Recent Rainfall Variability Over a Dryland Ecosystem of North Western India

In this study, an attempt has been made to examine the recent rainfall variability by means of daily rainfall data of 33 well spread stations over dryland ecosystem of Rajasthan in north western India during 1961-2017. For trend analysis, Mann-Kendall, Sen’s slope estimator and simple linear regression test have been used (at 95% con�dence level). The results have shown a high interannual variability in rainfall occurrence varying from 277 mm (in year 2002) to 839 mm (in year 1975) with mean of 583 mm over this dryland ecosystem. Most of the rainfall de�cit years have occurred with El-Nino years. The mean annual rainfall has shown a marginal non-signicant upward trend over the ecosystem. The station-wise mean annual rainfall has revealed a signi�cant rising trend over Barmer, Churu, Ganganagar, Jaisalmer and Pratapgarh stations. Interestingly, three year running average has shown a cyclic pattern of rainfall over dryland ecosystem under the changing climatic conditions. The spatial pattern has exhibited that the mean annual rainfall decreases from east and south east (more than 850 mm) to west and north west (less than 400 mm), which is mainly associated with the presence of Aravalli Mountains spreading north east to south west in central Rajasthan. Remarkably, majority of stations positioned in western parts of dryland ecosystem have shown increasing rainfall trends, whereas some stations located in eastern parts have recorded a non-signicant declining trend. The magnitude of signi�cant rising trend has varied from 5.34 mm/year (Pratapgarh station) to 2.17 mm/year (Jaisalmer station). Also, the frequency of heavy rainfall events has shown a positive trend with signi�cant increasing trends over Bharatpur, Jaisalmer and Pratapgarh stations, whereas Bundi station has shown signi�cant decreasing trend.


Introduction
Rainfall and temperature are vital climatic parameters, which have been frequently used to identify the alterations in global climatic conditions (Mayowa et al. 2015;Sa'adi et al. 2019).Global ocean's and land surface temperature trend has revealed a warming of 0.85°C (ranging between 0.65°-1.06°C)during 1880-2012 (IPCC 2014).This rise in the surface temperature could result in changing rainfall patterns globally (Wang et al. 2016).For instance, Trenberth et al. (2007) have observed a rising trend in rainfall over Asia, Australia, northern Europe, North and South America, whereas decreasing trend over the Mediterranean area, southern Asia, Sahel, western and southern Africa.Likewise, Longobardi  Additionally, it has been well-recognized that the warming environment has enhanced the intensity of extreme precipitation more rapidly than mean precipitation (Kharin et  In India, a number of efforts have been made to investigate the spatial and temporal trends in rainfall and related extremes.For instance, Guhathakurta and Rajeevan (2006) have not observed any trend in south west monsoon season rainfall over Indian region, but noticed substantial regional differences.Naidu et al. (2009) have observed a rise in summer monsoon rainfall in south India, whereas a decrease in northern India.Nonetheless, for India as a whole, the summer monsoon rainfall has not shown any trend (Kumar et  The above review exhibits that researchers have extensively examined the trends and pattern of rainfall and related extreme events for the different parts of globe, including India.However, studies on trends and pattern of rainfall and related extreme events pertaining to the dryland ecosystem of India i.e., for Rajasthan state are less in number.Climate of the Rajasthan state is of monsoon type and receives the major part of annual rainfall during the south west monsoon.The insu cient and erratic rainfall in many parts of the state constitutes a threat to agriculture.The distribution of rainfall in dryland ecosystem of Rajasthan is highly uneven and changes signi cantly from region to region and year to year (Pingale et al. 2014).Also, the state has been characterized by highest chances of rainfall de ciencies and drought occurrence (Rathore 2005).
The region is highly dependent upon rainfall as it is an important source of groundwater recharge.The long-term climatic changes related to rainfall may affect the agriculture and water resources of the state (Yadav et al. 2018).For that reason, some investigations have been attempted to examine the rainfall and related extreme events for this state (Rao et Meena et al. 2019).However, these studies are restricted to the analyses of a particular rainfall event, district, western arid region and short period of data.No study has presented a comprehensive examination of rainfall and heavy rainfall events over entire dryland ecosystem of Rajasthan state in north western India.Therefore, a detailed analysis of recent rainfall variability for this dryland ecosystem is required.In the light of these facts, the main objective of this study is to examine recent rainfall variability over dryland ecosystem of Rajasthan state in north western India.

