Alaska Climate Research Center, Geophysical Institute, University of Alaska Fairbanks, Fairbanks, Alaska

Temperature and precipitation of Alaska: 50 year trend analysis

J. M. Stafford, G. Wendler, and J. Curtis

With 8 Figures

Received August 30, 1999Revised March 21, 2000

Summary

Temperature and precipitation records from 1949 to 1998were examined for 25 stations throughout the State ofAlaska. Mean, maxima, and minima temperatures, diurnaltemperature range, and total precipitation were analyzed forlinear trends using least squares regressions. Annual andseasonal mean temperature increases were found through-out the entire state, and the majority were found to bestatistically signi®cant at the 95% level or better. Thehighest increases were found in winter in the Interior region(2.2 �C) for the 50 year period of record. Decreases inannual and seasonal mean diurnal temperature range werealso found, of which only about half were statisticallysigni®cant. A state-wide decrease in annual mean diurnaltemperature range was found to be 0.3 �C, with substan-tially higher decreases in the South/Southeastern region inwinter.

Increases were found in total precipitation for 3 of the 4seasons throughout most of Alaska, while summer preci-pitation showed decreases at many stations. Few of theprecipitation trends were found to be statistically signi®-cant, due to high interannual variability. Barrow, our onlystation in the Arctic region, shows statistically signi®cantdecreases in annual and winter total precipitation. These®ndings are largely in agreement with existing literature,although they do contradict some of the precipitation trendspredicted by the CO2-doubling GCM's.

1. Introduction

In recent decades, increasing air temperature hasbeen found throughout much of the NorthernHemisphere; it has been most pronounced in the

Arctic and Sub-Arctic regions (IPCC, 1992). Thisrate of climate change at higher latitudes has beenattributed in part to increased levels of greenhousegases (Environment Canada, 1995), and ispredicted by the General Circulation Models(e.g., Weller et al., 1995). Support for thesepredictions is provided by Chapman and Walsh(1993), while Stone et al. (1992) state for a shortertime period (1958±1986) that a potential green-house warming signal is not yet distinguishablefrom natural interannual climate variations.

Decreases were found in precipitation duringthe past few decades over northern Alaska and theArctic (Curtis et al., 1998; Warren et al., 1999);increases in temperature have been detected overnorthern latitudes by Weller et al. (1995); furtherWalsh and Chapman (1990) found a rise in Arcticsea level pressures. Such climate changes couldhave potentially severe consequences for manyaspects of life at high latitudes. Fluctuations in seaice, found by Chapman and Walsh (1993), couldhave a substantial impact on the ®shing industryand local wildlife. Likewise, permafrost thawing(Osterkamp et al., 1987) could cause manyengineering dif®culties, in addition to alteringthe ecological balance in affected areas (Weller etal., 1995). Although many recent temperature andprecipitation trend studies have focused on theentire Arctic or Northern Hemisphere (Wallaceet al., 1996; Chapman and Walsh, 1993; Karoly,

Theor. Appl. Climatol. 67, 33±44 (2000)

1989), or on small regions within Alaska (Mageeet al., 1999; Curtis et al., 1998), our studyexamines the trends for the entire State of Alaska.

Alaska encompasses 1.5 million km2 of land(Searby, 1968) and spans approximately 20degrees of latitude (Fig. 1). Consequently, itsclimate is dominated by large seasonal variationsin the amount of solar radiation received at thesurface. Local topography and proximity tooceans are also important factors that determineAlaska's climate (Benson et al., 1983). Addition-ally, throughout the year, several permanent andsemi-permanent pressure systems of varyingintensity alter Alaska's atmospheric circulation.These systems include the Arctic and Siberianhighs, which affect the northern and interiorportions of the state, and the Aleutian low(Overland et al., 1999), which affects the southernand western portions (Martyn, 1992). The ElNinÄo-Southern Oscillation (ENSO) has also beenshown to cause large interannual variation inAlaska's climate (Hess et al., in press). Due tothese varying conditions and in¯uences through-out the state, we have divided Alaska into fourclimate regions: South/Southeastern, Western,Interior, and Arctic.

