Climatological Bulletin Vol. 27, No. 2, August/ao"! 1993

Bulletin climatologique

Canadian Meteorological and Oceanographic Society

La Societe Canadienne de Meleorologle et d'Oceanographie

Information for Contributors and Subscribers

As a publication of thc Canadian Meteorological and Oceanographic Society, the CLIMATOLOG ICAL

BULLETIN provides a medium of information on climatology. The Editorial Board gives special encouragement to the submission of manuscripts on applied climatology (e.g., agriculture, commerce, energy, environment, flsheries. forestry, health, recreation, transportation, and water resources), climatic change and variability, climate impact studies, climate model applications (including physical climatology), and regional studies (including ocean areas).

Authors may submillheir manuscripts to "Articles", "Research Notes" or "News and Comments". This should be indicated in the cover leiter accompanying the manuscripl. Articles and Noles are independent ly reviewed by at least two anonymous referees . News or comments arc reviewed by the Editor in consultation with the Editorial Board. Manuscripts are accepted in either English or French. An abstract (in both English and French) is required for Artides and Notes.

Contributors should submit manuscripts to Alexander H. Paul, editor, CLIMATOLOGICAL

BULLETIN, Department of Geography, University of Regina, Regina, Saskatchewan, S4S OA2. All manuscripts should be typed double spaced on one side of good qUality white paper, 28 cm X 21 .5 em, or its nearest equivalent. The abstmct, list of references, tables, and a lisl of figure captions should be typed double spaced on sepamte sheets. Comments (indudi ng book reviews and opinions) and news items should not exceed 1500 words. Furnish an original and three copies if possible, in the order listed below.

The TITLE PAGE should indude the full names of author(s), and professional affiliation(s). The ABSTRACT should be less than 250 words, and typed on a separate page. The T EXT of longer contribu tions should be typed double spaced on numbered pages, and divided into sections, each with a sepanHe heading and numbered consecutively. The section heading should be typed on a separate line. ACKNOWLEIXiEM ENTS are typed on a separate sheet immed iately following the text . If FOOTNOTES are required , they should be typed, double spaced , on a separate sheet under the heading "Notes" at the end of the text. R. EFERENCES should be arranged alphabetically by senior author's last name. The text citation should consist of name(s) of the author(s) and the year of publication, for example, Jones (1975) or (Jones, 1975). When there are two or more cited publications by the same author in the same year, d istinguishing letters a, b, etc., shou ld be added to the year. A reference to "in press" implies that the paper has been accepted for publication. Titles of periodicals should be given in full . FIGURE LEGENDS must be provided for each figure , and should be typed together, double spaced, on a separate sheet. ILLUSTRATIONS shou ld be numbered sequentially. Original drawings, lettering, and symbols should be large enough so that after reduction, the smallest characters will be al least 1.5 mm high. Each TABLE should be numbered sequentially_ Type all tables double spaced on separate sheets.

Authors should use the International System of units. but may show other unitS in parentheses. Authors should provide instructions in the margin for any special type required (Greek letters, capitals, bold face, etc.). .

Although page charges are voluntary, only those authors who submit payment to cover production costs are entitled to receive 50 free offprints. Payment of lesser charges docs not entitle the author to any free offprints. In all cases, the additional cost of printing colour plates or excessive altemtions made by the author at the proof stage must be paid by the author.

Correspondence regarding Society affairs, membership, Institutional subscriptions and back issues of the journal should be directed to the Corresponding Secretary, Canadian Meteorological and Oceanographic Society, P.o. Box 359, Newmarket, Ontario L3Y 4X7. Telephone (4 16) 898-1040, Fa. , (4 16) 898-7937,

Renseignements pour les collaborateurs et les abonnes Pu blicat ion de 1a Soc,ete ca oad .co ne de meteorologle ("I d 'occanographtc. Ie Bu iletln

chm,lI ologlquc orfre un moyen d' inro rmallon su r la cl!m'lIo1oSIe. Le comlte de rcd;lclton encourage en P:lrllcul!cr 101 SQu nllssion de manuscnts su r la climatologic applrquec (comme

I'agrlculture. Ie comme rce. I'cnergrc. I'environnem~nt. 13 pechene. la sylviculture. la ~anlc. les loi'i,r:',lcs lransports. et le~ ressources en eau), les changcmcnb ella vanablhtc du chmat. la prospeclt\ic chma to logtque. Its applications des modelcs du chma t (lnelus 1:.1 chmalOioglc ph YSiq ue), cl Ics elUdes regional (Ind us Ics oceans). II eSI pu blic grace a une sub\t~m ,on accordee par Ie gouvcrnerncnt canadlcn par I'lntermediare du Consc,1 de recherche!> en sCiences naturel1es

t:t cn genic.

Le~ au teurs pcuvcnt chol sir de so umetlre IClIrs manuscrils aux "A nic1 es". "Noles dc Recherches". Oll "No uvcl le~ el Commcntaircs". lls doivent l'mdiquer su r la lellTe

d ' accompag nemenl du mal1l1scri l. Les articles de reche rche et Ie, "NOles" ~o nl indcpendammenl

so umls:1 I'cxamen d',1lI moim deux appreciatclirs anonymes. Lc rcdaetclir en chef CXllllllne les

"Nouvelles ct Commcnlalres" conJomtement avec Ie com ll e de redaction . On acceptc les articles !>oil en fr:tn\als. soit en angl:u!>. II faut envaye r un T(:~ume. de preference en fram;als cl en anglai~ .

Les collabor:u eurs son l pries de sou mellre leurs man uSenl.!) a Alexander H . Paul.

rMacteur en chef. 81 I.Il11~ l U \1 AroLOl.IQl l. D epa n emcni dc geographic, UOIve r!oHe Regina. Regilla. Saska tchewan, S4S OA2. ]] fa ul dactylographlf:r taus les manuscnlS a double Interligne,

sur un cOle d'un papler blanc de bonne qua lite, de 28 !our 21.5 em ou des dimenSions I~ plus

r3pprochecs. II faul da clylograpilier it. double IIl te rligne, s ur de~ (cullle!> it pan. Ie re~lImc, la liste

des references, Ics tableaux el In IIs te des legendcs. Les commen lai rcs. les cntique!>, les o ptOlons el h.'" mtormutlons nc dOlvc nt pas dcpasse r I 500 mOb. FourOlr I'origillal Ct .!ol posSlbk, lrol<;

COPICS, scion 1;1 d l~posltlon SlII vante:

I \ PM.I 01 1 n RI dOli comprendre: no m. pre-noms des aUleur'!. alOM que It·s alTih;llion!> profC!>slonnel lcs.

II R( "I MI, dactylographic sur une page a part. ne dOli P;I'! compter plus de 250 mOls.

I I I! ... !! . II faut taper it double IIllerligne. sur des pages numcrotCes. Ie lexlC des articles plus longs et Ie dlviser en SCClio ns, chacune dOlce d ' une en-tet e a part el num cro tce d:lO:' I'ordre. ]] raU l

dactylographicr I'cn-Iclc de seclion sur une ligne it part .

II RI MI R( II MJ I\IS doivenl clre dactylographics sur une pagr::\ part . aprcs Ie leXle.

I I " ~o I I ... 1)1 !lAS 1)1. PACd doiven\ ctre daclylographices a double HlIerlignc, ~ ur une feuille a part. au-dcssous dc l'cn-lelC "NOlc~", 3 la rin elu leXle .

u" R( II RI M l !o dOl\'cnl ctre dlsposccs dans I'ordrc a lphabet lque. d'apr~s Ie nom de I'autcur

pnnclpal Lc) cHat Ions dOlvenl com prendre Ie ou les noms du ou de'! aUleurs r: t I'anncc de publication, comme Jones (1975) ou (Jones, 1975). Si r on cite dCII~ ou plusicur .. pubhcallons du

mcme aUleu r pendanl 13 mcme annce. il rau l aJouler a I'anncc des lettres qUI pcrmctlcnl dc les dlfrcrcnclcr. comme a, b. elc. La menllo n "3 rimpresslon" Impllq ue q u'on 3 acccPle de publicr la

commu nlC3 110n. II faul ,"d iq uer les lures des periodlques en cnlier. II ... I I (.t !'om'! clles soo\ requiscs pour chaquc figure. On dactylographlera ccs Icgcnd c!> cll!oocmb1c.

a double 1I11 erhgne, sur une reuillc a part. I I ... 1\ LI " -1 RA IIOJ\~ dOIVenl clre numerotees dans I·ordrc. L 'onglOal des dcs~lOs. do c;lracteres el

des s)lmbolc!. doi t ct re a:.sel grand pour que. apres reducllon, Ie caractere Ie plu:. petit m('~u re au maIO:. 1.5 mm dc haul.

I LS 11\8U "I '( dOlvenl etre numerotes dans rordre. La dactylographle pour taus les tableaux dOli

etre :1 double IOlerhgne sur des reullics it part. Les aulcurs dOlvcnl ullhser Ie systemc IIlternallollal de me!ourc, malS Ib pcuvent

lndiqur:r d'au lrc:, mesures entre parentheses. Les auteurs dOlvenl porter dans la marge des

Instru ctiom pour lo ut curacterc specia l voulu (lellres grecqucs , maj uscu les. c:t ractCres gras. etc. ).

Lcs auteurs d 'unicles et de notes de recherche doivent payer des Inus de publication de

50 S par page, cc qui donne droit a 25 copies graluites . On peut ncgoc ler un laux redu ll en C4l..') de diflicultcs financie rcs. mais I'auteur n'aura pas drOll it des tlrages g ratu its .

La correspondancc concemantles activtles dc la SOClcte. les sousc npllon:.de!> membres

e t des IIlslilulions etles numerosdeJa parus. devrai t etre adressceau Secret3Ire-corrc:.pondanl. Soc lcte canadicnne de meleorologlc e l d' oceanographie, C. P. 334. Newmarket, Ontano LJ Y 4X7 Telephone

(416) 898· 1Q40. Fax . (416) 898·7937.

Climatological Bulletin Bulletin Climatologique Vol. 27, No.2, August; aout 1993

28 FOREWORD ; AVANT-PROPOS

ARTICLES

29 Towards Defining Important Averaging Periods for Climate Normals James R. Angel, William E. Easlerling and Scoll W. Kirsch

45 Zonage du risque agroclimatologique durant la saison froide au Quebec meridional: I - froid hivernal Philippe Rochelle and Pierre-Andre DuM

63 Trends of Daily Maximum and Minimum Temperature in Canada During the Past Century W. R. Skinner and D W. Gullell

ISSN 0541-6256

Editor ; Redacteur en chef A.H. Paul

UNIVERSITY OF REGINA

Associate Editors ; Redacteurs associes 1. Dublin GA McKay AES, ATLANTIC REGION ATMOSPHERIC ENVIRONMENT SERVICE

S. Tabata L.A. Mysak INSTITUTE OF OCEAN SCIENCES McG ILL UNIVERS ITY

D. W. Phillips E. E. Wheaton ATMOSPHERIC ENVIRONMENT SERVICE SASKATCHEWAN RESEARCH COUNCIL

R. Leduc S.1. Cohen ENVIRONNEMENT QUEBEC CANADIAN CLIMATE CENTRE

27

Foreword / Avant-Propos

The review of topical and geographic coverage of contributions to the Bul/eli" in recent years will appear in the December issue.

Le Bulletill publicra Ie sommaire des sujets et des regions qui ont caracterise son conleou ces demieres anntes dans Ie Dumero de decembrc.

Alec Paul Ediror/ Redaclellr en chef

28 Cfimatoiogical Bulletin I Bulletin Climalo\ogique 27(2), 1993

Towards Defining Appropriate Averaging Periods for Climate Normals

James R. Allgell 1V1lliam E. Easterling'­and SCOII W. Kirsch)

(Origina! manuscript received 21 April 1993; in revised form 27 July 1993]

A8STf{ACT

111cre is con.~idcrable debate com:eming the dcvdopmenl and use of clim3te normals for

planning. in part becau$C mlilly activities lu'IVC variable planning horizons. For example, the utility industry often needs heating degrcc-day (H DO) normals which art: cap<lbk of predicting short- term (1-5 years) annual, winter (OltCember-Fcbruary) and extended winter (October-April) HOD conditions. In II response to this need, historical climate dala from 41 stations in Illinois were used 10 determine the optimal lengths of averaging periods for normals for use in prc<licting nDO values for 1-5 years inlo the future and for !he subsequent 5-yc'J.r IIvemge value. We did Ihis by examining all averaglng periods up lo

30 years in length, plu,~ the World Meteorological Organi:wlion normal, which is a special kind of average. The optimal averaging period is arrived at through evaluation of each mean by four mtthods. Each of the cvaluatioo method~ produced very similar rankings of the averages. indicating that the results of the overall evaluation arc based on the data anti nOt 01) the method uscd in the evaluation. In general, ihC' I I-year mean is best at predicting HOD values I year imo the fUlll rc, while rneatl~ of aboul20 years ill length are best at predicting 2-5 years ahead and for thc subscquent 5-year average. It is not appropriate to immediately I::xlrnpoJatc these results to other climate variables ant.! other regions of the world. More work in Ihis area is clC""J.rly warr.mtOO.

Plusieurs discussiOnS ont pr6si!ntement court ~ur Ie perfectionnemcnt cl lu tilisation des normalcs climatiques pour In planilicalion d'activitcs, en partie parce que res demieres s'clentlem sur une pcriodc de lemps variable. Par exemplt, I\ndustrie des strvices a lrequcmmelll ~in des normales de degrts-jours de chauITage (DJq, qui pcuvcnt prevoir Its OJC hivemaux (decembre :i fevricr) Il mlUeI~ 11 court termc (I 115 ans) CL tics hi\ICrs prolong6 (OClobre it avril). Pour rtpondre a ce besoin, on a ctudi~ Its donn&:s climutiqucs c.Ic 41 stations mcu!orologiques ik l'lllinoi.~ pour caleuler les ¢riodcs rnoycntlcs optimales des normalcs el prevoir Its DJC des cinq prochaincs annees et la

I A1roosphetlc SciCitta Division. lUinni$: SI:'I<: WldCt SUr\l\.')', Ch.1mpaign. IL 1 Department or Agncul!umt MClt\)fology. UniyCrsit~ or Nebl1lskll\ Uneolu, NE l DepllTlmCnL of VCOl!fllphy llo(l t'lannin~. Memphis StAle Uniwrst ty. M...mIl/1is, TN

J. R. Angel, w.E. £a.werling & s.w. Kirsch I Appropriate Climate NOl"mals 29

valeur moyenne de cinq ans substquenlc. Pour cc faire, nous avons examine tou tes ks

ptriodes moyennes, a1lanl jusqu'a 30 annecs, en plus des nonnalcs de l'OrglllUsalion

meteorologique mondiale qui utilise une moyeooe speciale. Naus avons determine la periode moycnnc oplimale a partir de chaque moyenne calculee, scion qualIC methodes

differentes. Chaque methode II produil un rang scmbl8ble des moyenoes, ce qui indique que I'evaluation globa1c depend des donnees et non de la methode de calcul ulili...ee. En gcneral, la moyenne de II ans prevoil Ie mieux les DJC de I'anntt suivanlc, alors que [a

moycn.ne de 20 808 prevoil Ie mieux res valcurS des 2 a 5 pr6chaineli annees o:t de hI moyenne de cinq ans subs~qucntc. Toutcfois. il nest pas conseill€! di:xtrapoler immediatement ces rbultats II d'autres variables climatiques CI tI d'autrcs regions du globe. II est evident qu'on a besoin d'ctudicr ce domaine plus en profondeur.