Study Area
The Rajasthan state, situated in north western part of India, is the largest state of country.It covers a widespread area of 3,42,239 km 2 occupying about 10.4 percent of the India's geographical extent.Geographically, the state extends between latitudes 23°30´ to 30°11´ north latitudes and 69°29´ to 78°17´ E longitudes (Fig. 1).The altitude of terrain in the state varies from 6 to 1698 m amsl.The state has a variety of topographic features such as the Thar Desert in the north west, sandy plains in the north east, the Aravalli Hills (spreading from north east to south west) in the center, and south eastern plateau.The Thar Desert is India's largest desert, which covers approximately 70 percent area of the state.
The state has matchless climatological characteristics in the form of low, highly variable, erratic and ill-distributed rainfall, low humidity, high wind speed, high evaporation losses, and extremes of diurnal, seasonal and annual temperatures.The Aravalli Hills play a major role in the climatic pattern of Rajasthan due to its parallel running along the direction of the south west monsoon.The Arabian Sea branch of monsoon winds hit the eastern slope of these hills and the south eastern part gets su cient rainfall, whereas the north western part remains dry (arid and semi-arid).However, the south eastern parts of the state are characterized by humid to very humid climatic conditions (Fig. 1).
The state is characterized by frequent droughts, lack of perennial rivers, scarce vegetal cover, nomadic population and dependency of human on animal rearing.Further, the state of Rajasthan has a predominantly agrarian society with agro-pastoral economy.Of the total cultivated area (approximately 20 million hectares), only 20 percent area is irrigated due to 1 percent of the country's total water resources availability in the state.The onset and duration of monsoon signi cantly controls the performance of rain fed kharif crops.The yields of crops are sensitive to rainfall and wetter years are associated with higher yields.
3 Data And Methods

Data collection and database preparation
In the present study, daily rainfall data of 33 stations has been used for 57-years period (1961-2017), which are well spread over the Rajasthan state (Fig. 1).The data has been acquired from the online portal of Department of Water Resources, Government of Rajasthan, Rajasthan, India (www.waterresources.rajasthan.gov.in).Apart from this, annual temperature data of 26 stations have been obtained from the Department of Agriculture, Government of Rajasthan, Rajasthan, India for the period 1979-2014.The geographical position (latitude and longitude) and elevation details of these stations have been given in Table 1.The major problem with the rainfall analyses' is the missing records.For that reason, out of more than 400 stations, only 33 stations that have smaller number of missing values, has been selected for this investigation.Further, in above dataset, the daily rainfall data is available from the year 1957; however, the year 1961 has been considered as base year, because the data has been found consistent for the selected 33 stations only after 1961.Few stations have missing rainfall observations which are lled-in before undertaking further analysis.The normal ratio method has been executed to estimate the missing values by choosing the rainfall amounts of three nearest stations.Distance matrix is structured for all stations based on their geographical positions to evaluate closeness of stations with each other.

Methods
After preparation of complete dataset, suitable statistical analysis such as sum, frequencies, percentage, mean, standard deviation (SD), coe cient of variation (CV), skewness (C s ) and kurtosis (C k ) have been performed for each station for the period 1961-2017.Then, the excess and de cit rainfall years and heavy rainy days have been identi ed as follows:

Identi cation of Excess and De cit Rainfall Years
In this study, a year has been identi ed as an excess (de cit) rainfall year, if rainfall is more (less) than one SD from mean (Pant and Rupa Kumar 1997).Statistically, an excess rainfall year can be expressed as: 1 and a de cit year can be expressed as: where, R i = rainfall amount in a year i, R m = mean rainfall and S d = SD of rainfall.

Identi cation of Heavy Rainfall Events
In this study, an event is considered as heavy rainfall event if rainfall occurs > 65 mm in 24 hours (Kumar and Singh 2018).The total heavy rainfall events recorded at a particular station have been summed up to obtain monthly, seasonal and annual events.The total number of heavy rainfall events recorded at different stations have been averaged to represent their frequency.

