Trend analysis of historical and future precipitation projections over a diverse topographic region of Khyber Pakhtunkhwa using SDSM

Climate change has adversely affected the development of countries by influencing the physical as well as socio-economic sectors irreversibly (Rasul et al. 2012; Gomez-Zavaglia et al. 2020). The major consequence of changing climate is the alteration in the distribution of climatic parameters including temperature and rainfall. The Intergovernmental Panel on Climate Change Fifth Assessment Report (AR5) revealed that mountain ecosystem is significantly vulnerable to these climatic changes (IPCC 2013). The fluctuations in precipitation distribution may affect crop production and water availability in the future (Araya-Osses et al. 2020). The temperature changes have transformed the hydrological cycle which has resulted in variation in rainfall spatially as well as temporally across the world (Sa'adi et al. 2017; Khan et al. 2020). During the past thirty years, the extent, duration, and occurrences of extreme climatic events have risen and exacerbated in the future due to an increase in anthropogenic activities (IPCC 2013). The variation in land and sea surface temperature has also altered the atmospheric moisture and ultimately these effects can be perceived in monsoonal rain in the south Asian region (Thomas et al. 2015). The occurrence of these climatic extremes is uneven over the world; some global regions are more susceptible to these changes and Pakistan in particular has experienced frequent floods, drought, heavy storms, and heat waves during the past two decades (Ali et al. 2019a).

At global level, the temperature fluctuation has changed the rainfall pattern in terms of trend and magnitude (IPCC 2014; Ullah et al. 2018; Moazzam et al. 2020, 2022; Khan et al. 2021) and the same phenomenon was also found in Pakistan. In the scenario of climate change disastrous impacts, Pakistan holds the fifth position in the world (Eckstein et al. 2019). Pakistan receives an ample amount of precipitation during two different winds patterns. During winter, most of the precipitation is received from western disturbances which originate from the Mediterranean Sea and enter into Pakistan causing rainfall over the lowlands, whereas snowfall occurs over the high altitudes. Contrary to winter, during the summer season, monsoon originates from the Arabian Sea and Bay of Bengal which causes precipitation in the eastern section of Pakistan. Similarly, winter depression nourishes the glaciers in the hilly regions, which recharge river systems in various parts of the Khyber Pakhtunkhwa province. It has been estimated that Pakistan receives almost 60% of the annual rainfall during the monsoonal season, i.e., July–September. In a country like Pakistan, about 60% of the total land surface is arid receiving less than 250 mm rainfall annually, whereas 24% of the land surface comprises the semi-arid region and receives 250–500 mm rainfall, and the remaining 16% fall in the humid-to-sub-humid region (Rasul et al. 2012). Such variations in precipitation have on one hand created massive flood events, while on the other hand, the occurrence of extreme drought situations in the country (Rasul et al. 2012). The Khyber Pakhtunkhwa province has experienced the century's worst flood of July 2010 (Rahman & Khan 2013) and several severe drought events (Rahman et al. 2018). Generally, temperature and rainfall are considered as suitable parameters for climate change prediction (Zhong et al. 2020). Therefore, the projection of future climatic patterns in Khyber Pakhtunkhwa is essential to develop adaptation strategies and policies to cope with extreme climatic events.

Researchers are using different statistical trend tests for detecting trends in meteorological and hydro-meteorological time series parameters and these trend tests are broadly classified into two main categories, i.e., nonparametric and parametric (Harwell 1988; Dahmen & Hall 1990; Zhang et al. 2006; Chen et al. 2007; Kaur & Kumar 2015; Rahman et al. 2021b). In these, the parametric is considered more powerful but needs normally distributed and independent data, which is rarely applicable on meteorological data (Ahmad et al. 2015). Working with meteorological and hydrological time series data, the commonly used nonparametric tests are the Mann–Kendall (Mann 1945; Kendall 1975) and Spearman's Rho (Lehmann & D'Abrera 1975; Sneyers 1991). Among these, researchers have used Mann–Kendall very commonly in studying trends in meteorological data (Hamed & Rao 1998; Shadmani et al. 2012; Gocic & Trajkovic 2013; Ahmad et al. 2015; Rahman & Dawood 2017; Phuong et al. 2020; Ullah et al. 2020; Rahman et al. 2021a; Rafiq et al. 2022a, 2022b), while in comparison, Spearman's Rho is less commonly used in detecting monotonic trends in time series meteorological data (Yue et al. 2002; Gocic & Trajkovic 2013; Rahman et al. 2017, 2021b).

