Future changes in seasonal temperature over Pakistan in CMIP6

The present study analyzed seasonal (i.e., Dec-Jan [DJF] and June – August [JJA]) temperature change for the near (2025-2054) and far future (2070-2099) under SSP245, SSP370, and SSP585 scenarios over Pakistan. The anomalies, Mann-Kendall trend tests, Sequential Mann-Kendall trend test (SQMK), and probability density frequency (PDF) analysis were used to investigate future mean temperature variations. The DJF season projected higher increase in temperature in the northern (3.8 C, 5.1 C and 6.5 C), followed by central regions (3.8 C, 4.9 C and 6.4 C) under SSP245, SSP370 and SSP585 scenarios, respectively. The central region is likely to record significant increase in JJA (3.0 C, 4.4 C and 5.4 C) mean temperature in far future under the given SSP scenarios. Compared to historical (PDF), the far future DJF temperature changes revealed significant higher warming over northern, central and then over southern regions under most of SSP scenarios. The southern regions are projected to possible rise in far future JJA Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 11 January 2021 doi:10.20944/preprints202101.0188.v1 © 2021 by the author(s). Distributed under a Creative Commons CC BY license. temperatures by 2.7 C, 3.3 C and 4.3 C, under SSP245, SSP370 and SSP585, respectively. The PDFs for JJA further verify the highest positive abrupt shift in temperature across the central region and then southern region. The future diverse seasonal temperature changes supports further examination of the associated mechanisms and factors responsible for temperature changes to address climate change.

Several studies have investigated the mean temperature projections across diverse regions and timescales to estimate the exposure of the environment and communities to eminent future warming scenarios (Grose et al., 2020;Almazroui et al. 2020a, b;Mishra et al., 2020). To illustrate this, Grose et al. (2020) found a higher magnitude of 6.5°C in CMIP6 models ensemble projections than CMIP5 (5.1 °C) for SSP585 (RCP8.5) by the end of the 21 st century over Australia. Over Africa, Almazroui et al. (2020a) projected mean annual temperature for 2030-2059 (2070-2099) by 1.2 °C (1.4 °C), 1.5 °C (2.3 °C), and 1.8 °C (4.4 °C) for the SSP1-2.6, SSP2-4.5, and SSP5-8.5, respectively. Recently, Mishra et al. (2020) in a bias corrected projections study projected an increase in maximum and minimum temperatures by 3 °C to 4 °C and 3.5°C to 5.5 °C under SSP585 by the end of the 21 st century over south Asia.
Pakistan is ranked among the most vulnerable nation (8 th ) to climate change impacts across the globe (Eckstein, 2019). A few model based future projections have focused on mean temperature change and variability over Pakistan (Iqbal and Zahid, 2014;Ali et al., 2015;Babar et al., 2016;Rehman and Ali, 2018;Almazroui et al., 2020b). Existing studies have mainly centered over the broader Asian region or focused on extreme climate based on climate indices.
For instance, Almazroui et al. 2020b conducted a study using CMIP6 multimodel ensemblebased projections for annual and seasonal mean temperature over South-Asian countries. The study established various changes for SSP126 for near (2030-2049), mid (2060-2069), and far (2080-2099) periods as 1.0 °C, 1.4 °C and 1.3°C; for the SSP245 were 1.1 °C, 2.0 °C and 2.6 °C and for the SSP585 were recorded to be 1.4 °C, 3.1 °C and 4.6 °C, respectively. Using CMIP5, Ali et al. (2019) projected an increase in frequencies and magnitudes for warm extremes and decreased cold extremes during the 21 st century. Further, studies Sajjad and Ghaffar, 2019a) projected a significant increase in the frequency of summer days (6 days increase) and a mean maximum (1.3 °C) and minimum (1.9 °C) temperature over the northern parts Pakistan.
Hydrologically, summer and winter seasons are crucial to Pakistan's agriculture, industry, and domestic use. About 51-80 % of the country's water demand is met by glacial and rainwater stored in the northern mountainous regions (Nabeel and Athar, 2019). The water resources are anticipated to deplete rapidly under the warming phenomenon in the coming years.
Rapid urbanization, deforestation, and industrialization in the southern region may increase the warming intensity, frequency, and span in the coming years. The aforementioned phenomenon could lead to of floods, flash floods, GLOFs, droughts, heatwaves across various spatiotemporal scales across Pakistan (McSweeney et al., 2008).
This study is designed to project the future changes in mean seasonal temperature over Pakistan based on CMIP6 multimodel ensemble. This work will be the first attempt in using CMIP6 models to project temperature changes and trends across Pakistan. The outcomes from this work will help to strengthen and improve the current climate change mitigation and adaptation strategies. The rest of the paper is organized as follows: Section two highlights model description and the techniques employed while section three delineates results of the study.
Section four is mainly about discussions and implication of the study. Finally, conclusion and recommendations are presented in section five.