I . INTRODUCTION

Determination of the optimal averaging period for climate normals is an open question in climate research, as is indicated by Ule work of several researchers in recenl years (e.g., Lamb and Changnon, 198 1; Dixon and S hulman, 1984; and Karl, 1988). Should averaging periods be selected in order to best capture the long-term character of weather for a period of, for t'.xample, 30 years, a.~ is the current World Meteorological Organization (W MO) practice (WMO, 1983)? Or should a short-term averaging period be used in order to best caplure the more immediate nature of, for example, thc next year 's weather, as was suggested by Court (1968) and Lamb and Changnon (198 1)? The answers to these questions are partially dependent on the intended usc of the results. Climate dependent economic activities have specific requirements for incorporating climate normals imo decision making. These TC4uircments vary widely between types of activities. Certain 3hrricultural decisions are based on the expected conditions of the next growing season, which may be best predicted by short-term averages. At the other extreme, decisions concerning large fixed capital investmcnts with long operating lifetimes, such as the sizing of water reservoirs, are based on the climate over several decades (White, 1980).

111e purpose of this paper is to explore Ihe question of the appropriate length of a climate avemge and the dependency of the length of the period on its inlended usc. As a vehicle in the exploration of these questions we usc the problem of determining the optimal normal for use by power companies in "weather normalization" of rcvenues and costs, an exercise many utiUtics undertake before seeking a rate change. The climate vmiable used most often by power companies in the weather normaJization process is the healing degtee-(jay total (HOD), which is thc variablc wc explore here. In the discussion that follows, we review relevant research, bricfly describe the information 11(,.'eds of the power industry, describe the data and methods used to assess the climate averages, and present the results of our analysis,

30 Oimatological Bulletin / Bulletin Climatologique 27(2), 1993

2. RELEVANT RESEARCH

Several authors have conducted research concerning cJirnate."normals" and how well these can predict fulli re conditions (e.g., CoUTt, 1968; Lamb and Changnoll , 1981; Dixon and Shulman, 1984; and Sabin and Shulman, 1985). Climate, until recently, has been considered stable, with only random variation around a mean value (Sabin and Shulman, 1985). This·average or mean value has come to be called the "nonna!. " This follows the traditional statistical concepts of centml tendency and implies that confidence in the mean increases with more observations (more years of data in this case). Therefore. stations with the longest records were thought to provide the best means. Court (1968) pointed out that thi.~ assumption has two faults: I) climate is now understood to vary across many time scales, which can cause the standard deviation of a climate measurement to increase through time rather than decrease; and 2) longer records are prolle to data inhomol:,>eneities eaused by changes in instrumentation, exposure, and time of observation (Mitchell, 1976; National Academy of Sciences, 1975; Karl and Riebsame, 1984; and Karl, 1988). To minimize these elIects, the World Meteorological Organization (WMO) adopted a 30-year mean that is updated every decade (WMO, 1967). Karl (1988) suggests that significant deeadal nuetuations arc often imbedded within the 30--ycar means which may limit their uscfulness.

Court (1968) argues that the best empirical test of reliability for climate means is their ability to predict future conditions. While Guttman (1989) states that climate normals are not appropriate for prediction, they will continue to be uscd until better forecasting methods become available. Using the Court ( 1968) premise several groups of rcscarchen; have tried to define the most appropriate averaging period for a normal. Lamb and Changnon (198 1) examined four stations in lIl inois and concluded that a 5-year running mean was best able to predict seasonal tcmperatures for the next year. D iXon and Shulman ( 1984), however. building on the Lamb and Changnon analysis, examined heating degrcc-<lays for six U.S. Slations and found that running means 10-30 years in length were bettcr able to predict the next year's HDD total.

Dixon and Shulman (1984) were critical of the method used by Lamb and Cbangnon ( 198 1) for two reasons. First, only selected means were used (5, 10, 15, 20, 25 and 30 years). They noted that in this approach , the individual best predictors are constrained to those means tested and therefore the selection and number of means influence the results. Second, although the 5-year mean was most freq uently the best predictor. it also produced the largest prediction elrol'S when it was not the best predictor. It is clear that these studies have called into question the appropriateness of 'using one average, usually the WMO 30-year norma l, for al l sit uations.

To date, no study has considered the problem of a normal's effectiveness a.~ a predictor for conditions more than one year into the future, a

J. R. Angel. WE. Easterling & s. W. Kirsch I Appropriate Climate Normals 31

capability required lor certain appUcations, such as utili ty weather normalizations. Additionally, little attention has been given to the spatial vdriability of optimal normals for various uses. These two problems are partially addresSl,'"(j herc in the context of HDD normals for thc mid-continent USA.

3, SI'EClrlCATION OF CLIMATE INFORMATION NEEDS OF UTILITIES

Utilities often usc climate normals to adjust ("weather normalize') sales revenues and costs to the Icvc1that would occur in a year of nonnal weather (called a test­yea I). all elsc being equal. These adjustments are viewed as necessary because spat't heating fuel needs, which make up a large part of most utilities' sales, are highly dependent on the severity of winter conditions as measured by a weather variable such as HDD's or averdge temperatures (Quayle and Diaz, 1980; Findlay and Spicer, 1988). The purpose oflhis process is to assure, ill the words of lhe Pennsylvania Public Utility Commission (1980), that "cxtraordinary occurrenccs such as abnormal weather conditions ... not be reflected in test­year levels." Abnormal weather conditions which affect sales, costs or profits are not a valid basis for seeking r<lIc increases. For example, it is not valid for a company to bruic a rate inCI-ease request on data for a year which had an unusually warm winter, which decreased consumption and sales revcnues. From the standpoint of the utility, decreases in revenues after normalization has been performed are likely to butJd a stronger supporting case for a rate change (Gillan, 1984).

It ha~ been pointed ou l that the most critical problem in weather normalizing sales revenues is determination of what constitutes "normal" weather in order to measure the dcgree of abnormality of actual weather conditions (New York Public Utility Commission, 1960; Gillan, 1984). Utili ties require a climate nonnal which accuratcly depicts the expected heating degree day total within their market area for periods ranging between one and five years inlo the future. Five years is the general maximum lifetime of a given set of rates in Illinois and is therefore the upper predictive limit used in this study (Illinois Commerce Commission, personal communication).

The traditional approach ill weather normalization has been to usc the 30-year mean, which is updated once each decade, of HDD's (hereafter called the WMO normal). Two potcntilll problems can occur in using the WMO normal to perform a weather normali7..ation, both relating to the now commonly accepted idea tltat climate varies on timcscales that are shorter and longer than 30 years. The first problem is that the WMO normals arc calculated only once evcry 10 years, at the beginning of a new decade. Thus it is quile possible that a hypothetical weather normalization performed in 1989 might have used the WMO normal constructed from \951-80 data. Because the uti lity and the rate

32 Climatological Bulletin I Bulletin Climalologiquc 27(2), 1993

commi~sion are both interested in projecting consumption and sales revenues based on "normal" weather, is it reasonable to use an avcrage which is up (0 9 years out of date and includes data from only the first year of the 1980's?

The second , perhaps morc important, problem is the question of whether a JO-year mean, which is updated once a decade, is the most appropriate normal to use in establishing what is the normal climate at any point in time. For example, the WMO normal for 1941-70, which would have been used in a 1970's normalization in Illinois, included the century's warmest temperatures and lowest H DO totals up to thut date and thus described a climatic regime different from the 1970's (Changnon, 1984). We hope to contribute to the answers to thesc two problems.

4 . DATA A.ND MnHOnOLOGY

In seeking the optimal averaging period for predicting 1-5 years into the future. we examine all possible averaging periods or normals (I to 30 year running means plus the WMO normal) by using a variety of evaluation techniques that are sen~itive to different aspects of prediction error. HOD's are accumulated for d ifferent periods within a year in order to explore dilferences in the results which may be related to the definition of the heating season. Specifically, we examined HDD accumulations for the whole year, winter (defined as December to February) and extended winter (defined as October to April).

The temperature data used in this study (which were converted to HDD'~ and summed over the periods discussed above) were obtained from the National Climatic Data Center (NCDC). Forty-one stations across Illinois were chosen to provide representative spatial coverage (Fig. 1). Data were obtained for the period 1901 -[984 (wilh lhe exception of Kankakee, where recording hcgan in 1917). Data for 1985-86 were obtained late in the study and were only used in constructing Figures I and 3. These 41 stations arc a subset of 61 higb quality stations discussed in Changnon (1979).

Four methods were ust.'(j to evaluate the predictive ability of the different averages: root-mcan·square..error (R MSE), mean-squarc--error (MSE). mean-absolute-error (MAE) and the "WINS" method (explained below). The results for each mcthod were ranked according to the predictive accuracy of the avemge (best to worst), and an assessment was made of the differences in the rankings from the four methods. An examination of the top-ranked means acrO$$ 1Ilinois lor each mcthod and for each prediction period ( J-5 years into the future, plus the 5 year average) provides a basis for choosing the best overall average.

JI?. Al1}!el, IV,E Easterlillg & S. I¥. Kirsch I AppropriOle Climate Norma/~' 33

FIGURE! Oinlllle Sl:lliOlU used;n this ~lUdy and iwlines nr mean ~nnuall l DO val~ (190 1-19&6).

34 Climalological Rullctin I Bulletin Climatologique 27(2), 1993

The form of these methods is as follows:

RMSE = [I / n I (6.x)2J' /2 ( I)

MSE == I/ n ! (.6.X)l (2)

MAE = I / nI I ~x l (3)

In each method, 6.X represents the difference between the mean HOD total and the HOD total for the year or period to which it was being compared, n represcnts the number of years over which comparisons were made. Because our longest mean is 30 yt:ars, our first comparison is for 193 1. The last comparison is for 1984, therefore n has a value of 54.

The fil131 method - the WINS method ~ is a modification of the method used by Lamb and Changnon (1981). It require,<; tabulating the number of times each mean has the lowest prediction error for a given year compared to all other means in a pair-wise fashion. This is done untit all possible unique pair­wise combinations of means have been examined for that year. It is referred to as the WfNS method because it is the Slim of the wins by each mean over all of the other means through time. This is similar to Slimming a baseball team's World Series championships (winning over all oUter teams each year) through time.

The I-JO year running means or normals, are first calculated for a base period (fot example, 18 years based on 1913-1930 data). This mean is then used in each of the four evalumion methods as the predictor of the HOD total for 1-5 years into the futu re (1931-1935), as well as the average for those five years. Then the base. period is shifted forward I year, the I K-year mean i..<;

recalculated (now based on 1914-31 data), and the mean is then used in each of the evaluation methods as the predictor of the next 5 years' (1932-36) HOD values, plus the sub~equcnt 5 year averdge value. This i~ done until we have moved Ulrough the entire period of 1931 to 1984. The WMO normal, unlike the olller averages, is only updated at the beginning of each decade and is then used over the next 10 years. Although the WMO did not adopt the 30-year mean until 1956, wc have retroactively calculated them back to 19JO (i.e. , the 1901-1 930 mean)(WMO, 1967).

5 . RI:SU 1..TS AND DISCUSSION

Background Climol%f(Y

A background climatology of Illinois HOD's is helpful in framing our analysi.'>. Figure I shows the avemge annual totals of HDD's for the period of 1901 to 1986. The time series plot (Fig. 2) for al1 sites averaged together reveals a downward trend in annual HDD's from 190 1 to the mid· 1930's. From the mid-

J. R. Allgel. W. E. Eas/erling & s. IV, Kirsch I Appropriu/e C/i"ltIte Normal..,· 35

Q 3 • ,;

f· ;<

~

~

> « " w w

" " w o

" z

~ w I

6000 "';"~"l·~:~~~~~~~~::'-.-p-o-cTh-,--------------r----..:-~-it-:--T-,-p-O-C-h-2-r--. ~'~'~-:.-'.~~~~~-'P-O-C-h~3-----------•• -;t;;;r:~f~;-: ----,

.~ ~ i~ ~ J:.~ We.' , ,

r"< ~~ :;~"'.,} ~~. ~ ~:~1 :<'d '{i);

~~ ~1, }~_t{ ~,,'" ilf' ,,,;;. :~ ,~1. ;;'.;:.~, .• ~ .... '. " - ;;:~ '8 ';; t\ ~ft 0' ~ :. ~~ ¥,:; ~\)

~: f:~ ~- ~ ~ f~~ ~-'~ 1;- ~ "1.:. k: ,- -.",'f.

~,~ {~ "to', _,,,,',i Y.~ *:§ k~ :;,.y.' ?kf} ~~

5300.L,.?_--"= __ ="-_-;-;:,;;;-..:;L--,;~_-:",,,,,-1·'!!J" --,;=_-:=,-_":"!,::·;~_-d 1900 1910 1920 '930 1940 1950 1960 1970 1980 1990

YEAR

5900

"

5800

FIGU RE 2 Smoothed, spatially a\~ annulll HO~ (olals through time for the 41 statiOIlS in the Study.

1930's to the mid-1950's, HOD's level off before beginning an upward trend to 1980. These Ihree distinct, visually discernable epochs have lengths of about 35. 20 and 25 years. The existcnce of these periods sum,oeslS that normals shorter than Ihe epoch's length would make better predictors. For example. a long-term average applied during the relatively warm 1950's would have incorporated the higher HDD values from the carlier, colder epochs. Consequently, it would over­estim3te the HOD values. A shorter averaging period would not incorporate as much of the earlier epoch and would bellcr predict the 1950's HOD regime.

Eva/ua/ion Method Resulfs

Spearman's rank order correlation analysis, a non-parametric lest, was used to assess how well the rankings of the averages by the four different methods match each other. Based on the 3 L possible positions, i.e., 31 observations, in the ranking (a rank of 1 being the "best" and 31 being the "worstj, the correlation matrix contains Spearman rank order correlation cocffieieOlS (rs) which arc all greater than or equal to 0.90 (all significant at a = 0.05) (Griffith a nd Amrhein, 1991). This suggests that the four methods produce very similar r:tnkings. Therefore the results of this study are attributable to the data and not to the methods used in quantifying the prediction error.

Due to tbe similarity between the rankings. only a representative snmple of the rankings produced using all of the methods will be shown here (Table I). Even though the r'dnkings in Table 1 are quite similar, the results for the 5-year ahead average H DO value illustrate !.bat the methods do not produce identical resultS. It is interesting 10 note that the very commonly used WMO normal does not perform well using any of the four evaluating methods.

Predictillg Annual HDD Totals One to Five tears Ahead

Given that the four evaluation mcthods produce similar results, and to save spat."C, we will di."cuss only the RMSE resulls from this point on. For predicting the next year ahead , the three top-ranked averaging periods are the 11-, 12-, and 10-year means, respectively (Table 2). llle next 10 values in the ranking are generally clustered uround the 20-year averaging period.