Trend Detection Methods
The non-parametric Mann-Kendall (MK) (Mann 1945;Kendall 1948) has been used to detect the trend in rainfall.This MK test has been found an excellent tool to examine the possible existence of signi cant trends in the time-series at various signi cance levels

Spatial Interpolation
To examine the spatial pattern of rainfall characteristics, the inverse distance weighting (IDW) interpolation technique has been used in this study.The IDW technique is a simple and widely used interpolation technique.This spatial interpolation technique considers the role of each input point by a standardized inverse of the distance from the control point to the interpolated point.This technique assumes that each input point has a local impact that diminishes with increase in remoteness.It weighs more on points nearer to the processing points, than those of far away.It means a rainfall or its derived quantity at any desired location is interpolated from the given data using weights that are based on the distance from each rainfall station and the desired location (Burrough and McDonnell, 1998).This technique gives a smooth pattern of rainfall along with undesirable troughs and peaks.The interpolation of rainfall characteristic using IDW technique has been done by ArcGIS 10.2 software.
Temperature is an important meteorological variable after rainfall.A signi cant change in temperature can pose a serious threat to agriculture, vegetation, habitat and biodiversity of a place (Mishra 2018).Therefore, in this study, a temporal distribution of annual mean, maximum, and minimum temperature has been undertaken, which revealed a non-signi cant increasing trend in most parts of the state (Fig. 4).These results have been found consistent with that reported by Roy ( 2015

Seasonal Variations
The mean seasonal rainfall during different seasons over dryland ecosystem of Rajasthan during 1961-2017 has been displayed in Table 2.
Approximately, 90 percent of total annual rainfall over the dryland ecosystem has occurred during monsoon season.Remaining rainfall (less than 10 percent) has occurred in three seasons i.e. winter, pre-monsoon and post-monsoon.Winter season has accounted only about 2 percent rainfall.Pre-and post-monsoon seasons have accounted remaining 8 percent (4 percent each) rainfall.The values of SD have been found very small during winter (SD = 5.35), pre-monsoon (SD = 7.98) and post-monsoon (SD = 9.61) seasons due to very less amount of rainfall than monsoon season (SD = 213.53)(Table 2).Interestingly, the dryland ecosystem as a whole has shown a marginal non-signi cant increase in rainfall amounts during winter, pre-monsoon and monsoon seasons, whereas post-monsoon season has shown a marginal non-signi cant decrease (Fig. 5).Like annual pattern, three year running average has exhibited a clear cyclic pattern in rainfall occurrence during monsoon season over Rajasthan.Likewise, all the stations have shown positive trends in mean rainfall except three stations namely Jalor, Pratapgarh and Udaipur during pre-monsoon season.Further, the detected increasing or decreasing trend in the occurrence of rainfall during different seasons has not been found statistically signi cant.

Monthly Variations
The mean monthly rainfall over dryland ecosystem of Rajasthan during 1961-2017 has been shown in Table 2. Monthly rainfall has exhibited a wide variation in its occurrence.Remarkably, only two months (July and August) have accounted approximately two-third of total annual rainfall, whereas December month has accounted only about 0.49 percent rainfall.Further, the SD and CV values have been found high during the July and August as these two months have received huge amount of rainfall with very high variability, whereas rainfall occurrence has been observed very less in rest of the months (Table 2).Further, the mean rainfall occurrence has been noticed less in April month, however, a rising trend has been observed in this month (Fig. 6; Table 2).The months of June and July have also shown insigni cant increasing trends, whereas September, October and November months have shown an insigni cant decreasing trend.Rests of the months have shown almost constant trends.