The Arctic region is the coldest and the driest,with a KoÈppen classi®cation of ET and frequent

high winds. Snow accounts for approximately75% of the total annual precipitation in the Arcticregion (Searby, 1968). In the Interior (Sub-Arctic)region, classi®ed as Dfc, the summer temperaturescan reach as high as 30 �C or higher, while thewinter temperatures are frequently below ÿ40 �Cwith common strong shallow inversions. TheWestern region is transitional between continentaland coastal, and is in¯uenced by cold interior air,¯uctuations in sea ice extent, and low pressuresystems from the Bering Sea. Severe winters bringhigh winds, and summer temperatures rarely reachhigher than 10 �C. For most of Alaska, precipita-tion is higher in summer and autumn than inwinter and spring. The reverse is true in portionsof the South/Southeastern region, although theregion is generally wet all year. The coastal/maritime climate moderates both winter andsummer temperatures in this region, and stationclimate classi®cations within the region rangefrom Cfb to Dfc.

Our regional divisions, shown in Fig. 1, werebased primarily on latitude, longitude, and geo-graphic boundaries. Two mountain ranges, theAlaska and Brooks ranges, provide the principleboundaries between the South/Southeastern andInterior regions, and between the Interior andArctic regions, respectively. Somewhat less nat-

Fig. 1. Meteorological stations, cli-matic zones and aereal relief ofAlaska

34 J. M. Stafford et al.

ural was the de®nition of the fourth, the Westernregion, which is not based on clearly establishedtopographic boundaries. This region is affected attimes by the continental climate of InteriorAlaska, at other times by the maritime climateof the adjoining Bering Sea. However, statisticalanalyses of the temperature showed that it did not®t well in either of two climatic zones, the South/Southeastern and Interior regions. These regionswere more uniform in regard to temperaturewithout the Western stations; hence, it deservedto be a zone by itself. Now, cross correlationsbetween the stations for each climatic zone gavethe relatively high mean annual values between0.8 and 0.9. For all regions, the stations were mostuniform in winter, and least in summer. This is tobe expected given the increased geographicalvariability of temperatures during the summer andfall, when solar radiation and cloudiness areprimary factors, as opposed to winter and spring,when there is little or no solar radiation through-out the state and cloudiness is at an annualminimum (Dissing and Wendler, 1998). Snow

cover also contributes to the uniformity of thetemperatures during the winter season. The Arcticregion was not included in this analysis, given thefact that Barrow is our only Arctic station. BarterIsland, the other long-term meteorological stationin the Arctic, ceased operating in 1988, and henceis not included in our study. Intercomparing theestablished climatic zones, the interior of Alaskadisplayed the least degree of temperature variationbetween the stations. This is understandable, asthis area is sheltered from both the South (AlaskaRange) and North (Brooks Range), hence theclimate is more uniform.

Figure 2 is a contour plot of the mean annual airtemperatures for Alaska and shows the 25 stationsused in this study. Over the oceans, lines wereextrapolated, given the absence of oceanic data. Itshould be noted that none of the stations are athigh altitude, thus alpine in¯uences are eliminatedin the ®gure. The isotherms run largely along linesof latitude and are approximately parallel the twomountain ranges, lending support to our choice ofregional divisions.

Fig. 2. Mean annual air tempera-tures (�C) for the State of Alaskabased on the 25 study sites. Notethat none of the stations aremountainous (highest elev.Northway at 527 m), thus elim-inating alpine in¯uences in the®gure

Temperature and precipitation of Alaska: 50 year trend analysis 35

2. Analyses

2.1 Data sets

Temperature and precipitation data sets used inthis study were compiled from the followingsources: Alaska Climate Research Center, NOAANational Data Center, National Weather ServiceForecast Of®ce, Western Regional Climate Cen-ter, and NOAA Local Climatological Data. Ourdata sets represent the 50 year period from 1949through 1998. The 50 year period of record re-presents the longest period for which relativelygood quality data are available for a large numberof stations. Of course, some location and instru-mentation changes have occurred throughout thisperiod; nevertheless, trends in climatic variableswithin regions were consistent, lending trust to thequality of the data.