The General Circulation Models (GCMs) provide climate projections at the global scale for policymakers to adapt better strategies for climate change. These are the mathematical models used at global and regional scales for future climate projections (Munawar et al. 2022b). The GCMs represent significant outputs at the global, hemispherical, and continental scales by incorporating global system complexity that incorporates energy interactions among the spheres (Hassan et al. 2014). GCMs are the large-scale gridded model data and when used at the small-scale local/regional level have certain uncertainties. Despite improvements in the GCMs efficiency to represent climate processes in better ways these uncertainties cater to produce better climate projection but remain a subject of ample concern at the regional/local level (Hussain et al. 2021). To cope with these uncertainties, downscaling techniques have been widely applied (Munawar et al. 2022a). GCMs' coarse spatial resolution limits their use for sustainable environmental management (Araya-Osses et al. 2020). The downscaling techniques are being employed to transform the GCMs coarse spatial resolution to fine spatial resolution for their application at the local/regional scale (Ali et al. 2021). Two widely used downscaling models, statistical downscaling and dynamical downscaling, are used to relate GCMs predictors and local climatic variables (Baghanam et al. 2020). Statistical downscaling is a flexible and computationally efficient approach to downscale GCMs at the local/regional level (Hamlet et al. 2020). Statistical downscaling develops a connection among GCMs large-scale predictors and local scale predictands without requiring the physical knowledge at the local region (Fowler et al. 2007). The Statistical Downscaling Model (SDSM) has been widely applied by the researchers to downscale GCMs for climate projections at the local/regional scale (Hasan et al. 2018; Tahir et al. 2018; Ali et al. 2019a; Munawar et al. 2021b, 2021a).

The study pertaining to variations in precipitations and their trend and projections using GCMs is of great significance for the sustainable agricultural activities and development of socio-economic sectors. In Khyber Pakhtunkhwa, out of the total geographical area a mere 1.6 million hectares (17.86%) is under cultivation including formerly Federally Administered Tribal Areas (FATA). In the study area, out of the total arable land, more than half is rainfed. Rainwater harvesting is getting uncertain due to changing climate, as in rainfed areas, agricultural production totally relies on timely rainfall occurrence. In order to effectively address these issues, there is a need to extend irrigation facilities to rainfed areas by storing storm/rainwater and supplying assured water on crop demand. Therefore, this study focuses on the rainfall pattern and trend based on the historical baseline period (1971–2018) as well as on the projections of precipitation using GCM and SDSM for the 21st century in Khyber Pakhtunkhwa province.

Generally, during winter months, lofty mountain ranges collect an ample amount of precipitation as snow that is considered a prime source of clean water for the perpetual rivers system. In the northern mountainous regions of Pakistan, an ample amount of rainfall has been collected, whereas dry climatic conditions are mostly observed in relatively plain areas of the KP because of insufficient rainfall spells. A slow decline in rain and snowfall has been experienced in the northern region of Khyber Pakhtunkhwa, which has ultimately affected the river flow and has adversely affected various sectors including agriculture, the major source of livelihood earnings. Pakistan's economy relies on agriculture and its share of the GDP is 38%, whereas 44% of the population directly relies on agriculture for their livelihood (Rahman et al. 2021b).

To calculate and analyze the rainfall variability, trend and its distribution in Khyber Pakhtunkhwa province, the temporal data concerning rainfall was provided by Pakistan Meteorological Department (PMD) Peshawar. In Khyber Pakhtunkhwa, there exists a total of 15 meteorological stations. Initially, the daily rainfall data were computed and converted to monthly, seasonal and total annual. For future projections of precipitation for the 21st century, GCM CCSM4 (Community Climate System Model version 4) has been selected following the methodology as presented in the author's previous research (Munawar et al. 2022a). The daily large-scale predictors of GCM CCSM4 for RCP4.5 and RCP8.5 and the daily predictors for NCEP (National Centers for Environmental Prediction) were downloaded from a website (http://climate-scenarios.canada.ca/). The spatial resolution of daily large-scale predictors of GCM CCSM4 is 0.9° × 1.25°. The daily large-scale predictors of GCM CCSM4 and 26 NCEP predictors were re-gridded at the same spatial resolution to exclude spatial differences for further use in the SDSM model. The statistical techniques applied in the study can be described as follows.

