Less rain and rainy days—lessons from 45 years of rainfall data (1971–2015) in the Kathmandu Valley, Nepal

Understanding spatiotemporal variability in rainfall patterns is crucial for evaluating water balances needed for water resources planning and management. This paper investigates spatiotemporal variability in rainfall and assesses the frequency of daily rainfall observations from seven stations in the Kathmandu Valley, Nepal, from 1971 to 2015. Daily rainfall totals were classified into five classes, namely, A (light rain, daily rainfall < 10 mm in a day), B (moderate rain, between 10 and 50 mm), C (heavy rain, between 50 and 100 mm), D (storm, between 100 and 150 mm), and E (large storm, > 150 mm). An ordinary kriging method was used for spatial interpolation using QGIS. We performed Mann–Kendall (MK) test in conjunction with Theil-Sen’s (TS) slope estimator to detect monotonic trends, their significance, and magnitude. We find that the mountain stations depict a decreasing rainfall trend for all seasons, ranging from − 8.4 mm/year at Sankhu to − 21.8 mm/year at Thankot, whereas a mixed pattern is found on the Valley floor. Since the surrounding mountains are the chief source of surface runoff across the valley, rivers, and rivulets are substantially affected by falling rainfall tendency. Both annual rainfall amount and the number of rainy days decreased in the Kathmandu Valley over the study period. We observe a significant reduction in rainfall after 2000. As springs and shallow groundwater are the primary sources of water supply in the Kathmandu valley, it is apparent that decreasing rainfall will have (and is already having) an adverse impact on domestic, industrial, and agricultural water supplies and the livelihoods of people.


