天气是加拿大城市交通的潜在危险

Natural Hazards28:319–343,2003.319©2003Kluwer Academic Publishers.Printed in the Netherlands.Weather as a Chronic Hazard for Road Transportation in Canadian CitiesJEAN ANDREY1,BRIAN MILLS2,MIKE LEAHY1and JEFF SUGGETT31Department of Geography,University of Waterloo,Waterloo,Ontario,Canada N2L3G1(E-mail: jandrey@fes.uwaterloo.ca;2Meteorological Service of Canada;3Synectics Transportation Consulting Inc.(Received:9July,2001;acceptd infinal form:3December2001)Abstract.Inclement weather creates a chronic hazard for Canadian travellers.Past studies indicate that road collision rates increase during precipitation,although the magnitude of the increase varies from study to study,partly as a result of variations in weather and driving conditions,but also because of differences in methods.The goal of the current study is to improve our understanding of the links between weather and travel risk in mid-sized Canadian cities by using a standard-ized method for analyzing data from six cities with different climates:Halifax-Dartmouth,Ottawa, Québec,Hamilton,Waterloo Region,and Regina.The study has four interrelated objectives:(1)To conduct a sensitivity analysis to determine the extent to which risk estimates vary depending on the criteria used to define precipitation events and‘normal’conditions;(2)To compare the relative risk of collision and injury during precipitation relative to‘normal’conditions;(3)To determine the extent to which weather-related risks vary for different Canadian cities;and(4)To explore any differences in collision characteristics between events and controls,especially as they vary from city to city. Results are based on a matched-pair analysis,using six-hour time blocks over a four-year period, 1995to1998.Results indicate only modest sensitivity to the criteria used to define precipitation events and‘normal’conditions.On average,precipitation is associated with a75percent increase in traffic collisions and a45percent increase in related injuries,as compared to‘normal’seasonal conditions,but risk levels vary depending on the characteristics of the weather event.Both sensitivity to specific weather conditions and weather-related accident profiles vary from city to city in ways that are not easily explained.Key words:Road-weather hazards,motor vehicle collisions,accident characteristics,precipitation, Canada,urban,transportation1.IntroductionThis paper is a contribution to the Canadian Natural Hazards Assessment Project. One information resource for the project is a national disaster database developed and maintained by the federal Office of Critical Infrastructure Protection and Emergency Preparedness(EPC,2001).Transportation accidents represent over10 percent of the entries included in the national archive(Tudor,1997).Ship colli-sions,train derailments,and airplane crashes have claimed thousands of lives over the past century in Canada and are duly noted within the database.320JEAN ANDREY ET AL.The acute impacts of incidents such as the1979Mississauga,Ontario train derailment(evacuation of hundreds of thousands of residents);Swissair Flight111 plane crash off Peggy’s Cove,Nova Scotia in1998(229killed);and fog-related 82-vehicle crash along Highway401near Chatham,Ontario in1999(7dead,45 injured)(Gillespie,1999)are dramatic and often sensationalized through popular media.Such events elicit immediate and often costly responses by government agencies and other involved parties to probe and correct the disaster’s cause.Soci-ety and its institutions are similarlyfixated on the extreme event within the realm of natural hazards,for example the Pine Lake,Alberta tornado in2000.Regardless of how dramatic these disasters are,they are by definition extremely rare and the vast majority of Canadians are unlikely to encounter one.On the opposite side of the spectrum lie the chronic hazards–the frequent dangers of everyday living whose risks become‘accepted’(e.g.,lifestyle risks)or the type where risk is accumulated or dislocated such that cause and effect are isolated in time and/or space(e.g.,transboundary air pollution).No hazard typifies this corner of the spectrum as does the risk of a motor vehicle collision,including the subset that are attributable in part to ambient or road weather conditions.Over 5percent of the Province of Ontario’s7.9million licensed drivers were involved in a motor vehicle collision in1999(MTO,1999).Across Canada,2,927deaths and217,614injuries were caused by motor vehicle collisions in1998,roughly one injury for every137citizens(Transport Canada,1998).Motor vehicle collisions account for16percent of all injury admissions to acute care hospitals in Canada–injuries being