For predicting the second through ruth individual yelirs the 19- to 16-year means do best (Table 2). For predicting tJle average annuallOtal HOD's for thc subsequent five years following the end of the mean, the means ranging between 16 and 20 years in length arc ranked highest according to the RMSE method (Thble 2). Long (e.g., 30-year) and short (e.g., less than abou t 6 years) means all do relatively poorly. h can be concluded that an average of 11 years is best, according 10 the criteria outlined herein, for predicting the annual HDO totals for the next year ahead in Ill inois. For years 2-5 ahead, means between 19

J.R. Angel. W.E. Easterling & S. W. Kirsch I Appropriate Climate Normals 37

TAIlLF. , BesttQ worst rankmgs or the , - 10 3O-ytllr ml;llns and the WMQ nomla! in predicting the 5-yt:ar ahead ITICIn annual 1100 total , acc()(t/ing to the 4 !.:'1a)ualion methods-,

.... R~ISE MSiC MAE WINS

I (be$L) " " " " , " " " " ] 17 17

" 17

• ~ " , !II

] " " ,

" , " " 17 " , 10 10 '" " , " " " " , " " " • 10 , , " ,

" " " " " 12 " " " 10

" " " " " " " " " n

" II n " " " • , II " 17 ~ ~ 1 ~

" 1l 1l ,. " " " " " 1

'" 1 , " " " " " >1 WMO

" " " " " n " " " , l4 " JI) J<l " 2l , , , JI)

" ] ] ] " " • • • , " ] , J 4

" WMO , WMO , JO , WMO , ,

31 ("""fl;I) , ,

and 16 years are most appropriate, rcspechvely. And for predicling mean annual totals for the entire upcomi ng S·year period, an average of 18 or 19 years is best (supported by all of the methods, as shown in Tablc I).

PreJicJing WinJer and ExJellded Winter H DD Totals

The best predictor of the next winter's HOD total (Deccmber-February) is the 2O-year average (Table 3). The best predictor for years 2 through 5 and the 5· year average is also approximately tile 20-ycar mean (Table 3). A~ide from these differences, the winter season results are in general agreement with the annual results reported in the previous .section.

The resulls for the extended winter season (October to April) are also very similar to the annual results. The ! I·year mean is best for one year ahead predictions, while the 18· through 2O-year means are best for years 2·5 and for the 5-year average.

38 Climatological Bulletin I Bul letin Climatolowque 27(2), 1993

TABLE 2 Avcraging period, mnketl from hcstto worst, by the RMSE method, for predicting annual HOD's nnd the 5-year ahead mean.

I· )oar 2-)un 3-),,, ,,,, <",. ''''' ''''' ""' ... ~ ..... ...... .".ad ~,,~ ... ~"

llPt>;t) " " " " 16 18 , 12 18 " " " " , 10 " '" 16 " " • '" '" " 18 " '" , " " 10 '" " " 6 " 10 " " " 21 , " "

, , • 10

• 9 n 21 8 '" IS , 22 , • " " " 10 21 " " " 21 , " " " n 22 " " 12 " " " n ,

" " 2J " 23 " 2J " " 16 " " " " " Jj IS 24 " " 10 " " ,. '" " 10 " • " • " 12 , ,

" " " " " 12 " " " " " " " " '" '" " • , 26 JO , 21 ,

" " JO " " n " " 28 " 26 '" " " JO '" " '" " " JO , " '" " JO

" , • 6 6 6 6

" • , , , , , " G , WMO , , , "

, , , WMO WMO , "

, , J , , WMO

'" , WMO , , l l J)(w.mtl WMO , , ,

TABLE 3 Summary of the. lop ranked means, Ulling lhe RMSE method, for predicting 1-5 yca~ ahead annllal, winler, anU exten<Jed winlt;r HOD's and lbe 5-year llhead a~erngc UOD value ror lhe same periods.

Deambtr-F.btuuy Ortoon-April YH' "nlin: Y .. " (winlu) (.,.\fnoUd "inl..-I ,

" lQ " , " n " , 18 " " , " lO " , 16 " 18

s.yellJ 18 " " ~"

Temporal and Spatial VariabifilY of the H DD Pattern

A comparison of thc 1968-86 spatial pattern of HOD's in Ill inois with thc pattern over the state in the entire period covered by our data shows evidence of

J.R. Angel, WE. Easterling & S. W. Kirsch I Appropriar€ G imare Normals 39

a shift southward in lIle mean HDD pattcrn (Fig. 3). This illustrates that the spatial pattern of mean H DO's based on different averaging periods can vary. This figure also serves as further evidence that the mean for the entire period of record is not adequate for purposes which have short-tenn planning horizons (e.g .• utility weather normali1.ations).

Minimal spatial vadability in the pattern of the best predictive mcan for a I year ahead prediction is found across Ill inois (Fig. 4). The I I-year mean is the most frequently occurriog optimal meao for most areas of tbe state. When this is not true for a station the optimal means tend to be either clustered around I J years, or about 20 years. For 2 stations, Kankakee (optimal mean = WMQ normal) and Carbondale (optimal mean = ) years) this is notuue; we have no explanation for tllese two anomalIes.

6. SUMMI\RY I\NDCONCLUSIONS

In summary, daily temperature data from 41 long-tcrm stalions in Illinois were used to calculate yearly, traditional winter, and extended winter HOD values. Using these totals in conjunction with four evaJuative methods allowed us to arrive at optimal predictive H DD means, as defined herein, for our JlCliod of record and for our area. Running mcans from I to 30 ycars in length, plus the WMQ normaJ, were used to predict HDD totals for 1-5 years into thc fu ture, plus the future 5-year mean. 111e averaging period that had the minimum predictive error was considered to be the 'best' mean within the conte:<t of our investigation. Based on the results of the four eval uative methods, the means were milked from best to worst in terms of predictivc abil ity.

In general our results show that means o f between about 10 and 22 years are the best predictors of 1-5 year ahead, and S-year average, HOD conditions. The 11- through I)-year means seem to do best for the shorter term prediction ( I year ahead) anti the 16- through 20-year means seem best at longer term predictions (2-5 years ahead and for the 5-year ahead average). The WMO normal generally did not penorm well.

The differences between o ur work and that of Lamb and Changnon ( 1981) are due to several factors. We examined all averaging periods up to 30 years in length, plus the WMQ normal, while they examined only selected means (5, 10, IS, 20, 25 and 30 years). Also, our evaluation methods allowed us to keep track of the performance of all of the means through time., not just the top-ranked mean. Addit ionally, we examined HOD values while they investigated mean seasonal temperature. Our results are in general agreement with those of Dixon and Shulman (1984).

In a limited fashion, we examined thc spatial stability of the optimal avcrages across minois. Nineteen of the forty-one stations across Ill inois conform to our general conclusion regarding Ihe superiodty of the II-year mean for predicting I year ahead. Most of tile rest of the stations, which do not agree,

40 Climatological Bulletin I Bulletin Climatologique 27(2), 1993

6000 - 1901-86

5000 ~, 968-86

F1GtJRF. 3 Map oflhe 1968-19116. 19-year averab'C annual HOD patICm and the 1901-1986. 87-)'ClLf Iwcrage pattern showing thcmifi southW'~n1 in HDD IOUll~ when avernb~d 0\'Cf the =nl. shorter period.

J.R. Angel, f-Y.E. i:.tuter/ing & S. W. Kirsch I Appropriate Climate Norma!!,. 41

" .. " '"

" " ,. "

" 13

" '" . .. " "

" 2. ,. ,. "

" " " " " "

" •

" "

" " "

, "

"

FIGURE 4 Distribulion of the Lop ranked meuns (annual HOD "a1ues), usinglhc RMSI! method, for predicting I year ahead across Illinois.

42 Oimatological Bulletin I Bulletin Climatologique 27(2), 1993

have one of the other top-ranked means as their optimal mean (generally a mean of about 20 years),

It is not appropriate to immedilltely eKtmpolate these results to other climate variables and other regions of the world. We feel lbe value of this work rests on one major point, that in many ca~es it would be appropriate for other groups to repeat at least some portion of Ihis analysis fo r their situation so as to let them choose the best average or normal for their intended use. This work also suggestS that periodic reappmisal of the dala for an area is warranted because the length or the most appropriate normal may change through time.

ACKNOWLEDGMENTS"

We thank Peter Lamb, Stan Changnon and two anonymous reviewers ror their comments and su~'Cstions. We also thank John Brother aod Linda Riggin for drafting the figures. This research was supported by a grant from the Illinois Department of Energy and Natural Resources (STI LENRE R44WET HER.180).

REFER ENCrS

Court. A., 1968. Clinwtic Normall' as Prt.'tlielOrs: ParIS 1- V. Sci. Rept Air Forre Cambridge Research Lab., Bedford , MA, Contract AFI9(628)-5J76. (INTIS AD-657 ]58, AD-686 163, AD-672 268. AD-687 137, AD-687- 138).

Dixon, K.W. and M.D. Shulman, 1984. A Statistical Evaluation of the PrcdictiveAbilitics of Climatic AVl!ragc.~. Journal oj Applied Meleorology. 23. 1542-1552.

Easterling, w.E., J. R. Angel, and S.w. Kirsch, 1990. 71le Appropriate Use of Climalie Informalioll in 1IIinois NO/rlruf Gas UlililY Wealher NormafizOIiOlls. Report of Investigat iOll No. 112, llIinois Sta te Water Survey, Champaign, IL, 56pp.

Findlay, B.F. and L, Spiccr, 1988. Impact afClimate Wanning on Residential Consumption of Natwal Cas in Canada. ("/imill%gieal Buflf'lill. 22:2, 3-1:1.

Griffith, D.A. and C.G. Amrhcin, 1991. SUl/iSlfeal Analysi.·;jor Geographers. Prentice Hall, Englewood ClifTs, NJ.

Guttman. N. B .• 1989. Statistical Dc.~riplOrs of Climatc. Hul/erill of rhe American Mereor%gira/ Socie/y, 70, 602-607.

Karl, T.R. and WE. Ricbsame, 1984. The Identification of 10- 10 20- Year Temperature and Precipit.'ltion Fluctuntions in Ihe Contib'UOUS Uniled States. Journal oj

Glimare alld Applied Meteorology, 23, 950-966. Karl, T.R .. 1988. Multi-year Fluctuations ofTempcrature and Precipilation: The Gmy

Areu of Climate Change. Climate Change, 12, 179-197. Lamb. P..I . and S.A. Changnon, J r., 1981. On the ~Best" Thmperature and Precipitution

Nonnals: The Illinois Situation. Joumal of Applied Metcorology. 20. 1383-1390.

Mitchell, J.M., Jr. , 1976. An Overyiew of Oimatic Variability and Its Causal MechunisrrtS. QllalcnUlry FWsearch, 6, 48 1-493.

1 R. A ,,~el, WE. Easlerlin~ & S. ~¥. Kirsch I Appropriate Climate Normals 43

National Academy of Sciences, 1975. Ullderstamfing Climalic Chnnge: A Program/or Choll/(c. U.S. Committee for the Global Atmospheric Research Program, Washington, D.C., 239pp.

Quayle, R.O. and H.E Diaz. 19SO. Heating De1,'Tee Day Data Applied to R<!sidenlial Heating Energy Consumption. JOllrnal of Applied ''''Jeleorology, 19, 241-246.

Sabin, T.E. and Shulman, 1985. A Statistical Evaluation of the Efficiency of the Climatic Nanna! as a Predictor. Journal of Dimotology, 5, 63-77.

WMQ 1967, A NOle on Climtllologil'ol Normals. Gt.'fleva, Secretarial of the WMo. WMQ 1983. Guide 10 Climatological Practices. Geneva, Secretariat oflhe WMO. White, R.M., 1980. Anatomy of Climate Risk: in Climate Ilnd Risk: Proceedings of a

Conference sponwl"w by the Center for Advanced Engineering Study at M.I.T., vi, L, S. Pocinki, R.S. Greely alld L. Slater, Eds., The Mitre Corp., 130pp.

Climatological Bulletin J Bulletin Climatologique 27(2), 1993

Zonage du Risque Agroclimatique Durant la Saison Froide au Quebec Meridional: I - Froid Hivernal

PhifljJpe Rocheflel

Ol

Pierre-A ndre DuM2

lMa nuscri t re9u Ie 28 octobre 1992; en fonnc rCvisce Ie 10 juin 1993]

le but de cette etude est db.primer. a I'aide de variable<; c]imllliques, ta mCrtacc de domrnage !lUX planlcs perennes exer<.it! par Ie froid hivcmal el d'en dtlinir Ie palron de variation spaliale au Quebec meridional (C1nada). Les causes de dommugc idcnlifi&!s el tcs variables climatiques choisics pour les decrire ant etc: l'intcnsile du froid hivemaJ cxprimee par III temperature minimlile annuelle; l'iotcraction entre la durCe et l'intcnsitc du froid hivcrnul exprimee par la temperature minimale ;ournaliere moycrtnc du mois Ie plus froid; el I'uction du fmid sur Ics plan tes basses expnmoc par la difference, en jaup-;,

entre la duree de III pcriode au un froid iorerieur all egal a _15°C peut scvir et celie ou une couverture de neib'e est presente.

Les cartes de wnagc produitcs ont pernus d~dentifier ics gradients sp&tiaux de i'intensitc de la mCilIlCC excrete par chaque cause de dommal,>e. Elles pcrmcttent d\Ltil iscr l~nformation obtellue d'cxptricllccs en cOllditions contr61~ (inlcnsitc du [roid) pour detenniner Ie:; atre!; de rusticitc de certaines plantcs. A I'aidc de ccs cartes, des observations Sllr la survic tJ l'hiver faites en un point {)Cuvent etre gener'dlirees u l'ensemble de wne a laquelk l'endroi( apparticnl.

1\.11STRACT

'rhe purpose of this study is to determine the spatial pattern of the climatic risk of da!Th1f,>e to perennial plauts from their exposure to low wintCf air temperatures in southern QUi:bt!e, Canada. The causes of plant damab>e and the c1inmtie variables used to describe·

their intensity wcre: the cold inlcnsity expressed by the annual minimum temperature; the interaction betwecn cold intensity and duration Ci\.p~d by the mean daily minimum temperature of the coldest month; and the action of cold on short plants expre...sed by Ihe diffcrence, in days, between the length of the period in which an air temperature eqUnllO

or lower than -15°C can occur and the Il:ngth of the period of snow cover.

I untre de ('('ch<'ochcs sur Ic~ ttrN:.'; 1':t les r~u= biologiques, Direction de la nx:hcrchc.. Agricullure Canada, Oltaw:i (Ontario) KIA OC6.

2 Departernenl de phylolOllic, Faculte dCll scit:nccs de J'agrieulturc.ct de l'ulimt11lution, UnivcrsitC Laval, SaL1ttc-l'oy (Qucbt:c) (j I K 7P4

P. Rochette &P.-A. Dllbl! J Froid flivem ol au Quebec Meridional 45

The maps produced show u1e spatial gmdienls of the climatic risk for each call';e of

plant damage. They allowed usc of information gained in experiments under controlled environmental conditions (cold intensity) to define the hamincss wnes of certain planl~. 1l)ey also permit the genemli7.ation of observations on winter sUrviV',lI at one location to Ihe wholc woc to which Ihe site belongs.

INTRODUCTIoN

La rigucur de. l'hiver constitue une menace serieuse a la survie des plantes perennes au Quebec meridional (Andrews, 1987; Quammc, 1987). La plupart des regions agricoles de la province sont en e(fet soumises annuel1ement a des temperatures inf6rieures a -30 DC CI certaines d'entn! elles son! aussi cxposCcs a des pluies et des degels mverna ux frequents (Environncmcnt Canada, 1982).