Introduction
Rainfall, one of the major components of the hydrological cycle, influences life on earth (Shrestha and Sthapit 2015). A sound understanding of spatiotemporal rainfall variability is vital for evaluating water balances at various scales, which in turn are prerequisites for effective water resources planning and management (Wong et al. 2009;Thapa et al. 2017). An assessment of rainfall variability is a frequent practice in hydrology and has important applications in hydrologic modeling, water resource assessments, agricultural planning, flood frequency analysis (Buytaert et al. 2006), flood hazard mapping, climate change impacts, and other environmental assessments (Ngongondo et al. 2011).
In particular, studies on rainfall variability have significant roles in urban hydrology, where the hydrologic response is sensitive to rainfall distribution in both space and time, due to the dominant impervious surfaces, medium-sized catchment and high spatial heterogeneity of urban land use (Emmanuel et al. 2012;Cristiano et al. 2017). Further, the study of rainfall variability in urban areas is necessary to learn about the impact of global climate change as well as the influence of local urbanization and development on rainfall (Karki et al. 2017).
In Nepal, the summer monsoon rainfall (from June to September), which accounts for approximately 80% of the total annual rainfall, is governed by the South Asian Monsoon originating from the Bay of Bengal (Panthi et al. 2015). Significant spatiotemporal variations in the rainfall pattern have been illustrated in previous studies (Ichiyanagi et al. 2007;Karki et al. 2017;Bohlinger and Sorteberg 2018;Dahal et al. 2019). The summer monsoon is more active in eastern and central Nepal, whereas winter rainfall, caused by western disturbances originating from the Mediterranean Sea (Shrestha and Sthapit 2015;Talchabhadel et al. 2018a), is more active in western Nepal (Ichiyanagi et al. 2007;Sigdel and Ikeda 2012). The climatic regime varies from subtropical near Nepal's southern border, to warm and cool in the hills, to cold on the mountains; this climatic variability occurs in less than 200 km moving from India-Nepal border northward (Shankar and Shrestha 1985;Chalise 1994;Karki et al. 2016). Diverse topography over the latitudinal distance of around 200 km causes a high variation of rainfall throughout Nepal due to intense orographic lifting and subsequent rain shadowing. Interestingly, two significant rainfall peaks appear over the southern slope of the Himalayas across Nepal due to its unique topographical setting; the first peak appears along 500-700 m above sea level (masl), and the second peak appears along 2000-2200 masl (Shrestha and Sthapit 2015;Talchabhadel et al. 2018a). The substantial fluctuation in rainfall is the reason for the water crisis during the drier months and extreme precipitation events resulting in floods, landslides, and other water-induced disasters in monsoon months (Tuladhar et al. 2020).
To date, many studies have analyzed and observed satellite rainfall data in Nepal. Bohlinger and Sorteberg (2018) analyzed the trends in monsoon rainfall and extreme events in Nepal. A total of 291 distributed stations across Nepal were analyzed for a period of 50 years by Talchabhadel et al. (2018b) to know the spatiotemporal variability in precipitation extremes. Duncan et al. (2013) investigated long-term spatiotemporal rainfall data of Nepal from 1951 to 2007 at a fine spatial resolution to assess both annual and seasonal cumulative rainfall, timing of onset of monsoon, and severe weather events. The study by Panthi et al. (2015) compared seasonal rainfall trends on different agro-ecological zones of the Gandaki River Basin and used Asian Precipitation-Highly Resolved Observational Data Integration Towards Evaluation (APHRODITE) precipitation product for analyzing decadal anomalies. The same precipitation product was used in the study by Pokharel et al. (2020) which helped in understanding east to west patterns in annual precipitation and extremes.
However, there are limited studies dealing with the assessment of rainfall patterns in the Kathmandu Valley and the associated Bagmati River Basin which has been facing severe water quality and quantity crisis in recent decades. Also, the majority of extreme events taking place in Nepal is in localized areas, so assessing the rainfall variability in spatial extent is vital (Pokharel et al. 2020). Shrestha and Sthapit (2015) identified the temporal trend of rainfall in Bagmati River basin using time series data for the period of 1981-2008. Dhital and Kayastha (2013) performed frequency analysis of future rainfall and peak flood events in the Bagmati catchment. A study on rainfall intensity and distribution in the Kathmandu Valley was done by Pokharel and Hallett (2015). Karki (2015) studied the daily rainfall pattern of summer monsoon in the Kathmandu Valley. Tuladhar et al. (2020) analyzed local variability in rainfall distributions as well as long-term trends of monthly and annual rainfall in the Bagmati River catchment.
The Kathmandu Valley (hereinafter Valley) is the most populated urban center in Nepal. Uncontrolled urban expansion in the Valley has increased water demand and also reduced groundwater recharge potential by lowering surface infiltration capacities (Davids et al. 2018). Springs originating from the upper portions of the mountainous watersheds surrounding the Valley along with groundwater are the primary sources of freshwater supply in the Valley. Summer monsoon rain is the main source of these springs, surface water, and groundwater recharge (Shrestha and Sthapit 2015). Degradation of both quality and quantity of surface water supplies led to excessive extraction of groundwater (Shrestha et al. 2012b). To deal with the current water crisis, the Kathmandu Valley Water Supply Management Board (KVWSMB), a government agency responsible for water supply and sanitation management in the Valley, is planning to augment groundwater recharge in the Valley. Urban flooding due to short-duration intense rainfall coupled with increased built land use has become an emerging issue in the Valley. Improving the understanding of temporal and spatial structures of rainfall is, therefore, integral to the sustainable livelihood of people. Variation in rainfall patterns may significantly affect the springs, streams, and groundwater of surrounding headwater catchments, the lifeline of water supplies for downgradient urban dwellers, industry, and agriculture. Thus, the assessment of the annual rainfall pattern and its seasonal variations is critical for sustainable water resource management and planning. The major objective of our study was to improve our understanding of the rainfall in the Valley by: 1. Assessing the frequency of rain events and rainfall accumulated by events of various intensity 2. Investigating and quantifying the spatial and temporal trends in annual and seasonal rainfall variability The present study includes an investigation of spatio-temporal variability of rainfall in the Valley from 1971 to 2015, using data available from the Department of Hydrology and Meteorology (DHM), Government of Nepal.

Study area
The Valley lies between 27°32′13″-27°49′10″N latitude and 85°11′31″-85°31′38″ E longitude (Fig. 1). The Valley watershed has an area of approximately 587 square kilometers (km 2 ) (Davids et al. 2018). The Valley is surrounded by hills: Phulchowki in the South East, Chandragiri/Champa Devi in the South West, Shivapuri in the North West, and Nagarkot in the North East. The Valley is a roughly circular intermontane basin with an approximate diameter of 25 km and an average altitude of 1350 m above sea level (masl), while the surrounding hills reach as high as approximately 2800 masl in elevation Thapa et al. 2017). The Valley consists mainly of alluvial plains, alluvial and colluvial fans, fluvial and lacustrine terraces, and steep to very steep sloping mountains. The Valley lies in a semitropic zone and is characterized by a warm and temperate climate (Karki et al. 2016) having a rainy season from June through September.