the leading cause of death among Canadians(and in most developed countries)aged1–44years(Health Canada et al.,1999,p.243).CCMTA(2000) reports that the cost of casualty collisions to the Canadian health care system ex-ceeds$10billion per year.Property damage sustained during all collisions in1999 is estimated to be in excess of$1.25billion for Ontario alone.Considering these statistics,if even a very small percentage of motor vehicle collisions is attributable to natural hazard agents(i.e.,weather-related),it seems that a significant effort should be made to discover how Canadian travellers are vulnerable to weather conditions and how such vulnerabilities may be overcome through preparedness,advances in vehicle and highway engineering and other be-havioural measures.National summaries of accident statistics for1998indicate that16percent of fatal collisions and18.5percent of personal injury collisions occurred during adverse weather conditions(rain,snow/freezing rain/hail/sleet, fog/mist/smog/dust/smoke).Ontario collision data for1999support these estimates but also suggest that over25percent of property-damage-only collisions occurred during weather that impaired visibility(MTO,1999).Thus there appears to be a compelling rationale to evaluate chronic weather-related transportation hazards.To date only modest attention has been paid to this subject.Existing studies fall into four broad categories:Based on a$5800/collision cost estimate derived from Doherty(1994,p.2)and MTO(1994).WEATHER AS A CHRONIC HAZARD FOR ROAD TRANSPORTATION IN CANADIAN CITIES321 1.Efforts to establish the relationship between weather and the physical operat-ing environment,such as the effect of precipitation on road friction or driver visibility(see for example Andrey and Olley,1990;Henry,2000);2.Studies that propose and/or evaluate guidelines,technologies or managementsystems that are intended to mitigate the effects of weather(e.g.,Huebner et al.,1999,on guidelines for reducing hydroplaning);3.Case studies that detail extreme weather events and their impacts on trans-portation(e.g.,Rooney,1967;de Freitas,1975;Earney and Knowles,1974;Changnon,1979;Mills et al.,2002);and4.Safety research that estimates the extent to which collision/injury risk isaffected by‘adverse’weather.It is this latter theme that is explored in this paper.1.1.RESEARCH GOALThe goal of the research is to contribute to our understanding of the relationships between precipitation and road safety in urban areas of Canada,especially as they may vary from city to city.There are four interrelated objectives:1.To conduct a sensitivity analysis to determine the extent to which risk estimatesvary depending on the criteria used to define precipitation events and‘normal’conditions;2.To compare the relative risk of collision versus injury during precipitationrelative to‘normal’conditions;3.To determine the extent to which precipitation-related risks vary for differentCanadian cities and different weather types;and4.To explore any differences in accident characteristics between events andcontrols,especially as they vary from place to place.1.2.RESEARCH CONTEXTBoth the professional safety community and the public at large recognize that weather affects the safety of motorists.Weather that reduces road friction,im-pairs visibility and/or makes vehicle handling more difficult is seen to create a safety threat.On the other hand,human error plays a role in virtually all roadway collisions,and driver behaviour(e.g.,increased caution or reduced speed)can com-pensate for some or all of the risk associated with adverse weather.Thus the best way to obtain accurate estimates of weather-related collision risks is through the analysis of collision data.Although a number of weather-related factors affect the operation of motor vehicles,including glare,fog and other obstructions to visibility,heat stress and wind,most of the related literature deals specifically with precipitation in the form of rain or snow,as summarized in Table I.Based on these studies,the following conclusions can be drawn.322JEAN ANDREY ET AL.Table I.Literature review summaryRain or wet road conditions Snow or slippery road surface Other weather factorsCollision risk Collision risk Fog:Collision severity Haghigh-Talab(1973)Codling(1974)Musk(1982)Codling(1974)Jovanis and Delleur(1983)Satterthwaite(1976)Mende(1982)Fog:Collision frequency Sherretz and Farhar(1978)Andrey(1989)Smeed(1953)Bertness(1980)Andrey and Olley(1990)Storie(1984)Smith(1982)Suggett(1999)Edwards(1996)Mende(1982)Norrman et al.