Bias correction of daily precipitation measurements

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 104, NO. D6, PAGES , MARCH 27, 1999 Bias crrectin f daily precipitatin measurements fr Greenland Daqing Yang Schl f Gegraphy and Gelgy, McMaster University, Hamiltn, Ontari, Canada Shig Ishida and Barry E. Gdisn Climate Research Branch, Atmspheric Envirnment Service, Dwnsview, Ontari, Canada Thil Gunther German Weather Service, Berlin, Germany Abstract. A methdlgy fr crrecting the Hellmann gauge-measured daily precipitatin fr wind-induced undercatch, wetting lss, and trace amunt f precipitatin was presented and applied at 12 climate statins in Greenland fr years The results shw the fllwing: (1) daily crrectins f the biases increase the gauge-measured annual precipitatin by mm fr the 4 years (abut 25-90% f the gauge-measured yearly ttal) in Greenland; (2) f the biases, wind-induced undercatch is the surce f greatest errr; wetting lss and trace amunt f precipitatin are als significant biases in the nrthern Greenland regins f lw precipitatin; (3) mnthly crrectin factrs (crrected/measured precipitatin) differ by statin and, at an individual statin, by type f precipitatin; (4) cnsiderable intra-annual variatin f the yearly crrectins has been fund in Greenland due t the fluctuatin f wind speed, air temperature, and frequency f snwfall. These results suggesthat annual precipitatin in Greenland is much higher than previusly reprted, particularly in the suthern regins f high precipitatin; and the latitudinal precipitatin gradient may als be greater ver Greenland. These results will have a significant impact t water budget and glacier mass balance studies in Greenland. 1. Intrductin Paper number 1998JD /99/1998JD precipitatin measurements; and (3) intrduce a reference methd f slid precipitatin measurement fr general use t calibrate any type f precipitatin gauge [Gdisn et al., 1989]. The reference methd fr the true snwfall measurement It has lng been realized that systematic errrs such as windinduced undercatch, wetting, and evapratin lsses in precipitatin measurement affect all types f precipitatin gauges [Gdisn et al., 1981; Sevruk, 1982; Tablet et al., 1990]. The was critical fr this intercmparisn. After reviewing all psneed t crrect these biases especially fr slid precipitatin sible practical methds (bush shield, duble fence shield, frest measurement has nw been mre widely acknwledged, as the clearing, snw. bard, dual gauge system), the WMO Organizmagnitude f the errrs and their variatin amng gauges ing Cmmittee fr the intercmparisn designated the ctagbecame knwn and their ptential effects n reginal, natinal nal vertical duble fence, surrunding a shielded Tretyakv and glbal climatlgical, hydrlgical, and climate change gauge, as the intercmparisn reference (DFIR) [Gdisn et studies were recgnized [Gdisn and Yang, 1995; Yang and al., 1981; Glubev, 1985a, b; Gdisn et al., 1989]. The DFIR Gdisn, 1998; Walsh et al., 1998; Desbls and Desalmand, was perated at 19 statins in 10 cuntries arund the wrld 1995; Legates, 1995; Legates and Willmrt, 1990; Hanssen-Bauer during the study. and Frland, 1994; Grisrnan and Easterling, 1994; Grisrnan The Hellmann gauge is ne f the mst widely used precipand Legates, 1994; Karl et al., 1993; Grisman et al., 1991]. itatin gauges at abut 30,080 lcatins arund the wrld. It is T assess natinal methds f measuring slid precipitatin, the standard gauge in 30 cuntries [Sevruk and Klemrn, 1989]. including past and current prcedures, autmatic systems, and The Hellmann gauge is a nnrecrding gauge fr bth rain and new methds f bservatin, the Wrld Meterlgical Orga- snw measurements. It is 43 cm high, with an rifice area f nizatin (WMO) initiated the Slid Precipitatin Measure- 200 cm 2. There are varius versins f the Hellmann gauges. ment Intercmparisn Prject in 1985 [WMO/CIMO, 1985; The designs f the Hellmann gauges are very similar [Sevruk Gdisn et al., 1989]. The intercmparisn was designed t and Kleinre, 1989]: they all have the same rifice area. A metal crss is installed in the Danish Hellmann gauge t prevent (1) determine wind-induced errrs in natinal methds f snw t be blwn ut f the gauge. In different cuntries the measuring slid precipitatin, including wetting and evapra- Hellmann gauges are placed at the height f m abve tin lsses; (2) derive standard methds fr crrecting slid the grund, and in sme cuntries, they are equipped with a Cpyright 1999 by the American Gephysical Unin. wind shield (such as Nipher type shield and Tretyakv shield) 6171 [Sevruk and Klemrn, 1989]. Many experimental studies n the Hellmann gauges have been cnducted [Sevruk, 1982; Gunther,

2 6172 YANG ET AL.' PRECIPITATION CORRECTION FOR GREENLAND c::: 60 ¾ 40._ ß. 20 c) D E ] I I I I I I I Wind speed at gauge height (m/s) Figure 1. Relatin f daily catch rati f snw fr the Hellmann gauge versus daily wind speed at gauge height, Harzgerde, Germany. 1993; Milkvic, 1989; /ltlerup et at., 1997]. The Hellmann gauges were tested in the Internatinal Cmparisn f Natinal Precipitatin Gauge with a reference pit gauge [Sevruk and Hamn, 1984]. The crrectin methd fr wind-induced undercatch f rainfall measurement has been develped [Sevruk, 1981; Sevruk and Hamn, 1984] and applied t the archived precipitatin data in Switzerland [Sevruk, 1993]. The WMO Slid Precipitatin Measurement Intercmparisn prvided the pprtunity t develp and evaluate the imprved crrectin prcedures n a daily r 6 hurly timescale fr a number f precipitatin gauges cmmnly used arund the wrld [Gdisn et at., 1998; Gdisn and Yang, 1995; Yang et at., 1995; Yang et al., 1998a]. This paper presents a summary f the methdlgy f crrecting Hellmann gaugemeasured daily precipitatin fr biases f wind-induced undercatch, wetting lss, and trace amunt f precipitatin. Applicatin f the prpsed methdlgy was made at 12 climate statins in Greenland fr fur example years ( ). As a result, the magnitude f the biases was quantified and their seasnal and spatial variability was demnstrated. Discussins land [Aaltnen et at., 1993]. The wetting lsses fr bth the Tretyakv gauge and the Hellmann gauge were crrected when the vlumetric methd was used fr the gauge readings [Aaltnen et al., 1993]. The evapratin lsses were nt crrected since experiments shwed that they were very small [Aaltnen et al., 1993; Glubev et al., 1992]. When wind speed at the gauge height was nt measured, it was reduced frm a higher height. The DFIR data were crrected fr windinduced lsses using the prcedure derived by Yang et al. [1993]. Blwing snw events in the intercmparisn data were carefully identified and eliminated frm further analysis. The intercmparisn data cllected at the fur sites fr mre than three winter seasns were cmpiled; they represented