Study Area
Pakistan is situated on the western edge of South Asia, between latitudes 23°-37.5° N and 61°-78° E longitudes, covering an area of 880,940 km 2 (Figure 1). The country's terrain and geography is mainly rugged, reaching higher altitudes (8216 m high K2 peak) in northern and northwestern mountainous ranges. The Himalaya-Karakoram-Hindukush ranges are known as the water towers and lifeline of South Asia due to its vast glaciers, fresh water lakes, rivers and tributaries (Archer and Fowler, 2004). The central and southern regions are mostly hilly on the western sides, and the eastern and southeastern areas consist of Indus Plains (Sarfaraz, 2014). The majority of the area's climate is arid (60 %) and semiarid, including the vast deserts of Thal, Thar, and the Cholistan in the south; and hyper-arid lands and plateau in the southwestern part of Balochistan province (Farooqi et al., 2005).
Temperature over Pakistan ranges from as low as <0 °C in northern regions to the soaring temperatures of >35 °C over southern regions (Nasim et al., 2018). In recent decades, the seasonal trend of temperature has increased over northern Pakistan in all seasons (Asmat et al., 2018;Ullah et al., 2019a). The strong, frequent and prolonged heatwaves have persistently increased over southern, central and eastern parts of Pakistan (Ullah et al., 2019b;Sajjad and Ghaffar, 2019). Pacific Decadal Oscillation (PDO), Atlantic Multidecadal Oscillation (AMO), El Niño and La Niña phenomenon under various ocean-land-atmosphere mechanisms also influence the regional and global land surface temperature (Guan et al., 2017).

Models' Calibration and Standardization
This study utilized an ensemble of six models from CMIP6 datasets that are verified (Karim et al., 2020) to accurately simulate the observed temperature over Pakistan. The individual model description and relevant information is given in Table 1. The model's data for three SSPs: SSP2-4.5, SSP3-7.0 and SSP5-8.5, was accessed from the CMIP6 websitehttps://esgf-node.llnl.gov/search/cmip6. All the sub-members for each model were first merged then standardized for standard units and calendar time. Since all models had varying horizontal resolutions, they were regridded to a common grid of 1.4°×1.4° using the nearest neighbor interpolation technique (Vermeulen et al., 2017). This approach gives a good representation of hilly terrain (Mallika et al., 2015).
Further, monthly multimodel ensemble (MMEs) in the three SSP scenarios were generated by taking the mean of utilized models for each scenario. It is well understood that one most common applications of ensembles is to separate the external forcing from internal variability (Milinski et al., 2019). According to Frankcombe et al., 2015, models' realizations show internal variability and forced response time series. This study considered only the first realization (r1i1p1f1) for each model, based on data availability at the time of carrying this study. According to Maher et al. (2018) there is no clear consensus on how large and ensemble should be for any given application or the questions to be addressed. To understand the changes in temperature across diverse geography, the study area was divided into northern (70°E-78° and 33.6° N-37.5°N), central (60°E-75°E and 28.5°N-33.5°N) and southern (60°E-72°E and 23°N-28.5°N) regions.
To understand the temperature changes with time progression, two time periods i.e., 2025-2054 (near future) and 2070-2099 (far future), were defined from the monthly-scale MMEs. The two-time scales projected the Dec -Feb (DJF) and Jun -Aug (JJA) season mean temperature changes and trends across Pakistan.

Spatio-temporal changes in future projections
Usually, the future warming estimates and uncertainties in models and spread in projections is conveyed in probabilistic estimates like mean and range (Brunner et al., 2020). The annual temporal projections for 2015-2100 over the whole, northern, central, and southern regions of Pakistan were determined by subtracting annual projection values from the historical  value. The seasonal temporal projections over 2015-2100, near future, and far future were calculated by subtracting future series values for each specific season from the historical mean value for each specific season. The spatial JJA and DJF projections changes over the whole of Pakistan for the two timelines were determined by subtracting future seasonal mean from historical (1985-2014) seasonal mean value for each pixel.