L'action nuisibJe du climat hivernal sur les plantes perennes peut s'cxcrcer de plusieurs f~ons et la vulnerabilit6 des plantes par rapport a chacune d 'eHes peut differer (Sakai & Larcher, 1987). La variation spatialc de l'inteosile des causes climatiques de dommage prises une a une a pourtunt ete peu etudit!e et seules les zones dc rusticile des arbres et arbustcs a vocation ornemenutle Ont

ele defi nies dans Ie cad re d'une etude canadicnne (Ouellet & Sherk, 1967). Dcvant In complex ite du phenomene de In survie a 111iver,

Steponkus (1978) souligne la necessit! d'ooenter la recherche genetique visum a augmenter la rusticite des plantes agricoles perennes, vers I'elude de la resistance de ces plantes it chacun des stress auxquels d ies sont confrontees, plutot qu'a la somme de leur action respective. La determination de zones d 'intensite de menace par rapport a ehacunc des causes de dommage apparaJt done elre une information utile, sinon necessaire, a l'utilisation des resultats de la recherche en conditions controlees, One leUe etude de 2.0nage a done ete eotreprise da ns Ie but de completer I ~nformalion contenuc dans Ie zonagc de la rusticitc de Ouellet & Sherk (1%7).

L'etude a ete realisec en trois etapes. Les causes de dommages suivantes ont d 'ahord ete identifiees: les gels hatir:~ automnaux, les froids hivcrnaux depassam Ie niveau d'cndurcisscment de la plante, les gels tardifs printaniers. la prise des mcitles dans la glace et Ie d&:haussement Les variables climatiques exprimant la menace de dommagc aux pla ntes ¢rcnncs cxercecs par ces causes ont cnsuitc etc ehoisics ct, dans un dernicr temps, leur patron de variation spatiale a ete determine au Quebec meridional. Cel article presente les resultats relatifs au froid hivernal.

RF.VUF. Il f. un i'iRATuR I:

R esistance aI/froid hivernal

Lc gel du contenu cellulaire provoque presque toujours la mort de la cellule vegetale (Salisbury & Ross, 1985). Les causes probables de celie mort impliquent

46 Climatological Bulletin I Bulletin Climatologique 27(2), 1993

des dommages a la structure internt et aux membranes par la lonnation de cri~lau x. de glace (Levitt, 1980). La survie des organes vegetaux au froid hivernal repose done sur leur capacite a eviler Ie gel intr'dcclluJaire el a loierer les conditions internes creces par les m~canismcs de eet evilemcnt. Les mccarusmes d 'evilement ob!>CrvCs dans la nature sont la tolerance au gel cxtracctlulaire et la surfusion profonde du contcnu de la cellule.

Gel eXlraceJlulaire

Lc gel de !'eau dans les espaces exlraccllulaires engendrc un gradient de pression de vapeur d'eau qui provoquc la migration d'une partie de I'eau du protoplasme vcrs I'extericu r de la cellule, a la surface dcs crislaux, Si Ie refroidis.~ment est lem, Ie debit de la sortie de Peau entrrune un ab.'lissement du point de congelation du liquide intraccllulaire suffisamment mpide pour que Ie gel y soil evite (Sakai & Larcher, (987).

Les dommagcs causes par Ie froid hivernul aux tissus vegelaux ulilisanl ce mecanisme, peuvent eire engendres par la deshydratation de leurs ccllulcs (Levilt, 1980), par une alteration des [ipides des membranes ccllulaircs (Rajashekar et 01., 1979) et par les stress m&:aniques e.'\ercCs sur les parois par la pcJ1c de volume des cellules (Sakai & Larcher, 1987). Ccs boult:versements physiologiques ne senl cependant pas 10icres avec la rncme efficacite par toutes les pluntes qui y sont exposees. Ainsi, I'avoine nc resiste pas a des temperatures inlcrieurcs a environ _14°C (Gusta eta!' , J983), alars que les tiges de certains feuillus des fore ls boreules survivcnt a Icur immersion dans l'a'lOle liquidc (-196°C) (Sakai, 1965). Parmi les plantes agricolcs pCrennes, on observe un gradient chez le~ cereules d 'hiver ou Jc seigle est plus resis tant que Ic ble, suivi de l'orgc cl de I'avoine (Gusta et af. , 1983). Du Cole des plantes fourrageres, differents seuils de resistance sont aussi observes scion les e.~peces. Lcs Icgumineuses sont genemlement plus scnsibles quc Ics graminees, parmi lcsquelles Ie brOllle et la neole des pres sont Ics plus resistantes (Paquin, 1984; Gudleifs.~on ef 01., 1986: Sakai & Lal1;her, 1987).

Surjllsion profonde

Un liquide est en etat de surfusion lorsqu'il n'est pas cristallise a une temperature infericufe it son point de congelation. Une surfusion "legere" pennet a un liquide d 't':vitcr [a cristallisation a des temperatures ICgCrement SOlIS Ie point de congelation alors qU\lIlc surfusion "profondc" lui permet de demeurer liquide a des temperatures beaucoup plus basses. Chez certaines plantes, 1a surfusion profonde constituc Ie seu! mode de survie au froid hivernal pour Ics ccllulcs de certai ns tissus (Sakai & Larcher, 1987). La limite de la protection varie selon les csp<kes ci leur dcgre d'cndurcisscOlcnt, et peut altcindre jusqu 'a -47°C (Levin, 1980).

P. Rochelle & /?-A. DuM I Froid Hiverllololl Quebec Meridional 47

Pa rmi les organes ou tissus qui resistent aux gels hivernaux par la surfusion profonde, a ll notc Ics cellulcs dc rayons de parenchyme de plusieurs arbrcs fcuillus du nord-est de l'Amerique du Nord (Quamme el 01., 1972; George et al., 1974a; Gll.'i1a et Ill. , 1982), Ics bourgcons latents de la vigne (Quamme., 1986), les bourgeons noraux de I'azal&: el du rhododcndron (George et al. , J974b; Graham & Mullin, 1976) ct de nomhreux au lres arbres (lshikewa & Sakai, 1982).

L'exposition, mcmc rclativcmcnt breve, de ces parties de plante a des temperatures infcrieures a leur limite de surfusion enlrainc la mort des cellules par In crislallisatioll du Iiquide intraccllulaire. causaOl ainsi des hJessures plus ou mains graves a la plante (Quamme c( al., 1972; Ashworth el af., \983). Burke ct a1. (1976) COllcl ucnt que la mort des cellules de rayon de parenchyme, suite a In cristallisation de leur contenu, compromet les chan(~cs de survic des arbrcs qui en sont victimes en raison de In concordance entrc les a ires d'adaptarion de ccs derniers et les isothermes de la temperature minimale moyenne hivemale.

On re{rouvc asscz scuvenl Ics deux mecanismes d'evitement du gel intracelluJairc chC7_ differems lissus dune rneme plante. C'est Ie cas, entre a utres, dc plusieurs angiospermes ligneuses dOni les cellules des rayons de parenchyme surfusionnenl alors que cellcs du cortex tolercn! Ic gel e.x tracellulaire (George e(

al. , .I 974a). D'autres groupes de plantes perennes, comme les plantes herbacees, ne survivenl aux plus fro ldcs tcmperatures dc l'hiver de nos regions que grace a leur toJemnce au gel extracellulaire (Li, 1984).

RelatiOIl entre les dam mages et Ics conditiOl1s climatiques

En hiver, Ie froid sera fatal aux tissus vegetaux 51 son inlensite dcpassc leur degr<> d 'cndurcisscmcnt. Son dfet se manifestera cependant difTcremmcn t scion Ie mode d 'cvitement du gel inlr.\cellulairc utilise. En effct, I'observation de certains dommages est assoe-iee a des seuils de temperature relativement bien definis. Cht Ie eas des tissus qui scnt Ie siege de sunusiol1 profonde au la eristallisalion spontanee du eontenu eellulaire et In marl subsequente de In cellule se manifestent des que la temperature s'abaisse sous III li mitc maximalc de surfusion de leurs cellules (Quamme et al., 197 2; Ashworth et al. , 1983). On observe Ie meme phenomene pour les tissus dont les membranes cellulaires sont endommagecs el perdenl leur pcrmeabilite selective larsqu'une temperature critique est alteinte (Rajashekar el al. , 1979).

Chez certains tissus tolerant Ie gel extracellulaire et qui sont exposes it la deshydralation de leurs cellules, la duree du stress joue alors lin role plus important. II a ele observe que la luzerne pellt resister, lorsqu'adequatcmL"t1t endurcie. a une exposition de quelques heures a _26°C rnais qu'elle !o.wa detruite par des temperatures inferieures it + 10" C si Ie rroid dure pl usicUT!! jours (Paquin, 1984; Paquin el al., 1987). D'autres etude.~ ont permis d 'observer une reponse semblablc ehez les cereales d'hivcr (pomeroy el al. , 1975; Gust'l el ai., 1982).

48 Oimatological Bulletin I Hulk tin Climatologique 27(2), 1993

Aucune experience n 'a ete mentt.. a notre eonnaissance, pour cluJier 1'elTet dc l~nteractiOIl entre la duree et 11nlcnsitc du frciid sur Ie dommage cause nux tissus de plantes ligneuses tolerJ.nt Ie gel exlracelJulaire. 11 est ccpcndalll raiSODnablc· de poser 111ypothese qu 'j ls puissenl eux aussi etre endommage.~ a des scuils de temperature plus cleves lorsque la duree de I'exposition Rugmcntc.

Cctte breve revuc des m6canismes de resis tance des tissus vegttaux au froid hivcmal a permis de distinguer deux gtoupes principaux de planles. Le premier comprcnd les plantes dont Ics dommages resultent d'une exposition :\ une t(.:mpCraturc infelieure a un scuil bien defini. JI cst (;ompose principalcment des vegetuux doni certaines ccUules sont Ie siege de surfusion profonde. Le second regroupe les plantes dont les tissus vegetaux sont elldommages suite a une interaction entre l~ntcnsitC et la duree de leur exposition au froid hivernal. Ce sont ceux qui tolt!rent Ie gel extraceUulaire. Le choix des variables climatiques exprimant la menuce que rcprcsentc Ie froid hivernal sur ccs deux groupe.<; de planLCS ticndra comple de eette distinction.

MATtHtEL ET M!hHODHS

Le tcrritoire couvert par cette etude se limite it celui oU: se pratique I'ah'lieulture dans la province de Quebec (Canada).

Lcs observations mctcorologiqut.'S utilisees soot relies des temperdlures journalieres minimale el Illaximale mesurees SOllS abri. de memc que celle~ des precipitations SOlIS forme de neige et de pluic. L'c~tude visant a dctcnniner la variation spatiale des conditions climatiques auxquel1es soot soumiscs Ics plantes perennes a porle sur une pCriodc de I I aos (1972 a 1982). Les dossiers de plus de 200 stations du re.~eau c1imatologique du Mi.nistere de l'Environnemcnt du Quebec ont alars ete utilises (figure 1). l.cs variables climatiques 011.1 ete calculees pour ehaque saison froidc . L'annee c1imatologiquc concernant une sai.~on froide donnee s'etend du premier aoOt au 31 juillct dc J'annee suivante.

Vne analyse dc frcqucncc moycnne d'observation de tcmperatures minima hivernales depassam certains scuBs a egaJement etc Ca.ite. Scules les stations ayant un dossier d'au moins 20 nntleeS completes sur la periode 1950-19K2 ont etc rctenu~ pour cette etude.

Jours t/'ellllcigemefli

Ouellet (1969) a propose plusieurs Olodeles de regressions multiples de variables climatiques pour cstimer Ie nombre annuel moyen de jours ou une couche de ndge d'au moins 2,5 centimetres rccouvrc Ie sol (JeN) dans la moiti~ orientale du Canada. Nous avons retenu I'equation #1, pour raquelle les variables d'entree neccssaircs a SOn utilisation etaient disponibles a un plus grand nombre de

P. Rochelle &. P-A. /)ube I Froid llivema/ all Quebec Meridional 49

...

'.

".

•

ONTARIO

".

'"

" , . .,

. , • •

...

ETATS· UNIS

roo

NOUVEAU­BRUNSWICK

o 50 100 ISO ",..,

FIGURE I Slatiorl.'\ meteorologiques utilist:cs cl territoire couVer l par ('etude des conditions agrociimRliques de In saisan froidc.

".

50 Climatological BulLetin I Bulletin Climatologique 27(2), J 993

stations. L'ajustcment des observations utilisCcs par cette droite de regression a ete carrtcterise par un eeart~lype de I'crreur dc 11 ,6 jours el par un eoefficent de correlation de 0,972. ElJc s'cxplime comml..' suit:

.leN ~ 284,5-2, 111 TMN + 0,2135 N - 2,180TMM - 10,2 I'M -0,1547 SSGO II)

Oll: TMN CI TM M sont Ics temperatures maximah..-s moyennes des rnois de n()vembre et de mars (0 F);

Nest la hau teur rnoyenne de la chute dc neige annuelle (pouees); PM cst la hauteur moyenne de pillic tQmbl!e en mars (pOliCes); SSGO est la longueur rnoyenne, en jours, de la saisan sans gel (OC C).

La cartographie II etc teali&:-e a I'aide du logieiel "SYMAP" (Dougenik & Sheenan, 1975).

Rt.S UJ;JATS ET DISCUSS ION

En raison des reponses difTercnlcs des tissus vegetaux aux froids de l'hiver selon It;:uJ Illude d'evitcmcul du gel ili tlacdlulaiJe et sciOli leUl scllsibilile relative aux changements physiologiques engendJts, nous avons retenll deux variables climatiques pour exprimer l'intensile de la menace de dommagc aux piantcs cxeltte par Ie froid hivernal: "" a temperature minimale annuelle" (TMA) et "Ill temperature minimale journaliere du mois Ie plus froid" (TMMF).

TemperQ/ure Mil/imale Amwelle (TMA).

Sakai & Weiser (1973), Gcorgeel 01. (1974a) et Quamme (1976) ont deja mis en c.vidcnce In coincidcnce entre les isothermes de TMA et les aires de distribUlion de plusicurs plantes dont lcs organes les plus sensibles all fro id uti1iscn! la $urfusion profonde. NOllS avons done choisi la meme variable climatiquc pour cxprimcr I'effet de ['intensite du froid hivcrnal sur lcs ti~sus vegetaux dont 111 biesslirc cst a.~sociee a une temperature sellil rclativement bien definie. Sa variation spatiale dans Ic tcrritoirc sous etude est presClltee ;\ la figure 2.

Lcs temperatures minima1es extremes les plus froides sont obscr­vees au Lac St~Jean , en Abil ibi~TCmis(;aminglle r l dans Ics Laurentides au nord de Montreal. A l'oppose, la rebtJon connaissant Ics pointes de froid les moins prononcees est celie dc la c6te gaspesienne et du golfe St~Laurent jusqu'au sud de Rimousk..i. On peut aussi y eonstater que la region de Quebec appartient a la meme classe que Montreal bien qu'elle soit situ(.-c plus au nord el qu'eUc en soit scparec par line zone plus froide. L'infiuencc maritime excrcCe par Ie fJcuve qui s'c.largil :\. cel end roil est probablement rcsponsable des conditions J'(:lutivement demenles observees Ii Quebec.

