Rising Precipitation Extremes across Nepal

25 Pages·2017·2.29 MB·English

climate
Article
Rising Precipitation Extremes across Nepal
RamchandraKarki1,2,*,ShabehulHasson1,3,UdoSchickhoff1,ThomasScholten4
andJürgenBöhner1
1 CenterforEarthSystemResearchandSustainability,InstituteofGeography,UniversityofHamburg,
Bundesstraße55,20146Hamburg,Germany;[email protected](S.H.);
[email protected](U.S.);[email protected](J.B.)
2 DepartmentofHydrologyandMeteorology,GovernmentofNepal,406Naxal,Kathmandu,Nepal
3 DepartmentofSpaceSciences,InstituteofSpaceTechnology,Islamabad44000,Pakistan
4 SoilScienceandGeomorphology,UniversityofTübingen,DepartmentofGeosciences,
Rümelinstrasse19-23,72070Tübingen,Germany;[email protected]
* Correspondence:[email protected];Tel.:[phone]
AcademicEditor:ChristinaAnagnostopoulou
Received:10November2016;Accepted:6January2017;Published:13January2017
Abstract: As a mountainous country, Nepal is most susceptible to precipitation extremes and
related hazards, including severe floods, landslides and droughts that cause huge losses of life
andproperty,impacttheHimalayanenvironment,andhinderthesocioeconomicdevelopmentof
thecountry. Giventhatthecountrywideassessmentofsuchextremesisstilllacking,wepresenta
comprehensivepictureofprevailingprecipitationextremesobservedacrossNepal. First,wepresent
the spatial distribution of daily extreme precipitation indices as defined by the Expert Team on
ClimateChangeDetection,MonitoringandIndices(ETCCDMI)from210stationsovertheperiod
of1981–2010. Then,weanalyzethetemporalchangesinthecomputedextremesfrom76stations,
featuringlong-termcontinuousrecordsfortheperiodof1970–2012,byapplyinganon-parametric
Mann−KendalltesttoidentifytheexistenceofatrendandSen’sslopemethodtocalculatethetrue
magnitude of this trend. Further, the local trends in precipitation extremes have been tested for
theirfieldsignificanceoverthedistinctphysio-geographicalregionsofNepal,suchasthelowlands,
middlemountainsandhillsandhighmountainsinthewest(WL,WMandWH,respectively),and
likewise,incentral(CL,CMandCH)andeastern(EL,EMandEH)Nepal. Ourresultssuggestthat
thespatialpatternsofhigh-intensityprecipitationextremesarequitedifferenttothatofannualor
monsoonalprecipitation. Lowlands(TeraiandSiwaliks)thatfeaturerelativelylowprecipitationand
lesswetdays(rainydays)areexposedtohigh-intensityprecipitationextremes. Ourtrendanalysis
suggeststhatthepre-monsoonalprecipitationissignificantlyincreasingoverthelowlandsandCH,
whilemonsoonalprecipitationisincreasinginWMandCHanddecreasinginCM,CLandEL.Onthe
otherhand,post-monsoonalprecipitationissignificantlydecreasingacrossallofNepalwhilewinter
precipitation is decreasing only over the WM region. Both high-intensity precipitation extremes
and annual precipitation trends feature east−west contrast, suggesting significant increase over
theWMandCHregionbutdecreaseovertheEMandCMregions. Further,asignificantpositive
trendinthenumberofconsecutivedrydaysbutsignificantnegativetrendinthenumberofwet
(rainy)daysareobservedoverthewholeofNepal,implyingtheprolongationofthedryspellacross
thecountry. Overall,theintensificationofdifferentprecipitationindicesoverdistinctpartsofthe
countryindicatesregion-specificrisksoffloods,landslidesanddroughts. Thepresentedfindings,
incombinationwithpopulationandenvironmentalpressures,cansupportindevisingtheadequate
region-specificadaptationstrategiesfordifferentsectorsandinimprovingthelivelihoodoftherural
communitiesinNepal.
Keywords: Nepal; spatial precipitation pattern; precipitation extremes; consecutive dry days;
high-intensityprecipitation
Climate2017,5,4;doi:10.3390/cli5010004 www.mdpi.com/journal/climate
Climate2017,5,4 2of25
1. Introduction
Precipitationextremesareoneofthemajorfactorsthattriggernaturaldisasters,suchasdroughts,
floods and landslides, which subsequently cause the loss of property and life, and deteriorate
socioeconomicdevelopment. Undertheprevailinganthropogenicwarming,theprecipitationextremes
areobservedtobeintensifiedglobally,exacerbatingtheexistingproblemsoffoodandwatersecurity
aswellasdisastermanagement[1–11].
Consistentwiththeglobalpattern[4,10],theworld’smostdisaster-proneregionofSouthAsia[12]
has also experienced an overall increase in precipitation extremes [11], though such a pattern is
heterogeneousacrosstheregion[5,11,13,14]. Forinstance,studieshavefoundariseinthesummer
monsoonalprecipitationextremesovercentralIndiaandnortheasternPakistan[15–20]. Incontrast,
precipitationextremesfeatureafallingtrendoversouthwesternPakistan[20],theeasternGangetic
plainsandsomepartsofUttaranchal,India[21]. Further,contrarytotheextremesobservedatlow
altitudesorovertheplains,extremesobservedinthehigh-altitudemountainousregionsexhibitquite
anoppositesignofchangeduetotheinfluenceoflocalfactors,andarethuslesspredictable[22].
SituatedinthesteepterrainofthecentralHimalayanrange,Nepalislikewisemoresusceptibleto
thedevelopmentsofheavyrainfalleventsandsubsequentfloodinganddroughtsseverelyimpacting
themarginalizedmountaincommunities,aswasimpressivelyillustratedbyrecentevents.Forinstance,
thecloudburstof14–17June2013inthenorthwesternmountainousregionneartheNepaleseborder
killedaround5700peopleandaffectedmorethan100,000,extensivelydamagingthepropertyinboth
NepalandIndia[23]. Aheavyraineventof14–16August2014likewisecausedmassiveflooding
andtriggeredanumberoflandslides,resultinginhugelossesoflifeandproperty,affectingaround
35,000 households [24]. Similarly, one of the worst winter droughts of the country in 2008/2009
reduced yield of wheat and barley by 14% and 17%, respectively, leading to severe food shortage
in 66% of rural households in the worst hit far- and mid-western hill and mountain regions [25].