Our choices for the 25 stations were based onthe quality and continuity of the data. Data weregathered in the form of monthly maxima, minimaand mean temperatures, and monthly total pre-cipitation (rain plus water equivalence of snow).For temperature, no month was used if more thanten days of data were missing (no more than threedays for the transitional months: March, April,September, and October, which could have rapidshort-term warming or cooling). For precipitation,no month was used if even one day was missing.In the ®nal temperature data set, less than 1% ofthe months were missing (with 2% being thehighest value for a single station), and approxi-mately 1% were missing from the precipitationrecord (with 7% being the highest value for asingle station). In the case of missing tempera-tures, the mean value for the month over theperiod of record was calculated and inserted inplace of the missing value. Missing precipitationdata were treated somewhat differently. A nearbystation was used as a proxy by calculating themean difference between the stations for themonth in question, adjusting the value for theproxy site accordingly, then inserting it in place ofthe missing value.

2.2 Variables and statistics

Several climatic variables were analyzed in thisstudy: mean, maxima, and minima temperatures,diurnal temperature range, and total precipitation.

Temperature and precipitation are two majorclimate variables which display a relatively highreliability and availability in climatic records.Monthly values were averaged or summed (fortemperature and precipitation, respectively) toarrive at seasonal and annual values. Seasonswere de®ned using the standard meteorologicalde®nition (i.e. winter�Dec±Feb, spring�Mar±May, etc.).

For each variable, seasonal and annual timeseries graphs were plotted for our 50 year periodof record. A linear least-squares regression linewas then ®tted to each time series, and the changewas calculated from the end points of theregression line. Throughout this study, �TMean

refers to the change over the past 50 years in themean temperature, DTR refers to the change overtime in diurnal temperature range, and �P refersto the change over time in the total precipitation.In the case of the precipitation, percent changes(�P%) were also calculated. Finally, two-wayanalyses of variance (pure model I ANOVA's withreplication) were carried out to check for seasonaland/or regional effects in �TMean, DTR, and�P%. The Arctic region was omitted from theANOVA's due to its lack of multiple sites. For thesake of establishing equal sample sizes among theregions, ®ve stations were randomly sampled fromthe 14-station South/Southeastern region. Threedifferent samples were taken, and the tests wererepeated for each sample.

3. Results

3.1 Mean temperature

Figure 3 is a plot of the mean annual maxima,minima, and mean temperatures for Talkeetna,AK. Notice that the increases are generally linearover time, with scatter due mainly to interannualvariability. The increase in mean temperatureshown here is signi®cant at the 99.9% level. Notethat the temperature increase of the minima ismore pronounced than the maxima, a resultexpected due to increased greenhouse gases.Talkeetna is presented as an example, and in Figs.3 and 4 contour plots for all of Alaska arepresented, showing the �TMean annually andseasonally, respectively. Figure 4 shows that theincreases are highest in the interior of the state(2.2 �C), with a secondary but less extensive high

36 J. M. Stafford et al.

Fig. 3. Annual mean maxima, minima, and mean temperatures for Talkeetna, AK from 1949 to 1998 with linear least squaresregression lines. Maxima and minima trend lines are converging, resulting in a decreasing diurnal temperature range.Temperature increase and decrease in range are both statistically signi®cant at the 99.9% level

Fig. 4. �TMean (�C) for annualtemperature over the 50 yearperiod from 1949 to 1998. High-est increases are found in theInterior region (2.2 �C), with asecondary high point in the South/Southeastern region (2.0 �C).Contours are derived using asimple interpolation between sta-tions and does not consider terrain(i.e. land or water)

Temperature and precipitation of Alaska: 50 year trend analysis 37

in the southeast (2.0 �C). The average value for theInterior region was 1.9 �C, while the Westernregion showed the lowest average increase of1.3 �C. Similar trends are shown on the winter andspring maps (Fig. 5a and 5b): winter, 4.1 �C and2.2 �C, spring, 2.9 �C and 1.7 �C, for Interior andWestern regions, respectively. In the summerseason (Fig. 5c), the South/Southeastern regiondisplays the lowest average temperature increase(0.9 �C), but the �TMean is low throughout theentire state. Autumn (Fig. 5d) is the only season inwhich negative temperature trends are found, andall of the trends were of very small magnitude.The state-wide average �TMean for autumn isÿ0.2 �C.