Data
Daily rainfall data for the period between 1971 and 2015 were obtained from the Department of Hydrology and Meteorology, Government of Nepal (DHM 2019). Although there are 20 rainfall stations in the Valley with elevation ranging from 1212 m at Khokana, Lalitpur to 2163 m at Nagarkot, Bhaktapur, the present study considered the daily rainfall data only for seven spatially distributed stations due to incomplete time-series data for the study period in other stations. Stations located at the foothill of the mountain are classified as mountain stations (Sankhu, Thankot, Godavari) while those located at valley floor are termed as valley stations (Panipokhari, Kathmandu Airport, Khumaltar, Bhaktapur). Additionally, available data of the Bagmati River Basin (BRB) for the period 1986-2015 of 41 stations, including those stations that lie in the Valley, were analyzed for a broader perspective on basin level seasonal and annual  Table 1 provides a brief description of the Valley's rainfall stations used, including types of station, location, elevation, and year of installation. Missing data were replaced with the daily rainfall data of the nearest station with similar characteristics (topography, elevation).

Distribution pattern of daily rainfall
Rainfall was classified into five rainfall rate classes and was categorized based on the amount of daily rainfall ( Table 2). The amount of rainfall contributed in percentage (% of total rainfall) by various rainfall rate classes for various stations of the Valley was calculated. Similarly, the frequency of rainy days in percentage (% of total rainy days in a year) by various classes of rainfall rate was also calculated. A combo-chart of a bar graph and a line graph was prepared to demonstrate the relationship between daily rainfall and rainfall frequency of various rainfall rate classes. Further, the temporal distribution of rainfall amount of various rainfall rate classes was analyzed for five sub-periods.

Local and valley wide rainfall distribution
The total rainfall days were categorized into two major categories: (a) Local rainfall and (b) Valley-wide rainfall. When all the rainfall stations record daily rainfall > 0.1 mm (mm) per day, it is termed as Valley-wide rainfall. When one or more rainfall stations record daily rainfall > 0.1 mm per day but remaining stations record rainfall < 0.1 mm per day, it is termed as local rainfall. When all the stations record 0 mm rainfall per day, it is termed as Dry and the number of maximum consecutive dry days is termed as maximum consecutive dry days (MCDD). The number of maximum consecutive rainfall days were classified into Maximum Consecutive Local Rainfall Days (MCLRD) and Maximum Consecutive Valley-wide Rainfall Days (MCVRD). Local Rainfall Index (LRI) is the ratio of rainfall contributed by Local rainfall to the annual rainfall whereas Valley-wide Rainfall Index (VRI) is the ratio of rainfall contributed by Valley-wide rainfall to the annual rainfall. Local and Valleywide rainfall distribution was analyzed using line plots.

Annual and seasonal rainfall series
The average daily rainfall was analyzed on both a monthly and seasonal scale. The seasons considered were March to May as pre-monsoon, June to September as monsoon, October to November as post-monsoon and December to February as winter. The time series was divided into five sub-periods: (a) 1971-1980, (b) 1981-1990, (c) 1991-2000, (d) 2001-2010, and (e) 2011-2015. However, depending on the data availability, the time series were divided into four sub-periods for the Bagmati River Basin i.e., 1986-1990, 1991-2000, 2001-2010, and 2011-2015. The average values for different sub-periods for different seasons were then calculated. Ordinary kriging method, one of the common advanced geo-statistical methods was used for the interpolation of the seasonal and annual rainfall average (Buytaert et al. 2006). In the study area, using this method, the amount of rainfall at sampled locations was used as a reference for predicting rainfall amount at un-sampled locations and this method yields better outputs compared to Inverse Distance Weighting (IDW) i.e., traditional deterministic method (Adhikary et al. 2016). Spatial maps for different seasons for each sub-period were prepared using the Quantum Geographic Information System (QGIS). A box-plot of mean annual rainfall  was prepared to analyze the temporal variation of rainfall in the Kathmandu Valley. The

Trend analysis
Mann − Kendall (MK) test was used to determine the significance of temporal trends whereas Theil − Sen's (TS) slope estimator was used to quantify the magnitude of trends (Mann 1945;Kendall 1975;Talchabhadel et al. 2018b). The World Meteorological Organization (WMO) recommends the non-parametric MK test for assessment of trends in meteorological data as it is simple, robust and insensitive to missing data and outliers (Ngongondo et al. 2011).