Mann-Kendall test
Mann-Kendall (MK) trend test used to investigate and its changes (Mann, 1945;Kendall, 1975. The Mann-Kendall test has been commonly used in numerous trend studies (Iqbal and Zahid, 2014;Amin et al., 2017;Ongoma et al., 2018a;Ali et al., 2019;Ayugi et al., 2020;Ayugi and Tan, 2019) for temporal trend analysis. MK trend does not require normal distribution in datasets to handle outliers and missing values. The test hypothesis states that the trend does not exist (H0), and the trend does exist (H1) in a time series. MK correlation coefficient or Kendall's Tau establishes trend in a time series. The strength of the trend is defined by the magnitude of the MK test statistic, where greater magnitude exhibits more robust trends, and lesser magnitudes show a weaker trend.

2.5.Trend slope estimation
The JJA and DJF season spatial trends were calculated using the Sen's Slope estimation technique to see the trend changes. The Sen's Slope estimates the true slope for a linear trend in a time series (Sen, 1968).

Abrupt Changes in Trends
The Sequential Mann Kendall (SQMK) test is applied to detect trend turning points (Mann, 1945). The abrupt changes are estimated by the progressive (u) and retrograde (u') series in time series. The u values counted from first to last while u' values counted from last to first value of time series. The intersection of u(ti) and u'(ti) shows a potential turning point. If an intersection occurs at threshold level ±1.96 (95% confidence level) of z statistic, such a trend turning point is considered significant. The decreasing values of u(ti) and u'(ti) indicate a negative trend, while the increasing values represent a positive trend in the time series. When at least one of the reduced variable values is greater than ±1.96, a new trend beginning is considered (Chatterjee et al., 2016). The JJA and DJF temporal seasonal datasets for near future and far future are treated with SQMK to detect the abrupt change in the series.

Probability Density Function for mean temperature distribution
This study also focuses on the likelihood of an outcome (density distribution) for changes in MME based seasonal mean temperature changes/ dispersion from the mean value for different spatio-temporal scales. In probability density function (PDF), an integral function is applied on continuous random variables with many intervals to find the chance of occurrence for different values in the intervals (Ongoma et al., 2018b;Watterson, 2008). In this study, kernel smoothing density distribution was utilized following work by Farooqui and Soomro (1984), where kernel smoothing had an excellent ability to define and classify likely changes and rare events. The DJF and JJA seasonal data for near and far future timelines were treated for the probability density estimation over different spatial scales of Pakistan.

Annual temporal projected changes in mean temperature
The projected changes in mean temperature over different Pakistan regions show a warming tendency across all regions ( Figure 2). The projected change for the whole of Pakistan   (ECS) i.e., temperature response to CO2 doubling, and found it higher in CMIP6 models with values spanning from 1.8 to 5.6 K across 27 GCMs and exceeding 4.5 K in 10 models. This insignificant increase is primarily due to more substantial positive cloud feedback on warming from reducing extratropical low cloud coverage, water content, and the albedo process.
According to Rangwala et al. (2010), a higher warming tendency over higher altitudes may be due to increased downward long wave radiation, surface specific humidity, and absorbed solar radiation under a decrease in snow cover extent.

Projected JJA and DJF temporal changes in mean temperature
This is projected increase in mean temperature during 2015-2100, near future and far

Spatial temporal changes for near and far future
The spatial distribution of changes in projections for DJF and JJA mean temperature for the near future and far future is shown in Figures

Future seasonal temporal trends
The long-term annual and seasonal temporal trends for the near and far future are summarized in Table 2    Although the southern regions project a higher increase in trends (0.001 to 0.04 °C/year), a portion in northeastern region shows a slight decrease in trend (near to 0.01 °C/year). This decrease in trend is also in accord with its temporal trend (Table 2)      Inclusively for all the regions, the near and far future DJF mean temperature shift (far future especially) PDFs under all SSP scenarios is found to be highest over northern Pakistan.
This projects to the probable sharp increase in DJF temperature, and it also conforms to with the temporal projections (Figure 3b), temporal trends (Table 2)