Signi®cance tests of the individual regressionlines for each station yielded mixed results. In thecase of mean annual temperature (Fig. 4), all butone of the stations showed a signi®cant increase atthe 95% level or better. St. Paul Island was theexception, but this is not surprising due to theextremely strong oceanic moderation of theclimate there. For the winter season, all but fourof the sites showed signi®cant regression lines(exceptions were Kodiak, Wrangell, Cold Bay and

St. Paul Island, which are all relatively coastal).All stations showed signi®cant regressions forspring, except for the Western region where noneof the regression lines were signi®cant. Forsummer, approximately 66% of the stationsshowed signi®cant regression lines, and the non-signi®cant sites were scattered throughout theSouth/Southeastern and Interior regions. None ofthe stations showed signi®cant regression lines forautumn, most likely due to shallow slopes, ratherthan excess scatter in the data. A two-wayANOVA of the �TMean showed a signi®cant(99% level or better) seasonal effect as well as asigni®cant regional effect. The interaction termwas also signi®cant at the same level. Each of thethree repetitions of the ANOVA (as describedabove) yielded the same results, showing homo-geneity among the stations in the South/South-eastern region.

3.2 Diurnal temperature range

By subtracting the mean monthly minimatemperatures from the corresponding maximatemperatures, then averaging them annually and

Fig. 5. As with Fig. 3, but forseasonal temperatures. Winter(a) shows the highest increases,followed by spring (b) andsummer (c). Negative values(decreases in temperature) areshown for autumn (d)

38 J. M. Stafford et al.

seasonally, we obtained mean diurnal tempera-ture ranges. These ranges were tested in the samemanner as the mean temperatures. In Fig. 3,notice that both the maxima and minimatemperatures are increasing, but that the mini-mum temperature is increasing slightly morerapidly than is the maximum. Consequently, theannual mean diurnal temperature range forTalkeetna, AK is decreasing with time. Thisdecrease is signi®cant at the 99.9% level. As withmean temperature, DTR was calculated annuallyand seasonally for each of the 25 stations. Winterdisplayed the most pronounced decrease indiurnal temperature range (Fig. 6). From theDTR at the individual stations, regional meanswere calculated (Table 1).

Winter is the only season in which all regionsshowed a decrease in diurnal temperature range;the South/Southeastern region had the highestvalue, with a decrease of 1.8 �C. However, Fig. 6shows that this region is not uniform, and forKodiak- and Baranof- Islands, the diurnal tem-perature range increased. For all other seasons, aswell as for the annual values, there are bothpositive and negative DTR values present amongthe regions. The state-wide annual average DTR isÿ0.3 �C.

Upon testing the regression lines for signi®-cance, we found that only about half of thestations showed signi®cant trends in the DTR, atthe 95% level or better, for any given season.Winter, for example, showed 11 signi®cant regres-

Fig. 6. DTR (�C) for winter overthe 50 year period form 1949 to1998. The greatest decrease canbe seen in the South/Southeast-ern region (ÿ2 �C), but increasesare also seen in the same region

Table 1. DTR (�C) from 1949 to 1998: Annual and seasonal means for each region

Annual Winter Spring Summer Autumn

South/Southeastern ÿ0.9 ÿ1.8 ÿ0.6 ÿ0.8 ÿ0.5Western ÿ0.5 ÿ0.9 ÿ1.1 0.2 0.0Interior 0.0 ÿ0.6 0.5 0.1 0.2Arctic 0.4 ÿ0.9 ÿ0.2 1.8 0.7

Temperature and precipitation of Alaska: 50 year trend analysis 39

sions and 14 non-signi®cant. Two of these signif-icant trends, Kodiak and Little Port Walter, wereincreases while the rest were decreases. Only fourstations, Anchorage, Juneau, Seward, and Talk-eetna, showed signi®cant decreases for all sea-sons. Kodiak and Little Port Walter both showedsigni®cant increases for all seasons. The non-signi®cance of so many of the DTR values, maybe explained by the fact that normally the maximaand minima trend lines show parallel slopes,accentuating the �TMean, and minimizing theDTR. Therefore, we suggest that the non-signi®cance is due to the small magnitude of theslopes, rather than to a large amount of scatter. Atwo-way ANOVA of the DTR was performed, butno signi®cant regional or seasonal effects weredetected.