Suchintenseprecipitationandextremedryness(droughts)negativelyimpacttheyieldofbothcash
and cereal crops [26], and in turn, the livelihood of around 60% of the total Nepalese population
directlydependentonagriculture[27]. Therefore,analyzingtheprecipitationextremesandtheirtime
evolutionunderprevailingclimaticchangesisofparamountimportanceforensuringfoodandwater
securityinNepalanddevelopingaregion-specificdisastermanagementstrategy.
As compared to the rest of South Asia, studies on the observed precipitation extremes over
Nepal are rare and sporadic, lacking a comprehensive picture across the country. For instance,
computingvariousextremeprecipitationindicesfromonly26stationsfortheperiodof1961–2006,
Baidyaetal.[28]havefoundanincreasingtrendintotaleventsandheavyprecipitationeventsfrom
most of the stations. In contrast, analyzing only a subset of extreme precipitation indices used in
Baidya et al. [28], from the daily interpolated gridded precipitation of APHRODITE (1951–2007)
Duncanetal.[29]concludesthatthemonsoonalandannualprecipitationextremesareunlikelyto
worsenoverNepal. Sinceprecipitationextremesaremorelocalizedandcanbesmoothedoverwhen
limitedgaugedataisinterpolated[30],employingAPHRODITEmayhavesignificantlyaffectedthe
computationofextremestatisticsasithardlyincorporatesanyrealobservationsfromtheearly1950s
westofKathmandu[31]. Recently,analyzingtheextremeprecipitationindicesonlywithintheKoshi
RiverbasinineasternNepalforthe1975–2010period,Shresthaetal.[32]havereportedanoverall
increase in the precipitation total and intensity, though such trends were statistically insignificant.
ThesediscordantfindingsofchangesintheprecipitationextremesoverNepalmaybeattributedto
employingvaryingdatasets,analyzingdifferenttimeperiods,andfocusingondistinctstudyregions,
hencelackingacountrywidepicture.
In view of these limited countrywide studies with contrasting findings—and because an
understandingoftheextremeprecipitationeventsiscrucialtosocioeconomicdevelopment—,this
study presents an exploratory analysis of the widely adopted daily precipitation extremes across
thewholeofNepalbasedonthemaximumnumberofhigh-qualitylong-termstationobservations.
Forthis,theextremeindicesfromtheExpertTeamonClimateChangeDetection,Monitoringand
Climate2017,5,4 3of25
Indices(ETCCDMI)alongwithafewadditionalindicesarecomputedfromthedailyprecipitation
observations. First, we have analyzed the spatial distribution of the most relevant precipitation
extremesaswellastheseasonalprecipitationpatternsfrom210surfaceweatherstationsacrossNepal
over the most recent period, 1981–2010. Moreover, time evolution of the computed precipitation
extremes has been analyzed by ascertaining the monotonic trends from the long-term continuous
record available at 76 stations over the 1970–2012 period. In this regard, a robust non-parametric
Mann−Kendall [33,34] trend test along with the trend free pre-whitening (TFPW) procedure has
beenapplied. Thetrendsatthelocalstationsarefurtherassessedfortheirfieldsignificanceoverthe
distinctphysio-geographicregionsofNepal,inordertoestablishthedominantpatternsofchangesin
precipitationextremes.
2. StudyArea
Nepalisamountainouscountrythatstretchesbetween26.36◦ N–30.45◦ Nand80.06◦ E–88.2◦ E,
encompassing an area of 147,181 km2 and an elevation range of 60–8848 m above sea level (asl).
Alongthecardinaldirectionsandaltitude,thecountryisdividedintofivestandardphysiographic
regions, such as Terai, Siwaliks, Middle Mountains, High Mountains, and High Himalaya [35,36],
whichaccordingtoDuncanandBiggs[37]arefurthercategorizedintothreebroaderzones,namely,
Lowlands(TeraiandSiwaliks),Mid-MountainsandHills(MiddleandHighMountains)andHigh
Mountains(HighHimalayas). Owingtosuchuniquephysiographicalandtopographicaldistribution,
thecountryfeaturesavarietyofclimatesthatrangefromthetropicalsavannahoverthesouthern
plains to the polar frost in the northern mountains within a short horizontal distance of less than
200km[38]. Thepopulationdensityishighestintheeasternlowlandsregion,followedbytheregions
ofcentrallowlands,middleandeasternmiddlemountains,andwesternlowlands,respectively,while
suchdensityislowestfortherestofNepal.
TherearefourseasonsinNepal,namely,pre-monsoon(March–May),monsoon(June–September),
post-monsoon(October–November)andwinter(December–February). Pre-monsoonischaracterized
byhot,dryandwesterlywindyweatherwithmostlylocalizedprecipitationinanarrowband,whereas
themonsoonischaracterizedbymoistsoutheasterlymonsoonalwindscomingfromtheBayofBengal
andoccasionallyfromtheArabianSeawithwidespreadprecipitation. Post-monsoonreferstoadry
seasonwithsunnydaysfeaturingadriestmonth,November. Winterisacoldseasonwithprecipitation
mostlyintheformofsnowinhigh-altitudemountainousregions. PrecipitationoverNepalisreceived
by two major weather systems; the southwest monsoon greatly impacts the southeastern parts of
the country during the monsoon season while the western disturbances predominantly affect the
northwestern high mountainous parts during the winter season [39–43]. Similar to the monsoon
season, precipitation during pre- and post-monsoon seasons is also generally higher towards the
east[44]. IntheMarsyangdiRiverbasin(MRB)ofNepal,observedprecipitationatthestationsbelow
2000mismostlyreceivedintheformofrainwhileatthestationsabovethisheight,snowfallaccounts
for17%±11%oftheannualtotals,wheresuchafractionriseswithaltitude[45].