Distribution pattern of daily rainfall
Our study showed two distinct types of daily rainfall distributions in mountain and valley. We show the rainfall amount and frequency (number of rainy days) contributed in percentage (% of total annual rainfall amount and rainy days) for various rainfall classes in Fig. 2. 60% of total rain events were of class 'A' (light rain), however, these events contributed only 17% of total rainfall amount. Similarly, the frequency of class 'B' (moderate rain) was about 37%. Interestingly, it contributed around 63% of the total rainfall amount, indicating the majority of rain across the valley is contributed from moderate rain. In contrast, class 'C' (heavy rain) only had a frequency of around 3%, and it contributed about 15% of total rainfall. Also, storm and large storm events of classes 'D' and 'E', respectively, were very infrequent (< 0.5%) contributing about 2% to total rainfall. Overall, the result highlights that more than half of all rainy days were light rain, but their contribution was less than one-fifth of total rainfall. These rainy days with slight rainfall are critical for agriculture, environment and vegetation. Around 80% of the total rainfall amount was contributed by moderate and heavy rain classes whereas the frequency of these two classes was about 40% of the total rainy days. And the remaining few % contributed to storm and large storm events. Such heterogeneous patterns of rainfall indicate that the extreme events are very local in nature. Even though these instances occur only around 10 days in a year, they bring more than 15% of the total annual rainfall amount.
The frequency of moderate rain was almost equal to the frequency of light rain in mountain areas, while on the valley floor, the frequency of moderate rain was lower than the frequency of light rain. These rain classes maintain the water content and soil temperature for a longer period of time. Also, these light to moderate rainfall are infiltrated into groundwater. In contrast, heavy rainfall and storms normally route downward as a surface runoff.
The frequency of heavy rain is higher in mountain areas (4.5%) than on the valley floor (2.5%). Similarly, the rainfall contribution of the heavy rain class is higher in mountain areas and lower on the valley floor than light and moderate rain classes. These results show a higher frequency of heavy rainfall events and fewer light rainfall events in mountain areas, resulting in a higher amount of rainfall in mountain areas than on the valley floor. Normally, mountains act as barriers for moisture transport on its leeward side. The rainfall increases with an increase in altitude in the windward side (Shrestha et al. 2012a;Shrestha and Aryal 2010). As the valley is a bowl-shaped which is completely surrounded by the hills, there is a different windward and leeward effect for the moisture coming from any direction. The country's rainfall system is dominated by a summer monsoon system where the moisture from the Bay of Bengal enters from the southeastern to northwestern of Nepal (Shrestha et al. 2012a). At the same time, the winter rainfall is dominated by a westerlies system from northwestern to southeastern regions of the country. Also, when two systems from the Arabain Sea and Bay of Bengal occasionally converge resulting in an extreme rainfall event, the wind and moisture transport are totally different (Bookhagen and Burbank 2006). Additionally, the topographical undulation of the country adds another complexity. The country has two distinct mountain ranges, i.e., southern frontal mountains and northern elevated mountain range, which provide a doublepeak rainfall while looking cross-sectionally from south to north (Shrestha et al. 2012a). The valley lies between these two mountain systems that creates a unique microclimatic zone across the Valley.
In a study by Karki (2015), about half of total rain events had rainfall rates between 0.1 and 10 mm/day (Class A) which accounted for just 13% of the total rainfall. On the contrary, the frequency of rainfall events declined to 17% for a rainfall rate between 30 and 90 mm/day (class B and C) but contributed to 46% of the total rainfall. Heavy rainfall events i.e. rainfall rate more than 90 mm/day (class D and E) were less frequent and contributed to less than 2% of the rainfall received. The findings from the study are similar to what we observed i.e. higher amounts of rainfall were received in the mountain parts (higher frequency of rainfall events, i.e., between 10 and 100) compared with the Valley floor which might be because of the orographic effect.