The decrease in the DTR for Alaska over thelast 5 decades is in general agreement with resultsfound for the Northern Hemisphere (Karl et al.,1993). They found that the decrease in the DTR isapproximately equal to the increase in meantemperature. In our case the observed warmingwas somewhat larger than the decrease in theDTR. Further, the maximum temperature increasewith time was found in the interior, while themaximum decrease in DTR was found in theTalkeetna area, south of the Alaska Range.Hansen et al. (1997a, b) examined the sensitivityof climate models to a wide range of radiativeforcings, and Rebetez and Beniston (1998) studiedthe decreases in sunshine duration and DTR forSwitzerland. The latter found a good correlationbetween increased cloudiness and decreased DTRfor low lying stations. There is no systematicstudy on cloudiness trends of Alaska, however, innorthern Alaska (Barrow) (Curtis et al., 1998) andInterior Alaska (Fairbanks) a decrease in cloudi-ness has been observed. The observed decrease inDTR is small for both climatic zones. In contrastto this, South Alaska (Anchorage) experienced anincrease in cloudiness. The mean cloudinessincreased for the 50 years from 68% to 72%.This is also the region with the largest decrease inDTR.

Nevertheless, to explain a decrease in DTR forInterior and Northern Alaska with decreasingcloud amount, other radiative forcing mechanism,most likely increased CO2, changes in theintensity of Arctic haze (Shaw, 1995) and theconcentration of ozone have contributed to the

observed decrease in DTR. Another possibleculprit might be the atmospheric circulation. ForNorthern Alaska, we observed a temperatureincrease with a decrease in cloudiness, a trendwas also well pronounced in winter and spring,when clouds tend to warm the surface. Weexplained this counterintuitive result by changesin circulation (Curtis et al., 1998). A higher amountof water vapor in the atmosphere could be res-ponsible for the observed temperature increase ata constant cloudiness; such an increase in watervapor would also decrease the DTR.

3.3 Precipitation

Figure 7 is a smoothed contour plot similar to Fig.2, but for mean annual total precipitation. Note thedramatic difference between the northern portionof the state (< 200 mm) and the high point in thesoutheast at Little Port Walter (5577 mm). Of the25 stations used in this study, Barrow and LittlePort Walter represent the extremes of precipita-tion, with the latter receiving about 50 times theamount recorded at the former. The majority ofthis difference results from the large amount ofmoisture brought into the southern coastal regionsby low pressure systems. Coastal uplift in thesouthern portion of the state further increases theprecipitation there, while decreasing the amountof moisture that can reach into the Interior andArctic regions. Precipitation values for the Interiorregion range from approximately 200 mm to over400 mm, small values to be sure, yet not as smallas the 114 mm observed in the Arctic. Time seriesplots for total precipitation were generated, andthe changes were expressed as percentages of themean for the 50 year period. Annual and seasonalmeans were calculated for each region (Table 2),although we observed much variability among thestations. Precipitation in most of Alaska hasincreased over the past 50 years, with the largestincrease (25%) found in winter in the Westernregion. Only in the Arctic region is a decrease inprecipitation found throughout the year (ÿ36%for the total annual amount), while the Interiorgenerally displays the smallest changes. Theabsolute measurements of precipitation, especiallyat Barrow in winter, where the wind is moderate tostrong, is often underestimated. This has beenshown by Clagett (1988), who used Wyomingwindshield gauges. However, the data set from the

40 J. M. Stafford et al.

standard 8-inch precipitation gauge is all that isavailable for Barrow for the entire 50 year periodused in this study. Given that the instrumentationhas been consistent, we have much more con®-dence in the trend analysis than in the absoluteprecipitation values.