P. Rochctle & P'-A . Dt/be I Froid Hivemal all Qllebec Meridiollal 51

ONTARIO

'" ,00 '''' km

ETATS·UNIS

is" ,,- ",.

zone

2

3

4

5

NOUVEAU· BRUNSWICK

menace

t- Iaible -26.6 - -29.6 • -29,7 - -32.6

-32.7 - -35.6

-35.7- -38 6

• + lorte -38.7 - -45.3

FIGURE 2 Variation spatiale de 11ntensite du froid hivernal au Quebec meridional (1972-19K2). L l ntensiLe relative de I. menace excrete par IIntensite du froid hivernal est ""primee par 1. temperature minimale .nnuclle (TMA).

52 Climatological Bulletin / Bulletin Climatologique 27(2), 1993

TABLEAU 1 t- rtqucncc d'observalitHl de IcmperalUfC.'l minimalcs anoudlc! infcrieufC.'l iI cerlaios seuils pour hi period .. 1950- 19112 (%).

NU KE CC) ZON!! S,TAT. ." ·20 ·n ·M ·26 ." • )0 ." • J< .,. .38 ~ . ~, ... ... ... .", , • '00 " " " 7) " "

, ; , • " 0 " 0 0 0 , ,. '00 "" "'" '" " " " " '" , , , 0 " 0 0 " J " '00 '00 '00 '00 " ,.

" " " " " ; 0 0 0 0

• 17 '00 '00 "" '00 '00 '00 " " 80 " " 10 • 0 0 " 0 J ,. '00 '00 '00 '00 '00 '00 '00 '00 " " " "' " " • , 0

De nombrcult tr..tvaux ont ete mene.'t pour estimer la limite de la resistance des organes des plantes pCrcnncs it unc breve exposition au froid (Pellet ef ul., 198 1; Sakai, 1982: Sakai & Larcher, 1987). Leurs resuitats, ajoutes ala connruSAAnl.:e des frtqucnces d'observation de temperatures inferieurcs a certains seuils (tableau I), IlOUS permeltent de conna'itre les eham:es de Slirvie des organes de res plantes dans les dilTcrcntes zones definies par Je.'t conditions moycn nes.

Des observations nous indiquent que Ics bourgeons flomux sont les organcs les moins resistants des rhododendrons (Sakai, 1982). Chez les especes etudit'CS les plus I'ustiques, la limite de leur resistance est d'environ -30°C. L'cxamen des frequcnces des temperatures minimales annuclles indique que, memc en zone I, ces bourgcons seraient tues un hiver surcing. Panni les organes vegetatifs du rhododendron. Ie xyleme est In partie la moins rtsistallte. II est cndommagc a 40" C chez les espCccs les plus rustiques, indiqualll ainsi que ccs derrneres pourraient survivrc aux froids sevissant dans la zone 3. Leur nornison y scrait cependant compromise par Ie gel frequent des bourgeons.

Chez. les especcs les plus rustiques du genre Rosa, Ie xyleme est la partie la plus scnsible au froid et peut resister a une tempef"dlure de -4()<' C a -45° C (Sakai, 1982; Sakai & Larchcr, 1987). EI1es peuvent done survivre a une exposition aux fro ids hivcrnaux de la zone 4, mais la protection d'une couche de neige leur est necessaire en zone 5. Lcs cspcces cultivees a des fins ornementales sont beaucoup plus sensibles el suecombcnt lorsque la tempCr.lture s'abaissc sous -25°C. II est done IIcccssaire de proteger ces planleS de.'t froids hivernaux pour en assurer la survie dans la majeure partie de la province.

Quamme (1976) a eludie la relation entre la temperature annuelle nunimale moyennc et la repartition de la production commerciale de pommes et de poires dans Ie sud de l'Ontario. Lcs ccllules des rayons de parenchyme du xyh:'!me de-res arbres fruiticrs survivent a l'hiver par surfusion profonde. Leur limite d'endurcissemenl est comprise entre -37 ct 4()OC pour Ie pommier ct entre ·33 el ·39"C pour Ie poirier.

Quamme ( 1976) n'a observe aucune production commcrciale de pommcs dans les regions au In temperature rnirUmale anlluelle est infericure a ·34,4°C. Au Quebec, cene limite correspond environ a celie entre les zones

P. Rochette & P.·A. Dube I Froid Hivernal llll Quebec Mlm'r/i{mai 53

3 ct 4 de la figu re 2. L'examen de [a frequcnce des temperatures minimalcs (tahleau I) monlre que les probabilites d 'observer une temperature inferieure a -38° C dans les :wne.~ 5, 4 el 3 sont respectivement de 82%, 36% el 11% alors qu'cHes sont de 600/0, 10% et 3% pour un seuil de -400C. On constatc done qu 'une temperature annuclle minimalc moyenne d'environ _35°C correspond a des fn':quences eJevCes de dommagc au xyleme des pommier'S qui expliqueraient I'absence de production commcrciale de pommes dans les seclellrs qui y sonl exposes.

Dans la meme etude (Quamme, 1976), la production commerciale de poires a ele assodee aux l-egions ou la temperature minimale annuelle moycnnc cst comprise enlre _23,3°C et ·28.9°C. Cc Icrritoirc correspond environ ala zooe I de la figure 2. Dans ceUe zone, les frequeoces d'observations de temperatures infcrieures ou egales it -32°C, -34°C et -36°C soot rcspcctivement de 6%, 3% el 1%. Le froid hivernal [I 'y est done pas U11 critere limilant la production de poircs. Malgre la sUivic des poiriers a l"iYcr de la zone I, la production eommerciale de poire.~ n'e..~t cependanl pas possible sur la eote gaspesienne en raison de la fraicheur des temperatures estivales qU'on y enregislfe.

Tempera/urI'. Minimale Joumaliere dll Mois Ie plus Froid ( TMM F).

La menace nee a J'interac(i()n entre la dum et I ~ntensite du froid hivemal pom les organes tolcmnt Ic gel extracdlulairc a etc exprimt'C par TM M F. Le patron de variation spatiale de ceUe variable est presentc a la figure 3.

Les regions ou 1a menace est la plus fo rte son! celles de l'Abitibi­TemisC3minguc, du lac 5t-Jean ainsi que des haules terres des laun:nlidc~ et des Appalaches au nord--est de la rivicrc Chaudierc. A I'oppose, la vallee du SI­Laurent en amant dn lac SI-Pierre, les Cantons de rEst, la rive sud du St­La urent en aval de Quebec et la cote Gaspesienne sont celJes ou Ie dommugc a~soc-ie it celte cause c\imalique est Ie moins probable.

Ouellet et Sherk ( 1967) ont observe que plus de SOOtb de la variation de la fn!quence de mortalitc hivcrnale des arbrcs ct urbustcs ornc.JncnlllUX au Canada pouvait etre cxpliqucc par la TMMF. La rcsscmblancc entre la figure 3 et Ie zonage de Ouellet et Sherk ctail donc previsible et suggere que I ~nrormation presentee par ce demier est pl us pertinente aux tisslIs vegetaux toU:rant [e gel extracell ulaire que ccux. ou ;1 y a surfusion profonde.

Couvertlae de Ileige

L'analysc des Icmper'dtures minimales hivem a[es sevissant au Quebec monlre qU'el1es sont, dans toules lcs regions. inlcricures aux limites de la resis tance au froid de la plupart de.~ plantes herbacees agrieoles (Sakai & Larcher, 1987). La survie de ee..~ dernieres est donc. dans une large part, lite it ['efIicacite avec

54 Climalological Bulletin I Bulletin Climatologique 27(2), [993

~ '" '" ~

u. W " '" ci W ~- ':' ':' ~V I I I I ~- <D '" '"

.., '" .; w ~ ~ ;;

. l< "'U ~§ ~ .. ~ ><n y

~ ,Q " "'z c

'" 1'£

0'" ~ + + za: m

~ '"

.., ., '" ~

'" ..

g

FIGURE 3 Variation spaLiate de l'interaction entre la duree et l'intcnsite du froid hivemal au Quebec meridional (1972-1982). L'intensite relative de 1a menace exercee par l'interacllon entre la duree et i'inlensite du froid hivemal est exprimee par la lemperature journaliere minimale du mois Ie plus froid (TM M F).

R Rochelle & R-A. DuM I Froid Hivernal au Quebec Meridional 55

laljuclle la ncige les isolem des basses temperatures de !'air lImbiant. C'est Ie cas des espece.~ du genre Rosa (R:~ashekar & Burke, 1978) chez les plantcs ornement:des, de la fraise (Boyce, 1985) pour les cuhlJrcs fruitieres ct de la luzerne (Smith, 1964) et du ble d'hiver (Fowlcr et ul., 1976) pour les grandes cultures.

La variation spatiale du nombre de jours de eouverture de neigc, caleulee selon I'equalion ( I) de Ouellct (1969), est illustrCc ilia figure 4. lc nombre moyen de jours d 'enneigemcnt varie de 85 a 168 jours entrc 1a plaine de Montreal, ou iJ est Ie plus faible, ct ['Abitibi-Temiscamingue et I'ouest du lac St­Jean, ou il ~t Ie plus important.. En general, on note egalcrncnt une relation netle entre la duree de I'enneigement et In latitude.

La survie a l'hiver d'UJl grand nombre de plantes herbacCcs ct de certaincs plantcs ligneuses basses est done lice a In presence dune eouche de neigc au sol. La durie de la pres~ncc d'une couverturc de neige n'est ccpr..'tldant pas la stule variable impli4u~e. La longueur de la periodc durant laquclle Ie froid hivernal eOllstitue une menace doit elle aussi eire eonsiderCc. C'est pourquoi nous avons retenu la difference entre "Ia duree moycone, en jours, de In periode s'Ccoulant du premier au demier jour au la temperature mininmlc journaliere etai! plus petite ou egalc a _15°C duranlla saison froidc"ct"le nombre moyen de jours d'enneigement" (DNF) comme variablc supph~mentairc pour exprimcr Ie degre de sCverite avec lequelles froids hivernaux agisscnt ricllemcnt sur ce groupe de plantes.

Le patron dc variation spatiale de DNF cst presente a la figure 5. Si nous posons I'lypothese quc lcs deux rerindes considen!es sont, en moyenne, ccnu-ees sur la meme dale, la difference obtenue cxprime alors Ie nombre moyen de jours ou les plantes risquenl d'elTC exposees directement a de basses tcmpenttures. On eonslate (figure 5) que la difference esl positive dans la majeure partie du territoire eludit. La periode de couverture de neige cxcede celle ou Ie froid sevit de 14 a 52 jours dans les regions de Quebec, du Bas*Saint-Laurcnt et de la Gaspi=sie. On peut en dCduire que les plumes herbacCeS y som, en general bien protegees el que Ie froid ne rcpreseJlle Ime Ci.lUse importantc de dommagc qu'en certaincs annees exccptionnelles.

II en va tout :lutremcnl pour la plainc de Montreal ct la vallee de l'Outaouais ou des froids infcrieurs ou egaux a-15° C peuven! :revir en moyenne de quatee a 13 jours en nbscnce de couverture de ncigc. Cetle region cst celie ou la survie des pianles hcrbactes aux froids de l'hivcr esl la moins probable scion les resultats de notre elude. La zone 2 rcpresente, elle aussi, une regioo relativemcnt peu favorable si on considere. qu'une ann~ sur deux, elle presente des risque.~ d'cxrosition dc.~ rlanlcs basses a des lemperatures pOlcnliellement dommageablcs.

La consideration de la periode 01\ Ie froid ...evil n'a pas modifie Ie palron de 1a variation spaliaic relative, dcfini par Ie nombre moyen de jours de couverture de neige, sur la rivc sud du fleuve en amont de Quebec. Elle a

56 Climatological Bulletin I Bulletin Climatologiquc 27(2), 1993

ONTARIO

50 100 1SO kin

ETATS-UNIS

NOUVEAU­BRUNSWICK

"""

2

3

5

JCN (jours)

64 - 101

102-117

11.8-134

135 -150

151 - 168

FIGURE 4 Variation spatiaJe du nombre dejour1i de couverlurc de ncige au Quebec meridional (1912- 1982).

P. Rochette & P'-A. DuM I Froid Hivernal au Quebec Meridional 57

•

,.""

1 2 ONTARIO

3

50 100 1SO ~

ETATS-UNIS

lB'

NOUVEAU~ BRUNSWICK

menace

.Iaible

I ,.Ione

ON' (jours)

-12.5 - ·3.5

·3.4 - 5.4

5.5 - 14.4

14.5- 23.3

23.4 - 52.1

FIGURE 5 Variation spatialc de I'aetion du froid hivemal sur les plantes basses au Quebec meridional ( 1972-1982). t'lntensitc relative de la menace cxcrcee par I'action du froid sur los plantes basses cst cxprimee par la di(fcrcm:c, en jour'S. entrc la durCc de la pcriode ou un froid infcriclir ou ega! :\ -ISoC pcut sovir ct Ie nombre annuel de jours de couverture de ncige (DNF).

58 Climatological Bulletin / BuUetin Climatologiquc 27(2), 1993

cependant mis en relief des conditions relativcmcnt plus severes dans I'Outaouais. I'Abitibi-Temiscamingue, Ie lac St-Jean et les Laurentides au nord de Montreal. Par conlrc, SUI la rive sud du Oeuve en avaJ de Quebec, les conditions sont devenucs rclativcment plus c\cmentes.

CONCLlfS10 NS

CeUe elude repose sur l'hypothese quc les meeanismes de i'Csistance des vegetaux au froid sont suffisanunent bien conous pour qu'on puisse identifier In au les variables climatiques qui e.xpriment 1a menace c.:l(ertee par Ie froid hivcrnal sur les plames. Lcs r~sultats de recherche pubHes dans cc domaine indiquent qu~l cxistc unc relation direete entre,

I) lcs dommages aux cellules qui sonl Ie siCb'C de surfusion profondc ct la temperature minimalc annuellc;

2) la " ,,,;e de plus;e"" plantes basses" la p"'"noe dune ,ouch, d, "d!\".

Les effets de I~nleraelion cntre la durc'e et I~ntensite du froid sur la survie des tissus Oll s'observe Ie gel cxtracellulairc ont ete moins bien etudics, particulierement chez.lcs plantcs ligneuses. A III lumiere des eonnuissanccs actuelles, la tempCf"dture journaliere minimuJc moyenne du mois Ie plus froid notls appan,it loutcfois la variable climaliquc la plus adequate pour cxprimer ['cffct du froid hivernall\ur ce groupe de plantes.

La carte montrant la variation spatiale de la temperat ure annuelle minimale permet de delinir .l'aire de survie t\ 111iver des tis..~us des plant.cs qui resistent allx temper.ltures hivernales par surfusion profonde et dont la limite d'cndureissement est eonnue. Celie montrant Ie nombrc de jOllfS durant lcsquels Ie froid scvit en absence de eouverture de neige dcvrait reprt':senter asscz fidcicmcnt les regions Oll peuvent survivrc les plantes herbaeccs et les plantes lignetL~e.~ basses de fa ible rusticitc. La eomplcxite de la relation entre Ie climal et Ie dommage aux. plantcs nous crnpeche de fa.irc unc interpretation dirccte des 7.Onagcs concernant la tolerance au b'Cl cxlracellulaire ct de la perle d'endurcis~ent. Ces derniers prCscntent cependant les difTerellces spatialcs de l'intensile relative de la menace associce :\ ees causes de dommage. lis permetlent done d'idenlifier les regions ou Ie risque de dommngc cst Ie plus eleve et de generaliser des observations faites en point it I'ensemble de la zone a laquelle cct endroit appartient.

