5.1. Spatial Distribution of Mean Seasonal and Daily Precipitation Indices
In addition to the seasonal mean precipitation distribution, spatial patterns of mainly the high-intensity- and frequency-related extremes (R95/R99 percentile precipitation, RX1day and WD, R10, R20, respectively), which are relevant for water resources, as well as flood and agriculture management, are computed from 210 stations over the period of 1981–2010 (Figure 3
and Figure 4
The monsoonal precipitation dominates annual precipitation with contribution of around 80% of the annual precipitation, whereas precipitation during winter, pre- and post-monsoon seasons contribute only 3.5%, 12.5% and 4.0%, respectively [90
]. Therefore, monsoonal precipitation distribution (Figure 3
b) is similar to the annual precipitation distribution (Figure 1
a and [38
]), indicating three high-precipitation pocket areas around Lumle, Gumthang and Num that receive more than 3000 mm of monsoonal precipitation. On the other hand, dry leeward regions are in Mustang, Manang and Dolpa, featuring the lowest precipitation amounts of less than 150 mm. These findings are consistent with Böhner et al. [42
], who reported that the regions north of high mountains are drier (<500 mm) in general and, depending on the alignment of surrounding mountain ranges, river valleys and the valleys lying in middle mountains and hills are also relatively drier (500–1000 mm) than surrounding mountain slopes and ridges. Such a pattern is more pronounced in the lower reaches of the river valleys, which lie between the southern frontal mountains and northern elevated mountain range. Such comparatively dry valleys are distinct in the surroundings of 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 (Figure 3
b). On the other hand, as demonstrated earlier based on satellite data [47
], precipitation is high (1500–2500 mm) on the windward side of both mountain ranges that lie north and south of such river valleys, creating double-peaked rainfall bands from south to north. In contrast to the river valleys lying within the northern elevated mountain range and southern frontal mountains, the majority of river valleys in Pokhara (near Lumle) and the surrounding region receive abundant (>1250 mm) monsoonal precipitation.
The precipitation in pre-monsoon, monsoon and post-monsoon seasons, more or less, follows the same spatial pattern in terms of representing three-peak precipitation pocket areas, as well as an east to west gradient. Pre-monsoon precipitation, mostly associated with thunderstorms, is very low in CH and the western half of the lowlands. There is a variation from less than 100 mm in WL, CH and in some areas on the leeward side of high mountains in the eastern region, to more than 700 mm in precipitation pocket areas and eastern mountains. Post-monsoon is the transition season between monsoon and winter. Therefore, following the retreat of monsoon from west to east, the western half of the country remains very dry, receiving below 40 mm of precipitation, whereas it is more than 200 mm in the eastern mountainous region. In contrast to other seasons, winter precipitation is higher over WM (>200 mm) and in a few isolated wet pockets in the central and eastern high mountainous regions. Winter precipitation is lowest over eastern lowlands (<20 mm) and feature a clear west to east as well as north to south gradient (Figure 3
d). Our seasonal spatial maps are broadly similar to the previous studies for all of Nepal [29
]; however, the inclusion of a large number of stations with the improved spatial interpolation technique has resulted in a much finer and realistic representation.
Contrary to the distribution of annual precipitation (Figure 1
a), suggesting three rainfall pockets in Nepal, high values of extreme percentile thresholds are found in the lowlands (Terai and Siwalik), while low values are found in the highland regions (Figure 4
a,b). For instance, annual mean of R99 percentile threshold in the lowlands is around 150 mm, whereas it is only about 30 mm in the highlands. Interestingly, the spatial pattern of R99 is quite similar to R95, and to some extent, to ever-recorded RX1day (Figure 4
c). The high values of extreme thresholds in lowlands are associated with fewer heavy rainfall events (Figure 4
a,b), while their relatively lower values observed in the middle mountains—despite the high annual precipitation (Figure 1
a)—are due to weak but persistent rainfall [45
]. The ever-recorded one-day maximum precipitation (RX1day) is found lowest in the central highlands followed by western highlands, whereas it is highest in the central lowlands followed by eastern and western lowlands. These spatial patterns are broadly similar to RX1day mapping using ground observation [92
] and 90th percentile threshold mapping using TRMM data set [67
]. The overall high values of R95, R99 and RX1day over lowlands suggest that these regions are more prone to soil erosion, landslides, flash flooding and subsequent inundations.