ClassifyingtheprecipitationregimesofNepalbasedontheshapeandmagnitudeofmonthly
precipitationfrom222stations, Kansakaretal.[35]haveillustratedthattheprecipitationpatterns
are mainly controlled by the orographic effect of the complex central Himalayan terrain and the
east−westprogressionofthesummermonsoon. Thus,owingtotheintricateinteractionamidthe
weathersystemsandtheiralterationbytheextremetopographicalvariations(highmountains,valleys
andrivercatchments),spatialdistributionofprecipitationinNepalishighlyheterogeneous(Figure1a).
Forinstance,theannualprecipitationvariesfromlessthan200mmforthedriestregions(Mustang,
Manang,andDolpa,locatedattheleeward-sidenorthoftheAnnapurna)toabove5000mminand
around the Lumle region. Two additional wetter regions with annual precipitation greater than
3500mmareNumandGumthang. Ontheotherhand,regionsoflowprecipitationtypicallyalsoreside
intheleeward-sideoftheKhumbu,Everestandotherhighmountainousregions[38,41,46]. Alongthe
altitudinal extent of the central Himalayan region (~74◦–88◦ E), the Tropical Rainfall Measuring
Climate2017,5,4 4of25
Mission(TRMM)precipitationdataset(1998–2005)indicatestwoparallelsignificantpeakprecipitation
Climate 2017, 5, 4 4 of 25
zones[47];thefirstzoneisatthemeanelevationof~0.95km(meanreliefof~1.2km)whilethesecond
aolntietuisdainta~l2 g.1rakdmie(nmt oeaf nprreecliiepfitoafti~o2n kbmet)w. Beeanrr 1o0s0e0t aanl.d[4 455]0h0a mve aslutigtugdeset,e wdhaewreeaask ina ldtietuepd irnivalerg rvaadllieeynst
wofitphr esctiepeipta tsiloonpebse,t wsuecehn a1 0g0r0aadniedn4t 5o0f0 rmainaslttiotrumd ei,sw vheerrye asstrionndge. eIpn rgiveenrervaall,l ethyseyw hitahvset efeopunsldo pthese,
msuacxhimaugrmad pireenctipoiftaratiionns taolromngi sthvee rriydsgterso nangd. I an sgtreonnegra ela,stht−ewyehsat vriedfgoeu-ntod-rtihdegem parxeicmipuimtatpiornec gipraitdaiteinotn.
Sailmoniglarth reesruidltgse wsaenred oabsttarionnegd efraosmt− wstaetsitorni-dbgaes-etdo -oribdsgeervparteiocinpsi tiant idoinffgerreandti ernetg.iSoinms iolaf rNreespuallt swwheerree
pobretaciipnietdatiforonm pesataktsi oanr-obuansded 25o0b0s–e3r6v0a0ti omnms i nanddi fdfeercernetasreesg ifounrsthoefr Nweiptha linwchreearseep irne cailptiittuadtieo ninp heaigkhs
maroouunndta2in5 0r0e–g3io60n0s m[4m8–5a1n]d. decreasesfurtherwithincreaseinaltitudeinhighmountainregions[48–51].

Figure 1. (a) Annual precipitation distribution (mm) over Nepal with precipitation pocket areas and
Figure1.(a)Annualprecipitationdistribution(mm)overNepalwithprecipitationpocketareasand
dry areas delineated, and the meteorological stations used for the daily extreme analysis overlaid; (b)
dryareasdelineated,andthemeteorologicalstationsusedforthedailyextremeanalysisoverlaid;
The three broader physiographic zones of Nepal and nine sub-regions (WL—Western Lowlands,
(b)ThethreebroaderphysiographiczonesofNepalandninesub-regions(WL—WesternLowlands,
WM—Western Middle Mountains and Hills, WH—Western High Mountains, CL—Central
WM—WesternMiddleMountainsandHills,WH—WesternHighMountains,CL—CentralLowlands,
Lowlands, CM—Central Middle Mountains and Hills, CH—Central High Mountains, EL—Eastern
CM—CentralMiddleMountainsandHills,CH—CentralHighMountains,EL—EasternLowlands,
Lowlands, EM—Eastern Middle Mountains and Hills, EH—Eastern High Mountains).
EM—EasternMiddleMountainsandHills,EH—EasternHighMountains).
In addition to the horizontal and altitudinal precipitation gradients, a large seasonal
Inadditiontothehorizontalandaltitudinalprecipitationgradients,alargeseasonalprecipitation
precipitation gradient (~factor of 4) has also been observed over a short horizontal distance of ~10 km
gradient(~factorof4)hasalsobeenobservedoverashorthorizontaldistanceof~10kminMRB[52].
in MRB [52].
Precipitationdoesnotfeaturelong-termtrendsatseasonalandannualscales,exceptalocalized
Precipitation does not feature long-term trends at seasonal and annual scales, except a localized
trendinsomepartsoftheKoshiRiverbasin[44,53,54]. However,itfeaturesasignificantrelationship
trend in some parts of the Koshi River basin [44,53,54]. However, it features a significant relationship
with the Southern Oscillation Index (SOI) [55]. Changes in the monsoonal precipitation regimes
with the Southern Oscillation Index (SOI) [55]. Changes in the monsoonal precipitation regimes
indicatetheextensionoftheactivemonsoondurationmainlyduetosignificantlydelayedwithdrawal,
indicate the extension of the active monsoon duration mainly due to significantly delayed
thoughtheonsettiminghasbeenobservedunchanged[56,57].
withdrawal, though the onset timing has been observed unchanged [56,57].