REM ERCIEM ENTS

Cene etude a cte assistee finaneicrcmenl par Ie Ministere de l'Agrieuliurc, des Pecheries et de l'Alimentation du Quebec. Les autcurs aimeraient remercier M.

f. Rochelle & f.-A. Dube I Froid Hivernal (IU Qllebec Meridional 59

Oliva Couture pOUt son assis tance dans Ie traitemcnt des donnees et Ie trace

original des cartes e l M . Etienne Girard pour la realisation final e des cartes.

Andrews, CJ . 1987. Low-temperat urc stress in field and forage crop production: An

overview. -Can. J Plant Sci., 67: 11 21-1 1)3.

Ashworth , E.N., o.J Rowse & LA BiUmyer. [983. T hc freel.jng of water in WOOdy

tissues of apricot and peach and the relationship to frccling injury. -Am.

Soc. Hart. Sci .. 108(2):299-303. Royce, B.R. 1985. Cold 3ccljmatations and winter injury in stn,wbcrry phi lliS. -Cahicr de

comt\rcnces, SymPQSium sur III culture dc In fr'dise, 14 et 15 Nov., Univ. Laval, Que .. pp. 135-1 42.

Burke, M.J .. I..V. Custa, H.A Quamme. C.J. Weiser & P.H. Li. 1976. Freeling and inju ry in planll;. -Ann. Rev. Plant Physio!.. 27:507-528.

Dougenik, J A. & D.E. Sheenan. 1975. Symap uscr'S manuaL Harvard Univ., Grad.

School Design, Cambridge, Mass., 65 pp ..

Environnement Canada. 1982. Nomlilies climatiques au Canada, 1951-1980. Vol.1-8, Serv. F. llv. Alm_, Envil'On. Can .. Ottawa.

Fowler, D.B., L. V Gustn, K.E. l1owren, E.n Mallollgh, a s. Ikan & R.N. Mciver. [976.

Potential for wheat production in Saskatchewan. -Can. J. Plant Sci., 56:45-50.

George. M.E, M,J . BUrkc, H.M . Pellet & A.G. J ohnson. 1974a. Low temperature e:< otherm~ and woody pl ~nt distribution. -f-'l oTtsdem:e, 9:5 19-522.

George, M.E. MJ. Burke & c.J Weiser. 1974b. Supercooling in overlVintering azalea

flower huds. -Plant PhysioL, 54:29-35.

Graham, P.R . & R. Mullin. 1976. T he determination of lethal free?i ng temperuturtS in

buds lind stems of deciduous azalea by II freeling curve methot!. -1. Am. Soc. Hortic. Sci., 101 :3-7.

Gud leifsson, B.E" CJ. Andn:ws & H. Bjornsson. 1 9~6. Cold hardiness and ice tolerance of pasture grasses grown and testcd in controlled environments. -Can. J. Plant Sci., 66:601·608.

Gusta, L. V., 0.1:\. Fowler & N. J Tyler. 1982. factors influencing hardening and survival in winter wheat. Ill : Plallt cold hardiness and free~jng stress, Li, P.H. & A

Sakai (Ms.), Vol. 11, Academic Press, Lo nd rcs et New York. IIp. 23-40.

(justa, L.V., N.J. Tyler & T. Hwei.Hwang Chen. 1983. Deep undercooling in woody taxa

growing norlh of the -40"C isotherm. ·Plant PhysioL, 72: 122- 128. Ishikewa, M. & A Sakai. 1982. Characteristics offreering avoidance in comparison with

freezing tolerance: a demonstmtion of extraorgnn (tee-ling. In: Plant cold

hard illcss a nd freez.inS stress, Vol II , Li. I~ H . & A. Sakai (Eds), Academic Press. Lond rcs Cl New York, pp. 325-340.

Lcvitt, J 1980. Re.~ponse.~ of plants to environmentll! ~tresseS. Vol. I. Chilling, rrcc~jng and high temperatu re stresses, 2""'" Ed, Academic Press, London et New York, 497 pp.

60 QimaloJogicaJ Bulletin / Bulletin Climatologiquc 27(2). 1993

Li, P,H. 19~4. Subzero temperature stress physiology of herbaceous phillIS. -Hort. Rev" 6:373:417,

Lindsey, A.A, & 1 E. Newman. 1956. Usc of official \\lCather d:lln in spring time temperature analysis of an Indiana phcnological rteord. -Eeo!., 37:812-823.

Ouellct, C E. 1969. Estimation dll nombrc dejours d'cnneigement par des variables macroclimntiqucs dans In moilie ori(;n t:lle du Canada. ·Naturalistc can. (Rev. Eco!. Syst.) 96:637-650.

Ouellet, CE. & L.C Sherk. 1967. Woody ornemental plant zonation. 3. Suitability map for the probable winler survival of ornamental trl'CS lind shmbs. -Call . J. Plant Sci., 47:35 1-358.

Paquin, R. 1984. Innuence of environment on cold hardening and winter surviV'.d of forage plants and cereals; proline as a metabolic marker of hardening. III: Being alive on land. N.S, Margaris, M, Arianoustou-Faraggitaki & W.e. Ocche1. Ms, Pub. KluWl:r Academic Publisher.! Group. Roston. pp. 137-154.

Paquin, R. 1985. Survic t\ I'hiver de.~ plantcs fourmb~res I!t des c~reales sous ks climats nordiques, en particulicr 1I11 Qll~hcc: progrbl et PCnipt"ctivt:S. -

PhYlUprOleetion, 66: 105-139. Paquin. R.o M . .Befnier-Cardou & Y. Ca.~tongiJay, 1987. InOuence de Ibumiditc du sol, de.

b lem¢rature et de la duree du gel sur la survic de Iii IU7.erne. -Can. 1. Planl Sci .. 67:765-n5.

Pellct, H., M. Guarhart ct M. Diaf. 198 1. Cold hardiness capability of wo<H.Iy orncmenllli plant taxa. J. Am. Soc. Hort. Sci,. 106:239-243.

Pomeroy, M.K., C.J. Andrews & G. Pedal-.. 1975. Cold hardening and dchardening responses in winter wheat and winter barley. -Can. 1. Plant Sci., 55:529-535.

Quamme, H.A. 1976. Relationship of the low temperature cxothcrm 10 apple and pear production in North America. -Can. 1. P!emt Sci., 56:493-500.

Quamme, H.A. 1986. Use of thermal anulysis to mt:asure free ling resistance. of grape buds. -Can. J. Plant Sci., 66:945-952.

Quamme. 1l.A. 1987. Low·temperaturc slress in cannuian horticultural production: An ovtfview. ·Om, 1. Plant Sci .. 67: I 135- 11 49.

Quamme, H.A., C. SwshnolT & c..I. Weiser. 1972, The relationShip of e.>;olhcmlS 10 cold injury in apple stcrn tissues. -J. Am. Soc. Hor!. Sci. , 97(5):608-6ll

Rajashckar, C. & Mj, Burke. 1978. TIle occurrence of deep undereooling in the b,'CtlW\ PJ'rt/j', ' 'l'WI/IS ct Rosa: A preliminary report. Ill: Plunt cold hardiness and free1.ing slresses. Vol II , Li P.H . & A. Sakai CUds), Academic Press, Londres et New York, pp. 213-225.

Rajashekar, C, L.Y. QUSfa & M.J. Burke. 1979_ Membrane structure transition: Prob.1ble relation to frost damabocin hardy herbaceous species. 1;/: LoY( temperature stress in crop plants: the role orlhe mL'fllbrane, Lyon, J. M., D. Gmham & .I.K, Raison (Eds), Acndemic Press, Londres ~t New York, pp. 255-274.

Sakai, A. 1965. Survival of plaut tissue at supr:r-Iow tcmpcraturdl, Il \. Rehltion between dfcetive pre. f~ng temperatures and degree of frost hardiness. -Plant PhysioL 40:882-887.

P. Rochelle & Po-A. Dube I Froid Hivemal au Que.bec MeridiOllll1 61

Sakai, A. 1982. Fr'CC'(.ing resistance of ornamental trees and shrubs. -J. Am. Soc. Horl., !07:572-SRt.

Sakai, A. & W: L.1rchcl:- 1987. Frost survival of plants. Springer-Verlag, Berl in, Hddelberg et New York, 321 pp ..

Sakai, 1\. & C.l Weiser. 197). Preeling resistance of In.:es in North America with reference to tree regions. -Ecology, 54: I 18-126.

Salisbury, EB. & c. w. Ros~. 1985. Plant physiology. 3~me M., Wadsworth Publ. , Belmont, Ca., 540 pp ..

Smith, D. 1%4. Winler injury and the survival of forage plants. -Herbage I\bstr., 34(4):203-209.

Steponkus, P.L. 1918. Cold hardiness and freezing injury of agronomic crops. -Adv. Agron., 30:5 1-98.

62 Climatological Bulletin I Bullelin Climatologique 27(2), 1993

Trends of Daily Maximum and Minimum Temperature in Canada During the Past Century

W R. Skinner! and D. W. Gulleu!

[Original manuscript received 23 December [992; in revised form 7 August 1993] ABSTRACT

Mean monthly maximum, minimum and mean sunace temperatures arc analysed for 131 loc~tions in Canada for the period 1895 to 1989. These data were ohtllincd from the Historical Canadian Climate Database (HeeD). There has been a statistically signific:.nl rise in mc.1n annual national tcmperlllUres over the past century of about L 1°C. Annual

me:m maximum ood minimum national temperature U'Cnds since IR95 show a significant increase in mosl ly-daytimc maximum K'Illpcralures of about O.7°C and a highly

significant increase in mostly-nighttime minimunl temperatures of about L50C. Mean maximum and minimum temper<llure trends since [895 arc also examined seasonally and regionally and indicnte significant increases in minimum tempenttures mainly during the spring and summer seasons in the western, northwestern !lnd eastern continental n:gions or the country. Few slgniticant regional increases 3rc evident in maximum temperatures during uny season for this period. Significant decre:t!ics in mean temperature range, an important measure of climate "ariability, occur in most regions during all seasons. Mean maximum and minimum national temperature trends for the pa~t 4() years, since 1950, arc abo cJ;amit'll..'d. Thcy indicatc no 5ignificam increases in either minimum or maximum temperatures annually or seasonally. Mean maximum and minimum regional temperature trends since 1950 indicate significant increases in both minimum and maxlmum temperatures mostly during the spring season in the western and northwestern

ri!b>lOnS of the country.

les moycunCS mensuclles des temperatures de surface maximaJes, minimalcs ct moyennes

som anll lys&s ['Our 131 stations du Canada pour la ~riode aliant de 18951\ 1989. Ces donn6:s provicnncnt de la Base- de donnCes Climatiques cumldienne~ historiques

(BOCCH). Ce.demier sieele, il y a eu nne hau!\.'le statistiqucment importante des

temper'.! tu~ Ilationales annuelles moytnncs, hausse d'cn"iron 1.1 (l C. Depuis 1895, les tcndances de la moycnnc annucHe des tcmperatures nationalcs maximales <:t minimalcs

tndiquent une netic augnlentatioll, d'ellviron 0,7 °C, dans Ics tcmjXmtures muximalcs

I Qima!e Change Dtloclion Division, Atmospheric Environmenl Service, 4905 Uufferin Stl'l:Cl. Downsview, Ontario, CANADA M3H 5'1'4

W R. Skillner & D IY. Gillie/{ I Tempera/liTe 7ren(is ill Canada 63

enregi~trtes surtout pt'ndant Ie jour et une haussc tres signifir.:ative, d'environ 1,5 "C, des temperatures minimales enregisuie5 sur tout pendant la nuiL l)epuis 1895, les tendances des temperatures maximales ct minimales moycnnes sont aussi examinees par saison et region et rtvelent d'importantcs hausscs de la temperature minirnaJe su rvcnant surtout pendant les saisons du printemps et de l ~te dans les regions int~rieure;<; ouCSI, nord-oucst el est de la partie continentale du pays. Pendant cctte Ilcriode, it n'apparait guere de hausses regionales importantes dans Ics tcmperaturcs maximaJcs. Les baisses significativcs de la gamrnc de temperaturc moyenuc, importante mesurc de la variabilite elimutique, se presenlcnt dans la pluparl des regions et pendant toutes Ics saisons. En outre, on ex.amine lcs tc:m.lanees de la moyenne des temperatures nalionales maximaJes el minimales de.~ 40 uernieres amH~es, depuis 1950, ElIe5 ne rCvelenl aueunc haussc impoTlante, annueUe ou saisonnihc. des temperlllures millimales ou maximaies. l.es tendunces de la moycnne des temp6mtures regiollules maltimales et miniJlJulcs, depuis 1950, indiqucnt d'import.1ntes hausscs dans les temperatures lant minimales que maximaJes, surtout pendant la saison dll printcmps dans l'ouesl et Ie nord-oucsi du pays,

INTRODUCTION

Global mean surface temperature has increa~ed by O,3°C to 0.6" C over the past onc hundred years with the five global-averaged wannest years ol:Curring in thc decade of the 1980's (Houghton e l al. , IlJ9O). The magnitude of this warming is generally consistent with climate model predictions, but is also within the bounds of natural climate variability. It is believed by many researchers that this warming has been to some extent caused by increasing concenlr'dtions of radiatively active trace greenhouse ga.~es (GHG's) in the atmosphere, including CCh, CH4, CFC's and N20. Global climate systt.'Ill modcls predict global average surface temperature increases ranging between 1.5°C and 4.5°C for a doubling of C0 2 (Houghton el al. , 1990). However, regional tempemtul'c change is predicted to be even larger in middle and higller latitudes and smaller in the tropics.

The diurnal characteristics of the observed warming over the paSt century bave only recently been examined on a broad spatial sc.ue for a large number of stations (Karl ef 01 .• 1991). 'fhis study indicated that, for three large countries in the Northern Hemisphere, the contiguous United States, !.he Soviet Union and the People's Rcpublic of Ouna, most of the national warming ovcr the past for ty years can be attributed to an increase in mostly-nighttime, mean minimum surface temperatures. Mostly-.daytime, mean maximum surface ICIJlpcratures, show little or no warming. The result has been an observed decline in mean daiJy tcmpernture range over a large portion of the Northern Hemisphere landmass. The most likely cause for this i.~ increasing amounts of observed cloud cover, although thc effects of increases in other lower tropospheric scauering agents have been suggested (Houghton et al., 19cx). Other factors, such as urbani7_ation effects and changes in atmospheric

64 Climatological Bulletin I Bulletin Climatologiquc 27(2), 1993

circulation, must also be considered. It therefore becomes increasingly necessary to analyse the entire global land record, both regionally and seasonally, of maximum and minimum temperatures.

