Observed and Predicted Precipitation Changes in Pakistan Using Ground Observations, Satellite Data and Model Projections with Special Focus on Winter and Pre-Monsoon Precipitation

Fasiha Safdar (  fasiha.safdar1@gmail.com ) NUST: National University of Sciences and Technology https://orcid.org/0000-0003-1717-443X Muhammad Fahim Khokhar NUST: National University of Sciences and Technology https://orcid.org/0000-0003-4489-6593 Fatimah Mahmood NUST: National University of Sciences and Technology Muhammad Zeeshan NUST: National University of Sciences and Technology Muhammad Arshad NUST: National University of Sciences and Technology


Introduction:
Precipitation is a natural meteorological phenomenon and is a vital component of water cycle as well as energy balance of the earth; it also contributes a signi cant part in the development of global and regional climate (Kumar et al., 2017;Zhang et al., 2016;Rahman et al., 2012). Precipitation has a considerable spatial and temporal variability which renders studying and modelling rainfall as one of the most complicated phenomenon; thus highlighting the importance of precise precipitation measurement inputs for accurate hydrologic forecasts (Kumar et al., 2017). Accurate estimates of precipitation with good resolution, both spatial and temporal, are of utmost importance various hydrometeorological and water resource management applications like study of crop yields, extreme weather events, ood prediction and monsoon climate variability, particularly in data-scarce areas and regions where there is a strong struggle to utilise sparse hydrological resources (Rahman et al., 2012;Zhang et al., 2016a;Cao, Zhang and Wang, 2018). Reliable statistics of rainfall also have a vital part in model initialization, data assimilation and model veri cation for numerical weather prediction (Prakash et al., 2015).
The three main data sources of precipitation are point measurements from rain gauge based on ground stations, radar weather observations and remotely sensed satellite-based precipitation estimates (Kumar et al., 1992). Among these, the most accurate source of rainfall measurement are the ground-based rain gauges, but these stations are often scarce or unequally dispersed due to geographical, climatic, nancial or other limiting factors. Hence they are not representative for various locations, especially in the developing countries with less number of weather stations. Similarly, coverage of weather radar-based estimates for rainfall is also not dense and representative in many parts of the globe. Therefore, satellite-based precipitation estimates present numerous bene ts over weather radar and ground-based observation, for example long time and continuous coverage, easy data acquisition, suitable temporal and spatial resolutions and less disruptions due to variability in climate and terrain of the area (Kumar et al., 1992;Zhao et al., 2017).
As an attempt to increase precision of satellite based precipitation measurements, Tropical Rainfall Measuring Mission (TRMM) was launched by NASA and Japan Aerospace Exploration Agency in 1997, along with precipitation radar (PR) and TRMM microwave imager (TMI) (Kummerow et al. 1998). Following the launch of TRMM, microwave-based precipitation retrieval techniques progressed immensely, succeeded by launch of numerous high-resolution multi-satellite products of precipitation and a wealth of data has been obtained since then (Semire et al., 2012). However, multi-satellite precipitation products are also impacted by random and systematic errors which can vary with season and location, occasionally limiting their usage at both regional and global level (Xu, Niu and Shen, 2014;Liu, 2015;Prakash et al., 2015).
Rapidly changing climate puts an important in uence on precipitation in various parts of the world, as rainfall patterns are affected by changing climatic parameters like greenhouse gases and temperature, which ultimately may manifest in extreme events of either ooding or drought (Cubasch,  . Pakistan, in lieu of being located in South Asia is projected to be one of the most vulnerable and worst hit counties by the dire consequences of climate change, reasons being the geographical features and reliance on agriculture and natural resources for livelihoods (Islam, Sultan and Afroz, 2009). Water distribution and consumption predominantly relies on monsoonal rainfall, but the natural rain is unevenly distributed in time and space resulting in oods due to erratic rainfall in one region while drought in another region of the country (Hussain, Nabi and Wu, 2021).
In India, the largest country of South Asia, almost one-third of total annual precipitation falls in winter (December, January and February) through east-ward moving extra-tropical winds system called 'western disturbances' (Dimri et al., 2016) whereas summer monsoon rainfall system, the most important rainfall system which affects Indian sub-continent from June to September, is a component of global circulation system which also assists in regulating the temperature of Earth (Safdar et al., 2019). Similarly, annual precipitation received in Pakistan can also be divided into two main seasons, monsoon precipitation in summer months and winter precipitation, extending from December to February (Salma, Rehman and Shah, 2012). Various studies over South Asia and Pakistan have reported that rainfall in monsoon has declined after 1970s (Ashfaq et  Winter rainfall is important for the Indian sub-continent because the projection of in ow of water in summer for rivers like Indus is directly linked with the volume of winter precipitation in Himalayas Karakoram Kindukush (HKH) region. Hence, precipitation measurements and forecasts are crucial for winter and successive spring season for agricultural practices in the Indus plains which are predominantly dependent on irrigation as well as in rain fed northern and northwestern fruit growing regions of Pakistan where winter and spring rain has a critical impact (Ahmad et al., 2015). Any noticeable variation in water supply from these northern upstream reservoirs along with below average downstream rainfall results in damages to the food security and huge losses for Pakistan's economy (Akhtar and Athar 2019).
In this study, we investigate the seasonal precipitation variability over Pakistan, using three precipitation products from three different sources: gridded station data from Pakistan Meteorological Department (PMD), satellite-derived data from Tropical Rainfall Measuring Mission (TRM) and Community Climate System Model 4 (CCSM4) projections.