Astheprecipitationdistributionishighlyheterogeneousacrossthecountry,characterizingstrong
As the precipitation distribution is highly heterogeneous across the country, characterizing
north−southandeast−westgradients,thewholecountryisdividedintoninesub-regions(Figure1b)
strong north−south and east−west gradients, the whole country is divided into nine sub-regions
forregionalfieldsignificancestudy. Thelatitudinalextenthasbeendividedbasedonthedemarcation
(Figure 1b) for regional field significance study. The latitudinal extent has been divided based on the
ofthreebroaderphysiographicregions,whileforthelongitudinaldivision,longitudesof83◦ Eand
demarcation of three broader physiographic regions, while for the longitudinal division, longitudes
86◦Ehavebeentakenasthedemarcationpoints(asusedin[44]),yieldingwestern,centralandeastern
of 83° E and 86° E have been taken as the demarcation points (as used in [44]), yielding western,
regions,eachcontainingthreephysio-geographicregions. Forinstance,thesub-regionswithinthe
central and eastern regions, each containing three physio-geographic regions. For instance, the sub-
westernlongitudinalbeltarethewesternlowlands(WL),westernmiddlemountainsandhills(WM)
regions within the western longitudinal belt are the western lowlands (WL), western middle
andwesternhighmountains(WH).Thecaseforthecentral(CL,CMandCH)andeastern(EL,EM
mountains and hills (WM) and western high mountains (WH). The case for the central (CL, CM and
andEH)longitudinalbeltsissimilar.
CH) and eastern (EL, EM and EH) longitudinal belts is similar.
3. Data
3. Data
WehaveobtaineddailyprecipitationdatafromallavailablesurfaceweatherstationsinNepal
We have obtained daily precipitation data from all available surface weather stations in Nepal
that are being maintained by the Department of Hydrology and Meteorology (DHM), Nepal.
that are being maintained by the Department of Hydrology and Meteorology (DHM), Nepal. These
Theseobservationsareconsistentintermsofthemeasurementmethodastheobtainedstationsusethe
observations are consistent in terms of the measurement method as the obtained stations use the same
sametypeofUS-standard8-inchdiametermanualprecipitationgauges[58]. Thoughunderestimated
type of US-standard 8-inch diameter manual precipitation gauges [58]. Though underestimated as in
asinothertypesofgauges,thesegaugescanalsomeasuresnowwaterequivalent. Forthat,thesnow
other types of gauges, these gauges can also measure snow water equivalent. For that, the snow
depositedinthegaugeisatfirstmeltedbypouringhotwaterandthenmeasuredasnormalrainfall
deposited in the gauge is at first melted by pouring hot water and then measured as normal rainfall
measurement. In the DHM database, the longest precipitation record available since 1946 is from only
three stations, although there are records from 23 stations since 1947, and from around 40 stations
since 1950. Until 1956, the precipitation observations were available only from the stations located
Climate2017,5,4 5of25
measurement. IntheDHMdatabase,thelongestprecipitationrecordavailablesince1946isfromonly
Climate 2017, 5, 4 5 of 25
threestations,althoughtherearerecordsfrom23stationssince1947,andfromaround40stationssince
1950. Until1956,theprecipitationobservationswereavailableonlyfromthestationslocatedeastof
east of Kathmandu and particularly within the Koshi River basin. Afterwards, the number of
KathmanduandparticularlywithintheKoshiRiverbasin. Afterwards,thenumberofprecipitation
precipitation stations considerably increased, reaching up to 100, 190, 240, 250, 370 and 410 by 1961,
stationsconsiderablyincreased,reachingupto100,190,240,250,370and410by1961,1971,1981,1991,
1971, 1981, 1991, 2001 and 2010, respectively. However, all the available precipitation gauges do not
2001and2010,respectively. However,alltheavailableprecipitationgaugesdonotfeatureregular
feature regular data since the time of their inception [31]. For instance, among 450 stations that have
datasincethetimeoftheirinception[31]. Forinstance,among450stationsthathavebeenoperational
been operational until recently, regular data of varying lengths from only around 400 stations either
untilrecently,regulardataofvaryinglengthsfromonlyaround400stationseitherduetoshort-term
due to short-term discontinuity of the stations or their relocation are available (Figure 2). In order to
discontinuityofthestationsortheirrelocationareavailable(Figure2). Inordertoensureabalanced
ensure a balanced spatial distribution of stations across Nepal and employing the maximum common
spatial distribution of stations across Nepal and employing the maximum common length of the
length of the high-quality continuous data available from the maximum number of stations, we have
high-qualitycontinuousdataavailablefromthemaximumnumberofstations,wehaverestrictedthe
restricted the period of our daily extreme analysis to 1970–2012. Within such a period, the stations
periodofourdailyextremeanalysisto1970–2012. Withinsuchaperiod,thestationsthatfeaturedata
that feature data gaps for more than (1) a fortnight within a year; (2) four consecutive years or (3) for
gapsformorethan(1)afortnightwithinayear;(2)fourconsecutiveyearsor(3)forsixyearsintotal
six years in total were excluded from the analysis.
wereexcludedfromtheanalysis.
450
Precipitation Stations History
400
350
300
s
n
o 250
i
t
a
t 200
s
f
o 150
r
b
m 100
u
N 50
Stations with data
0
1940 1950 1960 1970 1980 1989 1999 2009
Year

Figure2. ThenumberofprecipitationgaugesandtheirageintheDepartmentofHydrologyand
Figure 2. The number of precipitation gauges and their age in the Department of Hydrology and
Meteorology(DHM)database.
Meteorology (DHM) database.