Canadian c1imato\ogislS have recently completed work on the construction of an historical data~et conlaining monthly, seasonal and a nnual mean maximum and mean minimum temperatures for 131 locations dating back to 1895 (Gullett, Skinner and Vincent. 1992; Gullett and Ski nner, 1992). These locations, and their spatial distribution, have been designated as represenlHtivc o f the Canadian land mass for climate ehanl,>e and climate variability studies. This report presents a preliminary analysis of both individual and collective results fou nd for a toln\ of eleven Canadian climate regions and provides valuable additional information on lhe nature of the observed warming of the Northern Hemisphere landmass.

DATA

The temperature data used in this study were cxtmcted from the Historical Canadian Climate Database (HCCD). The HeeD was constructed from the National Climate Data Archive (NCDA) of the Atmospheric Environment Service (AES), utilbjng climate stations that were selected on the ba~is of spatial dis tribution, length of record, data continuity, homogeneity as..~e.~sments and olher factors. The HeeD was assembled to provide climate researchers access to an initial but expanding dataset that has been ligorously quali ty controlled , assessed for homogeneity (Gullett el 01. , 1991), and adjusted, where necessary, to ensure regional representativeness. The data adjustments that were carried out had the efTect of removing some of the locally induced noise, or discontinuities, in the series resulting from changes in observing programme. instrumentation, site conditions, and other non-<:limatic efTects. These data adjustments resulted in more areally representat ive series that arc suitable for usc in regional scale analyses. l be development of the HeCD is fully described in an earlier report (Gullett t'l of. , 1992).

The HeeD currently consists of lempemture data only, for 131 Canadian climate location .. that arc as evenly distributed in spaee and time as possible, for the pCliod [895 to present. It consists of monthly, seasonal and artnual means of maximum, minimum and mean surface temperatures. T he mean is derived from calculating the a ... emge of the daily maximum and daily minimum temperatu res. The resulting dalabase contains long time series of "clean" data that arc suitable for regional scale and large area anaJyse.~. Actual temperatures and tempcmlure departures from the appropriate 195 1·80 reference period are permanently stored in the HeeD. In addi tion, both 1951-80 averages and standard deviations are retained.

W. R. Skinner & f1 IV. Gullett I Tempel"lllllrl' Trends in Callada 65

MI:I 110DS

The eleven Canadian climate region!; and the spatial dislribution of the 131 location!; examined in this study can be sct.'tl in Figure I. Table I describes the temporal and spatial characteristics of the 131 station dataset of maximum, minimum and mean temperature departure series cxtracted from the HeCD for this analysis. Individual departure series were combined into single monthly departure series, by region, by averaging departures for all locations in each region by month. Temporal differences arc partially overcome through the usc of temperature departul'l.."'S. This method as.~is t s in reducing biases which enter the time series because of a varying station network (Jones el al., 1986). In addition, monthly temperature ranges (max-min) and departures from the 1951-80 mean were calculated . Seasonal and annual mean departure series of max.imum, minimum, mean and range were similarly calculated for all eleven regions. The standard climatological seasons were used, wi nter being the average of December, January and February, spring the average of March, April and May, summer the-average of June, July and August , and autumn the average of Septcmber, October, and November. Rcgional series were then area-wcightcd to develop a national avel'3ge for Canada.

Tcmpcmture changes over time were calculated lIsing a line..1.r regression statistical model and expressed in ° C per speciJied time period. FutUIC work will concentrate on the application of non+lincar trends applied to timc series with both constant and nonconstant means (Woodward and Gray, 1993), in hopes of beller reflecting the observed temperature changes. Two periods were identified for analysis. The first is the entire 1895 to 1989 period of 95 years. The second is the most recent 40 year period from 1950 to 1989. Complete spatial coverage of locations in the HCeD is achieved by 1950. The 95 year period of analysis is referred to as a century, and all tempemture changes during this period were calculated from the regression equations on the basis of 100 years. All tempermure changes dUling the 40 year period were calculated fmrn the regression equations on the basis of 40 years, Standard statistical significance tests were performed on all results (J onC!), 1988: Karl et al., 199 1; Etkin, 1991), All linear trends found to be significantly different from One per year at the a=O.05 level are identified.

REsuns

Figure 2 shows the national annual mean temperature departures between 1895 and 1989. It ind icates a period of wamling (rom tbe begin ning of the record through the 1940's, foll()wed by a cooling period through the 1970's then a resumption of warming through the 1980's. A best-fit linear trend applied to the lime series indicates an increase of about 1.1 nC per century which is significantly different from onc per year at the 0'=0.05 level. An examination of lhe annual

66 Climatological Bulletin I Bulletin Oimatologique 27(2), (993

'~R'~":""""'''~' -, """" ....... Sl.""fORE.![ .. AACTIC "".t ...

STATIONS USED FOR ANALYSIS OF TEMPERATURE TRENDS

IN CANADA

,..., >100

- -= - '"

I<O~, ... """U~'''''.,., " "'<;TIC ~OVI'!T .... H 'lOqllG

FIGURE I Canadian dim;!te rt'gions mid climate station dislribution used fOf the calculation of regional and national <lll'a-avcmgcs.

Table 1 . HeeD d .... cl;"iption with numb"l;" 0' &tationll a dded io • climate region b y "" tinlt y e ar of each decade.

tll ... te Reg; on "'" 1910 1920 1930 1940 1950 '''' '''' , .. , Total

Pa,lIlc , , , , , , " South Be MDU"lta;ru; , , , , , , , " lukorv~crth Be "ounn; ",,· , , , , , , , , ,

Prairie , 1 , , , , , , , " North"~5tern Fores t , , , , 1 , , , , " Northu$tern F"re~t , , , , , , , " Gru\ Lakes/51. lallrcnce , 1 , , , , , , " ~tlanl lc "

, , , , , , , , " Mackonzl e , , , , , , ,

Arctic Turdrab , , " 1 , , 1 "

, " Arctic Mountains and flordsc

" " " ,

" 1 , , , N~ ticn.l Total " " " " " "

, 1 '" a ReQ'o" ... " year record 1901 - 191:19

b Re~;"" ... " yur record 1922-191:19

C Reg;"" ... " year record 19J.6-1989

W.R. Skillner & D W. Gullett I Temperature Trends in Canada 67

4.0c------------------ ---."

• , .;;;

2 .0

tl UI O.0E--.."i1 ~ E

-2.0 ".'," "'un Depi rlu res 01 Mnn Temperaturel

from the I U 1-1UG Averig e _. _ Sui-Fit LInea . Trend of I. l ·e pe l Cenlu, ~

(Trend Slgnl liCinll ~ Dineren! from O·C i t tile U % Levell

.4.0 /iTImr"1'rtm1t1Jmrrrnrp-l iiilli ill lilllm lii i!l ii l il l i ll iii l il i !!l i i fl l1 1I111l11 1)1f1 l1l l l ll l11mm

1890 1900 1910 1920 1930 1940 1960 1960 1970 1980 1990 2000

F IGURE 2 Canadian naliQ}1al annual deparlurc~ ("C) of daily mom tcmpernture for the period 1895 to 1989.

mean maximum temperature departures and the annual mean minimum temperature departures from 1895 to 1989 indicates similar patlerns. The allnual mean maximum temperature series, as seen in Figure 3, indicates an increase of about O.7D C per centu ry which is also significantly different from (}DC per year, A much higher increase in annual mean minimum tempemture departures, a.~ seen in Figure 4, of about I.Soe per century is also significantly different from we per ycar.

The annual temperature trends arc summarized in Table 2 nationally and lor the cleven Canadian c1imatc rC!:,>lons for the 1895 to 1989 period. Only two of the elcven regions show significant increases in maximum tcmpcratures for the full period, the Atlantic and Mackenzie regions. On the other hand, seven of the eleven regions show significant increases in minimum temperatures since 1895. Only the Atlantic, Yukon/ North Be and thc two Arctic regions do not show significant increa.~es in minimum lempemtures. Significant reductions in temperature range since J 895 arc evidcnt. in the seven western and L"cll trai regions of Canada. This may be attributed to the significant warming observed in the minimum temperatures and the non,significant changes in the maximum temperatures. An c.""\ception is the Yukon/ NOrth Be region where the temperatu re nmge over the fu ll period has remained basically unchanged due to the nearly identical, non-significant trends in both maximum and minimum temperatures.

Also included in Table 2 are the annual temperature trends for the 1950 to 1989 period. Since 1950, there have been no significant trends in either

68 Climatological Bullelil.l / Bulletin ClimalO10giquc 27(2), 1993

4.0r.:---------------------::J

2.0 -

·2,0 ••••• Mun Oep.o1 • .,n " I M.u:;rnu m Ta""p",atu, .. "0In 'h. 1951·1980 Average

•.• B.~I ..fit Unaa. T"nd 01 O.1"C pet Century tnend Significantly Ci tra,ent ',om O'C al lh . 95% Uoval)

-4. 0' Fmr111jnnnn'l"""m"lmTIT11'l'fTTTTTTTlpnm~nllnllnpniimjjmli"ll"l'"iI"i nllml Flif111' i 1111111111111 1

1890 1900 1910 1920 1930 1940 1960 1960 1970 19BO 1990 2000

FIGURE 3 Canadian national annual departures (0C) of daily maximum temperature for-the pt'riod J 8~5 to 19!!9.

4.0'c---------------------co

~ •• il

2.0

:: 0 .0 :

~

l -2,0

•. Mun Oep.ttulu 01 Mlnlmtlm Tempt tatu' n bom th e 1151 ·1110 Average BU1.f"il LiI1U t Tlend o f 1.SoC per Century ITr end SlgnlficlnUy Ollle, enilrom O' C al the iii '" Level l

-4.0 mlll1nprmmll nmmpmnmpnmmpmn;,;rnm;;;p;;;nmpnmmpmnmpnmmij 1890 1900 1910 1920 1930 1940 1960 1960 1970 19BO 1990 2000

FIGURE 4 Canadian national annuall1epartures ("C) or dail~ minimum icm!)l:roturc IQr 1M /lCriod 11195 LO 1989_

maximum or minimum annual tempcmtures, or in temperature range, for the national compilations. However, mean ma:.umum temperatures have risen significantly in the four continental western regions extending from the Prairies

~V.R. Skinner & D. IV. Glllletl I ThmperalUre Trends in CanadQ 69

l'able 2. Lin .... r b:end9 of mean ann"al temper ature in ·C f><Ir cent u ry fc~ the pariod 1895-1989 and ·C pa r 40 year .. for the period 1950-1 9 89 tor Canada and by cl Imate region.

1 89') -!9S~ 1950·19a9 1895·1989 19~0- 19119

t ' n,d. \~.ttOf1al' .. ,. "" 1. ~, ••• P""H!c .. , "" 0.6- .. , South Be H"......ui", -0 .1 ••• 1. 0' 1. I '

vuI;on/Nortn BC HOurtt8in.,b ••• 1.4" .., 1. 6" ~caido ••• .. ,. ).38

'.1 Wor th .ut~,.,. 1000v. t .., 1.3° 2 . 1· I. ' ~ortM~Uern for~ .. ••• .. , L OB ·0.2 ~rut \.._~sfS t. l ... r"",,~ .. , ·0.4 1.0· ••• _,l81>1,,,- D.~· -D.' ••• · O.S"

"at ~""I ' e 1.4° 1. 2° <'. I" ••• ~ '""t \e Turdr l c .., .., ' .1 .. , ~r ct i < ~O'SItel"" and fiords

d ·0 . 7 -0.5 -0.5 · 0 . •

o 1rffld .' lI" , f ,eanlly (~~O.OS) differenl I r ... O"t per speciH.d Ii .... ~r;od b Re~;on MP S9 yur r~co rd 1901· 1989 C ~cyion ho . 61! yeer record 1nl- l ?69

d ~"~j"" h.S 44 ),<,. r ,ee-ord 1946· '989

. a95 - 1989 11l50·'989

·O.s' U -O.S~ .. , . ,. l ' ·0.3

• •• · 0.2 . , .08 0.6"

·1.1" ·0.1 - 1.0~ ••• ·0.9° · 0 .4°

U 0.'· · 0 .1' 0. ' & .. , .., -0 .3 ·0.2

to the Yukon. Thcse increases have ranged rrom approximately 1.2c C to I.7cC per 40 ycars. Minimum temperatures, on {he other band, have increased in only the two mounlain regions or British Colu mbia and {he Yukon. The Atlantic region has actually seen a significant reduction. about .. 0.80 C per 40 years, in minimum tempcratures and a non-significant decrease in maximum temperatures. This has rcsult".'d in a significant increa~e in the temperature range in the Atlantic region oro.4° C per 40 years. Similar significant increases in temperature rangc are :$een in tbe Prairie and Mackenzie regions due to significant increases in maximum temperatures and n~n-si~nificanL incrcases in minimum temperatures. Thc Greal Lakes/ St. Lawrence region has seen a significant reduction in temperature range since 1950 due to a non-signi1icant reduction in maximum temperatures and no trend in min.imum temperatures.

Seaoronal trt::nd summaries are provided in Tables J to 6 ror winter, spring, summcr and autumn, respectively, ror the two specified periods. Since 1895, significant national trends in maximum temperature have oceurred in thc summer sea<;()n only, while significant national trends in minimum temperature, and temperature range, have occurred in all seasons. Sincc 1950, there have been no significant trends in either maximum or minimum lempemtllrc, or temperaturc mnge, in any sea~on ror the national compilations.

Over the past century, significant linear trends at the regional scale are evident in both maximum and minimum tempcratures and in temperature range in most sea.'IDns but wi th vary ing rcgional differences. The rrequency of occurrence or significant trends increa~es considerably over wider spatial areas, or collections or adjacent regions, during specific seasons. An examination or thc

70 Climatological Bulletin I Bui!ctin Climatolob>ique 27(2).1993

Table 3. Linear trends ot mean winter cewperature in ~C pa r c e ntury for th~ periods 189~-1989 and ·C per 40 yea rs tor the period 1950-1989 for Canada and by climate region.

1895· 1989 1950-19~9 1/19S-1989

"""" (~ational) 1.1 .. , I. ~ P&eifi c "-' '-' ••• Sn~th I t ~oo.nh;"" .. , .. , .. , T"onjNDttll BC ~"""tI;n.b U ••• ••• Prli'i' "-' ,., '-' ~ o.t~~~.\~rn F"ro$\ ••• 3.1' ~ . 2' Nortllustern forH' '0. 1 ·0.2 I.S· Grut L.kn/St. Lftw,,,nce- "-' -0.9 I. ' AII.nlie .. , ' 1.6' ••• ~ac"~nzi~ 2.1 ' LI· Z.6'

~rc t i c a'nll"c \.9" '-' 1.9' Arctic ~"'*' tain. _ fiord.1I .., .. , .. , • Ir"nd s;gniflc .... 'iV (<t.O.OS) <liHcrtnl fr ... O·C rer yea, b Reg,e" hu 89 yur r~ord 1901-1989

c H~g'Dn has U v",r rocord 19U-1969

"~~gl"" h'~ '~ VUr record 1946·1989

195Q-1989 1895-1989 1950- 1989

••• -0_6· .., • •• 0 .1B O .5~

'-' -0 .3 -0 .2

'-' .., .. , ,. , '0.6- ••• '-' " _l' ..,

'0.6 '1.",' ••• '0.6 ·0 . 6" -O. l ·Z.3" ,0. , "-' 2.5' '0.2 ••• ••• ••• 0 .5" .., '.1 -0. I

Table 4 . Linear trends of mean opring temperatura in ·C per century for the p~rlod8 1895-1989 and ·C per 40 years for the period 1950-1989 fo e Ca~ada and by climate region .