Figure 8 is a comparison of the total annualprecipitation at two of our study stations, the mostnorthern (Barrow) and a southern station (Seward)at approximately the same longitude. Barrowdisplayed a decrease of 36%, while Sewarddisplayed an increase in precipitation of 17% forthe 50 year period; the mean increase of theSouth/Southeastern region is 10%. Notice thelarge amount of interannual variability in both ofthe time series plots, as well as the opposing

slopes of the regression lines. At the 95% level ofcon®dence, the Barrow regression line is signi®-cant, while the one for Seward is not.

In the vast majority of cases, signi®cance testsof the trends in total precipitation yielded non-signi®cant results. Signi®cant decreases werefound for Barrow in the annual and winter totals.Yakutat is the only station for which signi®canttrends were found for every season; all of themwere positive. In light of this large amount ofnon-signi®cance in the individual regressionlines, it is not surprising that the ANOVA alsoyielded non-signi®cant values for the differencesamong regions and among seasons. Excessscatter, caused by high interannual variability isbelieved to be the primary cause of the non-

Fig. 7. Mean annual total pre-cipitation (mm) for the State ofAlaska based on the 25 studysites. Notice the dramatic con-trast between the northern coast(< 200 mm) and the high point inthe southeast at Little Port Wal-ter (5577 mm)

Table 2. �P% from 1949 to 1998: Annual and seasonal means for each region

Annual Winter Spring Summer Autumn

South/Southeastern 10 22 12 ÿ1 10Western 9 25 17 ÿ11 21Interior 7 3 6 1 19Arctic ÿ36 ÿ106 ÿ57 ÿ16 ÿ36

Temperature and precipitation of Alaska: 50 year trend analysis 41

signi®cance in the precipitation trends. Precipita-tion is also highly variable geographically (Fig.7), as it is dependent on many local factors suchas proximity to the ocean, relative position ofmountain ranges, and local topography. These alltend to increase the variance within subgroups inour ANOVA, and thus prevent signi®cant results.

4. Discussion

Our ®ndings regarding mean temperature are ingood agreement with recent research. Chapmanand Walsh (1993) also found warming to be mostpronounced in winter and spring, and least insummer. This pattern is also predicted by GCM'sbased on a doubling of CO2 (Weller et al., 1995).Karoly (1989) also found evidence that couldsupport anthropogenic climate forcing viaincreases in greenhouse gases, but he concedesthat there are other factors that could haveproduced similar results. Our ®ndings of decreas-ing temperature in autumn cannot be explained interms of greenhouse warming. The results of the

ANOVA suggest that the �TMean is not homo-geneous throughout the state, and that the regionaleffects differ by season.

Analyses of diurnal temperature ranges haveyielded mixed results, such as those seen in thisstudy, and in previous literature as well. The IPCC(1992) reported ®ndings of increases in the dailyminima temperatures, and little or no increase inthe daily maxima temperatures, resulting in adecrease in the diurnal temperature ranges for theArctic. However, ®ndings of increases in diurnalranges were also reported in the same publication.Our results are roughly consistent with theproposed pattern of decreasing DTR, althoughwe found several cases in which the decreaseresulted from decreases in both the maxima andminima temperatures, or from decreases inmaxima but not minima.

GCM's based on a doubling of CO2 predict thatincreases in Arctic temperatures will be accom-panied by increases in precipitation for bothsummer and winter (Weller et al., 1995; Environ-ment Canada, 1995). Our results partially support

Fig. 8. Comparison of time series plots (1949 to 1998), with linear least squares regression lines, of total annual precipitation(mm) for Barrow and Seward, AK. Note the large amount of interannual variability at both stations, as well as the opposingslopes of the regression lines

42 J. M. Stafford et al.

this prediction. We found precipitation increasesin winter for 3 of our 4 regions (South/South-eastern, Western and Interior), but we also founddecreases in summer precipitation for 3 of our 4regions (South/Southeastern, Western and Arctic).Our ®ndings of decreased winter precipitation onthe Arctic coast are consistent with ®ndingsreported by Warren et al. (1999) and by Curtiset al. (1998). Additionally, Tao et al. (1996) founda signi®cant inverse correlation between summertemperature and summer cloud cover over theArctic Ocean, lending credence to our ®nding ofdecreased summer precipitation throughout muchof Alaska.