TThhee qquuaalliittyy ccoonnttrrooll ooff tthhee ddaattaa ffrroomm tthhee ccoonnssiiddeerreedd ssttaattiioonnss wwaass ppeerrffoorrmmeedd uussiinngg tthhee RRCClliimmDDeexx
ttoooollkkiitt [[5599]],, wwhhiicchh ccaann iiddeennttiiffyy ppootteennttiiaall oouuttlliieerrss aanndd nneeggaattiivvee pprreecciippiittaattiioonn vvaalluueess [[44]].. AAfftteerr tthhee
qquuaalliittyy ccoonnttrrooll,, tteessttiinngg tthhee hhoommooggeenneeiittyy iiss tthhee mmoosstt iimmppoorrttaanntt sstteepp [[6600]] aass iitt iiddeennttiiffiieess tthhee vvaarriiaattiioonnss
tthhaatt ooccccuurrrreedd dduuee toto ppuurerleyly nonno-nc-lcimlimataicti cfafcatoctrosr, ss,uscuhc ahs afsauflatus litns tihnet hinestirnusmtruenmtse notrs cohranchgeasn gine sthine
mtheeamsueraesmuerenmt emnettmhoedth, oadg,gargeggraetgioanti omnemtheotdh,o dst,asttiaotnio nlolcoactiaotino, ns,tsattaiotino neexxppoosusurere aanndd oobbsseerrvvaattiioonnaall
pprraaccttiiccee [[6611]].. TThhee hhomomogoegneenietyit ytetsets wt wasa pseprfeorrfomremde fdorf oeraceha cshtastitoanti obny bmyomntohnlyth tilmy eti mseerieses ruiessinugs tihneg
RthHetResHt tteosotltkoito,l kwith,iwchh iccahnc asntatsitsattiicsatlilcya lildyeindteinfyt iftyheth meumltuipltliep lestestpe pchcahnagnegse sbbyy uussiningg aa ttwwoo--pphhaassee
rreeggrreessssiioonn mmooddeell wwiitthh aa lliinneeaarr ttrreenndd ooff tthhee eennttiirree ttiimmee sseerriieess [[6622]].. SSiinnccee tthhee ssttaattiioonnss oobbsseerrvvee llaarrggee
EEuucclliiddeeaann ddiissttaanncceess iinnc coommpplelexxm mouountnatianionuosutse trerarrinaisn,sw, ewhea vheavpee rpfoerrfmoermdeadr eala rteivlaetihvoem hoogmenoegietyneteitsyt,
twesitth, owuitthuosuintg uasirnegfe rae rnecfeertiemncees etirmiees [s6e3r,i6e4s] .[6T3h,6e4i]n. hTohme oingehnoemitoygoefnaesittya toiofn ah satsatbioeenn hdaesc bideeedn bdaesceiddeodn
bthaeseRdH otnes tthree sRuHltst,egstr arpehsuiclatsl, egxraampihniacatilo enxaanmdincoatiinocnid aenndce coofinkcniodwennceE NofS Oknoorwlonc aElNizSeOd porre cloipciatlaitzieodn
pevreecnitpsi.taTthioens teavtieonntss.f eTahtuer isntagtiaonnys infehaotmuroinggen aenityy winehroemexocgleundeeidtyf rwomereth eexacnluadlyesdis .frSoumch thster icatnsatlaytsioisn.
Ssuelcehc tsiotrnicct rsitteartiioanh asevleecytiieolnd ecdrittehreiac ohnatvien uyoieulsd,ehdo tmheo gceonnetoinuuso,uhsig, hh-oqmuoalgietynedoauilsy, ohbigshe-rqvuaatiloitnys dfraoilmy
oobnsleyr7v6atsiotantsio fnrosmfo ornthlye 7169 7st0a–t2io0n1s2 fpoerr tihoed .19T7h0e–s2e07162 psteartiioodn.s Thhaevsee b76e esntautisoends fhoarvteh beeceonm upsuedta ftoiorn thoef
cdoamilypuptraetciiopni toaft idoanileyx tprreemciepsit(aFtiigounr eex1tareamndesT (aFbilgeu1r)e a1nad asneda sToanbalel p1r)e acnipdi tsaetaiosonntaolt aplr.ecipitation total.
FFoorr ssuubb--rreeggiioonnss ccoonnssiiddeerreedd ffoorr ffiieelldd ssiiggnniiffiiccaannccee ssttuuddyy,, tthheessee ssttaattiioonnss eennssuurree aann aaddeeqquuaattee ssppaattiiaall
ddiissttrriibbuuttiioonn ooff ssttaattiioonnss aat tleleaasst taacrcorosss sththe emmididddlel emmouonutnatiani nanadn dlolwolwanladn sdusbu-rbe-greiogniosn. Ts.hTe hneunmubmerb oefr
sotfatsitoantiso tnhsatth faatllf ianll eianceha scuhbs-urebg-iroeng ioofn WofMW, WML,W, CLH,,C CHM,,C CML,, CELM, EanMd aEnLd aEreL 1a4r,e 41, 43,, 41,63, ,1116, ,1191 a,n1d9
9a,n rdes9p,ercetsipveelcyti.v Seinlyc.eS tihnec esttahteiosntas twiointhsiwn tithhei nwtehsetewrne satnedrn ceanntdracle rnetgriaolnrse garioen rselaarteivreellya tyiovuelnygyeor,u tnhgereer,
are more stations within the eastern region that fulfill the selection criteria. As for high mountain sub-
regions, long-term data was available only from CH, due to the low density of the high-altitude
station network. Hence, we limit our sub-regional analysis to seven sub-regions only.
Climate2017,5,4 6of25
therearemorestationswithintheeasternregionthatfulfilltheselectioncriteria. Asforhighmountain
sub-regions,long-termdatawasavailableonlyfromCH,duetothelowdensityofthehigh-altitude
stationnetwork. Hence,welimitoursub-regionalanalysistosevensub-regionsonly.
Table1.Listofmeteorologicalstations.