Annual Pattern
The spatial pattern of mean annual rainfall over dryland ecosystem in north western India during 1961-2017 has been displayed in Fig. 7a.The mean annual rainfall has exhibited an increasing rainfall gradient from arid and semi-arid to humid and very humid regions.The mean annual rainfall occurrence over entire arid and semi-arid regions has been found less than 400 mm, whereas it has been more than 850 mm over the south eastern parts located in the very humid regions.This decline in average annual rainfall from east to west is mainly associated with the presence of Aravalli Mountains spreading north east to south west in central Rajasthan (Yadav et al .2012).The Arabian Sea branch of monsoon winds hit the eastern slope of these hills and the south eastern part receives ample rainfall, while north western parts remain dry (Dutta et al. 2015).The moisture loss of monsoonal winds with distance also reduces the rainfall amounts towards westward from south east.The lowest mean annual rainfall has occurred over Jaisalmer station (196.81 mm), whereas highest over Banswara station (1070.18mm) during 57-years study period (Table 1).The spatial pattern of CV has exhibited almost opposite pattern to the mean annual rainfall (Fig. 7b).The higher (above 50 percent) values of CV have been noticed over the west and north west, whereas lower (below 35 percent) over east and south east.
Exceptionally, the CV value has been found very high at Barmer station (about 60 percent), whereas it is just about 26 percent at Bundi station (Table 1).The higher CV values in the arid and semi-arid regions of the state have indicated high annual variability of rainfall.Conversely, the annual variability of rainfall has been observed relatively less in the areas of high rainfall.Hence, lack of rainfall and high variability are responsible for frequent occurrence of drought conditions (once in 2-3 years) over arid and semi-arid regions.2001) have also projected that the western arid region of India could receive higher than normal rainfall in warming climate.This rising trend in mean annual rainfall over western arid region, which is considered to be the most drought prone, could be extremely bene cial for agriculture as well as recharging of groundwater.However, these results are in contrast with Pingale et al.

Seasonal Pattern
The seasonal pattern of rainfall over dryland ecosystem in north western India during 1961-2017 has been displayed in Fig. 8.During winter season, the entire state has received less than 20 mm rainfall.The northern and north eastern parts have received rainfall between 10 to 20 mm and extreme western parts receive less than 5 mm during this season (Fig. 8a).During pre-monsoon season, the pattern of rainfall has appeared almost identical to the winter season, however, rainfall has slightly increased over all parts of the state (Fig. 8b).The northern parts receive approximately 30 mm mean rainfall during this season.During these two seasons, the northern and north eastern parts of the state have observed some amount of rainfall, which may be attributed to the western disturbances and associated thunderstorm activity over this region (Bhardwaj and Singh 2018; Singh and Bhardwaj 2019).The rainfall pattern during monsoon and post-monsoon seasons has been found almost similar, with higher concentration over the eastern to south eastern parts of the state and a consistent decrease towards westward and north westward (Fig. 8c-d).During monsoon season, rainfall amount has ranged from approximately 900 mm in south eastern parts to less than 300 mm over western and north western parts.The entire state has received the rainfall from the summer monsoonal winds during this season.As about 91 percent of total annual rainfall has occurred in this season, therefore, the rainfall pattern of monsoon season is almost similar to the pattern of mean annual rainfall.The moisture loss of monsoonal winds with distance and Aravalli Mountains reduces the rainfall occurrence towards westward from southeast.During post-monsoon season, less than 15 mm of rainfall has occurred in western parts and about 30 mm over south eastern parts.Further, the spatial pattern of CV for different seasons has shown opposite pattern to the mean seasonal rainfall (Figure not shown).The variability of rainfall occurrence has been found comparatively high over the parts of less rainfall and vice-versa.The values of CV have been observed very less over all parts during monsoon season, which may be accredited to the higher occurrence of rainfall with summer monsoon.The values of CV have been found very high over entire state during other three seasons namely winter, pre-monsoon and post-monsoon seasons due to very less and erratic rainfall.
The spatial pattern of seasonal rainfall trends and magnitude during 1961-2017 detected by the MK test and Sen's slope estimator has been shown in Fig. 9.During winter season, all the stations in the study area have shown non-signi cant positive trends in mean rainfall except Nagaur (Fig. 9a).A signi cant increasing trend in rainfall has been detected in humid and semi-arid regions of the state during pre-monsoon season (Fig. 9b).The magnitude of increasing trend has varied from 0.60 mm/year (Churu station) to 0.12 mm/year (Pali station).During monsoon season, majority of stations located in arid region (western parts) of Rajasthan have shown increasing trends in rainfall (Fig. 9c).A signi cant rising trend has been detected for stations like Ganganagar, Pratapgarh and Jaisalmer during this season.The magnitude of increasing rainfall trend has differed from 5.57 mm/year (Pratapgarh station) to 2.02 mm/year (Jaisalmer station).These ndings have been found consistent with previous studies During monsoon season, several stations located over eastern parts have shown a negative change in rainfall (Fig. 10c), consistent with Mundetia and Sharma (2015).However, no station has detected positive percent change in rainfall during post-monsoon season for a period of 57-years (Fig. 10d).