Unlike R95, R99 percentiles, and ever-recorded RX1day, spatial patterns of WD, R10 and R20 are largely similar to the annual distribution of precipitation. WD is lowest (below 40 days) in the very dry region of Mustang while it is highest (above 180 days) in EH followed by CM, WH and WM regions (Figure 4
f). On average, WD is around 100 days over the lowlands and around 140 days over high and middle mountains. Similar to WD, R10 is lowest over Mustang, featuring only less than 11 days whereas R10 is above 110 days for the EH region (Figure 4
d). On average, R10 is around 70 days over most of Nepal. R20 is almost zero over the dry Mustang region and at the minimum over WH and CH as these regions rarely experience very heavy precipitation events. However, over EH and CM regions and particularly over the high precipitation pockets, R20 is more than 50 days (Figure 4
e). On average, R20 is around 40 days across the whole country. Interestingly, from Figure 4
, it can be clearly noted that the frequency-related extreme indices of WD, R10 and R20 feature quite an opposite north−south gradient compared to high-intensity-related extreme indices of R95/R99 percentile precipitation and ever-recorded RX1day, with few sporadic exceptions.
The large spatial heterogeneity of mean precipitation across the country in different seasons indicates the requirement of localized information for water resources management and planning. Terai, Siwalik and river valleys, for example, are more prone to flood disaster, while mountains and hills face a higher landslide risk. The R95 and R99 thresholds computed here can also be directly utilizable for fixing the thresholds for flood warnings in different regions of Nepal.
5.2. Trend Analysis
In addition to the spatial distribution of mean precipitation and extreme indices, their time evolutions have been analyzed in order to see how these indices are changing over time. For this, trend slopes from the individual stations, their field significance for the seven geophysical sub-regions of Nepal along with the summary of these trend features are shown in Figure 5
, Figure 6
, Figure 7
, Figure 8
and Figure 9
. The stationwise trends and percentages of the negative/positive trends along with their significance are also additionally included in Supplementary Materials Table S1
5.2.1. Seasonal Precipitation
All stations show a mixed pattern of increasing and decreasing trends for the pre-monsoon precipitation across Nepal (Figure 5
a). However, around 62% of the total stations feature a rising trend, where such a trend is significant only at around 12%. Most of these stations are mainly concentrated within EL, CL, WL and CH regions. Only 4% of the total stations suggest a significant decreasing trend in pre-monsoon precipitation (Figure 6
). The results of field significance analysis are also congruent with the stationwise trends, indicating a significant rise in pre-monsoon precipitation over WL, CL, EL and CH. These findings are mostly consistent with Duncan et al. [29
] who have shown a countrywide precipitation rise. For the field insignificant middle mountain regions, differences with Duncan et al. [29
] may arise due to employing distinct observational datasets and methodology. Since pre-monsoon precipitation is mostly accompanied with thunderstorms as evident heavily over EL [94
], rise in the pre-monsoon precipitation over lowlands and CH regions will increase the extremely intense thunderstorms. Further, an increase in the pre-monsoon precipitation indicates changes in the seasonality of the precipitation regime over such regions [95
The monsoon precipitation features a mixture of drying and wetting trends (Figure 5
b). About one-fifth of the total stations exhibit significant trends with around 11% negative and 7% positive trends (Figure 6
). The significant negative trends are concentrated in the central and eastern parts while significant positive trends are found in WM and CH regions. Field significance analysis more clearly suggests the rising and falling trends of monsoonal precipitation over the designated regions. For instance, increase in monsoonal precipitation is significant over the WM and CH regions whereas decrease is significant over CM, CL and EL regions, largely consistent with the signals observed at the local stations. Since the monsoonal precipitation is very important for summer crops (paddy, maize and millet), which constitute around 80% of the total national cereal production in Nepal [56
], decreasing monsoonal precipitation may significantly affect the yield of cereal crops, as a significant decrease in the yield of rice has already been reported for the years of below-normal monsoonal precipitation [97
Interestingly, most of the stations (92%) show a decrease in the post-monsoon precipitation, where such a signal is statistically significant at 41% of the total number of station. Decreasing post-monsoon precipitation is further suggested by the field significant decrease in all regions except for WL and CH (Figure 5
c). This is in agreement with the findings of Khatiwada et al. [98