Region ID Name Lat(◦) Lon(◦) Height(m)
106 Belaurisantipur 28.683 80.35 159
209 Dhangadhi(atariya) 28.8 80.55 187
WL
416 NepalgunjReg.off. 28.052 81.523 144
510 Koilabas 27.7 82.533 320
101 Kakerpakha 29.65 80.5 842
103 Patan(west) 29.467 80.533 1266
104 Dadeldhura 29.3 80.583 1848
201 Pipalkot 29.617 80.867 1456
202 Chainpur(west) 29.55 81.217 1304
203 Silgadhidoti 29.267 80.983 1360
206 Asaraghat 28.953 81.442 650
WM
303 Jumla 29.275 82.18 2366
308 Nagma 29.2 81.9 1905
402 Dailekh 28.85 81.717 1402
404 Jajarkot 28.7 82.2 1231
406 Surkhet 28.587 81.635 720
504 Libanggaun 28.3 82.633 1270
511 Salyanbazar 28.383 82.167 1457
703 Butwal 27.694 83.466 205
902 Rampur 27.654 84.351 169
903 Jhawani 27.591 84.522 177
907 Amlekhganj 27.281 84.992 310
909 Simaraairport 27.164 84.98 137
CL 910 Nijgadh 27.183 85.167 244
911 Parwanipur 27.079 84.933 115
912 Ramolibairiya 27.017 85.383 152
1109 Pattharkot(east) 27.1 85.66 162
1110 Tulsi 27.013 85.921 251
1111 Janakpurairport 26.711 85.924 78
701 Ridibazar 27.95 83.433 442
722 Musikot 28.167 83.267 1280
802 Khudibazar 28.283 84.367 823
804 Pokharaairport 28.2 83.979 827
807 Kunchha 28.133 84.35 855
808 Bandipur 27.942 84.406 995
810 Chapkot 27.883 83.817 460
814 Lumle 28.297 83.818 1740
CM
904 Chisapanigadhi 27.56 85.139 1729
1008 Nawalpur 27.813 85.625 1457
1015 Thankot 27.688 85.221 1457
1022 Godavari 27.593 85.379 1527
1023 Dolalghat 27.639 85.705 659
1029 Khumaltar 27.652 85.326 1334
1030 Kathmanduairport 27.704 85.373 1337
1115 Nepalthok 27.42 85.849 698
601 Jomsom 28.78 83.72 2744
CH 604 Thakmarpha 28.739 83.681 2655
607 Lete 28.633 83.609 2490
Climate2017,5,4 7of25
Table1.Cont.
Region ID Name Lat(◦) Lon(◦) Height(m)
1112 Chisapanibazar 26.93 86.145 107
1213 Udayapurgadhi 26.933 86.517 1175
1216 Siraha 26.656 86.212 102
1311 Dharanbazar 26.792 87.285 310
EL 1316 Chatara 26.82 87.159 105
1319 Biratnagarairport 26.481 87.264 72
1320 Tarahara 26.699 87.279 121
1408 Damak 26.671 87.703 119
1409 Anarmanibirta 26.625 87.989 122
1102 Charikot 27.667 86.05 1940
1103 Jiri 27.633 86.233 2003
1202 Chaurikhark 27.7 86.717 2619
1203 Pakarnas 27.443 86.569 1944
1204 Aisealukhark 27.36 86.749 2063
1206 Okhaldhunga 27.308 86.504 1731
1207 Manebhanjyang 27.215 86.444 1528
1210 Kuruleghat 27.136 86.43 341
1211 Khotangbazar 27.029 86.843 1305
EM 1303 Chainpur(east) 27.292 87.317 1262
1305 Leguwaghat 27.154 87.289 444
1306 Munga 27.05 87.244 1457
1307 Dhankuta 26.983 87.346 1192
1308 Mulghat 26.932 87.32 286
1309 Tribeni 26.914 87.16 146
1322 Machuwaghat 26.938 87.155 168
1325 Dingla 27.353 87.146 1169
1403 Lungthung 27.55 87.783 1780
1410 Himaligaun 26.887 88.027 1654
Note: WL(westernlowlands),WM(westernmiddlemountainsandhills),WH(westernhighmountains),
CL(centrallowlands),CM(centralmiddlemountainsandhills),CH(centralhighmountains,EL(eastern
lowlands),EM(easternmiddlemountainsandhills)andEH(easternhighmountains)).
Inadditiontoadailyextremeprecipitationtrendanalysis,wepresentspatialvariabilitymaps
ofmeanseasonalandofphysicallyrelevantdailyprecipitationextremeindicesfromthemaximum
numberofstationsthathavethedataforatleast20yearsintherecentnormalperiod(1981–2010)but
twostationsfromhighaltitude(>3500m)withonlyfiveyearsofdataarealsoincludedtorepresent
the high-altitude spatial pattern. Since these indices do not require the high-quality data, around
210precipitationstationswereconsideredforthisanalysis.
4. Methodology
4.1. PrecipitationIndices
Wehaveconsideredtheextremeprecipitationindicesthataredevelopedandrecommendedby
theExpertTeamonClimateChangeDetection,MonitoringandIndices(ETCCDMI),jointlyestablished
bytheWorldMeteorologicalOrganization(WMO)CommissionforClimatologyandtheResearch
ProgrammeonClimateVariabilityandPredictability(CLIVAR)(Table2). Basedonthecalculation
method,theseindicesfallintofourgroups[65,66],namely,absoluteindices,thresholdindices,duration
indicesandpercentile-basedthresholdindices.
Climate2017,5,4 8of25
Table 2. The description of ETCCDMI (Expert Team on Climate Change Detection, Monitoring
andIndices).
Category ID NameofIndex Definition Unit
HIP R95 Verywetdays Annualtotalprecipitationofdaysin>95thpercentile mm
HIP RX1day Max1-dayprecipitationamount Annualmaximum1-dayprecipitation mm
HIP RX5day Max5-dayprecipitationamount Annualmaximumconsecutive5-dayprecipitation mm
HIP R99 Extremelywetdays Annualtotalprecipitationofdaysin>99thpercentile mm
Numberofheavy
FP R10 Annualcountofdayswhenprecipitationis≥10mm Days
precipitationdays
Numberofveryheavy
FP R20 Annualcountofdayswhenprecipitationis≥20mm Days
precipitationdays
Maximumnumberofconsecutivedrydays
DWS CDD Consecutivedrydays Days
(precipitation<1mm)
Maximumnumberofconsecutivewetdays
DWS CWD Consecutivewetdays Days
(precipitation≥1mm)
EX WD Annualwet/rainydays Annualcountofdayswhenprecipitationis≥1mm Days
EX PRCPTOT Annualtotalwet-dayprecipitation Annualtotalfromdays≥1mmprecipitation mm
EX SDII Simpledailyintensityindex RatioofannualtotaltoWDinayear mm/day
Note:HIP:High-intensity-relatedprecipitationextreme,FP:Frequency-relatedprecipitationextreme,DWS:Dry
andwetspellorduration,EX:Extra.