Monthly Variations
The spatial pattern of mean monthly rainfall over dryland ecosystem during 1961-2017 has been demonstrated in Fig. 11.In the months of January, February, March and April, the northern and north eastern parts have recorded the rainfall occurrence.However, the rainfall amount during these months has been found almost negligible.In the month of May, rainfall pattern has been found almost similar to April, but the amount of rainfall has increased.After the month of May, rainfall pattern has changed signi cantly.From the month June to August, rainfall has increased consistently over entire state.During these months, rainfall occurrence has been found comparatively higher over humid and very humid parts and a decrease towards westward and north westward.Then, the rainfall has started to lessen rapidly from September to December over all parts.Moreover, the spatial pattern of CV has exhibited the opposite pattern to mean monthly rainfall (Figure not shown).The values of CV have been found lowest over entire state in July month, whereas highest in December month.

Trends in heavy rainfall events
The annual maximum daily rainfall in Rajasthan during 1961-2017 has been displayed in (Fig. 12a).It has shown a non-signi cant positive trend, which signi es that the severity of the rainfall events has enhanced over dryland ecosystem in north western India.The maximum daily rainfall has been recorded highest in 1981 (550 mm) and lowest in 1989 (97 mm).The station-wise analysis of annual maximum daily rainfall has shown the mixed trends in rainfall without any signi cant rising or declining trend.The mean frequency of heavy rainfall events has been exhibited in Fig. 12b.The gure has revealed a positive trend in the frequency of heavy rainfall events over Rajasthan.The maximum number of heavy rainfall events have been observed in the years 1983 and 2006 (about 2.51 events), whereas lowest in the year 2002 (about 0.40 events).Additionally, the station-wise analysis of heavy rainfall events has shown almost constant trend except Bharatpur, Jaisalmer and Pratapgarh stations, which shows a signi cant rising trend, whereas Bundi station has shown signi cant decreasing trend (Figure not shown).The spatial pattern of heavy rainfall events over dryland ecosystem has been shown in Fig. 13.The pattern of heavy rainfall events has been found almost similar to the pattern of mean annual rainfall increasing from arid and semi-arid regions to humid and sub-humid regions.The mean frequency of heavy rainfall events has been noticed relatively higher over the humid and very humid parts (more than 2.50) and decreases towards arid and semi-arid parts (less than 1.0).

Conclusions
The distribution of rainfall over dryland ecosystem of Rajasthan is highly uneven due to its location and topography.Rainfall changes signi cantly from region to region and year to year having highest chances of de ciencies resulting occurrence of drought.But the studies pertaining to rainfall variability and trends over the dryland ecosystem are far less in number.To ll this research gap, an effort has been made in this study to examine the recent rainfall variability and trends for the period 1961-2017.The results of this study have revealed a signi cant rising trend over most of the arid and semi-arid regions except very humid region.The state has also recorded high interannual rainfall variability coupled with rainfall de cit years during El-Nino period.Remarkably, July and August months have accounted approximately two-third of total annual rainfall over the dryland ecosystem.Furthermore, the frequency of heavy rainfall events has shown a positive trend.The maximum number of heavy rainfall events have occurred in the years 1983 and 2006 (about 2.51 events), whereas lowest in the year 2002 (about 0.40 events).Apart from this, an increase in mean, minimum and maximum temperature has been detected in most parts of the state though statistically non-signi cant.
These above results show de nite signs of change in rainfall and temperature over the state of Rajasthan, which can duly affect the climate as well as physical and economic landscapes.The overall increasing trends in rainfall may really be advantageous in the light of prevailing water scarcity problems over the dryland ecosystem.Interestingly, an increase in rainfall amount is anomalously due to increase in rainfall intensity, which produces large runoff volumes over a short period of time.Conversely, high rainfall variability may affect the net cultivated area.About four million ha of culturable land has either been left uncultivated or has sowing been delayed during the year 2002 over the dryland ecosystem (Bhuiyan et al. 2006).Meanwhile, an increase in mean, maximum and minimum temperatures will lead to enhancement of aridity, which is most likely to affect the evapotranspiration, vegetation cover, crop growth and resultant agricultural productivity.
Furthermore, the ndings of this study are important for water management as the water is key to human survival and all kinds of vegetative growth over the dryland ecosystem.Since the state holds only 1 percent of the total water resources of the country, the increasing trends of rainfall can hold good for recharging of groundwater resources.The increased availability of the rainwater resource will lessen the usual waterde cit scenario in arid and semi-arid regions due to statistically signi cant positive rainfall trends.Apart from this, the increasing trends in rainfall and temperature at most of the stations conspicuously reveal a changing climate over the dryland ecosystem.A persistence of climate change in future over the ecosystem may result in damages to both soil and water resources, leading to deserti cation and decrease in agricultural productivity.The changing climate may also result in crisis of drinking water along with the scarcity of water for irrigation.Finally, the outcome of this study may prove to be useful to policy makers, hydrologists and water resource managers dealing with climate change in the state of Rajasthan for sustainable development and planning of water resources.