], who have also indicated a decreasing post-monsoon precipitation over the Karnali basin in western Nepal for the 1981–2012 period. Over the same period, a significant decrease in precipitation during the post-monsoon dry months of November and winter season month December was also noticed in the Gandaki river basin of Nepal [57
]. The observed decreasing post-monsoon precipitation may adversely affect the production of paddy crop, as it enters into sensitive stage of spikelet formation, fruiting and ripening, requiring more water during the post monsoon season [100
Similar to post-monsoon precipitation, most of the stations feature a negative trend (68%) for winter precipitation over Nepal. However, such a negative trend is statistically significant at only 4% of the total stations, mainly lying within the WM region. Field significance analysis also suggests a significant decreasing trend for the winter precipitation over WM (Figure 5
d). Khatiwada et al. [98
] have also reported a decreasing winter precipitation over the Karnali basin in western Nepal for the 1981–2012 period. Based on GPCP and satellite-based datasets, Wang et al. [101
] have likewise identified declining winter precipitation over the western region of Nepal in recent decades. Consistently, weakening influence of the western disturbances over the central Himalayas has also been found [102
], and in line with this, decreasing winter precipitation has been reported in the adjoining western Himalayan region in India [103
]. Further, Wang et al. [101
] have attributed this decline to three main factors: (1) decadal trend towards negative phase of arctic oscillation in recent decades that has created a local mass flux circulation with descending branch over western Nepal; (2) the Indian ocean warming, and; (3) the anthropogenic aerosol loading. It is pertinent to mention that the winter precipitation, though low in volume, plays an important role in meeting the water demand of the winter crops and in feeding the rivers through accumulating their headwaters with snow that melts during the dry pre-monsoon season [42
]. Particularly for the western hills and mountainous regions where food production is largely dependent upon rain-fed agriculture, decreasing winter precipitation may affect the winter crop production [101
] of wheat, barley and potatoes, a major crop of the hills and mountains. Moreover, the decreasing winter precipitation in the region, where winter precipitation is substantially higher than in other regions, could also lead to a reduction in pre-monsoon season river flows, which are largely dependent on the snow and glacier melt during the dry season.
5.2.2. High-Intensity-Related Precipitation Extremes
The stationwise trends of R95 (total annual precipitation from the days of a year featuring >95 percentile precipitation) and RX1day and RX5day (annual maximum 1-day and 5-day precipitation) indices, along with their field significance, are shown in Figure 7
The analysis reveals a mixture of equally increasing and decreasing trends in R95 across Nepal with only one-fifth of the total stations featuring significant trends. For instance, around 16% of the total stations show a significant positive trend in R95 while only 4% show its significant negative trend (Figure 6
). The stations featuring a statistically significant rise in R95 are mainly concentrated within the western part of the country. Rising trends in R95 at local stations are found field significant over the WM and CH regions, while falling trends in R95 are field significant over the CM and EM regions.
For RX1day indices, about 60% of the analyzed stations feature falling trends and around one-tenth of the total stations show such a falling trend as significant. A large number of stations showing falling RX1day are located in CM and EM regions. In contrast, stations in the western region (WL and WM) have a higher number of increasing trends for RX1day. Similar to RX1day, a higher number of negative trends in RX5day is observed within CM and EM regions, while more positive trends are found in the western regions (WL and WM). Field significance results are largely in agreement with the stationwise trends. Positive trends in RX1day and RX5day are field significant over the WM, CL and CH regions, while negative trends are field significant over the CM region and RX1day decrease is additionally field significant over the EM region. Such a pattern of change in RX1day is consistent with the previous studies [32
] that also report a decreasing trend of RX1day from most of the stations, where such a trend is particularly significant above 100 m (asl) within the Koshi River basin—a basin spanning mainly over the eastern and partially over the central region of our study area.
In summary, all three indices of R95, RX1day and RX5day feature a field significant rising trend over the WM and CH regions, whereas over the CL region, only the latter two are field significant. Decreasing trends in R95, RX1day and RX5day are found field significant over the CM region, while the former two are additionally field significant over the EM region. Coherence amid field significance rising trends of all three intensity-related extreme precipitation indices together with the dominance of stationwise significant rising trends over WM and CH regions indicate that precipitation extremes might be more intense in the near future.