Incomplete agreement between the CO2-dou-bling GCM's and the observed atmosphericchanges forces us to search for other possiblesources of variability that could either mask oraugment the effects of anthropogenic greenhousegas increases. Zhang et al. (1996) proposed onesuch source when they reported cycles in Arctictemperature and precipitation on a roughlydecadal scale. Similar cyclical patterns werefound by Juday (1984). Urban heat islands canenhance the possible effects of greenhousewarming (Benson et al., 1983), and can accountfor up to 20 percent of the observed warming incertain locations (Magee et al., 1999). Addition-ally, interrelationships and feedback between seaice extent and the corresponding air temperaturesare another source of variability in the climaticrecords. Sellers (1969) found a substantial feed-back at work in the Northern Hemisphere, and healso found that albedo changes in the SouthernHemisphere could have potentially severe reper-cussions in the Arctic. However, Ingram et al.(1989) caution us that the method used to estimatethe surface albedo feedback can alter the apparentstrength of the mechanism.

5. Conclusion

Increases in mean annual air temperature over the50 year period from 1949 to 1998 were found forall stations in this study, with the highest increaselocated in the Interior (2.2 �C). Seasonally, theincreases were highest in winter and lowest insummer. Autumn was the only season in whichtemperature decreases were observed. The major-ity of the annual and seasonal increases werestatistically signi®cant at the 95% level or higher.

None of the autumn decreases were statisticallysigni®cant.

Diurnal temperature range has generallydecreased over the period covered in this study,with a decrease of 0.3 �C found in the state-wideannual mean. Approximately 50% of the stationsshowed statistically signi®cant trends for anygiven season, but not all of these followed theexpected pattern of slowly increasing maximatemperatures and more rapidly increasing minimatemperatures.

In the precipitation record, the majority ofstations showed increases for winter, spring, andautumn. Decreases in summer precipitation werefound for many stations, which, although notstatistically signi®cant, can be at least partiallycorroborated with existing literature. Statisticallysigni®cant precipitation decreases were foundfor the Arctic region in the annual and wintertotals.

Acknowledgments

This study was supported by a grant to the Alaska ClimateResearch Center at the University of Alaska by the State ofAlaska and by a National Science Foundation REU Grant tothe Geophysical Institute. Blake Moore helped with thegraphics and Gunter Weller read the manuscript: thanks toboth of them. Further, our thanks go to the two to usunknown reviewers of this paper; their kind and knowledge-able comments improved the manuscript.

References

Benson CS, Bowling SA, Weller G (1983) Urban climates inAlaska. Environments 15(2): 23±26

Chapman WL, Walsh JE (1993) Recent variations of sea iceand air temperature in high latitudes. Bull Amer MeteorSoc 74(1): 33±47

Clagett GP (1988) The Wyoming windshield ± an evaluationafter 12 years of use in Alaska. Proceedings of the WesternSnow Conference, Kalispell, Montana, 1988, ColoradoState University, Fort Collins, Colorado, 113±123

Curtis J, Wendler G, Stone R, Dutton E (1998) Precipitationdecrease in the western Arctic, with special emphasis onBarrow and Barter Island, Alaska. Int J Climatol 18:1687±1707

Dissing D, Wendler G (1998) Solar radiation climatology ofAlaska. Theor Appl Climatol 61: 161±175

Environment Canada (1995) The State of Canada's Climate:Monitoring Variability and Change. SOE Report No 95-1.Report for the Minister of Public Works and GovernmentServices Canada, 52 pp

Hansen J, Sato M, Lacis A, Ruedy R (1997a) The missingclimate forcing. Phil Trans, Royal Society London, B 352:231±240

Temperature and precipitation of Alaska: 50 year trend analysis 43

Hansen J, Sato M, Ruedy R (1997b) Radiative forcing andclimatic response. J Geophys Res 102(D6): 6831±6864

Hess J, Scott C, Hufford G, Fleming M (2000) El NinÄo andits impact on ®re weather conditions in Alaska. Int JWildland Fire (In Press)

Ingram WJ, Wilson CA, Mitchell JFB (1989) Modelingclimate change: an assessment of sea ice and surfacealbedo feedbacks. J Geophys Res 94(D6): 8609±8622