Page 4/26
The study area for this research is Pakistan shown in Figure.1, a country located in South Asia at coordinates of 24.3539 to 35.9186° N and 61.74681 to 75.16683° E making it a country of the temperate zone. Generally, climate of Pakistan is arid being hot in summers while cool or cold in winters. There is a large deviations between extremes of precipitation and temperature at different locations because of varied landscape and terrain from north to south. Study of Pakistan's precipitation patterns is important not only due to its distinctive geographical location and complex geomorphology, but also because of being part of the South Asian monsoon system that has a huge in uence on the global and regional water cycle and climate change (Rasul and Chaudhry, 2010).

Ground stations data
Pakistan is divided into 5 climatic zones according to its climate and geographical features, each with their distinctive  well they represent the reality on ground and future precipitation estimates have been studied to observe the estimated precipitation trends till the end of this century.

Statistical indices and validation methods
The statistical methods employed for the current study are brie y explained below; i. Mann-Kendall Test Mann-Kendall test has been used in this study to investigate whether a time series for precipitation has a monotonic upward or downward trend. It is extensively used in studies pertaining to climatic parameters (e.g. Latif  Where se is the square root of the variance of S (var).

ii. Sen's Slope Estimator
Sen's slope estimator has been used for calculating the slope of the trend line (Sen, 1968   sample t-test has been used to check the signi cance of variation in winter and pre-monsoon precipitation. Table 3 gives an overview of winter precipitation while Table 4 gives an overview of variability of pre-monsoon precipitation in Pakistan. There is a decrease observed in seven out of eleven years for winter precipitation during 2008-2018.    Table 5 and monthly maps in Figure 9, with TRMM slightly underestimating rainfall in May, November and December. RMSE and MBE values also depict good coherence between the two datasets, which makes TRMM_3B43 a good alternative for ground stations for precipitation studies over Pakistan and similar regions with limited ground-based observation.  ( Figure 4 and Table 2). The pre-monsoon rainfall in Zone C has also decreased unlike Zone A and B. This decrease in winter and pre-monsoon precipitation in Zone C can have an adverse effect on crops which rely mainly on rain water as the area has already been hit by two repeated droughts in the recent decades (Rasul, 2008;Naz et al., 2020).
Zone A and Zone B are primarily effected by winter and pre-monsoon rains ( Figure 10)

Spatial shift of winter and pre-monsoon precipitation in Pakistan
An attempt has been made to map the area affected by winter and pre-monsoon precipitation in Pakistan. A related study by Safdar et al., (2019) drew a conclusion that area impacted by summer monsoon has decreased in Pakistan, but there is no study that highlights the spatial extent and shift of winter and pre-monsoon precipitation over Pakistan. TRMM monthly rainfall product TRMM_3B43 satellite data has been used to plot the area that has received rainfall according to two set criteria: a) spatial extent of precipitation >1 mm/day and b) spatial extent of precipitation>2.5 mm/day as shown in Figure   11. All the pixels above the threshold values are used to discriminate such changes.
The average area of years 2010 -2018 have been compared to the average of available years of TRMM data i.e. 1998-2018 for this purpose. The results are presented in Figure 11 and Figure 12 whereas Table 6 gives the summary of the area increased or decreased. In terms of precipitation amounts, the area of winter precipitation has also decreased during the last 9 years as compared to the baseline years. This decrease is 66,000 km 2 during 2010-2018 for precipitation greater than 2.5 mm/day (75 mm/month). Whereas the area effected by pre-monsoon precipitation greater than 2.5 mm/day (75 mm/month) has increased by around 60,000 km 2 during 2010-2018. This nding complements the trend of decreasing precipitation in the winter season and increasing in the pre-monsoon season in zones A and B of Pakistan, according to the ground observation of PMD. been extracted over Pakistan and has been investigated for its accuracy over the region by comparing it to ground based observations (PMD data), and precipitation changes has been calculated for two time spans 2040-2049 and 2090-2099. As evident from Figure 13 and Table 7, the model generally over-estimated precipitation over Pakistan except over Zone A for RCP 4.5 scenario and zone D for RCP 8.5 scenario. Table 7 depicts the error analysis of the observations and forecasts.  The projections from CCSM4 for RCP 4.5 estimate an increase in average precipitation over Pakistan by 1.41 mm/day and 1.28 mm/day for 2040-49 and 2090-99 periods respectively as shown in Table 8. The rise in precipitation is 0.68 mm/day and 0.55m/day in RCP 8.5 for these periods. Most of this increase is seen to be occurring in Zones A and Zone C till 2049 and also for Zone E for years 2090-2099.

Conclusions
There is a decreasing trend in winter precipitation in all zones of the country with a signi cant decrease in western mountains i.e. Zone C of the country. Maps depicting monthly precipitation over Pakistan (1998-2018) through satellite and ground observations for all the months (a -k). Comparison between PMD ground data and TRMM_3B43 satellite data monthly precipitation for selected stations is shown by the same color in the legend for both the data sets. Both the data sets show a strong correlation with Pearson's correlation "r" values ranging from 0.89 (November) to 0.97 (July and August) Figure 10 Distribution of average winter and pre-monsoon precipitation in Pakistan from 1998-2018 Figure 11 Spatial extent of winter rainfall in Pakistan; There is a decline in area equating to 34,474 km2 for precipitation ≥ 1mm/day and a fall of 66,318 km2 in area with precipitation ≥ 2.5 mm/day Figure 12 Spatial extent of pre-monsoon rainfall in Pakistan; There is an increase of 107,365 km2 in area with precipitation ≥ 1mm/day and a decrease of 60,163 km2 in area with precipitation ≥ 2.5 mm/day