On the other hand, decreasing field significant trends of all three indices over the CM region and of two indices, R95 and RX1dy, over the EM region indicate to some extent the weakening of intense precipitation extremes over such regions.
Since our analysis period is only until 2012, occurrence of extreme events during 14–17 June 2013 in Uttarkhanda, India and the bordering areas of western Nepal, and the 14–16 August 2014 event in western Nepal indicates the continuation of this pattern in the western region of Nepal. Further, Cho et al. [105
] attributed the increase in extreme precipitation events like that of Uttarkhanda and the bordering region of western Nepal in recent decades to an amplification of an upper tropospheric mid-latitude shortwave trough pattern in the northern region of South Asia due to the increase in greenhouse gases and aerosols. In general, this kind of amplification in association with west–northwestward migration of monsoon low creates the highly favorable environment for vigorous interaction of tropical (monsoon) and extra-tropical (mid-latitude) circulation resulting in extreme precipitation events in the western Himalayan region [106
]. Thus, the western mountainous region of Nepal lying at a higher latitude and being an adjacent region of western Himalayas could have greater influence of such mid-latitude wave train pattern, whereas opposite or no influence of that pattern can occur towards eastern region. However, the shortwave train pattern was analyzed only for June and no conclusions could be made for whole monsoon season during which extreme precipitation events occur.
In addition, physical mechanism responsible for enhancement of monsoon precipitation and extremes towards the central and eastern region of foothills of Himalayan region are normally associated with break/active monsoon condition in mainland India/Himalayan foothills (north ward migration of monsoon trough towards foothills of Himalaya from its normal position) during which interaction of southward migration of extratropical westerly troughs (dry air subsidence in Indian subcontinent at mid-to-upper troposphere) and weak monsoon circulation takes place [107
]. The detail analysis of changing pattern of break monsoon situation and other physical mechanism responsible for changing extreme precipitation pattern over the whole monsoon season for Nepal is still lacking and it could be far from simple as the complex interaction of global, synoptic scale weather systems and topography takes place in the monsoon dominated region producing localized extreme precipitation.
The increasing intense precipitation over western mountainous region indicates higher risks of soil erosion and landslides in the fragile mountainous regions which are extremely vulnerable to these disasters due to manmade activities—deforestation for settlement, road network and agricultural activities—or natural causes like earthquakes. Lacking adaptive capacities of the remote mountains and hills further aggravate the situation. Additionally, increasing intense precipitation events in these regions consequently increases the risk of floods and inundation in densely populated river valleys and southern lowlands, destroying life and property, and causing damages to the agricultural land thereby impacting the socioeconomic development.
Nevertheless, absence of increasing trend of extreme intense precipitation in the central region where land is highly ruptured with recent earthquake still requires suitable adaptive measures to reduce the risk of landslides in the regions, as continuous but low intensity and even normal extreme threshold precipitation values of rainfall can easily trigger such disasters in the region.
5.2.3. Frequency-Related Precipitation Extremes
Since days with 10 mm and 20 mm precipitation events are quite common during the monsoon season over many parts of Nepal (Figure 4
d,e), these events which are typically defined as heavy and very heavy precipitation in fact represent only moderate precipitation events over large areas of Nepal. Our results show a mixed pattern of stationwise increasing and decreasing trends for R10 (annual number of days with ≥10 m) across Nepal. However, around 57% of the total stations show a negative trend, which is significant at around 16% of the stations mostly concentrated in the central and eastern regions (Figure 6
). Field significance test results also suggest the same, exhibiting significant negative trends for the CM, EM, CL and EL regions (Figure 8
a). The pattern of change in R20 (annual counts of days when Precipitation is ≥20 mm) likewise indicates a mixed response with approximately equal numbers of rising and falling trends (Figure 8
b). However, R20 is significantly decreasing for the regions of CL, EL and WM and significantly increasing for the region of CH. Further, coherent decrease in R10 and R20 over CL and EL clearly suggests a decrease in the number of moderate precipitation events over the respective regions.