(2000)Brodsky and Hakkert(1988)Carson and Mannering(2001)Temperature extremes:collision Andrey(1989)Khattak and Knapp(2001)frequencyAndrey and Yagar(1993)Baranowska and Gabryl(1981) Changnon(1990)Collision frequency Nofal and Saeed(1997) Suggett(1999)Hanbali and Kuemmel(1993)Andreescu and Frost(1998)Hail:Collision frequency Collision frequency Knapp et al.(2000)Pike(1992)Smeed(1953)OECD(1976)Injury frequency Assorted other weatherScott(1983)Balutikof(1983)Orne and Yang(1972)Storie(1984)Fridstrom and Ingebridsten(1991)Roer(1972)Mercer(1986)Fridstrom et al.(1995)Thomas and Gallon(1988)Edwards(1996)Sankar et al.(1995)Khattak and Knapp(2001)Levine et al.(1995)Andreescu and Frost(1998)Collision severityCodling(1974)Collision rate Mende(1982)Andrey(1989)Evans(1991)Injuries and fatalitiesClissold(1977)FatalitiesThomas and Gallon(1988)Fridstrom and Ingebridsten(1991)Fridstrom et al.(1995)Fridstrom et al.(1995)Edwards(1996)Edwards(1996)WEATHER AS A CHRONIC HAZARD FOR ROAD TRANSPORTATION IN CANADIAN CITIES323•Collision risk usually increases during precipitation,from negligible amounts to several hundred percent,although the typical estimate in more rigorous studies is50to100percent.Variations would seem to be due in part to dif-ferences in methods and weather conditions,but may also reflect urban/rural or regional/contextual differences in sensitivity.•There is considerable evidence that snowfall has a greater effect than rainfall on collision occurrence,although snowfall-related collisions tend to be associated with fewer fatalities than other collisions.There is a debate in the literature about whether snowfall is associated with a net increase or net decrease in injuries.•Based on available evidence,collision risk increases appear to be greatest for freezing rain/sleet and thefirst snowfalls of the season,and lowest for light drizzle or snowflurries.There also appears to be a positive relation-ship between precipitation intensity and collision risk,but few studies have examined weather conditions in detail.•There appear to be differences in the types of crashes that occur during inclement weather as compared to other times,but there has been limited research on this topic and thus it is difficult to know to what extent the results are generalizable and to what extent differences are specific to a particular study/location or weather type.The precipitation-related literature provides a good basis for predicting the dir-ection of the outcome,but does not allow accurate estimates of the magnitude and details of the effects for different places.This is because road collisions are the result of many interacting factors,and most of these vary from one setting to another.Thus additional empirical research is required,especially studies that allow meaningful regional comparisons to be made and understood.2.MethodsThis study is based on analysis of weather and collision records for a four-year period(1995-1998)for six mid-sized urban areas:Halifax-Dartmouth,Québec City,Ottawa,Waterloo Region,Hamilton,and Regina.The six study areas,as summarized in Table II and illustrated in Figure1,represent different climatic regions.Halifax,on the east coast,receives large quantities of both rainfall and snowfall.Québec receives less rainfall than Halifax,but more snowfall.Regina receives only one-quarter as much precipitation as Halifax,and is colder in winter than any of the other study areas.The three Ontario stations are similar to one another,although Ottawa is colder and receives more snowfall than the other two stations.Hamilton and Waterloo are located in close proximity to one another,but winters in Hamilton are milder due to its lakeside location.Thus the study sites provide an opportunity to compare weather-related driving risks both across and within climatic regions.324JEAN ANDREY ET AL.T a b l e I I .1961–1990A v e r a g e c l i m a t e s t a t i s t i c s f o r c a s e s t u d y c i t i e sH a l i f a x -Q u ´e b e c O t t a w a W a t e r l o o H a m i l t o n R e g i n aD a r t m o u t hC i t y R e g i o n C l i m a t e n o r m a l s t a t i o n n a m eH a l i f a x Q u ´e b e c A O t t a w a W a t e r l o o -H a m i l t o n A R e g i n a AI n t ’l A I n t ’l A W e l l i n g t o n A C l i m a t e n o r m a l s t a t i o n l o c a t i o n 44◦53 N 46◦48 N 45◦19 N 43◦27 N 43◦10 N 50◦26 N (L a t i t u d e /l o n g i t u d e )63◦31 N71◦23 W75◦40 W80◦23 W79◦56 W104◦40 WP o p u l a t i o n (1000s ),1996325672763418624194T o t a l a n n u a l p r e c i p i t a t i o n (m m )1473.51207.7910.5917.0890.4364.0T o t a l a n n u a l r a i n f a l l (m m )1222.7881.3701.8773.6743.3280.0T o t a l a n n u a l s n o w f a l l (c m )261.4337.0221.5158.0152.4107.4D a y s w i t h m e a s u r a b l e r a i n f a l l a12811711011410762D a y s w i t h m e a s u r a b l