Annual rainfall series
Annual rainfall had high spatial variation across the Valley. The range of annual variation in rainfall can be seen in  Table 2 Fig. 3. The mean annual rainfall was 1610 mm (1880 mm in mountain areas and 1410 mm on the valley floor) for the study period. Along with spatial variability, the temporal variation was also substantial. The highest observed rainfall was 3425 mm in Sankhu station in 1978 where the mean annual rainfall was 1865 ± 247 mm for 1971-2015. And, the lowest observed rainfall was 827 mm in Khumaltar station in 1992 where the mean annual rainfall was 1237 ± 67 mm for the study period. Despite a substantial year to year variation, we observed spatial and topographical insights that the mountain stations tended to receive higher rainfall than the valley floor, meaning the valley is drier than mountains. Sankhu station, located in the north-eastern part of the valley at the elevation of 1449 m, received the highest annual mean of 1946 mm, whereas, Khumaltar station, located in the central part of the valley at the elevation of 1350 m, received the lowest annual mean of 1228 mm (Fig. 4). Similar findings were observed in a study by Karki (2015), where the stations located in the mountains received an average of 1978 mm rainfall, while that of the valley floor received 1151 mm on average. Also, Karki et al. (2017) reported that valleys are drier compared with surrounding mountain ridges, for instances in Tamor at Mulghat, Arun at Leguwaghat, DudhKoshi at Kuruleghat, Tamakoshi at Manthali, Sunkoshi at Nepalthok, Pachuwarght and Dolalghat, Bheri at Rakam and Dunai, Karnali at many upper river valleys (Jumla, Thirpu, Nagma, Gamshreenagar, etc.), Seti at Dipayal, Mahakali at Binayak and all other river valleys that lie north of tall mountain ranges.
The northern part of the Valley receives most of the rainfall; when the moisture-laden air entering the Valley sinks into the southern part of the Valley floor, the air warms resulting in less rainfall in that area compared with the mountainous northern parts (Karki 2015). The nature of rainfall distribution in the valley suggests that local factors like topography, elevation, latitude, aspect, etc. play an important role in the spatial distribution of rainfall (Prajapati et al. 2021).

Seasonal rainfall series
Around 80% of the total annual rainfall occurs in the monsoon season from June to September. Pre-monsoon contributes around 14% and post-monsoon and winter contribute around 6% of the total annual rainfall. The pattern of the spatial distribution of seasonal rainfall was similar to the annual rainfall as mountains received more rainfall than the valley floor. The difference of monsoonal rainfall among the time series spatial maps (Fig. 4) for different sub-periods indicated a significant reduction in rainfall after 2000. The reduction of rainfall in mountain areas is quite higher compared with the valley floor. Similarly, Fig. 4 showed a decrease in post-monsoon rainfall with time. The decrease in monsoon precipitation and subsequent increase in evapotranspiration results in low flow in rivers (Sharma and Shakya 2006). In a study by Sharma and Shakya (2006), the river flow decreased in the monsoon season; however, there was no significant change in the pre-and post-monsoon. Figure 5 shows the temporal distribution of Local rainfall days (LRD) and Valley-wide rainfall days (VRD) in the Kathmandu Valley. From Fig. 5b, it can be seen that for the first two decades (1971-1980 and 1981-1990), the average VRD was 52 and it increased to 63 during 1991-2000. Then, it decreased to 47 and 40 in 2001-2010 and 2011-2015, respectively. For the first two decades (1971-1980 and 1981-1990), the average dry days (DD) was around 165 which increased to 181 in 1990-2000 (Fig. 5a). Then, it decreased to 171 and 178 in 2001-2010 and 2011-2015, respectively, seen in Fig. 5a. The average MCDD was increasing in the last few decades. Although there was some decrease in LRD (120) in 1991-2000, the average LRD remained almost constant in other decades (146). But, the average VRD was decreasing and the average DD was increasing with time with an exception in 1991-2000 ( Fig. 5a and b). The declining VRDs in the recent period highlight that the rainfall pattern is shifting towards more spatially heterogeneous. Valley wide agriculture and their yield are greatly affected by such shifting rainfall patterns.

Local and valley-wide rainfall distribution
The average MCVRD decreased from 7. 5 (in 1971-1980 and 1981-1190) to 6.5 (in 2000-2010 and 2011-2015) with an exceptional increase up to 10 in 1991-2000 (Fig. 5c). The  (Fig. 5d). It shows that the contribution of valley-wide rainfall to the annual rainfall is decreasing with time, whereas the contribution of local rainfall to the annual rainfall is increasing with time. The results indicate a micro-climatic feature is more pronounced in recent times.