In Equations (1) and (2), are the records of the temporal data in the years and j, ‘’ is the duration of the temporal data, is significantly the number of ties for the th value, whereas ‘’ is the number of tied values. In the calculation, the positive and negative output value of indicates increasing and decreasing trends in the rainfall temporal records. In rainfall time series data, a noteworthy trend was found when the value is greater than which is basically the critical value of ‘’ from the standard normal table that is 1.96 for a 95% confidence level.

In Equation (5), denotes the rank of the observation in the time series data, while ‘’ denotes observations total number. The resultant positive value expresses the rising trend whereas the negative values specify the declining trend in the rainfall temporal data. When the H_o will be rejected and contrary to this H₁ will certainly be accepted, which point out that there exists a significant trend in data. Moreover, is the critical value of ‘’ from the t-student table. In the present research, 48 years of rainfall data were used, and the t value for 95% significance value is 2.012.

The statistical downscaling method (SDSM) is applied on CCSM4 using the combination of stochastic weather generator (SWG) and multiple linear regression (MLR) to build a relationship between observed and modeled data (Munawar et al. 2021a). The MLR combined NCEP predictands and local predictors by applying empirical/statistical association to generate regression parameters that simulate precipitation at monthly and annual time scale. NCEP and GCM predictors along with regression parameters are processed using SWG (Mahmood et al. 2015). The selection of NCEP predictors is the key to success for SDSM process and this selection is made using the absolute partial correlation coefficient (abs. Pr) (Munawar et al. 2021b). SDSM is calibrated using the conditional sub-model for precipitation due to its complex nature and uneven distribution. The NCEP predictors used for precipitation are summarized in Table 1.

The SDSM performance evaluation for the period (2006–2018) for the observed station data and simulated biased corrected downscaled data for RCP 4.5 and RCP 8.5 is presented in Table 2.

The study region has basins of some significant rivers such as Indus, Chitral, Swat, Panjkora, Kurram, Tochi, Gomal, Kunhar and Kohat Toi. Hence, the current region is crucial in the context of hydrology and agriculture. Similarly, both hydrology and agriculture possess an intimate association in the scenario of rainfall variability patterns. Indeed, rainfall is a recurrent process, particularly in areas holding diverse physiography. In Khyber Pakhtunkhwa, the rain pattern is extremely variable in space. For example, at Malam Jabba, the highest average annual rainfall (1,732 mm) was documented closely followed by Balakot (1,567 mm) and Dir (1,400 mm) meteorological stations, respectively, during the study period. These stations lie at very high altitude and at the same time collect rain from both the monsoon and western disturbances. The comparatively low average annual rainfall was verified at Drosh (562 mm), Chitral (448 mm), Bannu (361 mm), and D.I. Khan (314 mm) meteorological stations (Table 1). The D.I. Khan and Bannu are located in comparatively plain areas experiencing arid and semi-arid climate where slight rainfall is collected during the monsoon months, whereas Drosh and Chitral meteorological stations lie in a rain-shadow area at higher altitude where little rainfall occurs.

The meteorological stations such as Balakot and Kakul are located in the Himalaya region, while Saidu, Malam Jabba, Kalam, Dir, Timergara Pattan, Drosh and Chitral are in the Hindu Kush region. Contrary to this, Peshawar, Bannu, Cherat and D.I. Khan are positioned in the plain areas of the province. The meteorological station, i.e., Parachinar is found in the Koh-e-Safed mountain ranges. In Khyber Pakhtunkhwa, 60% of the annual average rain is received from the monsoon winds in the summer months from July to September. Furthermore, with the advent of the monsoon that enters the province from the eastern direction, the eastern sections, particularly the mountain systems, such as the Siwalik, lesser Himalayas and a few portions of the Hindu Kush ranges, collect enough rain. Therefore, in the scenario of the seasonal distributional map, the meteorological stations that received the highest amount of rain are Balakot, Pattan, Kakul, Kalam, Saidu, Malam Jabba, Dir and Timergara. The lower-most summer rainfall-collecting meteorological stations are Drosh and Chitral due to great differences in terms of monsoon winds. Moreover, the winter and spring months mostly collect rain from the western depressions. The specific winds that come from the Mediterranean Ocean finally enter in Khyber Pakhtunkhwa province from the western side via Afghanistan. Thus, the stations that are situated in the western section collect relatively more rain as compared with the other seasons of the year. Additionally, the rainfall occurrence is high at Malam Jabba, Dir, Kalam, Kakul and Balakot as these stations are located at comparatively high elevations in the study area.