Theabsoluteindicesincludetheannualmaximumof1-dayand5-dayprecipitationreferredtoas
RX1dayandRX5day,respectively. Thresholdindicescompriseofnumbersof(1)heavyprecipitation
days(R10)and(2)veryheavyprecipitationdays(R20). R10(R20)denotesthecountsofdaysinayear
whenprecipitationwas≥10mm(≥20mm). Thedurationindicesencompasstheconsecutivewetand
drydays(CWDandCDD),whichdescribethemaximumlengthsofconsecutivewetanddrydays,
respectively. Unliketheabsolutethreshold-basedindices,percentile-basedindicesbetterrepresent
the spatial aspects of precipitation extremes, accounting for spatial heterogeneity of precipitation.
However, different researchers have used different percentiles as thresholds for defining such
extremes. Forexample,Krishnamurthyetal.[18]andBookhagen[67]haveusedthe90thpercentile,
Duncanetal. [29] and Casanueva et al. [68] have used the 95th percentile and Alexanderetal. [4],
KleinTanketal.[5],Donatetal.[14],andSheikhetal.[11]haveusedthe95thand99thpercentiles.
Here,thepercentile-basedthresholdindicesincludetwo;theyearlytotalprecipitationfromverywet
days(R95)andextremelywetdays(R99)observingabovethan95thand99thprecipitationpercentiles,
respectively. SinceR95andR99indicesarebasedonthelong-termpercentilethresholds,theymay
differfromstationtostationbutnotonaninter-annualscale. Threeadditionalindicesthatdonot
fallinanyoftheabovecategoriesaretheannualwetdays(WD)describedastheannualcountof
dayswhenprecipitationwas≥1mm,annualtotalprecipitationfromWD(PRCPTOT)andthesimple
dailyintensityindex(SDII)definedastheratioofPRCPTOTtoWDwithinayear. Theseindiceswere
computedusingtheRClimDextoolkit[59]whiletheregionalizationofthephysicallyrelevantindices
hasbeenperformedusingGlobalKriginginterpolation(See[38]fordetails).
Theconsideredextremeprecipitationindicesaremainlytheindicatorsoffloodsanddroughts
orindirectindicatorswhichincombinationwiththemainindicatorsprovidevaluableinformation.
For instance, R95, R99, RX1day and RX5day indices characterize the magnitude of very intense
precipitation events that trigger flash floods and landslides. R10 and R20 assess the frequency of
heavyandveryheavyprecipitationevents. Ontheotherhand,CDDandCWDindirectlyprovidethe
indicationofdroughts,whichisquiteimportantfortheagriculturalactivities[68]. TheSDII,PRCPTOT
andWDindirectlyprovideusefulinformationwhencombinedwithotherextremeindices. Therefore,
wehavepresentedanddiscussedtheresultsinfourbroadcategories,namely,high-intensity-related
precipitationextreme(HIP)indiceswhichconsistofabsoluteandpercentile-basedthresholdindices,
frequency-relatedprecipitationextremesorthreshold-based(FP)indices,dryandwetspellorduration
(DWS)indicesandextraindices(EX)(Table2).
Climate2017,5,4 9of25
Inadditiontothecomputationofextremeprecipitationindices,thetemporalchangesinthese
indicesandseasonaltotalprecipitationhavealsobeenassessed. Forthis,trendsintheconsidered
indices (except R99) were analyzed using the Mann−Kendall (MK) trend test [33,34] while the
magnitudeoftrendwasestimatedusingtheTheil−Sen’s(TS)slopemethod[69,70]. Trendassessment
for R99 was possible, but no trends were found for more than half of the stations due to large
inter-annualvariabilitythatresultedinzerovaluesformorethanhalfoftheR99timeseries. Theywere
therefore,excludedfromtheanalysis.
4.2. TrendAnalysis
4.2.1. Mann−KendallTrend
TheMKtrendtest[33,34]hasbeenwidelyusedtoassessthesignificanceofatrendinthetime
seriesasthetestdoesnotrequirenormallydistributeddatasets[71,72]andcancopewithmissing
datarecordsandextremes.
4.2.2. Theil−Sen’sSlope
Ifalineartrendispresentinatimeseries,thetrueslopeoftheexistingtrendcanbecomputed
usingthenon-parametricTSapproach. Thistestiswidelyusedandrobustasitislesssensitiveto
outliersandmissingvaluesindata[69,70].
4.2.3. Trend-FreePre-Whitening
TheMKtestisbasedontheassumptionthatthetimeseriesisseriallyindependent. However,
oftenthehydro-meteorologicaltimeseriescontainaserialcorrelation[73–75],affectingtheMKtest
results. Forinstance,existenceofapositive(negative)serialcorrelationinatimeseriesoverestimates
(underestimates)thesignificanceoftheMKtest(e.g.,[74,76]).Tolimitsucheffectofserialcorrelationon
theMKtestresults,severalpre-whitening(PW)procedureshavebeenproposed[71,75–77]including
trend-free pre-whitening (TFPW). TFPW more effectively reduces the effect of a serial correlation
presentwithinthehydro-meteorologicaltimeseriesontheMKtestresults[78–80]. Here,wehave
usedTFPWasproposedbyYueetal.[74].
InTFPW,theinitialstepistoestimatethetrueslopeofatrendusingSen’sslopemethod,unitize
thetimeseriesbydividingeachsamplewiththesamplemeanandde-trendthetimeseries. Thelag-1
auto-correlationisthenestimatedandremovedifexistingandthetimeseriesissubsequentlyblended
backtothepre-identifiedtrendcomponent. Finally,theMKtestisappliedtopre-whitenedtimeseries
(see[74,75]forfurtherinformation).
4.2.4. FieldSignificance
Acertainregioncanfeatureanumberofstationswithpositiveornegativetrends. Thus,thefield
significancetestisusedtoidentifyregionswithtrendsofaconsistentsign,independentofstatistical
significanceoftheindividualstationtrends[81,82].
Variousmethodsareavailableforassessingthefieldsignificanceoflocaltrends[71,73,74,83–86].