a great variety f climate, terrain, and expsure. Relatin f catch rati as a functin f wind speed and air temperature has been develped fr the Hellmann gauges [Yang et at., 1994; Gunther, 1993; Yang et at., 1999]. The Hellmann gauges used in Greenland were equipped with a mdified Nipher wind shield. This gauge and shield cnfiguratin has been tested at Harzgerde in Germany durf the crrectin methds and results were made by cmparing the WMO intercmparisn [Gunther, 1993]. Data analysis isn f this wrk t ther studies. It is expected that the bias indicates that wind speed is the nly factr with the statistically crrectin, such as demnstrated by this study, will significantly significant relatinship t the gauge catch, and air temperature imprve the accuracy and hmgeneity f precipitatin data des nt have a statistically significant effect n the gauge and will have an impact n climate mnitring and hydrlgical mdeling. The results f this study will als be useful fr catch, when precipitatin is classified int snw, rain, and validatin f the GCM/RCM simulatin in the high latitudes mixed. It is knwn that an unrealistically large variatin f the [Watsh et at., 1998; Kattsv et at., 1998]. catch rati will be intrduced by the small abslute differences between gauges fr small precipitatin events. T minimize this effect, nly thse intercmparisn data when the DFIR 2. Crrectin Methd measurement was greater than 3.0 mm were used fr the sta- During the WMO Slid Precipitatin Measurement Intercmparisn the Hellmann gauges were tested against the DFIR reference at fur statins in Finland, Russia, Germany, and Cratia. Extensive experiments n the wetting and evapratin lsses f the Hellmann gauges and many ther natinal gauges have been cnducted at the Jkiinen Statin in Fintistical analysis. Figure 1 shws, as an example, the daily catch rati f snw fr the Hellmann gauge versus daily wind speed at the gauge height. A wide range f wind speed and catch rati (CR) has been sampled ver several winter seasns at the experimental site; this may ensure that the derived crrectin methd is

3 . YANG ET AL.: PRECIPITATION CORRECTION FOR GREENLAND 6173 mre likely t be applicable t a wide range f climate cnditins. The scatterplts f the CR wind fr rain and mixed precipitatin generally have similar patterns as fr snw. The scatter f the CR is smaller fr mixed precipitatin and rain events, and their catch ratis are higher (in cmparisn t snw events), particularly at high wind speeds, indicating less effect f wind n gauge catch f rain and mixed precipitatin. Regressin analyses derived the best fit relatin f daily catch rati (CR, %) as a functin f the daily wind speed (Ws, m/s) at the gauge height fr varius precipitatin types. Snw CR = Ws Ws:, (0 <- WS <_ 6.5 m/s), (n = 43, R:: 0.50), (1) Mixed precipitatin Rain CR = Ws Ws:, (0 <_ WS _< 6.5 m/s), (n = Thule (.AFB)! (¾ O6mrkshvn "'" ' '72N:. - ",' "',10W,,,'gedesm nde ".-'"" '"., HOls :e. lndb.. r9 '.,., ', '. " Frede ikshab ",...' "'.. ' k NCrssarssua ',.....,'%U½ h _;._.... ". ''... P(i. s CXri db' ' CR = Ws Ws:, (0 -< WS -< 6.5 m/s), 48, Statistical tests shw that these relatins are significant at 90-95% cnfidence level. The perfrmance f the crrectin equatins (1)-(3) was further checked independently by using all the intercmparisn data (withut the DFIR > 3.0 mm limitatin) at the Harzgerde WMO intercmparisn statin. The results demnstrated that the differences between the ttals f crrected precipitatin and the DFIR measurements were 6% fr snw and less than 3% fr bth rain and mixed precipitatin. Therefre these crrectin equatins wrk generally well at the experimental statin and they shuld be used fr test crrectin f the Hellmann gauge-measured precipitatin data in reginal statin netwrks. 3. Applicatin f the Methd Twelve climate statins in Greenland were chsen fr this study. These statins spread alng the castalines and represent varius climatic cnditins (Figure 2). The site and instrumental infrmatin (gauge height, shielding infrmatin, anemmeter height) is given in Table 1. The daily data f temperature, wind speed, precipitatin, snwfall, and snw depth n the grund fr the perid f were btained frm the U.S. Natinal Climatic Data Center. The mean yearly values f temperature, gauge-measured precipitatin, and wind speed at the 12 sites were summarized fr the 4 years in Table 2. Crrectin f gauge-measured precipitatin shuld be made fr trace precipitatin, wetting lsses, evapratin lss, and wind-induced errrs caused by the wind field defrmatin ver gauge rifice [Sevruk and Hamn, 1984]. Since the wind field defrmatin affects the ttal gauge catch, including bth the wetting and the evapratin lsses, we mdified the general mdel [Sevruk and Hamn, 1984] fr precipitatin crrectin t + + X?p + (4) Where Pt is the crrected precipitatin, P q is the gaugemeasured precipitatin, AP w and Ap, are wetting lss and [i.r. S l ct d cli t st ti.s i. Or.la.d fr this study. evapratin lss, respectively, AP t is the trace precipitatin, and K is the crrectin cefficient (usually K > 1) fr windinduced errrs. The methd f determining each f the terms in (4) is given belw Trace Precipitatin Fr the Hellmann gauge, a measurement f precipitatin f less than 0.10 mm is beynd the reslutin f the gauge measurement, and thus it is recrded as a trace precipitatin event. Officially, all f the trace precipitatin is treated quantitatively as a zer event which cntributes nthing t the mnthly ttals. Hwever, the day during which trace precipitatin was recrded is cunted as a precipitatin day. Climate data cllected during in Greenland indicate that the annual gauge-measured precipitatin varied greatly frm mm in the nrthern regins t mm in the suthern cast. The percentage f snw in annual precipitatin decreased frm 70-80% in the nrth t 40-50% in the suth cast, with the temperature increasing frm nrth t suth. The ttal number f precipitatin days (e.g., the sum f the days with trace precipitatin and measurable precipitatin) ranged frm 120 t 230 at the 12 statins (Table 2). The average numbers f trace precipitatin days varied frm 60 t 125 fr the 4 years. Gegraphically, mre trace precipitatin days were recrded in the nrthern Greenland regins and trace precipitatin recrdings there can make up 50 t 70% f the annual ttal f precipitatin days. Fr example, 88 f the 136 precipitatin days were trace days at Danmarkshavn and the values were 125 ut f 203 at Thule. Seasnally, mre trace precipitatin days were reprted in summer than in winter in nrthern Greenland, but the percent f trace precipitatin days t the ttal number f days f precipitatin is much higher in the relatively dry winter seasn than in the wet summer seasn. It has been reprted that in nrthern Alaska and the Nrth Ple regins, abut 80% f the winter snwfall recrd cnsisted f trace [Yang et al., 1998b; Bensn, 1982; Clny et al.,