e f r e e z i n g 1615n /a n /a n /a14p r e c i p i t a t i o n aD a y s w i t h m e a s u r a b l e s n o w f a l l a637664654956A v e r a g e J a n u a r y t e m p e r a t u r e −5.8−12.4−10.4−7.3−6.2−16.5(◦C )A v e r a g e J u l y t e m p e r a t u r e (◦C )18.319.120.819.920.819.1D a y s w h e n m a x i m u m 60100857162109t e m p e r a t u r e l e s s t h a n o r e q u a l t o 0◦CD a y s w h e n r e d u c e d v i s i b i l i t y 122n /a n /a n /a n /a28r e p o r t e d (v i s i b i l i t y <1k m )S o u r c e s :M e t e o r o l o g i c a l S e r v i c e o f C a n a d a ,1961–1990C a n a d i a n C l i m a t e N o r m a l s ,h t t p ://w w w .c m c .e c .g c .c a /c l i m a t e /n o r m a l s /e p r o v w m o .h t m S t a t i s t i c s C a n a d a (C e n s u s o f P o p u l a t i o n )h t t p ://c e p s .s t a t c a n .c a /e n g l i s h /p r o fil /P l a c e S e a r c h F o r m 1.c f m a M e a s u r a b l e =0.2m m (r a i n f a l l ,f r e e z i n g p r e c i p i t a t i o n )o r =0.2c m (s n o w f a l l ).WEATHER AS A CHRONIC HAZARD FOR ROAD TRANSPORTATION IN CANADIAN CITIES325Figure1.Location of six study areas.(Source of base map:Government of Canada).2.1.MATCHED-PAIR ANALYSISThere are two main approaches for documenting the sensitivity of road accidents to weather.Thefirst uses the general linear model to explain temporal variations in collision occurrence as a function of weather and other variables.The second is based on temporal comparisons of collision frequency or characteristics(Andrey and Olley,1990).The most common type of temporal comparison,and the approach adopted for the current study,is referred to as matched-pair analysis.In a matched-pair ana-lysis,collisions are compared between events and corresponding controls(see for example Codling,1974;Sherretz and Farhar,1978;Bertness,1980;Mende,1982; Smith,1982;Andrey,1989;Andrey and Yagar,1993;Suggett,1999;Khattak and Knapp,2001).The goal is to identify event-control pairs that are similar in all aspects(e.g.,study area,time of day and day of the week),except for the one factor being studied(precipitation),so that the effect of that factor on the variable being studied is isolated.When studying smaller temporal units,the most important extraneous factor that needs to be controlled is traffic volume,which in turn affects the frequency of collisions.326JEAN ANDREY ET AL.To control forfluctuations in traffic volume,periods that were one week apart were matched by time of day and day of week.For example,a Tuesday afternoon period when rain was present would only be matched to a Tuesday afternoon,either one week prior to or one week afterward,when rain(or any other adverse weather) is not occurring.To determine the relative risk of a collision,the total number of collisions in the event periods was divided by the total number of collisions in the control periods.Thus,if20event-control pairs were selected,characterized by rain during the events and the absence of any adverse weather during the controls, and50collisions occurred during the events and40collisions occurred during the controls,the relative collision risk would be50divided by40,or1.25.Injury risk was calculated in a similar way,except each injured person was counted separately.Therefore,if in the above example,the50event collisions resulted in10injuries and the40control collisions resulted in5injuries,the relative injury risk would be10divided by5,or2.00.It is important to note that this calculation does not determine the absolute risk of a collision,since the collision and injury rates are unknown.The absolute risk could be estimated if the traffic volume was known at several representative points throughout each of the study areas.It would be expected that the absolute risk would produce a slightly different conclusion than the relative risk since less traffic would likely be on the roads during severe weather(Hassan and Barker, 1999).Thus the method used in this study should produce conservative estimates of weather-related risks.2.2.WEATHER DATAThe Meteorological Service of Canada(MSC)is the main source of weather data used to define events and controls.More specifically,three variables were used for this purpose:six-hour precipitation amounts,and instantaneous hourly obser-vations of both weather condition(e.g.,thunderstorm,snow,fog,blowing snow) and visibility.One of the challenges of the analysis is to match standardized observed weather conditions with those recorded on specific collision reports.Issues that were con-sidered include the location of weather observing stations relative to where a collision occurred,and both the accuracy and completeness of weather data.For all six