Declarations Figures
Map showing the position of study area in India and geographical location of selected rainfall and temperature stations.
Spatial distribution of heavy rainfall events over dryland ecosystem of north western India (1961-2017).
and Villani (2010) and Altava-Ortiz et al. (2011) have shown a decreasing trend in average annual precipitation over Mediterranean basin and nearby regions.More recently, Adler et al. (2017) have not detected any signi cant trend in the global mean precipitation, however, a rising trend over tropical oceans and a declining trend over certain mid-latitudes areas has been detected.Nicholson et al. (2018) and Caloiero et al. (2018) have detected a signi cant downward trend in annual rainfall over West Africa, North Africa, and eastern Mediterranean.Besides, several other attempts have been made to examine the possible in uences of changing climate on spatial and temporal rainfall trends (Loo et al. 2015; Mayowa et al. 2015; Xiao et al. 2016; Hu et al. 2017; Sein et al. 2018; Biasutti 2019; Haag et al. 2019; Sa'adi et al. 2019; Gebrechorkos et al. 2019).
al. 2013; Boucher 2013; Berg 2013; Fischer and Knutti 2016; Myhre et al. 2019).Hartmann et al. (2013) have observed that the occurrence of extreme precipitation events has increased over larger land areas than it has decreased in the second half of 20th century.Therefore, extreme precipitation has increasingly become an illuminating subject of research during the period of changing climate (Cammarano and Tian 2018, Dahal et al 2018).Researchers have continuously made efforts to examine the extreme precipitation trends in relation to changing climate (Dore 2005; Alexander et al. 2006; Fischer and Knutti 2014; Asadieh and Krakauer 2015; Westra et al. 2013).Apart from this, increasing trends in extreme rainfall events have been detected over different parts of the world, for example, Europe (van den Besselaar et al. 2013), North America (Villarini and Vecchi 2012; Donat et al. 2013; Easterling et al. 2017), South America (Donat et al. 2013, Skansi et al. 2013), south eastern South America (Wu Y and Polvani 2017), Australia (Laz et al. 2014; Herath et al. 2017), South Asia (Sheikh et al. 2015) and south east Asia (Ge et al. 2019).Interestingly, most of the land areas have observed a rise in extreme rainfall events in summer season excluding Europe, which showed an increase during winters (Hartmann et al. 2013; Pinskwar et al. 2019).
al., 2010; Jain and Kumar, 2012), while Ghosh et al. (2009) have shown mixed trends in rainfall occurrence over different regions in India.Kumar et al. (2010) have shown a decline in annual and monsoon rainfall, while a rise in winter, pre-and post-monsoon seasons for India as a whole.However, these trends are statistically insigni cant.Jain et al. (2013) have not detected any signi cant trend in annual rainfall occurrence over north east India, although seasonal trends have been detected.However, several researchers have shown a substantial rise in number and magnitude of extreme rainfall events over the Indian region (Goswami et al. 2006; Joshi and Rajeevan 2006; Krishnamurthy et al. 2009; Pattanaik and Rajeevan 2010).Conversely, Guhathakurta et al. (2011) have detected a decline in extreme rainfall events over central and north India, whereas an increase over peninsula, eastern and north eastern parts of India.Recently, Mukherjee et al. (2017) have noticed a signi cant rising trend in number of extreme rainfall events and ascribed it to increasing anthropogenic warming.Apart from this, numerous efforts have been done on the spatial and temporal pattern of rainfall and related extreme events for different states of India in recent times such as Kerala (Krishnakumar et al. 2009; Pal and Al-Tabbaa 2009; Pal and Al-Tabbaa 2011; Nair et al. 2014; Thomas and Prasannakumar 2016), West Bengal (Chatterjee et al. 2016; Ghosh 2018; Kundu and Mondal 2019), Andhra Pradesh (Rao et al. 2009; Patakamuri et al. 2020), Madhya Pradesh (Duhan and Pandey 2013; Jana et al. 2017); Orissa (Patra et al. 2012), Uttarakhand (Kotal et al. 2014; Nandargi et al. 2016), Jammu and Kashmir (Kumar et al. 2009), Punjab (Gill et al. 2013), Assam (Deka et al. 2013), Gujarat (Lunagaria et al. 2015; Dave and James 2017; Dave et al. 2017), and Maharashtra (Ratna 2012; Sonar 2014; Ingle 2018).