The month of February expresses diverse results as in a few of the meteorological stations a rising rainfall trend has been seen; however, some of them lie in the northeastern part which indicates a negative rainfall trend and at the same time is insignificant. The single significant rising trend was observed at Parachinar, whereas the significant decreasing trend in rainfall was documented at Pattan meteorological station. In the month of March, the significant decrease was noted at Saidu and Balakot and the significant increase (positive trend) was verified at Parachinar meteorological station.

In April, at Parachinar meteorological station using SR results, the significant increasing trend was observed, whereas the same results were not documented as significant using the MK test. In May, the significant rise in rainfall was once again found at Parachinar meteorological station; however in June, the significant increase in rainfall was documented at Peshawar, Saidu and Parachinar. The Balakot meteorological station results showed the significant decrease in the amount of rainfall in July while in August, the same results were observed in Kakul. The Kakul and Balakot falls in the Himalayas in the northeastern section of the Khyber Pakhtunkhwa, where from monsoon, an ample amount of rainfall occurs. In SR results, the Pattan meteorological station observed a significant decline in rainfall in September, whereas in the months of October and November, the Parachinar meteorological station recorded a significant rise. The worrying condition was observed in December where all meteorological stations recorded a negative trend, though the significant declining trend was only noted at Drosh from the already acquired results of SR and MK (Figure 2).

The annual rainfall records show the trend of significant decline based on MK and SR results at Kakul, Balakot and Kalam meteorological stations, whereas a significant rise was noted at Parachinar station (Table 4; Figure 4). At Pattan station, there was recorded a significant decreasing trend by getting the results from SR; however, by applying the MK trend test, the trend value was recorded as negative and insignificant in the case of annual rainfall which might be because of the short period data as this meteorological station was established in the year-2010.

The projections for the summer and winter seasons depicted an increasing trend till the mid-century but the latter half of the century projected a decline. These seasonal changes and shifting patterns may cause flooding and drought in the province and to cope with these catastrophes better water management strategies for the 21st century. The variations in precipitation patterns are more abrupt and spontaneous for RCP-8.5 than the RCP-4.5. The shifting seasonal patterns of precipitation may also influence the precipitation forms. The lower precipitation during the winter season is directly related to the decrease in snow accumulation that affects the glaciation process of the lofty mountains of the province. This factor of deglaciation directly disturbs the hydrosphere of the province as the major rivers flowing under the influence of glaciers from the lofty mountains to the plain valley area of the province. The shift in precipitation forms directly influences the cryosphere and hydrosphere of the province and indirectly affects the socio-economic activities of the low-lying population of the province. These seasonal climate change projections for the 21st century may help policy makers to adapt and mitigate climate changes.