case cities,the only suitable stations were located at airport sites which range from less than5km(Waterloo)to about35km(Halifax)from the city centre.While the MSC provides detailed information on weather conditions at observing stations, collisions occur throughout the cities in question.Therefore,collision reports were used to verify the representativeness of the station’s data for the city as a whole.Only a small subset of stations within the MSC weather observing network monitors the variables required in this analysis at the appropriate hourly temporal scale.Six-hour total precipitation variables were not available for Québec Airport from1996to1998.Hourly weather conditions and six-hour total precipitationWEATHER AS A CHRONIC HAZARD FOR ROAD TRANSPORTATION IN CANADIAN CITIES327 variables were not measured at Waterloo-Wellington Airport(1700–0500local time)and Hamilton Airport(2200–0500local time)stations over the study period. Attempts were made to estimate missing values using neighbouring station data; however the accuracy achieved was only deemed sufficient for temperature vari-ables.This limited the definitional criteria that could be used to define event-control pairs for comparing all six locations.2.3.COLLISION DATACollision data were obtained from the national collision database maintained by Transport Canada(TRAID3).Data were obtained for all reportable collisions that occurred within the six study areas for the years1995–1998.Separatefiles were provided for each urban area.The following elements were obtained:•Collision severity(i.e.,fatal,injury or property damage only),•Number of persons killed and injured,•Date and time(including day of the week),•Number of vehicles and persons involved,•Type of collision and collision configuration(e.g.,head-on collision between two motor vehicles),•Roadway characteristics(e.g.,urban/rural;intersection vs.non-intersection;divided vs.undivided;asphalt vs.gravel;state of repair;alignment,traffic control,speed limit),and•Weather,road surface and light condition.2.4.EVENT AND CONTROL PERIODSThe event and control periods used in the analysis were six hours in length in keeping with the six-hour precipitation summaries provided by e of a six-hour period allowed for dailyfluctuations in traffic volume to be addressed,yet kept the number of records to a manageable size.Various selection criteria were used for precipitation events–some based on both MSC weather data and collision reports of weather,and others based on collision reports only,as detailed in Tables III and IV.Because of weather data limitations,the entire range of scenarios was applied to only three of the study sites–Halifax-Dartmouth,Ottawa and Regina.Risk comparisons for all six cities were based on Scenario9,which requires collision data only.Precipitation events were divided into two types(rain and snow)as defined below:•A rain event was defined by any of the following precipitation types:thunder-storms,rain,rain showers,and/or drizzle.•A snow event was defined by any of the following precipitation types:snow, snow showers,snow pellets,snow grains,ice crystals,ice pellets,ice pellet328JEAN ANDREY ET AL.Table III.Selection criteria for event-control pairsEvent and control P HOP HOV Collision OtherOptionsEvent1≥0.4≥3n.a.≥50Event2≥0.2≥3n.a.≥50Event3≥0.2n.a.n.a.≥50Event4n.a.≥3n.a.≥50Event5≥0.2n.a.n.a.n.a.Event6n.a.n.a.n.a.≥50Event7n.a.n.a.n.a.≥75Control10≤1≤10Yes aControl20≤1≤10Yes bControl3n.a.n.a.n.a.0P:Six-hour precipitation total(water equivalent,mm).HOP:Num-ber of hourly observations of precipitation.HOV:Number of hourlyobservations of visibility<0.5km.Collision:Percentage of colli-sions that occurred during precipitation based on collision reports.a No police reports of icy pavement and no measurable precipitationfor6hours before the control.b No police reports of icy pavement.Table IV.Event-control scenariosEvent definition Control definitionoption optionScenario111Scenario221Scenario331Scenario432Scenario541Scenario651Scenario761Scenario871Scenario963Scenario1073showers or freezing rain.Note that any period during which both rain and snow occurred was classified as a snow event.It should also be noted that holidays or weekends associated with a statutory holiday were excluded,since traffic patterns would likely be altered during these periods.Holidays excluded were New Year’s Day,Easter,Victoria Day,St.Jean-Baptiste Day(Québec only),Canada Day,August Civic Holiday,Labour Day, Thanksgiving,Remembrance Day,Christmas and Boxing Day.Exclusion of other special days and events was considered,but these likely only have a very local effect on traffic,and therefore would not affect risk estimates.2.5.DATA