The mean annual rainfall over dryland ecosystem of Rajasthan during 1961-2017 has been demonstrated in Fig.2.The time-series has exhibited a high interannual variability in mean annual rainfall (SD = 221.10,CV = 37.90) varying from 277 mm (in year 2002) to 839 mm (in year 1975) with a mean of about 583 mm (Fig.2; Table2).The lowest (highest) rainfall in the year 2002 (1975) may be attributed to the prevalence of strong El-Nino (La-Nina) conditions(Mahala et al. 2015;Bhardwaj et al. 2019).Similarly, C k and C s are the measures of data peakedness, atness and symmetry and their values for normal data distribution is zero or near zero.The positive values of C k reveal the peaked distribution and vice versa.In the study area, Banswara (very humid) and Nagaur (semi-arid) stations have high C k , whereas Bhilwara (humid) and Churu (semi-arid) have low C k .All the stations except Bharatpur and Sawai Madhopur (humid region) have positive C s , which suggests that there is an increment in the amount of rainfall than normal in all over the study area.It has been observed from the analysis that most of the rainfall de cit years such as1965, 1969, 1972, 1987, 2002 and 2009  have occurred with El-Nino years (Table3)(Mahala et al. 2015;Bhardwaj et al. 2019).The El-Nino years have disturbed the Indian monsoon badly and affected the agriculture system speci cally in the Rajasthan state which is completely dependent on south west monsoon for arrival of rainfall.However, the mean annual rainfall has shown a marginal non-signi cant upward trend over the dryland ecosystem.Interestingly, three years running average has shown a cyclic pattern of mean annual rainfall under the changing climatic conditions.The mean annual rainfall has revealed a signi cant rising trend in arid and semi-arid regions except Pratapgarh stations situated in the very humid region of the state, which has been found consistent with Singh et al. (2001) (Fig.3).Apart from this, most of the stations located in semi-arid, sub-humid and humid regions have shown increasing non-signi cant trends, whereas eastern and south eastern parts have shown non-signi cant decreasing trends.Rest of the regions have shown almost constant trends in the mean annual rainfall.
), Sharma et al. (2017) and Mehta and Yadav (2019).Pingale et al. (2014) and Roy (2015) have attributed these warming trends to increasing anthropogenic activities, namely, land use/cover changes, industrialization and urbanization.Generally, a rise in temperature is directly associated with rising evapotranspiration rates which further results in increasing rainfall trends.Previously, several such studies have revealed similar results in different parts of India (Singh et al. 2008a, b; Jain et al. 2013; Pingale et al. 2014; Roy 2015).In a study over the dryland ecosystem, Mishra (2018) have concluded that a change in climatic conditions leads to a signi cant change in rainfall, temperature and evapotranspiration rates, which in turn adversely affects the physical as well as economic landscapes of an area.
(2014)   andSingh and Kumar (2015), who have shown a decline in mean annual rainfall over western parts of the state.Conversely, some stations located in the eastern parts of the state have recorded a non-signi cant decreasing trend, consistent with Kumar et al. (2010),Mondal et al. (2015) andMundetia and Sharma (2015).However, no station has shown statistically signi cant declining trend in mean annual rainfall during 1961-2017.Additionally, the arid and semi-arid regions of the state have observed the maximum percent change in rainfall than humid and subhumid regions (Fig.7d).Jaisalmer station has shown the highest percent increase (63 percent) in mean annual rainfall during the study period 1961-2017.
like Basistha et al. (2007), Kharol et al. (2013), Mondal et al. (2015) and Meena et al. (2019), and in contrast with Pingale et al.(2014).During post-monsoon season, majority of stations have shown non-signi cant negative trends (Fig.9d).Further, the spatial pattern of percent change in rainfall amount during different seasons for the study period 1961-2017 has been shown in Fig.10.During winter season, humid and very humid regions of the state and only three stations in arid region (Ganganagar, Churu and Sikar) have observed a positive change of more than 25 percent (Fig.10a).During pre-monsoon season, a substantial change in the pattern of percent change in rainfall has been observed (Fig.10b).The northern and north eastern parts have observed a positive change of more than 60 percent for the period 1961-2017.