Trend analysis
The MK test was used to test the significance of monotonic trends in the whole time series to the rainfall frequency and rainfall amount for annual rainfall, different rainfall rate classes, and seasonal rainfall. The test was performed at 10% significance level. Spatial distribution of long-term rainfall trends and their magnitudes based on the MK test and Sen's slope for annual rainfall is shown in Fig. 6. All stations (except Kathmandu Airport) showed a falling trend for annual rainfall. All the mountain stations showed falling trends, ranging from 8.4 mm/year at Sankhu to 21.8 mm/year at Thankot, though the trend at Sankhu was not significant. A rising trend of 4.3 mm/year was observed at Kathmandu Airport station located at the central part of the Kathmandu valley. The magnitude of falling trends in the valley floor is lower compared with the mountain areas. Panipokhari and Khumaltar showed falling trends of 0.5 mm/year and − 1.3 mm/year, respectively. Sankhu (− 0.7 count/year), Panipokhari (− 0.9 count/year), and Godavari (− 0.9 count/ year) stations showed statistically significant falling trends, whereas Bhaktapur and Kathmandu Airport station had a statistically insignificant falling trend of − 0.2 count/year for annual rainfall frequency. Thankot (0.1 count/year) and Khumaltar (0.3 count/year) stations which are located at the western part of the Kathmandu valley showed rising trends for annual rainfall frequency; however, only the rising trend at Khumaltar station was statistically significant. The trend of rainfall frequency is mostly dependent on the rainfall frequency of class "A," whereas rainfall of classes "B" and "C" dominates the trend of total annual rainfall.
The spatial distribution of trend results of rainfall frequency and rainfall amount of different rainfall rate classes are shown in Fig. 6. In the rainfall frequency of class "A," statistically significant trends were identified in five stations with two rising (Sankhu and Khumaltar) and three falling trends (Godavari, Panipokhari and Kathmandu Airport). The rainfall frequency of class "A" at Sankhu station showed a Fig. 6 Station-wise trends for rainfall frequency, and rainfall amount for (a) Annual rainfall, (b) Rainfall class "A," (c) Rainfall class "B," (d) Rainfall class "C," and (e) Rainfall class "D" over the period of 1971-2015 (significance at 0.1). The unit of trend is mm/year and count/year for rainfall amount and rainfall frequency, respectively falling trend (− 0.3 count/year), whereas Bhaktapur station had a constant trend. The falling trends in rainfall frequency (ranging from − 0.3 count/year to − 0.6 count/year) and rainfall amount (ranging from − 13.9 mm/year to − 6 mm/ year) of class "B" in mountain stations were statistically significant. Two stations (Bhaktapur and Khumaltar stations) in the valley floor had a falling trend in rainfall frequency (− 0.2 count/year and − 0.1 count/year) and rainfall amount (− 6.1 mm/year and − 0.5 mm/year) of class "B," whereas Kathmandu Airport station had no trend in rainfall frequency (but rising trend of 2.3 mm/year in rainfall amount). Panipokhari station had a rising trend in rainfall frequency (0.1 counts/year) and a statistically significant rising trend in rainfall amount (4.5 mm/year) for rainfall class "B." All the stations in the Valley floor had no trends for rainfall frequency of class "C." Thankot and Godavari stations showed statistically significant decreasing trends, but Sankhu station showed a statistically insignificant rising trend for rainfall frequency and rainfall amount of class "C." However, mixed patterns were found in the trend of rainfall amount of class "C" on the valley floor. Rainfall classes of "D" and "E" had negligible trends for both rainfall frequency and rainfall amounts in all stations. For total annual rainfall, all the stations (except Kathmandu Airport station) showed a falling trend of which 3 stations were statistically significant. Godavari station located at the southern part of the valley showed a statistically significant consistent falling trend for both rainfall amount and frequency of rainfall classes "A," "B," "C" and for annual rainfall.
The MK test was employed to test the significance of monotonic trends of seasonal rainfall for the study period. All the mountain stations showed a falling trend (average 14 mm/ year) for rainfall in the monsoon season of which 2 stations (Thankot and Godavari) were statistically significant. For the monsoon season on the valley floor, half of the stations (Panipokhari and Kathmandu Airport) showed a rising trend (0.8 mm/year and 2.9 mm/year) and rest (Bhaktapur and Khumaltar) showed a falling trend (6.1 mm/year and 2.9 mm/year). Similarly, a mixed pattern was found for winter and pre-monsoon season. Interestingly, all stations showed a falling trend ranging from 0.2 mm/year to 1.2 mm/ year for the post-monsoon season. All the mountain stations showed a falling trend for rainfall in all seasons (except Godavari in pre-monsoon season and Sankhu in winter season), whereas a mixed pattern was found on the valley floor. Kathmandu Airport station showed a rising trend for all seasons except post-monsoon season. Significant falling trends in monsoon season have resulted in a significant falling annual trend at Thankot, Godavari and Bhaktapur stations. In the analysis of rainfall data from 1989 to 2009 by , in the pre-monsoon, an increasing rainfall trend was observed for almost all stations of the Bagmati river basin, including Godavari and Kathmandu Airport, which lies particularly in the Kathmandu Valley. Similar falling trend was observed during the post-monsoon period . Even when we see the findings of time-series rainfall analysis for the countrywide rainfall by Karki et al. (2017) using data from 1970and Duncan et al. (2013) from 1951-2007, similar trends were witnessed except for a few areas which might be because of differences in the data analysis methods used.
The analysis of rainfall trends for different seasons shown in Fig. 7 revealed that monsoon rainfall is decreasing in the mountain region. It showed that the recent decades are drier than in earlier decades. The falling trend of 11 mm/ year of class "B" contributed much to the decreasing monsoonal rainfall. The natural springs of mountains are the major source of water supply in the Kathmandu Valley. The decreasing trends of monsoon rainfall along with total annual rainfall in mountain areas may have a significant impact on these spring sources, resulting in reduced flow and even drying up of springs. The northern part of the valley floor showed an increasing trend of monsoon rainfall. Similarly, the rainfall frequency and rainfall amount of classes "B" and "C" were in a rising trend but class "A" was in a falling trend in the northern part of the valley floor. This part of the valley is mostly covered by high built land-use with a lower water infiltration rate. The analysis indicated that heavy rainfall was more frequent than light rain, which might cause high surface run-off in the urban area. The falling trend of rainfall frequency or the number of rainy days might affect the livelihood of the people as a result of increased temperature. In Kathmandu Airport station, the annual rainfall is increasing but the number of rainy days is decreasing, indicating more heavy rainfall events. The detailed information of Fig. 6 and Fig. 7 is tabulated in Supplementary Material I; Kathmandu Valley. Similar findings were revealed in a study by , where a decreasing trend in the frequency of rainfall days was observed while intensity of maximum precipitation was increasing in the Kathmandu Airport station. Alteration in the rainfall pattern, accompanied by changes in temperature and discharge, might result in water quality deterioration, decline in agricultural yield and hydropower generation, effect in aquatic biodiversity, etc. . Furthermore, the decreasing post-monsoon and winter rainfall might have a negative impact on seasonal agricultural production like paddy rice, wheat, barley, potato, etc. (Panthi et al. 2015).