5.2.4. Dry and Wet Spells
The analysis of CDD (consecutive dry days) indices suggests a widespread increase in the dry period over the whole country (Figure 8
c). Around 80% of the analyzed stations exhibit an increase in CDD over the period of 1970–2012, though this trend is significant only at 22% of the total (Figure 6
). In line with the stationwise trends, field significance trends are also statistically significant for all the sub-regions, except for CH. This finding is consistent with previous studies of Shrestha et al. [32
] and Sigdel and Ma [104
] over Koshi basin of Nepal. A similar increase in CDD has also been reported for the Songhu River basin in China [108
] and across Bangladesh [109
], as CDD is mostly related to the large-scale weather systems rather than localized systems [68
The CWD (consecutive wet days) also reveals a mixed pattern (Figure 8
d). However, the field significance analysis suggests a significant decreasing trend for CL and EL regions but a significant increasing trend for EM and CH regions. The CWD changes are mainly observed during the monsoon season; hence, such changes do not necessarily corresponds to changes in CDD, which are observed mostly in the dry seasons.
It is worth mentioning that duration and occurrence-related indices of CDD and CWD can only indirectly characterize the drought, which is a complex phenomenon and depends upon many other factors besides precipitation. Nevertheless, increasing CDD observed over the study area is consistent with the most widespread and worst drought observed in recent decades across the country [99
]. Rise in CDD clearly indicates the prolongation of the dry period across the country, implying certain changes in the seasonality of prevailing precipitation regimes [95
]. Further, this increase in the dry period can negatively impact crop yield and hydropower generation and can, moreover, elevate respiratory-related health problems in Nepal by increasing the concentration of particulate matter in the air. Since the frequency and scale of the forest fires in Nepal and other regions are also strongly related to the length of dry spells [112
], the rise in CDD will aggravate such events, endangering wildlife and causing huge socioeconomic losses.
5.2.5. Extra Indices (PRCPTOT, SDII and WD)
One-fifth of the total number of stations exhibit significant trend changes in PRCPTOT. Around 13% of stations suggest a significant falling trend while 7% of stations suggest a significant increasing trend (Figure 6
). The majority of statistically significant negative trends are concentrated in the central and eastern parts while higher numbers of significant positive trends are observed in the western parts of the country (Figure 9
a). Interestingly, the stations above 29° N reveal an increasing trend while those below this latitude features a decreasing trend. The field significance test also indicates a significant decrease in PRCPTOT over CM, CL, EM and EL regions and a significant increase over WM and CH regions. The increasing trend in WM is consistent with Baidya et al.’s [28
] findings for the western region. The spatial pattern of trend changes in PRCPTOT is quite similar to that of the monsoonal precipitation as it dominates (about 80% of) the total annual precipitation in Nepal. Decreasing PRCPTOT over the eastern region that covers most of the Koshi River basin within Nepal is consistent with the reports of significant precipitation decrease over the Koshi River basin during the 1994–2013 period [50
] and over the middle mountains and hills during the 1975–2010 period [32
]. Based on the coarse resolution Global Precipitation Climatology Project (GPCP) dataset, Yao et al. [113
] have also provided evidence of decreasing precipitation over all the Himalayas and an increase in the eastern Pamir regions during 1979–2010.
Notably, WD (wet days/rainy days) is decreasing (at 67% of the stations) across most of Nepal with a statistically significant decreasing pattern at one-third of the stations. Additionally, the individual stations’ trends are also consistent with the field significant trends for all sub-regions, except for CH (Figure 9
c). These findings are further consistent with the regional pattern of decreasing number of wet days over the whole of Southeast Asia [114
The simple daily intensity index (SDII), defined as the ratio of PRCPTOT to WD, exhibits a positive trend (60% of the stations) across Nepal with statistically significant increasing trends at 14 stations (18%) and significant decreasing trends at four stations. Unlike significant decreasing trends in PRCPTOT over EM, EL and CL regions, in accord with Shrestha et al. [32
], SDII features a significant increasing trend. An increase in SDII over such regions is mainly due to the higher decrease in WD than in PRCPTOT (Figure 9
b). On the other hand, the reason for the increase in SDII over the western region is due to an increase in PRCPTOT but decrease in WD. Rising trends in R95, RX1day and RX5day over the WM region further indicate rising high-intensity precipitation extremes over the region. In contrast, decreasing trends in PRCPTOT, R95 and RX1day over the EM and CM region somewhat reinforce a decreasing trend of high-intensity precipitation extremes. Additionally, significant decreasing trends in WD, R10 and R20 indices over EL and CL regions indicate a decrease in the moderate daily precipitation events.