Discussion
Until recent past, Pakistan has experienced a steady rise in mean temperature under the global warming phenomenon at various spatio-temporal scales (IPCC, 2014). The change in temperature reaching extreme levels has caused negative impacts on the ecosystems, human socioeconomics, and broadly the human wellbeing on large scale across Pakistan (Chaudhry et al., 2009). The impact of temperature increase is noted in other weather parameters/phenomena such as on precipitation under the atmospheric circulation like ENSO (Loo et al., 2015;Boucharel et al., 2011) on a global scale. The warming tendency of temperature is projected to continue with even higher magnitudes at seasonal and annual timescales over most Pakistan Almazroui et al., 2020b). The new CMIP6 models-based climate projections for the 21 st century show warmer temperatures than CMIP5 based projections despite identical instantaneous radiative forcing (Wyser et al., 2020).
The current study focused on projected changes in DJF and JJA seasonal mean temperature over northern, central, and southern Pakistani regions for near and far future timescales. The current study projections and trend analysis identify the winter (DJF) mean temperature in the near, and far future under SSP245 (2.0 °C and 3.6 °C), SSP370 (2.1 °C and 4.8 °C) and SSP585 (2.5 °C and 6.2 °C) scenarios will significantly increase across all Pakistan.
The summer (JJA) mean temperature in near and far future under SSP245 (1.7 and 2.9 °C), SSP370 (1.8 o C and 3.9 o C) and SSP585 (2.1 °C and 5.9 °C) is projected to increase across the central and southern region although with lesser rate of change than in DJF. Several studies (Almazroui et al., 2020b;Athar and Latif, 2018;Babar et al., 2016;Iqbal and Zahid, 2014;Sajjad and Ghaffar, 2019) found higher rates of increase in future winter and summer temperature Recently, Almazroui et al. (2020b)  On the other hand, the increase in temperature over winter will exceed that of summer season.
The seasonal changes in temperature across diverse spatio-temporal scales may have severe implications for the environment, agriculture, water security, economy, and Pakistanis' wellbeing (Fatima et al., 2020;Khan et al., 2016). Over the years, Pakistan has experienced temperature rise, drought, pest disease, health issues causing changes in lifestyle . Increased warming may also result in rainfall events, intensifying hydrological cycles in many Pakistan regions, leading to floods and droughts .
The northern region projects the highest rate of significant positive increase in winter mean temperature (Figure 3) under all SSP scenarios than in the summer (Figure 4) in comparison to the southern regions. The summer mean temperature rise is also insignificantly increasing over the northern region under SSP370 and SSP58. These results are the findings of related recent studies (Almazroui et al., 2020b;Asmat et al. 2018;Ikram et al., 2016;Rehman and Ali, 2018) which projected the highest temperature rise in the northern regions of Pakistan.
The far future JJA season however, shows cooling trends over the northeast most part of the country. This is possibly linked to the enhanced aerosol effect on model parameterization and runs (Ikram et al., 2016). Most of the northern region comprises of high-altitude terrain with renowned mountain ranges of Hindukush, Karakorum, Himalayas, and Pamirs. These mountain ranges being globally one of the most sensitive ecosystems to climate change have, in recent years shown significant warming across the region and beyond (Fatima et al., 2020;Fowler and Archer, 2006). The phenomenon of elevation-dependent warming (EDW) is observed and projected as the significant phenomenon leading to abrupt warming at higher altitudes (Rangwala et al., 2010). The EDW phenomenon is significantly increasing during winters and annual scale due to mechanisms involving snow-albedo positive feedback, nighttime cloud cover, higher midtroposphere water vapor, aerosols' concentration, and land-use changes at higher altitudes (Pepin et al., 2019;You et al., 2017You et al., , 2020. The future temperature surge under the aforementioned phenomenon of EDW will pose severe impacts on the hydrology (Hasnain, 2014), ecosystems (Schickhoff et al., 2016) and human development (Shrestha et al., 2015) of the HKH region. . Kraaijenbrink et al. (2017) found that warming under RCP4.5, RCP6.0 and RCP8.5 scenarios in Asia will result in almost disappearance of ice by 49 ± 7 %, 51 ± 6 % and 64 ± 5 %, respectively, by the end of the century. This is likely to cause rapid melting and depletion of glaciers under rapid warming will create serious implications like flash flooding, soil erosion and GLOFs over northern regions in near future and severe drought situation in far future. The resultant droughts and desertification across downstream regions will be more severe by the end of 21 st century (Abas et al., 2017;Ahmed et al., 2019b;Soncini et al., 2015). A remarkable water inflow into major reservoirs constructed downstream regions under glacier melting in the future will pose potential large-scale flooding during future spring, autumn and winter season (Haider et al., 2020).