Seasonal and annual rainfall series for the Bagmati River Basin
Similar seasonal and trend analyses were performed for a larger scale, i.e., in the Bagmati River Basin (hereinafter BRB), analyzing data from 41 stations. The basin is divided into three parts, covering Kathmandu Valley in the upper, mountain and hills in the middle and Terai plain in the lower parts of the basin (Babel et al. 2014). In the upper part of the basin, there is a critical water quality and quantity issue year-round, especially in winter, moving downward into the middle and lower part, landslides and monsoonal floods are prevalent, respectively. In between 1986 and 2015, the average rainfall was found to be 1676 mm. The highest observed rainfall was 3746 mm in Makwanpur gadhi (Index No. 919) in the year 1999. From Fig. 8, it can be clearly seen that the annual rainfall has been decreasing over time with 1775 mm between 1986 and 1990 and 1505 mm between 2011 and 2015. In the pre-monsoon season, precipitation patterns are almost constant and have not varied significantly over years. However, the monsoonal rainfall has increased from around 1390 mm between 1986and 1990to 1463mm between 1991and 2000. In recent years (2011-2015, the average rainfall dropped below 1200 mm. Since the lower part of the basin is Terai plain and has high agricultural potential, most of the people's livelihood is connected to crop production. The declining monsoonal precipitation directly affects the river flow impacting the irrigation systems present. In the study by Sharma and Shakya (2006), due to less water availability in the lower part of the basin, the yield of crops, particularly rice and wheat, declined. In the BRB, the winter season received the least rainfall followed by post-monsoon, pre-monsoon, and monsoon seasons. On average, the driest month seen in the study period was November, while the wettest months were July and August. Also, it is clearly observed from Fig. 8 that recent years are dryer compared with past decades. The detailed information on the seasonal and annual rainfall trend in the BRB is tabulated in Supplementary Material 1; Bagmati Basin.
The findings in our analysis showed an alike trend with other basin studies of the country. A study by Dahal et al. (2019) showed alteration in precipitation patterns in the Rosi river basin and witnessed effects on the river flow and water insecurity in the river basin. Unlike the BRB, in the Gandaki basin, the monsoon precipitation increased with time. However, no significant trend was observed which suggests that the basin has a high annual and seasonal rainfall variability. Due to an increase in rainfall in the monsoon season, the basin is likely to face water-induced disasters (Panthi et al. 2015). On the other hand, decreasing post-monsoonal and winter rainfall affected winter crops production due to soil moisture deficiency (Panthi et al. 2015). In a study by Khatiwada et al. (2016), in the Karnali Basin, the average rainfall declined by 4.91 mm/year and similar effects on food and livelihood of people were observed. Similar effects might be observed in the mountainous part of the Bagmati river basin.
Analyzing seasonal rainfall patterns, particularly in the BRB as illustrated by Shrestha and Sthapit (2015), is essential in circulating data-driven findings to relevant authorities like government and non-government institutions for Fig. 7 Spatial distribution of long-term seasonal rainfall trend at stations in the Kathmandu valley over the period of 1971-2015 (significance at 0.1). The unit of rainfall trend is mm/year constructing water infrastructures like barrage wherever required, initiation for flood control or evacuation, irrigation purposes, etc. The aforementioned water-induced disasters are major threats to infrastructure, agricultural land, and livelihoods throughout the basin. Therefore, it is crucial to cope with the hydro-climatic variability to minimize livelihood impacts and improve water security.