IPCC (1992) Climate Change 1992: the SupplementaryReport to the IPCC Scienti®c Assessment. WMO/UNEP.Cambridge: Cambridge University Press, 200 pp

Juday GP (1984) Temperature trends in the Alaska climaterecord. Proceedings of the Conference on the PotentialEffects of Carbon Dioxide-Induced Climatic Changes inAlaska, Ag Exp Sta, Miscellaneous Publication 83-1,University of Alaska, Fairbanks, Alaska, 76±88

Karl T, Jones P, Knight R, Kukla G, Plummer N, RazuvayevV, Gallo K, Lindseay J, Charlon R, Peterson T (1993) Anew perspective on recent global warming: Asymetrictrends of daily maximum and minimum temperature.Bams 74(6): 1007±1023

Karoly DJ (1989) Northern Hemisphere temperature trends:a possible greenhouse gas effect? Geophys Res Lett 16(5):465±468

Magee N, Curtis J, Wendler G (1999) The urban heat islandeffect at Fairbanks, Alaska. Theor Appl Climatol 64:39±47

Martyn D (1992) Climates of the World. (Developments inAtmospheric Science 18). Amsterdam: Elsevier, 435 pp

Osterkamp TE, Gosink JP, Kawasaki K (1987) Measure-ments of permafrost temperatures to evaluate the conse-quences of recent climate warming. Geophysical Institute,University of Alaska. Report for Alaska Department ofTransportation and Public Facilities, 16 pp

Overland JE, Adams JM, Bond NA (1999) Decadal varia-bility of the aleutian low and its relation to high-latitudecirculation. J Climate 12(5): 1542±1548

Rebetez M, Beniston M (1997) Changes in sunshine dura-tion are correlate with changes in daily temperature rangethis century: An analysis of Swiss climatological data.Geophys Res Lett 25(19): 3611±3613

Searby HW (1968) Climates of the States: Alaska. ESSANo 60-49. Report for the US Department of Commerce,22 pp

Sellers WD (1969) A global climatic model based on theenergy balance of the Earth-Atmosphere system. J ApplMeteorol 8(3): 392±400

Shaw GE (1995) The Arctic Haze Phenomenon. BAMS,76(12): 2403±2413

Stone RS, Kahl JD, Serreze MC, Schnell RC (1992) Trendanalyses of tropospheric temperatures in the Arctic ± isthere evidence of greenhouse warming? Proceedings ofthe Third Conference on Polar Meteorology and Oceano-graphy, Portland, Oregon, AMS, Boston, Massachusetts,137±140

Tao X, Walsh JE, Chapman WL (1996) An assessment ofglobal climate model simulations of Arctic air tempera-tures. J Climate 9(5): 1060±1076

Wallace J, Zhang Y, Bajuk L (1996) Interpretation of inter-decadal trends in Northern Hemisphere surface air tem-perature. J Climate 9(2): 249±259

Walsh JE, Chapman WL (1990) Short-term climatic varia-bility of the Arctic. J Climate 3(2): 237±250

Warren SG, Rigor RG, Untersteiner N, Radionov VF, Bryaz-gin NN, Aleksandrov YI, Colony R (1999) Snow depth onArctic sea ice. J Climate 12(6) 1814±1829

Weller G, Lynch A, Osterkamp T, Wendler G (1995) Climatechange and its effects on the physical environment ofAlaska. Proceedings of the Arctic Science Conference,Fairbanks, Alaska, 1995, The Center for Global Changeand Arctic Systems Research, University of Alaska, Fair-banks, Alaska, 5±13

Zhang T, Osterkamp T, Stamnes K (1996) Some character-istics of the climate in northern Alaska, USA. Arctic andAlpine Research 28(4): 509±518

Author's addresses: J. M. Stafford (e-mail: jstafford@rock-etmail.com), Alaska Paci®c University, 4101 UniversityDrive, Box #429, Anchorage, AK 99508-4672; G. Wendler,J. Curtis, Alaska Climate Research Center, GeophysicalInstitute, University of Alaska Fairbanks, Fairbanks, Alaska99775-7320.

44 J. M. Stafford et al.: Temperature and precipitation of Alaska: 50 year trend analysis