4 6174 YANG ET AL.: PRECIPITATION CORRECTION FOR GREENLAND Table 1. Siting and Instrumental Infrmatin at 12 Climate Statins in Greenland Statin WMO Crdinates Statin Elevatin, Wind Sensr Identificatin N W m Height, m Gauge Height, m Wind Shield Danmarkshavn Thule (AFB) Scresbysund Egedesminde Sndre strmfjrd Hlsteindb rg Angmagssalik Gdthab Frederikshab Narssarssuaq Julianehab Prins Christian Sund ]. In sme winter mnths, n measurable precipitatin was reprted in the nrthern Alaska except trace amunts f snwfall. Similar finding was als reprted fr Nrthwest Territries (NWT) and Yukn f Canada [Metcalf et al., 1994]. W and Steer [1979] designed a methd f measuring trace rainfall in the high arctic and determined a mean rate f 0.01 mm/h. Unfrtunately, there were n data available fr the perid f trace precipitatin in the climate archive since trace precipitatin is nt measurable by an rdinary precipitatin gauge. Precipitatin bservatins in Greenland shw that a number f events f trace precipitatin are reprted in a single trace precipitatin day, thus it is nt unreasnable t assume that a trace precipitatin culd be a measurable amunt f mm. T be cnservative, trace precipitatin was crrected n a daily basis in this study, fr example, fr any given trace day regardless f the number f trace bservatins reprted, a value f 0.10 mm was assigned and added t the mnthly ttal. The averaged yearly crrectin fr trace precipitatin varied frm 6 t 13 mm at the 12 statins, r abut 5-11% f the gauge-measured annual precipitatin in the nrthern regins and less than 3% in the suthern regins. Unlike Alaska [Yang et al., 1998b], the trace crrectin in Greenland shws very similar values between summer and winter seasns. Fr in- stance, the averaged mnthly ttals varied between 0.8 and 1.3 mm fr the 4 years at Thule. On the average, the ttal crrectins at Thule accunt fr 5-25% f the gauge-measured mnthly precipitatin. The amunt f trace recrd is inversely prprtinal t the gauge-measured annual precipitatin fr the 4 years (Figure 3), as was reprted by Yang et al. [1998b] and Bensn [1982] fr Alaska. Thus crrectin fr trace precipitatin is imprtant especially in the nrthern Greenland regins f lw precipitatin. It is imprtant t nte that after identifying trace precipitatin days, the number f measurable precipitatin days exhibits a larger variatin acrss Greenland (Figure 4). Therefre we recmmend t separate trace precipitatin days frm measurable precipitatin days when analyzing precipitatin climate in the nrthern Greenland regins Wetting Lsses Wetting lsses are gauge specific and vary by precipitatin type and the number f times the gauge is emptied. Accrding t the wetting lss experiments, the average wetting lss f the Hellmann gauge was 0.14 mm per bservatin fr rain and 0.10 mm fr snw [Aaltnen et al., 1993]. In this study, wetting lss was calculated n a daily basis. Fr each precipitatin day, an averaged wetting lss was determined accrding t the type f precipitatin and added t the daily recrd. This is the minimum crrectin since generally mre than ne bservatin was made in a precipitatin day. Fr the 4 years, the mean annual ttals f the wetting lss crrectin were between 5 and 15 mm at the 12 statins in Greenland, r 5-6% f the gauge-measured annual precipitatin in the nrthern regin and 2-3% fr the ther regins (Table 3). At each individual climate statin, the annual amunt f the wetting lss crrectin is different between the 4 years, and the percentage f the annual crrectin t the gauge-measured yearly ttals is different between years as well. Table 2. Mean Temperature, Wind Speed, and Gauge Measured Precipitatin at 12 Statins in Greenland fr Temperature, C Statin Min Max Danmarkshavn Thule (AFB) Scresbysund Egedesminde Sndre strmfjrd Hlsteindbrg Angmagssalik Gdthab Frederikshab Narssarssuaq Julianehab Prins Christian Sund Wind Gauge Percent Speed, Precipitatin Trace Measured, f m/s Days Days mm Snw

5 ._ YANG ET AL.' PRECIPITATION CORRECTION FOR GREENLAND 'Uz 100 'G z r-, 80 E 6O 4O 2O ø rl O 0 ø n D 0 % Gauge-measured annual precipitatin (mm) Figure 3. Number f trace precipitatin days versus gaugemeasured annual precipitatin at the 12 climate statins in Greenland fr years Generally, there is a clear wetting lss increase with increasing gauge-measured annual precipitatin in Greenland, since mre precipitatin generally requires mre bservatins and mre bservatin leads t a higher wetting lss Evapratin Lss Cmprehensive assessment f evapratin lsses indicated that average daily lsses varied by gauge type and time f the year. Fr the Danish Hellmann gauge at Jkiinen in Finland, evapratin lsses in summer f mm/d and winter f mm/d were measured [Aaltnen et al., 1993]. Ideally, evapratin lss shuld be crrected. Hwever, because f its strng dependence n weather cnditin and its daily variatin and seasnal change which can be very site dependent, it is nt apprpriate t estimate the daily evapratin lss at climate statins by using the average evapratin amunt btained frm ther experimental sites. Therefre evapratin lss was nt crrected in this study Wind-Induced Errrs T crrect fr wind-induced undercatch, wind speed at gauge height is required. When wind speed is nt measured at the height f the gauge, it was estimated frm measurements at higher heights, using the lgarithmic wind prfile apprach. Rughness length is required by the wind prfile technique. Accrding t Sevruk [1982] and Glubev et al. [1992], Z 0 = 0.01 m fr a winter snw surface and Z0 = 0.03 m fr shrt grass in the summer are apprpriate average rughness parameters fr mst sites. In this study, Z0 = 0.01 m was used fr the cld perid frm September t May and Z m was assigned t the warm perid frm June t August in Green- land. It has been well dcumented that fr the same wind speed gauge, undercatch fr snw is much higher than fr rain [Gdisn et al., 1981; Yang et al., 1995, 1998b]. Classifying the type f precipitatin is therefre necessary in rder t apply the best wind-lss crrectin. In this study, type f precipitatin was classified int snw, mixed and rain. Fr the snw categry, blwing snw events were fund reprted n sme f precipitatin days in Greenland. Blwing snw cnditins n precipitatin days are a special case when crrecting the gauge-measured precipitatin data. It is pssible that under certain cnditins, any gauge can catch sme blwing snw. Since wind speeds are generally greater during blwing snw events, a larger crrectin fr "undercatch" culd be applied t a measured ttal already augmented by blwing snw. This prblem wuld be mst severe fr gauges munted clse t the grund. The Hellmann gauges in 3OO 25O [] measurable [] trace Figure 4. Number f precipitatin days in 1996 at the 12 climate statins in Greenland.