PROCESSING PROCEDURESBecause of the large amounts of data involved,automation procedures were de-veloped for both defining the event-control pairs and also estimating risk.To accomplish this,a series of macros for Microsoft Excel was programmed using the Visual Basic Editor,and a graphical user interface was built so that a variety of criteria could be used to define event-control pairs.3.Results3.1.RISK LEVELSThefirst part of the results focuses on precipitation-related risk for the two general types of precipitation events–rain and snow.The results are based on several hundred event-control pairs for each of the three cities for which there were com-plete weather records,i.e.,Halifax-Dartmouth,Ottawa and Regina.The number of event-control pairs and number of collisions and injuries that occurred during the event-control pairs are summarized in Table V.Tables VI and VII contain best estimates of the relative risk of collision and injury,respectively,during precipitation versus control periods for all three cities combined.Overall,the risk of collision increases by approximately75percent (relative risk ratios between1.55and1.89)during precipitation.There is also a significant,though smaller(approximately45percent)increase in the relative risk of personal injury during precipitation relative to‘normal’seasonal driving conditions.The remainder of the discussion on Tables VI and VII is organized around the following three themes:(a)the sensitivity of the risk estimates to the criteria used to define event-control pairs;(b)differences in the relative risk of collision by type of precipitation event;and(c)the debate on whether safety is reduced during snowfall.Table V.Sample size for3cities combinedScenario#Event-#Collisions#Injuriescontrol pairs(n)1863794229492927834330863116898613709412851064339725929835830936188016238611971170987637168818595122039146713169470910101179382816Table VI.Relative risk of collision for3citiescombinedScenario Rain Snow Rain and snowcombined1 1.67 2.53 1.892 1.66 2.44 1.863 1.61 2.35 1.784 1.60 2.27 1.765 1.66 2.44 1.866 1.43 2.00 1.557 1.61 2.34 1.788 1.54 2.18 1.709 1.54 1.94 1.6910 1.48 2.09 1.693.1.1.The Sensitivity of the Risk Estimates to Definitional CriteriaPrevious studies(e.g.,Smith,1982;Brodsky and Hakkert,1988)have demon-strated that risk estimates vary when research methods are altered.Thus a stand-ardized approach must be applied if one wants to compare risks for different situations.Authors who have employed the matched-pair approach argue strongly that events and controls should have contrasting weather.Thus considerable care shouldTable VII.Relative risk of injury for3citiescombinedScenario Rain Snow Rain and snowcombined1 1.43 1.84 1.522 1.42 1.80 1.513 1.38 1.75 1.464 1.37 1.71 1.455 1.42 1.80 1.516 1.30 1.65 1.367 1.39 1.76 1.468 1.39 1.55 1.439 1.32 1.54 1.3810 1.35 1.53 1.41be taken to ensure that weather events are of sufficient duration and spatial ex-tent such that the majority of event collisions occur during the specified weather ing similar logic,there should be no evidence of inclement weather during the control.The types of definitional criteria associated with Scenarios1 and2are ideal in this way,but require both hourly weather data from a nearby observing station as well as complete collision reports of the weather condition at the accident scene.While collision reports were typically available for the vast majority of collisions at each of the study sites,nighttime hourly weather data were not available for Waterloo and Hamilton.Thus,less restrictive definitional criteria were applied to the data,including two scenarios based only on collision reports (Scenarios9and10).As shown in Tables VI and VII,the results are similar,but the relative risk is highest for Scenario1when the contrast between event and control is greatest.As the restrictions are lessened and scenarios are more inclusive,risk ratios are lower, but only marginally,despite the large increases in sample size.From a pure risk assessment perspective,the results of scenario1are preferred.But from a regional comparison perspective,the results of scenario9are used to allow comparisons to be made across the six urban areas.3.1.2.Differences in the Relative Risk of Collision by Type of Precipitation Event Despite the many studies on weather-related risk referred to in Table1,only a handful of studies compare risks during rainfall and snowfall for the same location. Exceptions include the early work of the Transport Road and Research Laboratory in Great Britain(Codling,1974),a study on the Indiana Tollway by Jovanis and Delleur(1983),and Canadian studies by Andrey(1989)and Suggett(1999).。