The warming tendency across central Pakistan in this study yielded higher warming in the near and far future DJF season than in JJA under all SSP scenarios (Figures 3 and 4). The region consists of mostly plain terrain (Indus plains), deserts and hilly regions in western edges and over northern parts (Nawaz et al., 2019). This region accommodates most of country's population with vast glacier water irrigated lands and is major industrial hub of the country (Abbas, 2013). Over recent years, the winter warming rates have significantly increased across southern plains while summer mean temperature across northern and western hilly tracts of the region (Abbas, 2013;Nawaz et al., 2019). In the future annual and seasonal minimum temperature over the region is projected to increase significantly in southern Punjab region, conforming to our analysis (Fahad et al., 2016;Rasul et al., 2012;Sajjad and Ghaffar, 2019). The projected rise in temperature over central regions is caused by mainly caused by an increase in GHG emissions under chaotic hydrocarbon use in expanded industries, transportation system and other uses (Hussain et al., 2019). The population pressure, unplanned energy use, rapid and unplanned urbanization, improper waste management, unplanned industrialization, expanding transportation, change in agricultural practices and livestock management are further exacerbating the situation (Mumtaz et al., 2019). In addition to potential poverty rise under the reduction in agricultural productivity due to severe water shortage, the desertification and famine situation in the region and beyond is also inevitable (Shakoor et al., 2011).
A significant increase in DJF and JJA mean temperature, with a higher significant change rate (Table 2) in far future DJF than in JJA, is expected over southern Pakistan (Figures 3 and 4).
The near and particularly far future DJF (Figure 7) project a significant temperature over the southeastern and southwestern corners over the south. While the far future JJA trends (Figure 8) also project a considerable increase in temperature over the southwestern fringes of the country.
The southern region is already under extreme temperature changes in recent years, particularly over southeastern parts (Ahmed et al., 2019a;Ali et al., 2019). Moreover, Ikram et al. (2016), Rehman and Ali, (2018) and Sajjad and Ghaffar (2019) projected an increase in JJA mean temperature over southern Pakistan under RCP4.5 and RCP 8.5 after 2050 onwards to the end of the 21 st century. The southern region is mostly arid with plains and deserts in the centre, rugged mountains in northwestern fringes, and warm coastal regions (Rasul et al., 2012). The agriculture sector is the economic backbone, is severely under threat of backlash due to water scarcity under temperature rise and precipitation decreasing trends.
Further depletion of groundwater resources and others is eminent in future climatic conditions (Abbas et al., 2018). Over the southeastern and coastal regions of Pakistan, intense, prolonged-lasting, and frequent heatwaves, aridity and drought conditions are projected in the coming years (Haider and Adnan, 2014;Khan et al., 2019). The rising temperature in summer over southern regions may enhance the sand and dust storms intensity and frequency (Sajjad and Ghaffar, 2019).
The surface temperature severity to CO2 increase in the CMIP6 models has improved substantially (Tokarska et al., 2020;Forster et al., 2020). Such an increase is due to the varying representation of low cloud coverage and water content under strong warming with further enhanced sunlight absorption by CMIP6 models (Zelinka et al., 2020). Historical interdecadal mean surface temperature variability in CMIP6 models is greater and is associated with regional variability in tropical deep convective (forced) regions. This defines how, when, and where the warming effect will be felt since climate change progresses as a combination of internal and forced changes in the future (Parsons et al., 2020). The future warming estimates, uncertainties, and projection spread are better represented in mean and ranges (Brunner et al., 2020).
Developing future predictions based on models' independence and performance skills for historical simulations is quite skillful in obtaining future changes estimates (Brunner et al., 2020;Eyring et al., 2019). Climate predictions being uncertain due to initial conditions and computational representational representation of equations in individual models can be better presented by the multimodel ensembles (Palmer et al., 2005).
The impacts of anthropogenic activities on surface temperature are essential in understanding the present and future climate change, environment, and sustainable development (Siddique et al., 2020). In Pakistan, rapid growth in transportation, industrialization, urbanization, deforestation, waste, agriculture, livestock, and energy use are the main drivers of GHG emissions and will result in countrywide temperature rise . By 2050, GHG emissions contribution by energy and agriculture sector may rise to 59.1 %, 38.2 %, while for industry, land use/land use change-forestry, and waste management sector will be 5.4 %, 2.3 % and 1.6 % respectively (TCFD, 2018). Pakistan's population may grow up to 350 million by the year 2050, particularly in rural areas and around small cities (Survey, 2018). Due to promising economic opportunities, a large section population is bound to migrate to major commercial and industrial hubs (Abas et al., 2017;Abbas, 2013). Conclusively, future population pressure, fossil fuel-based industrialization, transportation, commercialization, and land use changes may intensify the GHG emissions and temperature rise over Pakistan. It is necessary to study the change and variability in drivers of GHG emission and temperature rise over Pakistan to understand future temperature warming-induced impacts in Pakistan. Moreover, it will also help develop a comprehensive, national, and regional level environmental, economic, disaster management, and climate change adaptation and mitigation policies.

Conclusion
The future seasonal mean temperature projections and trends are essential to identify its future spatiotemporal distribution and variability and its impact on the environment and humans.
This study employed an ensemble of models to reduce output biases and errors in mean

Conflicts of Interest:
In a unanimous agreement, all authors declare no conflict of interest in the present study.