6 6176 YANG ET AL.' PRECIPITATION CORRECTION FOR GREENLAND Table 3. Crrectins f Gauge-Measured Annual Precipitatin at 12 Climate Statins in Greenland fr Crrectins, mm Crrectin Factr P.q P, mm Wind Wetting Trace Sum mm Mean Min Max Danmarkshavn Thule (AFB) S cresbysund Egedesminde Sndre strmfjrd Hlsteindbrg Angmagssalik Gdthab Frederikshab N arssarssuaq Julianehab Prins Christian Sund Greenland were placed at 3 m abve the grund. T avid the pssible vercrrectin caused by high wind n precipitatin days, an upper value f wind speed has t be determined. Crrectins at higher wind speed are estimated by using this threshld wind speed [WMO/CIMO, 1993]. This is imprtant since the regressin equatins that are derived frm the Intercmparisn data are nly valid statistically fr the interval fr which they are develped and shuld nt be used fr extraplatin utside f this range. The threshld wind speed was set up at 6.5 m/s at gauge height fr the crrectin equatins in this study, since insufficient wind data greater than this threshld were cllected at the WMO sites. When daily wind speed at the gauge height was available, the daily catch rati (CR, %) fr the Nipher-shielded Hellmann gauges was calculated n a daily basis using the regressin equatins (1)-(3) fr snw, mixed, and rain. The wind-lss crrectin cefficient (K) was calculated as K - 100/CR. The average yearly crrectin fr the wind-induced bias at the 12 statins range frm 60 t 260 mm fr the 4 years (Table 3). Spatial variatin f the yearly crrectins fr wind-induced undercatch in Greenland can be explained by wind speed, percent f snw in annual precipitatin, gauge-measured precipitatin amunt. Generally, the abslute amunt f the yearly crrectin increased with increasing gauge-measured annual precipitatin frm nrth t suth. Hwever, the percentage f the yearly crrectin t the gauge-measured annual ttal decreased frm 50-75% in the nrthern regins t 20-40% in the suthern regins, because f the nrth-suth gradient f percentage f snwfall. higher wind lss fr snw than fr rain and due t the smaller amunt f abslute precipitatin in the cld seasn than in the warm seasn. It is even mre imprtant t realize the intraannual variatin f the mnthly CF values due t the fluctuatin f wind speed, frequency (r percentage) f snwfall, number f trace precipitatin events, amunt f gaugemeasured precipitatin and air temperature. Table 3 summarizes the yearly ttal crrectins fr windinduced errr, wetting lss, and trace precipitatin. Fr each statin the individual crrectins are stated and summed. The sum f the crrectins is added t the gauge catch (Pg) t btain the crrected value (p,). The verall annual crrectin factr (CF) is then the rati P,/P.q. Althugh the sum f the 50 ' 40 E g ½ 3-20 [] trace [] wetting lss [] w nd lss ßmeasured i, Thule air base, Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nv Dec 3.5. Mnthly and Yearly Ttal Crrectins Mnthly variatin f the crrectin shws the fllwing: ( 1 ) in each mnth thc abslutc mnthly amunt f wind-induced errr was always greater than wetting lss and trace precipitatin; (2) the cntributin f trace precipitatin and wetting lss t the ttal crrectin decreased frm nrth t suth, in the mst suthern regin (such as at Prince Christianssund), trace precipitatin was almst negligible; (3) mnthly precipitatin was increased by 30-50% due t the crrectins, and it was even dubled fr winter mnths in the nrthern regins f Greenland (Figure 5, fr example). It is imprtant t nte the seasnal variatin f the mnthly crrectin factr (CF, rati f the mnthly crrected t gaugemeasured precipitatin), i.e., the high values fr snw data in the cld seasn frm September t May and the lw values fr rain data in the warm seasn frm June t August, due t the Danmarkshavn, Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nv Dec Figure 5. Mean mnthly crrectins fr wind-induced errr, wetting lss, and trace amunt f precipitatin at Thule and Danmarkshavn Statins fr

7 YANG ET AL.' PRECIPITATION CORRECTION FOR GREENLAND 6177 abslute crrectins was highest fr Prince Christianssund, the highest annual crrectin factrs were generally in the arctic 30 statins. Fr the 4 years the maximum CF was at Danmark- g shavn and the minimum CF was at Angmagssalik. g 25 Studies f.precipitatin crrectin in nrthern Canada [Met-._g 20 calfe and Gdisn, 1993] and Alaska [Yang et al., 1998b] indicated that trace precipitatin and wetting lss crrectins 8 15 were imprtant in high latitudes. This study in Greenland 10 shwed that due t the higher undercatch f snwfall f the Hellmann gauge, wind-induced errr was the largest bias and 5 wetting lss and trace amunt f precipitatin were significant 0 errrs as well. In nrthern Greenland regins, trace crrectin is even greater than wetting lss (Table 3). a) Canadian Nipher gauge Daily crrectin (mm) 4. Discussin Different methds f crrecting gauge measurements fr systematic errrs have been develped and applied fr climatlgical and hydrlgical studies [Larssn and Peck, 1974; Grisman et al., 1991; Grisman and Legates, 1993; Grisman and Easterling, 1994; Grisman et al., 1996; Metcalfe and Gdisn, 1993; Metcalfe et al., 1994; Sevruk et al., 1993; Legates and Deliberty, 1993; Gdisn and Yang, 1995; Yang and Gdisn, 1998; Yang et al., 1998a, b; UNESCO, 1974]. The validity f these crrectin techniques fr different climate cnditins needs t be evaluated, and the cmpatibility f the crrectin results amng different gephysical regins als needs t be investigated. Bias crrectin was cnducted n a daily timescale in this study, using the crrectin methd derived frm daily experimental data f the WMO intercmparisn. Crrectin n a shrter