Conclusions and recommendations
The study investigated the spatial and temporal characteristics of rainfall in the Valley and highlighted two distinct types of rainfall distributions, each in mountain areas (annual rainfall of 1880 mm on average) and valley floor (1410 mm on average). We performed the trend analysis of daily, seasonal and annual rainfall patterns using the MK test. Daily rainfall analysis over the Valley clearly indicated the higher frequency of moderate and heavy rainfall events (rainfall class "B" and "C") and fewer light rainfall events (rainfall class "A") in mountain areas resulting in higher rainfall amounts in mountain areas compared with the valley floor. Light rain dominated more than half of the rainy days while it contributed to just one-fifth of the total rainfall. However, moderate and heavy rain, having low frequency compared to light rain, contributed to about 80% of the total rainfall amount. This heterogeneous pattern of rainfall, i.e., light rain-high in frequency but less in amount and moderate and heavy rain-high in amount but low in frequency, in the Valley is critical for river flow, agriculture, and their yield. The trend of rainfall frequency is mostly dependent on the rainfall frequency of light rain, whereas rainfall amount of moderate and heavy rain classes dominates the total annual rainfall.
If we see the annual rainfall trend for different subperiods, a significant reduction in rainfall was observed after 2000. With time, the amount of monsoonal and postmonsoonal rainfall decreased. There was a shift in rainfall pattern in the Valley with an increase in the number of dry days and a decrease in valley-wide rainfall days, which   for the Bagmati River Basin can significantly affect the cropping cycle in the Valley. The observations of this study depict a falling trend of an average of 14 mm/year in both monsoonal as well as annual rainfall in the mountains. In the valley floor, half of the stations showed a rising trend and rest showed a falling trend in monsoon season whereas all the stations except Kathmandu Airport showed a falling trend for annual rainfall. An analysis performed at the basin scale (Bagmati River Basin) also showed a similar decline in average annual rainfall, in recent years, from 1463 mm in 1991-2000 to 1200 mm in recent years (2011)(2012)(2013)(2014)(2015) leading to the aforementioned impacts on agriculture.
The study showed that both annual rainfall amount and rainy days decreased in the Valley over the study period. Our current study considered seven rainfall stations  which seem insufficient to assess the rainfall variability at small scales. Monitoring and understanding spatiotemporal rainfall distribution at small-scales is vital for local and regional water resource management; this study only represents an initial step towards evaluating this in the Valley.
New technologies (like X-radar) and practices have been developed to monitor the spatiotemporal variability of rainfall patterns at a higher resolution. Satellite-based rainfall estimation ) and use of the citizen science approach (Davids et al. 2018(Davids et al. , 2019 in rainfall measurement can be an alternative way to fill the spatial gaps and to cover the spatial heterogeneity.