WehaveadoptedthemethodofYueetal.[74]toassessthefieldsignificanceinthesevensub-regions
(Figure 1b) with sufficient data. In this method, the original station network has been resampled
1000timeswiththebootstrappingapproach[87],distorting(preserving)theauto-(cross-)correlationto
avoiditsinfluenceonthefieldsignificanceanalysis.TheMKtestisthenappliedtosynthetictimeseries
ofeachsite. Atthegivensignificancelevel,thenumbersofsiteswithsignificantupwardtrendsand
downwardtrends,respectively,havebeencountedusingEquation(1)foreachresamplednetwork.
n
∑
C = C (1)
f i
i=1
Climate 2017, 5, 4 10 of 25
(cid:3041)
Climate2017,5,4 10of25
(cid:1829) =(cid:3533)(cid:1829) (1)
(cid:3033) (cid:3036)
(cid:3036)(cid:2880)(cid:2869)
where, n, indicates the number of stations within a region of analysis and C refers to a count of
i
wstahteisreti,c anl,l yinsdigicnaifitecsa nthtetr ennudmsb(aetr 0o.1f lsetvaetilo)nats twheitshtainti oan r,ei.gTiohnis opfr oacneadlyusries haansdb eCein rreefperesa tteod a1 0c0o0utnimt oesf
sfotartiesatcicharlleys asmigpnliefidcannett wtroernkd.sR (aant k0i.n1g lethveelc) oartr ethspeo sntadtiinogn,1 0i.0 T0hviasl uperoscoefdcuoruen htsaso fbseiegnn irfiecpaenattpedo s1it0i0v0e
t(inmegesa tfiovre )eatrcehn dressainmapnleads cneentdwionrgko. rRdaenrkuinsign tghteh ceoWrreeisbpuolnld[8in8g] p1l0o0t0ti nvgalpuoess iotifo ncofuonrmts uolfa syigienlidfiscathnet
pemospitiirvicea (lnceugmatuivlaet)i vterednidstsr iibnu atino nass,cCen,daisn:g order using the Weibull [88] plotting position formula
f
yields the empirical cumulative distributions, Cf, as:
P(C ≤C(cid:48)r )=r/(N+1) (2)
f f
P (Cf ≤ C′rf) = r/(N + 1) (2)
where, ,istherankofCandNisthenumberofresamplednetworks. Theprobabilityofanumberof
where, rr, is the rank of C and N is the number of resampled networks. The probability of a number of
significantpositive(negative)trendsintheoriginalnetworkhasbeenestimatedbycomparingwith
significant positive (negative) trends in the original network has been estimated by comparing with
Cf ofsignificantpositive(negative)trendsobtainedfromtheresamplednetworks(Equation(3)).
Cf of significant positive (negative) trends obtained from the resampled networks (Equation (3)).
(cid:40)
(cid:1842) P, fo,rf(cid:1842)orP ≤0≤.50.5
PobsP=obsP = (PC (fC,of,b osbs≤ ≤ CC(cid:48)′rrff) )wwhheerere (cid:1842)P(cid:3033)f==(cid:3420)1(cid:3042)−(cid:3029)1(cid:3046)(cid:1842)o−bsP, (cid:1842)(cid:3042),(cid:3029)(cid:3046)Pob>s 0>.50 .5 (3()3 )
(cid:3042)(cid:3029)(cid:3046)obs(cid:3042)(cid:3029)(cid:3046) obs
AAtt tthhee ssiiggnniiffiiccaannccee lleevveell ooff 00..11,, iiff PPff ≤≤ 0.01.,1 t,hthenen ththe etrternendd oovveer raa rereggioionn wwaass ccoonnssidideerreedd ssigignnififiiccaanntt. .
AA ssiimmiillaarr aapppprrooaacchh hhaass bbeeeenn eemmppllooyyeedd bbyy PPeettrrooww aanndd MMeerrzz [[8899]] ffoorr aasssseessssiinngg tthhee ffiieelldd ssiiggnniiffiiccaannccee
ooff fflloooodd ttiimmee sseerriieess iinn GGeerrmmaannyy aanndd bbyy HHaassssoonn eett aall.. [[6633]] ffoorr hhyyddrroo--mmeetteeoorroollooggiiccaall ttiimmee sseerriieess iinn tthhee
HHiinndduukkuusshh--KKaarraakkoorraamm--HHiimmaallaayyaann rreeggiioonn ooff tthhee uuppppeerr IInndduuss bbaassiinn..
55.. RReessuullttss aanndd DDiissccuussssiioonn
55..11.. SSppaattiiaall DDiissttrriibbuuttiioonn ooff MMeeaann SSeeaassoonnaall aanndd DDaaiillyy PPrreecciippiittaattiioonn IInnddiicceess
IInn aaddddiittiioonn toto thteh esesaesaosnoanl amlemane apnrepcripecitiaptiitoanti odnistdriibsturtiibount,i ospna,tsiapla ptiaatlteprnatst eorfn msaoinflmy athinel hyigthhe-
ihnitgehn-siintyte-n asnitdy -fraenqdufernecqyu-erenlcayte-rde leaxtetrdemexetsr e(mR9e5s/(RR9995 p/eRr9c9enpteilrec epnrteicleippitraetciiopnit, aRtiXo1nd,aRyX a1ndday WanDd, RW1D0,,
RR2100,, Rre2s0p,ercetsivpeelcyt)i,v ewlyh)i,cwh hairceh raerleevrealnetv afnotr fworatwera terersroeusrocuersc, esa,s aws welle llasa sflfloooodd aanndd aaggrriiccuullttuurree
mmaannaaggeemmeenntt,, aarree ccoommppuutteedd ffrroomm 221100 ssttaattiioonnss oovveerr tthhee ppeerriioodd ooff 11998811––22001100 ((FFiigguurreess 33 aanndd 44))..

Figure 3. Spatial distribution of mean seasonal precipitation (mm) for (a) Pre monsoon; (b) Monsoon;
Figure3.Spatialdistributionofmeanseasonalprecipitation(mm)for(a)Premonsoon;(b)Monsoon;
(c) Post monsoon; and (d) winter season over the period of 1981–2010. Note: Legend scale of all four
(c) Post monsoon; and (d) winter season over the period of 1981–2010. Note: Legend scale of all
seasonal maps are different. .
fourseasonalmapsaredifferent.