timescale, such as hurly r 6 hurly is expected t prduce better results since wind speeds may vary thrughut the day and daily mean wind speeds may nt be representative f wind cnditins f precipitatin perids. T assess the errrs that are incurred in use f daily versus shrter time intervals, hurly r 6 hurly data are required. Hurly r 6 hurly Greenland data are nt available t this study. An alternative site at Reslute Bay, NWT, Canada, with similar climate cnditins as in Greenland (i.e., windy and cld with snw dminatin) was thus chsen fr the cmparative runs f 6 hurly versus daily crrectins. It was fund that fr the Canadian Nipher gauge used at this site, the errrs f daily crrectin in the selected mnths are small n a daily basis (Figure 6) and the mnthly results f daily and 6 hurly crrectins cmpared very well t each ther (Table 4). The Hellmann gauge is nt used in Canada, and crrectin factrs (CF) were calculated fr this gauge, using 6 hurly versus daily wind data at Reslute Statin. The results illustrate a visible scatter in the daily CF cmparisn (Figure 6), but the mnthly mean CF values f daily and 6 hurly crrectins generally agreed better, with the difference being less than 10% fr mst f the mnths (Table 4). Given the uncertainties f estimatin f wind speed at gauge height frm a standard height and classificatin f precipitatin types by surface air temperature, it seems reasnable t cnclude that the errrs f daily crrectins are essentially randm and they tend t cancel ut when mnthly ttals are cmputed. It has been lately reprted that blwing snw is imprtant t gauge crrectin in Alaska [Yang et al., 1998b]. Greenland climate data shw less blwing snw events due t lwer wind speed. It is hwever imprtant t pint ut that the threshld wind speed (6.5 m/s) fr crrecting wind-induced errr is nt 3, 2 s 2.00 LL 1 50 r,._) 1.00 Eb 0 50 b) Hellmann gauge, crrectin factr (CF) ø ½ O O O0 Daily crrectin factr (CF) Figure 6. Cmparisn f daily versus 6 hurly crrectins during May t July 1996 at Reslute Bay, NWT, Canada, fr (a) crrected daily precipitatin f Canadian Nipher gauge (in millimeters) and (b) crrectin factr f the Hellmann gauge. very lw fr the 12 statins in Greenland. Fr instance, Prince Christianssund is the windiest site f the 12 statins. The an- nual mean daily wind was 6.8 m/s at 8 m and 5.6 m/s at gauge height (f 3 m) fr the 4 years. At Prince Christianssund there were 41 days ut f 190 precipitatin days in the 4 years when daily wind speed at gauge height was greater than the threshld wind. Number f days when wind exceeds the threshld was much less at ther sites. It is therefre apprpriate t set up the threshld wind at 6.5 m/s and t apply the crrectin prcedures with this threshld wind in Greenland. It is interesting t cmpare this wrk t ther studies. In a Greenland arctic drainage basin (76øN) water balance study, Thmsen [1991] estimated the annual precipitatin crrectin factrs t be ; this estimatin generally agreed well with ur crrectin results (Table 3). Canada has cnducted preliminary tests in applying bias crrectin t its digital archive data in Nrthwest Territries [Metcalfe et al., 1994]. The results indicated that actual annual precipitatin can be % greater than the gauge-measured amunt in Canadian arctic regins. Table 5 gives the results f bias crrectin at five selected climate statins in the eastern Arctic regins f Canada. These results are cmpatible t the current Greenland wrk, since bth studies used the same reference instrument (i.e., DFIR) fr true snwfall estimatin. It is apparenthat the relative increase f annual precipitatin due t the crrectins were lwer at the Canadian statins, regardless f the higher percentage f snwfall at these sites; this is mainly because the Hellmann gauge catches much less snw than the Canadian

8 6178 YANG ET AL.: PRECIPITATION CORRECTION FOR GREENLAND Table 4. Mnthly Summary f Daily Versus 6 Hurly Crrectins fr Canadian Nipher Gauge and Hellmann Gauge, Reslute Bay, NWT, Canada, 1996 Mean Nipher Crrectin Mean Wind Number f Measured Temp, at 10 m, Precip Precip, Daily, 6 Hurly, øc m/s Days mm mm mm Hellmann Crrectin Factr Daily 6 Hurly May June July Sum/mean Nipher gauge at the same wind speeds (Figure 7). This cmparisn therefre clearly demnstrates that discntinuity f bserved precipitatin data exist acrss natinal bundaries. Effrts have been reprted t prduce precipitatin climatlgies fr Greenland. Brmwich et al. [1998] have evaluated recent precipitatin studies, including results f numerical analyses. Thmsen [1991] fund large spatial variatins f annual precipitatin acrss Greenland: a strng latitudinal gradient, with the annual ttal being less than 200 mm in the extreme nrth and mre than 1500 mm in the sutheastern Greenland; and als a significant lngitudinal variatin, with higher precipitatin at the cast and lwer precipitatin ( mm) n the ice sheets. Lenart [1991] prduced a precipitatin map fr the plar regins (nrth f 60øN), Greenland precipitatin annual ttals varied frm 800 mm in the suth cast t 150 mm in the nrthern regins. Ohmura and Reeh [1991] cnstructed new precipitatin and accumulatin maps f Greenland. Their results prvided a higher-reslutin distributin f precipitatin and accumulatin. The annual ttals f mre than 2500 mm n the sutheastern tip f Greenland and abut 100 mm in nrthern Greenland were reprted; a strng lngitudinal gradient in the suthern Greenland was als fund, with the eastern cast receiving cnsiderably mre precipitatin than the west cast. The mean annual precipitatin fr all Greenland was estimated t be 340 mm [Ohmura and Reeh, 1991]. The abve studies have generally revealed