeria.
The study evaluates the precipitation change at a seasonal scale over the five selected stations representing different climate in different regions of Algeria. When compared to the reference period, a significant increase in precipitation across all three future scenarios is projected during the summer except the western coastal region and extreme southern Sahara. The Northwestern of Algeria will be affected by a strong reduction in precipitation during the entire seasons, which contributed appreciably to magnifying drought over this region. Shifting in the wet season is shown in the Northeast in Algeria where the precipitation amount is projected to be greater during the summer and below in winter relative to the baseline period. Therefore, it can be concluded that in Algeria, winter is globally expected to become dryer as compared to the recent climate.
Consequently, the results obtained in this study are in quite agreement with previous studies conducted in the context of the north-African and Mediterranean regions. Since for a business-as-usual pathway and by the end of this century, the expected strong warming will obviously make heatwaves hotter, longer, and more frequent. Climate change is therefore expected to be felt more severely in Algeria given the high exposure fragility and low adaptive capacity. Enhancing resilience, engaging adequate mitigation measures, and integrating the climate information delivery in development policy and planning processes to both assess and monitor the impact of climate change, should be accomplished with great priority in the coming decades. Despite the ability of SDSM to simulate the climate, a comparative study considering several different climate model outputs is strongly recommended to provide greater confidence in the projected climate variables.

Declarations:
-ACKNOWLEDGMENTS: The author wishes to sincerely thank for the help provided by Dr.