In Khyber Pakhunkhwa, the highest annual rainfall was recorded in Malam Jabba (1,732 mm) followed by Balakot (1,567 mm) meteorological station, while the lowest recorded annual rainfall was in D.I. Khan (314 mm) and Bannu (361 mm) meteorological stations (Rahman et al. 2021a). This high variation in rainfall is attributed to the variation in altitude, where Malam Jabba is located at a higher elevation thus experiencing a humid climate, whereas D.I. Khan is located in the plain area where the climate is arid (Rahman et al. 2021b). In Pakistan, 60% of the rainfall occurs from the monsoon winds especially in the northeastern mountains, i.e., Siwalik, Lesser Himalayas and some portions of the Hindu Kush region receive more rainfall (Rasul et al. 2012). Therefore, the meteorological stations situated in the northeastern mountains (Balakot, Pattan, Kakul, Kalam, Saidu, and Malam Jabba) receive more rainfall in the summer season. Moreover, rainfall occurs in the region from the western depression during the winter and spring season. The monthly, seasonal and annual trend was analyzed using SR and MK trend models. The monthly trend test results showed a reduction in the amount of rainfall in January in most of the meteorological stations though only a significant trend was recorded at Balakot at 95% significance level. There was a significant increasing trend in rainfall in February and March at Parachinar meteorological stations, whereas a significant decrease was observed at Saidu and Balakot in the month of March (Table 3; Figure 2). The Parachinar meteorological station recorded a significant increase in the amount of rainfall in April and May which indicates that there is a consistent rise in the amount of rainfall which may be attributed to an increase in rainfall in this region due to some changes in the local climate. The Peshawar meteorological station observed a significant increase in the amount of rainfall in June which may be the result of the monsoon early onset (Ali et al. 2019b). A decrease in rainfall was observed in the meteorological stations located in the eastern mountains (Balakot, Kakul, and Pattan) in the months of July, August, and September, which indicates a decline in the amount of monsoon rainfall in this region. The same decline was observed in seasonal analysis (summer season) with a significant decrease in Balakot, Dir and Pattan meteorological stations (Figure 3; Table 4). Previous studies also showed a declining trend in summer monsoon rainfall in the South Asian region (Imran et al. 2014; IPCC 2014; Ali et al. 2019b; Safdar et al. 2019). Wang et al. (2019) examined and observed the existence of a significant negative relationship of Tibetan Plateau heating and monsoonal rainfall during July–August over Pakistan. Furthermore, the winter rainfall which is received from the west in Pakistan, originating from the Mediterranean and Atlantic Ocean, traveling through Iran and Afghanistan (Iqbal & Athar 2018), has indicated a decline over the northern mountainous region of Pakistan (Table 4; Figure 3). Decline in winter precipitation is also observed in a previous study conducted by Iqbal et al. (2019) and this change in winter precipitation is highly associated with changes in sea surface temperature (Khan 2004), while Malik et al. (2012) associated this variability to the shifting of 200-mb zonal winds over the area. The precipitation decreasing trend in winter is a warning of serious circumstances for a country having an agricultural based economy like Pakistan as the flow of water in the rivers found in the vicinity totally depends upon the snowfall in this specific season that supplies water in the rest of the months. The locale of the meteorological stations that holding decreasing trends value is considered the home of numerous glaciers. The precipitation was projected for the 21st century using de-biased downscaled GCM simulation for RCP-4.5 and RCP-8.5 and divided into four temporal divisions 2021–2040, 2041–2060, 2061–2080, and 2081–2099. The results showed an increasing precipitation trend, but the positive change was more obvious in the RCP-4.5 with 7.5% more than RCP-8.5 for the 21st century. A more positive increase was observed in the northeastern mountains as compared to the western part of the study area. The northern and southern parts of the province are under arid to semi-arid conditions where the precipitation is less than 500 mm. The aridity pattern is more dominant for RCP-8.5, especially for the latter half of the century. The precipitation patterns in the region have depicted that an increasing trend is dominant in the eastern part of the province in comparison to the western part, which has been also deduced in the previous study conducted by Ali et al. (2021).

The analysis concluded that there is high spatial variability in rainfall patterns which may be attributed to the local orographic impacts. The second major reason is the sparse distribution of meteorological stations that further poorly grip the spatial variability. The future precipitation projection for the 21st century depicted that monsoon precipitation patterns will dominate the region more than the westerlies. The north and south regions of the province are the more affected regions for the aridity as compared to the eastern and western boundaries that are influenced by the monsoon and westerlies wind patterns. The seasonal precipitation projections depicted shifting of precipitation patterns for the 21st century. The monsoon pattern indicated an early precipitation shifting towards the spring season and similarly, there has been a decreasing precipitation pattern for the winter and summer seasons. These shifting seasonal precipitation patterns may cause water issues for the province; therefore, adaptation and mitigation strategies are required to deal with these changing precipitation patterns.

The authors declare that they have no competing interests for the research.

No external funding agency was involved in this research.

GR designed methodology and conceptualization in writing and editing, AR supervised and reviewed the manuscript, SM worked on GCMs and SDSM, MF helped in data collection and writing while MD worked on spatial analysis and mapping. MM and SP critically evaluated and proofread the article.

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.