the basic features f precipitatin distributin ver Greenland, but the climatlgies derived are qualitatively very different. This discrepancy may be due t different data sets and analysis methds used fr the studies. Fr instance, Ohmura and Reeh [1991] used a cmbinatin f snw pits, ice cre data, and climatic bserva- tins in the cast regins fr their analysis, and they estimated annual precipitatin n Greenland ice sheets frm glacier ac- cumulatin. Thmsen [1991] cnsidered biases f gauge measurements and used bias-crrected precipitatin data in his wrk. Ohmura and Reeh [1991] prvided the climate nrmals f annual gauge-measured precipitatin fr the perid f 1950s- 1980s at 35 statins perated by the Danish Meterlgical Institute in Greenland. In cmparisn t these lng-term annual means, the annual precipitatin fr 1994 t 1997 was generally lwer, perhaps indicating tempral variability f precipitatin in Greenland. Given the intra-annual variatin f wind, temperature, and precipitatin, it is recmmended t cnduct the bias crrectin fr the entire recrding perid n a daily timescale [Yang and Gdisn, 1998]. Hwever, it is als useful t estimate the magnitude f the bias crrectin by simplified appraches. Table 6 gives an example f such effrts. The maximum and minimum values f yearly CF derived frm the daily crrectins fr was applied t the lng-term means at the selected sites in Greenland. The results indicate that the crrected lng-term annual precipitatin may vary frm mm fr the nrthern regins t mm fr the suthern casts. These results suggesthat annual precipitatin in Greenland is much higher than previusly reprted, particularly in the suthern regins f high precipitatin; this may imply a greater latitudinal precipitatin gradient acrss Greenland and als a higher mean precipitatin fr all Greenland. These bias crrectin results therefre will have a meaningful impact n water budget and glacier mass balance studies in Greenland. Legates and Deliberty [1993] reprted a general pleward increase f the percent f crrectin in winter seasns ver the cntinental United States due t the increasing snwfall prprtin. This study in Greenland shws the basic feature f increasing percentage f the annual crrectin frm the suth cast t the nrthern arctic regins. T achieve a better understanding f spatial variability f the crrectins, mre cli- Table 5. Cmparisn f Bias Crrectin Between Eastern Canadian Arctic and Nrthern Greenland Eastern Canadian Arctic, [Metcalfe et al., 1994] Crdinate Gauge Percent Elevatin, Measured, f P,, Statin N W m mm Snw mm Mean CF Statin Nrthern Greenland, (This Study) Gauge Percent Measured, f P,, mm Snw mm Mean CF Alter Eureka Reslute Clyde Cape Dyer Danmarkshavn 1.51 Thule (AFB) 1.58 Scresbysund 1.39 Egedesminde 1.35 Sndre strmfjrd

9 YANG ET AL.: PRECIPITATION CORRECTION FOR GREENLAND loo '-' 80._ O --e-- Canadian N pher Hellmann unsh Wind speed at gauge height (m/s) Figure 7. Cmparisn f the catch rati f snw as a functin f wind speed at gauge height fr the Hellmann gauges and the Canadian Nipher snw gauge. mate statins in Greenland, particularly thse research statins lcated in central Greenland ice cap, shuld be invlved and mre years f precipitatin data shuld be crrected. Further analysis f the crrected precipitatin in the circumplar cuntries is certainly necessary t evaluate the validity f the precipitatin crrectin prcedures fr the plar regins. Gridded prducts f the crrected precipitatin shuld als be develped fr validatin f the climate mdel simulatins [Walsh et al., 1998; Kattsv et al., 1998]. 5. Cnclusin The crrectin prcedures derived frm the WMO Slid Precipitatin Measurement Intercmparisn data set fr the Hellmann gauge have been applied at 12 climate statins in Greenland fr fur test years. Biases f wind-induced undercatch, wetting lss and trace amunt f precipitatin were crrected n a daily basis, and the gauge-measured annual precipitatin was increased significantly by 35 t 280 mm (abut 25 t 90% f the gauge-measured yearly ttal) at the 12 statins. Of the biases in precipitatin measurement, wind lss is the greatest. Wetting lss and trace amunt f precipitatin are als significant errrs in the nrthern Greenland regins f lw precipitatin. Seasnally, the crrectin is greater in win- ter and smaller in summer due t the increased effect f wind n gauge undercatch f snwfall. The mnthly crrectin factrs (crrected/measured precipitatin) in Greenland differed frm statin t statin. At an individual statin the mnthly crrectin factrs varied by type f precipitatin and by mnth even fr the same type f precipitatin, since these biases nt nly dependent n the wind speed but als n the wetting lss, trace amunt f precipitatin, and the actual gauge-measured precipitatin amunt. In additin, there is a cnsiderable intra-annual variatin f the magnitude f the crrectin, due t the fluctuatin f wind speed, air temperature, and frequency f snwfall. It is clear that the mnthly crrectin factrs are nt cnstant. Thus crrectin f the biases shuld nt be cnducted n a mnthly basis and the mnthly crrectin factrs btained frm ne intercmparisn statin shuld nt be used fr ther climatic and hydrlgical statins withut further detailed analysis n wind and snw climate. Bias crrectins shuld be cnducted n a daily basis fr the entire recrding perid fr each individual statin in an bservatinal netwrk. These crrectins require cnsiderable statin infrmatin (metadata) and additinal meterlgical infrmatin (i.e., wind at gauge height n precipitatin days, temperature, precipitatin type, and gauge-measured amunt f precipitatin) fr their implemen- tatin. The WMO Slid Precipitatin Measurement Intercmparisn