Cli ... , e ~cg 'or. ~e.,. ~ul_ """,n Min' ....

1895' 1989 1950' 198'1 18Q-5 -1969

C(Irn8d_ (hl;"""l) 0.' '-' '-~ Pa.c:H'c 0.' 0.9" 0.1' Soulh 8e ~",*,1ai~ ••• 1.6" 1.'"

Y"kotl/ North 8C "",,,,,.'n.b loS· 1.6" 1.0 Pr.irle ••• J.S· 1.S" ~OI"U'WMacn rQrlU ••• 1.1" 1.8~

~art~u.tetn Fatu t '-' I.' ,.4" GrUI Lake./SI . l""r..-.ce .., ••• l.l"

AI I"" t ic ••• '-' ••• ~.d.nd $ I.. '-' >.f' An tic lundr." '-' ·1.0 ••• Arc . l~ ",,,,,,,t.l,,,, an<! Figrdl;d ·2 .G$ '2.0· - 1.0

• " ..... ~i9n; f'c""tl 'l' (0:0.05) difforent I' .... OOt p<'r yeor b ReVi"" ~u 69 y~.r ,ecord 1901 -1 969

C hgi"" "as 60S v-ar record 1922 1969

d '_Vi"" h~S <.4 yeor recor d 1""6 1989

1!l'i0·1969 .. , 1.0~

1.6" 2.]-

~.8· 2.48

••• 1.0

••• I.,

' 1.2

- 1.4

"' ..... lIa~-~;n

189~ - 1989 1950~1969

-1. 1' '-' -O.~· - 0.1

- 1.1' -0.0 .. , · O.S ., .. , O·

1.9" - 0.2 - 1.Z" .., ' 1.0· - 0. 2

- 0.1 .., ' 1. 3D .. , '-' ..,

- 1. ,. _0.6°

temperl:lture trends from 189510 1989 indicates most signifi cant increases in minimum temperatures have occurred during the spring and summer seasons throUghOUl many neighhouring regions extending from the Pacilic coa..~t, through

W R. Skl,mer & f) II( Gullett J Temperature Trends in COllada 71

Table 5 _ Linear crende o f mean ~ummer temperatura in ·c per century for the period~ 1895-J989 and ·C per 40 years for the period 1950-1989 for Canada and by olimatD ragion.

Cli..,te ~ e5li"" ~e"" ~ .. i_ ~un " ;~I ....

1 895-1~ 1'?50-1989 UI95- 19S9

,~'" (HII;"",,' ) o_s- ... 1.S-

PKilic · O.lIa -0.2 1. 1·

SOUl h 8C """" l eIM - 0.6 .., U· yu .... "/Nt>'· Lh Bt N"",)\~",~b -0.1 ••• 1.0· Pr.(c[e .. , ... I. S" NGr lho<este,n for .. 1 ... '" ~.O·

Northeastern '~rUI ... ... o.a" OrUI L.~~s./SI. L~"r"""e -0.6 -0.2 o.e-_t[antic 1. ,~ ••• 0.6~

"Kk..". [e 1.0~ .. , 2.0· Arelic T..,.;Ir . o ••• .. , .. , Are!!c """"lai n< lind flordscl -0.3 ·0.1 •••

• lr""'" ", ~nd lcon«y (.·O.O~) clill ereot I r ... O· C per ~ar b ~eg[on ~a& 119 ye~r ,.cora 1901·19119

e ~egion ~n U yur 'oco,.." 1922· 1989

d RegIon ~ a. ~, year r.o~r~ 19lo6·1989

1~1J'19IW

'" .. , 0 . 9" ... .. , ••• .. , .., ... .. , ••• •••

"~an ~,.-~ in

11195- 191!9 1950-19S9

- 1.0· .., -1.v" - O.~

·LS· -0.7 -I. 1· '0.3 .1. 1" .. , ,1. 1" ·0.5 .0.5" .. , -1.4" · O.Sa

0.5" .. , · 1.0a .. , ••• 0.'"

·0.1 ·0 .1

T_ble 6 _ Linellr t r ende o f m ..... n .. utumn temper .. ture tn · C per century for the periods l895-1989 and ·C p9( 40 ye .. r s for the period 1950-1 989 for C~nada snd by climate region.

Cll .. t~ ~"'-IIM "U~ MA i_ Mean Nini_

11195· '98~ 1950-1989 1895· 19119

,~'" (Wati ONlI} .., -II.} 1.1 ·

Pacific .., • . < .., South Bt ~"""f&in. ••• .. , ••• ''''''''''Nor th ~c lIo ..... toi""b .., ••• ••• Prairie ·0.2 ·o .~ .., N""~""'I~rn f~roSl ·O.S -0.4 1.6&

North e .. t~rn fores t ' 0. 1 -0 .8 .., Gre.t l"""/~ I . l .~r.nc~ ••• ·1. 2~ O,S"

"" "ntic ••• -0 .9- -0. 3

N&c~."'le ••• ·0 .4 ,., Arctio lur<lr. c .. , ••• .., Arc tic Nount.i"" and 'l~td.<I · 0,1 ·0.1 ·0 . 6

• Trmel 5iflnlHcMtly (g'II.0~1 dl H erenl Ir ... O'C po. y.,~t b Rag !on hos 89 year r~,or" 190T ' 1989 , ~e~j"" ha. Ita yur re~ord 1922·1989

d Re9i on II .. 44 ~ear fecor~ 19~6'1989

1950·19119

••• .. , .., ..,

-0 . 6

'11.4

'0. 8 -O , i -I. 14

· 1. 0

·O.?

· 0 .2

~Un MaX·Nin

1895-1989 1950·1989

- 0.6" -0.1

-0 .3 .., ·0.8" -0.1 .., -0.1 • II. vi' .. , ·1.9~ · 0.0

' 0. r' 0 .. 1

·O.S" ·0 . 9"

0.30 .. , -0 .3 ••• .. , .., .. , .. ,

the western, northwestem and eastern continental interior regions of Canada. Spring increases range from O.7°C to 2.SoC per century while summer increases are from O.6°C to 2.(f'C ~r century. These have been accompanied by few

72 Oimatological Bulletin I Bulletin Climatologique 27(2). 1993

significant increases in maximum temperatures, occurring onJy in the Yukon/ North Be region during spring and in the Mackenzie and Atlantic regions during summer. Significant reductions in temperature range throughout these regions during these two sea-.ons an:: evident. In geneml, reducliollJi in spring and summer temperature ranges over the past cent.ury are attributable to the rising minimum temperatures and the relatively unchanging ma,umum temperatures. The only exception is during summer in the Atlantic region where a relatively larger increase in maximum temperatures and It smallc.r increa~e in minimum temperatures has resulted in a significant increase in temperature range of about O.5°C per century.

An examination of the temperature trelld~ from 1950 to 1989 indicates significant and comparable increases in both m.uimum and minimum temperal\lfC->; mainly during the spring season in the western and northwestern regions of Canada, from 0.9°C to 3.8°C for maximum temperatures and I.O°C to 2.8°e for minimum temperatures per 40 years. However, few significant changes are seen during this season in the temperature range. implying that thc recent rise in temperatures throughout this broad area has been diurnally unilOrtn. The exception is an increase in the Prairie region range of abou t I.O"C per 40 years due to a considerable increase in maximum temperatures and a smaller increase in minimum temperatures.

Figure 5 and Figure 6 show the spatial extent of the spring season significant (a=O.05) maximum and minimum tempcmturc trends, respectively, for the 40 yC<lf period 1950 to 1989. These maps were compiled from the individual 40-ycar temperature trends for the 13 J !ocations in the HCeD. Large similar areas of positive u-ends are evident in both maps, although the area of positive maximum temperature trend extends further into northern Ontario. The Arctic regions have expericnced morc spatially widespread significant decreases in ma.'timum temperature than in minimum temperature and therefore temperature range sillce 1950 during the spling seasoll. Significant decreases in bOt.ll ma.'tlmum and minimum temperatures have occurred during thc wimer and au tumn seasons in the Atlantic region as seen in Tables 3 a nd 6, respectively.

CONCLUS IONS

Observed lempemturc changes over two periods, 1895 to 1989 (one century) and 1950 to 1989 (40 years), were analysed both on (I national ba~i.~ and regionally for eleven Cunadian climate regions. Mean monthly maximum (daytime), minimum (nighttime), and mean surt·ace temperature lime series for 131 locations in Canada were analysed in this study with the use of be.st-fitlincar trends. All linear trends found to be significantly different from O"C at the a=O.05 level were identified,

Maximum and minimum temperature analysi.~ on a national ba~is suggests nighttime warming both annually and seasonally over the century. On a

~v. R. Skinner & D. W Gullett I 7emperalUr!' Trell(J.f ill Canada 73

FIGURE 5 Spatial extent ohpring SC'3!lon significant ((1'=0.05) maximum temper.lture trends for the period 1950 to 1989.

regional ba'\is, seasonal temperatures over the past century show significant increases in minimum temperatures mainly in the western, northwestern and eastern continental regions of lhe country while only a few significant increases arc evident in coincident maximum temperatures. The implication is for nighttime warming in these rq,tions.

Nationally, there is no evidence of a significant rise in either minimum or maximum temperatures, annually or seasonal ly, over the most recent 4O-year period. This is not surprising since lhis period was dominated by a period of hemispheric cooling, from the 1950's to the early 1970's, and of hemispheric warming, from the early 1970's through the 1980's (Jones, 1988), with opposite trends tending to cancel one another. A regional analysis is beneficial, however, since strong temperature trends become evident over large arcas of the country as was seen previously in the regional one century analyses.

Regionally, seasonal temperature changes since 1950 indicate significant increases in both minimum and maximum temperatures, but only in the spring season, and only in the western and northwestern regions of the country, Recent studies have provided evidence of this warming during the spring season through the effects on lakes in these regions, such as the errects of

74 Climatological Bulletin / Bulletin Climatologique 27(2), 1993

FIGURE 6 Spatial C)l.tcnt of spring sca.o;oo significant (<>=0.05) minimum temperature tfCllds for lhe period 1950 10 1989.

earlier snowpack removal through increased daytime heating (Schindler et at., 1990) and in the earlier break-up of lake iec cover in northwestern Ontario and in the Prairie provinces (Skinner, 1993). Corresponding inereases in minimum and maximum temperatures have resulted in no significant changes in the temperature range in these regions. Overall, the evidence does not support the hypothesis of nighttime wamung over the past 40 years in Canada.

The absence of significant trends in many regions, especially during winter when most GCM's tell us the greatest CO2-induced effect should become evident, might be e.'<plained in the selection of the standard climatological seasons. Recent evidence has shown that, when comparing the two normal periods 193 1-60 and 1961-90 for seven stations in the eastern Canadian Prairies, December temperatures have become cooler while January and February temperatures have increased (Raddatz el al. , /99 1). Further investigation and analysis may be warranted.

The observed diurnal temperature changes over the past century in Canada are not proof that an enhanced greenhouse gas effect is taking place. Canadian observations of mean temperature increase, however, are in keeping with the general rise in global mean tcmpcmturcs obscrvt'd over this period. On

W R Skinner & D. W. Gullett I Temperalure Trends in Callada 75

the other hand, ovcr the past 40 years, the Canadian landmass has expetienced significant changes in both minimum and maximum temperatures on regional and sea.-.onal spatial and temporal scales, but not on a national scale O[ o n an annual ba. .. is. No clear indication exists in thi s- analysis of en hanced nighttime warming over the past 40 years, however, 'hi .. may be a function of the application of the linear trend technique. Future activities will explore non~l inear

methods and will fOclls on attribution of observcd temperature changes as well a~ on comparisons with output from global circulation models forced with increa~ed greenhouse gases.

ACKNOWLEDGM ENTS

The authors would like to thank Krislina Czaja for her cartographic assistance.

REFERENCES

Elkin, D., 1991 : Winter and Summer Surface Air Temperature Trends in the Northern

Hl!misphere: 1950 to 1988. CiimaloioRical Bulle/in, 25, 182-/93-Gullett , D. w., W.R. Skinner and L. Vincent, 1992: Development of an Historical

Canadian Oimate Database for Temperature and Other Climate Elements.

Climato/ogiooi Bulk/in, 26, 125-1 31. GulleU, o.w. and W.R. Skinner, 1992: The Stale o/Cantu/as Climate: Temperature

Change in CQ/ladaJ895-J99/. State of the Environment Report No. 92-2. Minister of Supply and Services Canada, 36 pp.

Gulleu, o.w., L. Vincent, and L.H. Malone, 1991: Homogeneity Tesling of MomMy Temperature Series. Application of Multiple- Phase Regte~si()n Model~ with Mathematical Challgepoillfs. Canadian Climate Centre Re[lOrt No. 91- 10. Downsview: Atmospheric Environment Service, 47 pp.

Houghton, J.T. , GJ. Jenkins aud 1.1. EphrnulIls, 1990: Climate Change: 71le /PCC Scientific Assessment, Cam\;)ridge Univenity Press, Caml;lridgc, U.K. , 365 pr., (xii, xxv, 33).

Jones, P.o., S.C.B. Raper, R.S. Brddley, H. E Dia:.>., P.M. Kelly, and T.M.L. Wigley, 1986: Northern Hemisphere Surface Air Temper,u llre Variations: 1851-1984. JOllrnal 0/ Climate and Applied Meteorology, 25, 161-179.

Jones, P.o., 1988: Hemispheric SUiface Air Temperature Variations: Recent Trends and

an Update. 10 J987. Journal a/Clima/e, t, 654-660. Karl, T.R., G. Kukla, Y.N. Rasuvayev, M.J. Changery, K.G. Quayle, R.R. Heim, Jr .•

D.R. Easterling, Hnd C.B. !-u, 199 1: Global Wanning: Evidence for

Asymmelric Diurnal Temperalure Change. Geophysical R£search Lellen, 18, No. 12, 2253-2256.

Raddatz, R.L., 1. Maybank and G.B. Atkinson, 1991: Mean Daily Temperature Normals

from 190I-JO to 1961-90 on the. Eastern Canadian Prairies. Climatological BIIl/elill , 25, 118- 123.

76 Climatological Bulletin I Bulletin Climalologique 27(2), 1993

Schindler, D.W., K.G. Beaty, E.J Fee, D.R. Cruikshank. E.R, DeBruyn, D.L. rlndlay. G.A. Linsey, J.A. Shearer, M.P. Stainlon nnd M.A. Turner, 1990: Flfects of Oimatic Wanning on Lakes of the Central Boreal Forest. Science, 250, 967-970.

Skinner, W. R., 1993: Lake Ice Conditions as a Cryospheric Inw'cator for Detecting Climale Variability in Canada. SIIOW Walch 92, Delection Stmlegies!or Snow and Ice. ProcccdinKs from an International Workshop on Snow and Luke Icc Cove( and the Climate System, Niagara-on-the.Lake, Canada. R.G. Barry. B. E .. Goodison and E.P. LeDrew, cds., 204-240.

Woodward, W.A. and H.L. Gray, 1993: Global Warming and the Problem of Testing for Trend in Time Series Dab. JOllrntl1 of Climate, 6, 953-962,

~y. R. Skinner & D W Gul/ett I Temperature Trends ill Canada 77