Figure 3 Station
Figure 3

Figure 4 Station
Figure 4

Figure 5 Year
Figure 5

Figure 6 Year
Figure 6

Table 1
Basic information of selected stations and station-wise rainfall characteristics over dryland ecosystem of north western India.The double mass curves of all stations have not shown any breakpoint and displayed more or less a straight line which indicates the homogeneity in the data.Similarly, student's t test has not revealed any break point or statistically signi cant difference in the rainfall time series at 95% con dence level.Apart from this, the daily rainfall data has been summed up to obtain monthly, seasonal and annual values.Seasonal analysis has been done for four seasons namely (a) winter (January-February), (b) pre-monsoon (March-May), (c) Monsoon (June-September) and (d) post-monsoon (October-December), as per categorization of India Meteorological Department (IMD).
SD = Standard deviation, C V = coe cient of variation, C K = kurtosis, C S = skewness Further, data quality check is essential before the investigation, as outliers can extremely affect the rainfall trends and patterns (You et al. 2008; Shahid et al. 2016).To ensure the homogeneity, subjective double mass curve method and objective student's t-test have been performed for rainfall time-series.
(Yue et al. 2002)2006; Mohammad and Jha 2014).Also, the non-parametric Sen's Slope estimator(Sen 1968) has been used to detect the magnitude of the trend.It is closely associated to the MK test(Gilbert 1987) and provides a robust estimation of trend(Yue et al. 2002).Additionally, the parametric simple linear regression has also been used to detect the trend in time-series in this study.These three methods have been used widely for this purpose in the hydro-meteorological studies.A detailed discussion about these methods is available in(Deka etal.2013; Pingale et al. 2014; Jaiswal et al. 2015; Mayowa et al. 2015; Singh et al. 2020).These tests have been performed by means of XLSTAT 2017 software.
(Von Storch 1995; 2010;Luo et al. 2008;Pingale et al. 2014 variable Z has used to detect the trend and its signi cance level.The positive (negative) values of Z show rising (declining) trend in the time-series.In this study, the existence of statistically signi cant positive or negative trend has been considered at 95% con dence level.The MK test is a rank-based test, which have been used extensively for the veri cation of autocorrelation in the data series(Duhan and Pandey 2013).The test shows that if the lag-1 serial coe cients of the data are not statistically signi cant, then the MK test has been performed without any modi cation to the original time series(Karpouzos et al. 2010;Luo et al. 2008;Pingale et al. 2014).However, the existence of an autocorrelation can increase the probability of detecting a signi cant trend (Bayazit and Onoz 2008).Therefore, in the present study, MK test has been applied after pre-whitening for all stations and seasons to eliminate the effect of autocorrelation in the data series(Von Storch 1995;

Table 3
Excess and de cient rainfall years over dryland ecosystem of north western India These results are in correspondence with Narain et al. (2006) with respect to dryland ecosystem.The spatial pattern of annual rainfall trends and magnitude during 1961-2017 detected by the MK test and Sen's slope estimator has been presented in Fig. 7c.Interestingly, majority of stations (23 stations; 70 percent) in the Rajasthan state have recorded positive trend in mean annual rainfall.Among these 23 stations, ve stations namely Barmer, Churu, Ganganagar, Jaisalmer and Pratapgarh have observed statistically signi cant positive trends.These results have been found in agreement with previous studies (Pant and Hingane 1988; Singh et al. 2001).The magnitude of signi cant increasing trend has varied from 5.34 mm/year (Pratapgarh station) to 2.17 mm/year (Jaisalmer station).Majority of stations positioned in western parts of this dryland ecosystem have shown increasing trends, consistent with Kumar et al. (2010), Mondal et al.