has prvided new crrectin techniques fr a number f precipitatin gauges cmmnly used arund the wrld [Gdisn et al., 1998], such as the Canadian Nipher snw gauge [Gdisn and Metcalfe, 1992], the U.S. Natinal Weather Service (NWS) 8" standard gauge [Yang et al., 1998b], the Russian Tretyakv gauge [Yang et al., 1995], and the Hellmann gauge [Gunther, 1993; Yang et al., 1994; Yang et al., 1999]. These crrectin prcedures are recmmended fr test crrectin f the gauge-measured daily precipitatin in thse cuntries where natinal meterlgical r hydrlgical statin netwrks perate these gauges fr precipitatin bservatin. It is expected that the crrectin will prvide reliable unbiased estimates f the true amunt f precipitatin fr the natinal bservatinal netwrks and the crrectins will have an impact n climate mnitring. It is hped that mre effrt will be made by the natinal meterlgical services t apply the crrectin prcedures t their archived precipitatin data. It is believed that t d s will significantly imprve the accuracy and hmgeneity f precipitatin data ver large regins in the Nrthern Hemisphere. Table 6. Estimated Crrectin f Lng Term Mean Annual Precipitatin at 11 Sites in Greenland Mean Range f Crrectin Range f Crrected Annual Factr Precipitatin, mm Precipitatin, Data Statin mm Min Max Min Max Perid Danmarkshavn Thule (AFB) Egedesminde Sndre strmfjrd Hlsteindbrg Angm ag salik Gdthab Frederikshab Narssarssuaq Julianeh ab Prins Christian Sund

10 6180 YANG ET AL.: PRECIPITATION CORRECTION FOR GREENLAND Acknwledgments. This study wuld nt have been pssible with- Hanssen-Bauer, I., and E. J. Frland, Hmgenizing lng Nrwegian ut the c-peratin f the participating cuntries and the many b- precipitatin series, J. Clim., 7(6), , servers at the statins where the WMO Intercmparisn was cn- Karl, T. R., R. G. Quayle, and P. Y. Grisman, Detecting climate ducted fr a number f years. The authrs wuld als like t thank Dr. variatins and change: New challenges fr bserving and data man- Henning Madsen and Mr. Fleming Veijen, Danish Meterlgical agement system, J. Clim., 6(8), , Institute, fr prviding Greenland climate statin and instrumental Kattsv, V., V. Meleshk, and V. Gvrkva, Simulatins f snw/ice infrmatin. The authrs appreciate cnstructive cmments f tw and ablatin ver Greenland and Antarctic in atmspheric mdels, reviewers f the riginal manuscript. in Prceedings f a Meeting f Experts n Crysphere and Climate, edited by R. G. Barry, pp. B4.1-B4.10, WMO/TD-867, WCRP-102, Wrld Meterl. Organ., Geneva, References Larsn, L. W., and E. L. Peck, Accuracy f precipitatin measurements fr hydrlgical mdelling, Water Resur. Res., 0(4), 857- Aaltnen, A., E. Elmaa, A. Tuminen, and P. Valkvuri, Measure- 863, ment f precipitatin, in Prceedings f the Sympsium n Precipita- Legates, D. R., Precipitatin measurement biases and climate change tin and Evapratin, vl. 1, edited by B. Sevruk and M. Lapin, pp. detectin, in Prceedings f the 6th Sympsium n Glbal Change 42-46, Slvak Hydrmeterl. Inst. and Swiss Fed. Inst. f Technl., Studies, pp , Am. Meterl. Sc., Bstn, Mass., Bratislava, Legates, D. R., and T. L. DeLiberty, Precipitatin measurement biases Allerup, P., H. Madsen, and F. Veijen, A cmprehensive mdel fr in the United States, Water Resur. Bull., 29(5), , crrecting pint precipitatin, Nrd. Hydrl., 28, 1-20, Legates, D. R., and C. J. Willmtt, Mean seasnal and spatial vari- Bensn, C. S., Reassessment f Winter Precipitatin n Alaska's Arctic ability in gauge-crrected, glbal precipitatin, Int. J. Climatl., 10, Slpe and Measurement n the Flux f Wind Blwn Snw, 26 pp., , Gephys. Inst., Univ. f Alaska, Fairbanks, Brmwich, D., R. Cullather, Q. Chen, and B. Csath, Evaluatin f Lenart, W., Structural and measurement prblems f arctic atmrecent precipitatin studies fr Greenland ice sheet, J. Gephys. spheric water balance, in Arctic Hydrlgy: Present and Future Tasks, Res., 103, 26,007-26,024, Tech. Rep. 23, pp , Nrw. Natl. Cmm. fr Hydrl., Osl, Clny, R., V. Randinv, and F. R. Tanis, Measurements f precipitatin and snw pack at Russian nrth ple drifting statins, Plar Metcalfe, J. R., and B. E. Gdisn, Crrectin f Canadian winter Rec., 34(188), 3-14, precipitatin data, in Prceedings f the 8th Sympsium n Meter- Desbls, M., and F. Desalmand, Past and present precipitatins re- lgical Observatins and Instrumentatin, pp , Am. Metelated t climate changes--methdlgies and perspectives, Bull. f rl. Sc., Bstn, Mass., Am. Meterl. Sc., 76(7), , Metcalfe, J. R., S. Ishida, and B. E. Gdisn, A crrected precipita- Glubev, V., On the prblem f standard cnditins fr precipitatin tin archive fr the Nrthwest Territries, in Mackenzie Basin Imgauge installatin, in Prceedings f the Internatinal Wrkshp n pact Study, interim reprt 2, pp , Prceedings f the Sixth the Crrectin f Precipitatin Measurements, WMO/TD-104, pp. 57- BienniaI AES/DIAND Meeting n Nrthern Climate, Atms. Envirn. 59, Wrld Meterl. Organ., Geneva, 1985a. Serv., Yellwknife, NWT, Canada, Glubev, V. S., On the prblem f actual precipitatin measurements Milkvic, J., Preliminary results f the WMO field intercmparisn at the bservatin site, in Prceedings f the Internatinal Wrkshp Yugslavia, in Prceedings f the Internatinal Wrkshp n Precipn the Crrectin f Precipitatin Measurements, WMO/TD-104, pp. itatin Measurement, Swiss Fed. Inst. f Technl., St. Mritz, , Wrld Meterl. Organ., Geneva, 1985b. Ohmura, A., and N. Reeh, New precipitatin and accumulatin maps Glubev, V. S., P. Y. Grisman, and R. G. Quayle, An evaluatin f fr Greenland, J. f Glacil., 37(125), , the U.S. 8-inch nnrecrding rain gage at the Valday plygn, Rus- Sevruk, B., Methd f crrectin fr systematic errr in pint precipsia, J. Atms. Oceanic Technl., 49(5), , itatin measurement fr peratinal use, WM0-589, 91 pp, Wrld Gdisn, B. E., H. L. Fergusn, and G. A. McKay, Cmparisn f Meterl. Organ., Geneva, pint snwfall measurement techniques, in Handbk f Snw, ed- Sevruk, B., Methd f crrectin fr systematic errr in pint precipited by D. M. Gray and M.D. Male, pp , Pergamn, Tar- itatin measurement fr peratinal use: Crrectins in Switzerrytwn, N.Y., land, in Prceedings f the Sympsium n Precipitatin and Evapra- Gdisn, B. E., B. Sevruk, and S. Klemm, WMO slid precipitatin tin, vl. 1, edited by B. Sevruk and M. Lapin, pp , Slvak measurement intercmparisn: Objectives, methdlgy, analysis, Hydrmeterl. Inst. and Swiss Fed. Inst. f Technl., Bratislava, in Atmspheric Depsitin, IAHS Publ. 179, pp , Wallingfrd, England, Sevruk, B., and W. R. Hamn, Internatinal cmparisn f natinal Gdisn, B. E., P. Y. T. Luie, and D. Yang, WMO slid precipitaprecipitatin gauges with a reference pit gauge, WMO Instrum. Obs. tin measurement intercmparisn, final reprt, WMO/TD-N. 872, Meth. Rep. 17, 111 pp., Wrld Meterl. Organ., Geneva, pp, Wrld Meterl. Organ., Geneva, Sevruk, B., and S. Klemm, Types f standard precipitatin gauges, in Gdisn, B. E., and J. R. Metcalfe, The WMO slid precipitatin Prceedings f the Wrkshp n Precipitatin Measurements, edited intercmparisn: Canadian assessment, in WMO Technical Cnferby B. Sevruk, pp , Wrld Meterl. Organ., Geneva, ence n Instruments and Methd f Observatin, WMO/TD-462, pp. Tabler, R. D., N.H. Berg, D.C. Trabant, H. S. Santfrd, and P. A , Wrld Meterl. Organ., Genewt, Rechard, Measurement and evaluatin f winter precipitatin, in Gdisn, B. E., and D. Yang, In-situ measurement f slid precipi- Cld Regins Hydrl 9; and Hydraulics, edited by W. L. Ryan and tatin in high latitudes: The need fr crrectin, in Prceedings f the R. D. Crissman, pp. 9-38, Am. Sc. f Civ. Eng., Restn, Va., ACSYS Slid Precipitatin Climatlgy Prject Wrkshp, WMO/TD- 739, pp. 1-12, Wrld Meterl. Organ., Geneva, Grisman, P. Y., and D. R. Easterling, Variability and trends f ttal precipitatin and snwfall ver the United States and Canada, J. Clim., 7( 1 ), , Grisman, P. Y., and D. R. Legates, The accuracy f United States precipitatin data, Bull. Am. Meterl. St., 75(2), , Grisman, P. Y., V. V. Kknaeva, T. A. Belkrylva, and T. R. Karl, Overcming biases f precipitatin measurement: A histry f the USSR experience, Bull. Am. Meterl. St., 72(11), , Grisman, P. Y., D. R. Easterling, R. G. Quayle, and V. S. Glubev, Reducing biases in estimates f precipitatin ver the United States: Phase 3 adjustments, J. Gephys. Res., 101, , Gunther, T., German participatin in the WMO slid precipitatin intercmparisn: Final results, in Prceedings f the Sympsium n Precipitatin and Evapratin, vl. 1, edited by B. Sevruk and M. Lapin, pp , Slvak Hydrmeterl. Inst. and Swiss Fed. Inst. f Technl., Bratislava, Thmsen, T., Arctic hydrlgy in Greenland-permafrst hydrlgy and data technlgy, in Arclie Hyd vlgy.' Present and Future Tasks, Tech. Rep. 23, Nrw. Natl. Cmm. fr Hydrl., UNESCO, Wrht Water Balance and Water Re,surces f the Earth (WWB) (in Russian), 638 pp., Gidrmeteizdat, St. Petersburg, Russia, (Engl. Transl., 1978) Walsh, J. E., V. Kattsv, D. Prtis, V. Meleshk, Arctic precipitatin and evapratin: Mdel results and bservatinal estimates, J. Clim., I l (1), 72-87, W, M. K., and P. Steer, Measurement f trace rainfall at a high arctic site, Arctic, 32(1), 80-84, Wrld Meterlgical Organizatin (WMO)/CIMO, Internatinal Organizing Cmmittee fr the WMO Slid Precipitatin Measurement Intercmparisn, final reprt f the first sessin, WMO Rep., 31 pp., Wrld Meterl. Organ., Geneva, Wrld Meterlgical Organizatin (WMO)/CIMO, Internatinal Organizing Cmmittee fr the WMO Slid Precipitatin Measurement

11 YANG ET AL.: PRECIPITATION CORRECTION FOR GREENLAND 6181 Intercmparisn, final reprt f the sixth sessin, WMO Rep., 140 pp., Wrld Meterl. Organ., Yang, D., and B. E. Gdisn, Cmment n "Reducing biases in estimates f precipitatin ver the United States: Phase 3 adjustments" by P. Y. Grisman et al., J. Gephys. Res., 103, , Yang, D., J. R. Metcalfe, B. E. Gdisn, and E. Mekis, An evaluatin f the duble fence intercmparisn reference gauge, in Prceedings f the Eastern Snw CniC'fence, pp , Quebec City, Quebec, Canada, Yang, D., B. Sevruk, E. Elmaa, V. Glubev, B. Gdisn, and T. Gunther, Wind-induced errr n snw measurement: WMO Intercmparisn Results, paper presented at the Internatinal Cnference n Alpine Meterlgy, Lindau, Germany, pp , Yang, D., et al., Accuracy f Tretyakv precipitatin gauge: Results f WMO Intercmparisn, Hydrl. Prcess., 9(8), , Yang, D., B. E. Gdisn, J. R. Metcalfe, V. S. Glubev, R. Bates, T. Pangburn, and C. L. Hansn, Accuracy f NWS 8-inch standard nn-recrding precipitatin gauge: Result f WMO Intercmparisn, J. Atms. Oceanic Technl., 15(2), 54-68, 1998a. Yang, D., B. E. Gdisn, C. B. Bensn, and S. Ishida, Adjustment f daily precipitatin at 10 climate statins in Alaska: Applicatin f WMO Intercmparisn results, Water Resre: Res., 34(2), , 1998b. Yang, D., E. Elmaa, B. Gdisn, T. Gunther, V. Glubev, B. Sevruk, H. Madsen, and J. Milkvic, Wind-induced precipitatin undercatch f the Hellmann gauges, Nrd. Hydrl., 30(1), 23-31, S. Ishida and B. Gdisn, Climate Research Branch, Atmspheric Envirnment Service, Dwnsview, Ontari, M3H 5T4, Canada. T. Gunther, German Weather Service, Berlin, Germany. D. Yang (crrespnding authr), Inst. fr Glbal Change Research, SEAVANS Building Nrth, 7th flr, Shibaura, Minat-ku, Tky , Japan. (dyang@frntier.est.r.jp) (Received June 26, 1998; revised Octber 27, 1998; accepted December 3, 1998.)