A Demonstration That Large-Scale Warming Is Not Urban

Abstract

On the premise that urban heat islands are strongest in calm conditions but are largely absent in windy weather, daily minimum and maximum air temperatures for the period 1950–2000 at a worldwide selection of land stations are analyzed separately for windy and calm conditions, and the global and regional trends are compared. The trends in temperature are almost unaffected by this subsampling, indicating that urban development and other local or instrumental influences have contributed little overall to the observed warming trends. The trends of temperature averaged over the selected land stations worldwide are in close agreement with published trends based on much more complete networks, indicating that the smaller selection used here is sufficient for reliable sampling of global trends as well as interannual variations. A small tendency for windy days to have warmed more than other days in winter over Eurasia is the opposite of that expected from urbanization and is likely to be a consequence of atmospheric circulation changes.

Full text

A Demonstration That Large-Scale Warming Is Not Urban
DAVID E. PARKER
Hadley Centre, Met Office, Exeter, United Kingdom
(Manuscript received 14 March 2005, in final form 25 July 2005)
ABSTRACT
On the premise that urban heat islands are strongest in calm conditions but are largely absent in windy
weather, daily minimum and maximum air temperatures for the period 1950–2000 at a worldwide selection
of land stations are analyzed separately for windy and calm conditions, and the global and regional trends
are compared. The trends in temperature are almost unaffected by this subsampling, indicating that urban
development and other local or instrumental influences have contributed little overall to the observed
warming trends. The trends of temperature averaged over the selected land stations worldwide are in close
agreement with published trends based on much more complete networks, indicating that the smaller
selection used here is sufficient for reliable sampling of global trends as well as interannual variations. A
small tendency for windy days to have warmed more than other days in winter over Eurasia is the opposite
of that expected from urbanization and is likely to be a consequence of atmospheric circulation changes.
1. Introduction
There have been several attempts in recent years to
estimate the urban warming influence on the large-
scale land surface air temperature record. Jones et al.
(1990) found that the urban warming influence on
widely used hemispheric datasets is likely to be an or-
der of magnitude smaller than the observed century-
time-scale warming. Easterling et al. (1997) found that
global urban warming influences were little more than
0.05°C century
⫺1
over the period 1950–93. Hansen et al.
(1999) concluded that the anthropogenic urban contri-
bution to their global temperature curve for the past
century did not exceed approximately 0.1°C. Further-
more, they estimated the global average effect of their
urban adjustment during 1950–98 as only 0.01°C (see
their Plate A2), though their adjustment procedure re-
moved an urban influence of nearly 0.1°C in the con-
tiguous United States in this period. Peterson et al.
(1999) compared global temperature trends from the
full Global Historical Climatology Network with a sub-
set based on rural stations, defined as such both by map
metadata and by nighttime lighting as detected by sat-
ellites. They found that the rural subset and the full set
had very similar trends since the late nineteenth cen-
tury, and inferred that the urban influence on the full
set was therefore insignificant. Accordingly, only small
systematic errors were ascribed to urbanization in the
global warming trend estimates made by Folland et al.
(2001a) and in the Intergovernmental Panel on Climate
Change Third Assessment Report (IPCC TAR; Fol-
land et al. 2001b).
Nevertheless, controversy has persisted. Following
IPCC TAR, Hansen et al. (2001) upgraded the God-
dard Institute for Space Studies (GISS) analysis and
found that overall urban adjustments to data for the
contiguous United States required since 1900 exceeded
0.1°C but that the adjustment differed according to the
dataset used. Furthermore, adjustments for other
changes such as observing time, siting, and instrumen-
tation were typically as large as or larger than, and of
opposite sign to, the urban adjustment. Kalnay and Cai
(2003) used the difference between trends in observed
surface temperatures in the continental United States
and corresponding trends in the National Centers for
Environmental Prediction–National Center for Atmo-
spheric Research (NCEP–NCAR) reanalysis, which
does not use the surface temperature observations, to
infer 0.27°C century
⫺1
surface warming due to land use
changes since 1950. This was contested because the
NCEP–NCAR reanalysis did not include cloudiness
data and had errors in its surface heat budget (Tren-
berth 2004), and because the surface air temperature
observations used by Kalnay and Cai (2003) had not
Corresponding author address: Mr. David Parker, Hadley Cen-
tre, Met Office, FitzRoy Rd., Exeter EX1 3PB, United Kingdom.
E-mail: david.parker@metoffice.gov.uk
2882 JOURNAL OF CLIMATE VOLUME 19
JCLI3730

been adjusted for heterogeneities such as changes of
instrumentation or observing time (Vose et al. 2004).
Subsequently, Zhou et al. (2004) used an improved ver-
sion of the NCEP–NCAR reanalysis and a method
similar to Kalnay and Cai (2003) to estimate 0.05°C
decade
⫺1
urban warming in the winter during 1979–98
over southeast China where the surface air temperature
data were found to be homogeneous in regard to in-
strumental and other nonclimatic changes. However,
this urban warming was only about 10% of the total
warming found by Zhou et al. (2004), and the authors
point out that the urban warming effect will be smaller
in summer owing to greater cloud cover. Simmons et al.
(2004) were unable to find support for Kalnay and Cai’s
conclusions in the 40-yr European Centre for Medium-
Range Weather Forecasts (ECMWF) Re-Analysis
(ERA-40), which had almost as much warming over the
eastern United States near the 850-hPa level as in the
Jones and Moberg (2003) surface air temperature
analysis. Furthermore, Peterson (2003) found no statis-
tically significant impact of urbanization in an analysis
of 289 stations in 40 clusters in the contiguous United
States, after the influences of elevation, latitude, time of
observation, and instrumentation had been accounted
for. One possible reason for this finding was that many
“urban”observations are likely to be made in cool
parks, to conform to standards for siting of stations.
Peterson and Owen (2005) noted that the type of meta-
data (population versus night lights) made significant
differences to urban versus rural classification for sta-
tions in the United States, but that the omission of the
stations classified as urban by both schemes (popula-
tion ⬎30 000 within 6 km) made very little difference to
the overall warming trend in the United States Histori-
cal Climatology Network (USHCN) since 1931.
The influence of urbanization on air temperatures is
greatest on calm, cloudless nights and is reduced in
windy, cloudy conditions (Johnson et al. 1991). Al-
though the main effect of urbanization is on nighttime
temperatures, there is some evidence of lesser effects
by day (Arnfield 2003). Biases arising from poor instru-
mental exposure and micrometeorological effects are
also likely to be greatest in calm, cloudless conditions,
by day as well as by night (Parker 1994). The effects of
solar and longwave radiation inside thermometer shel-
ters, specifically the resulting temperature differences
between thermometers, their housing, and the air, are
minimized in windy weather (Lin et al. 2001). Time-of-
observation biases (Karl et al. 1986) are enhanced in
calm, cloudless conditions when diurnal temperature
ranges are large and reduced in windy, cloudy weather
when diurnal temperature ranges are small.
Accordingly, an analysis of trends of worldwide land
surface air temperature on windy, cloudy days and
nights is most likely to be free of urban biases, and also
to be less affected by the instrumental, siting, and pro-
cedural biases that had to be treated by Hansen et al.
(2001) and Peterson (2003) in their analyses of urban
warming. Comparison of such an analysis with trends
based on all data may yield a new estimate of the size of
the systematic errors to be accorded to urbanization in
estimates of global warming over land. Instrumental
measurements of wind strength at temperature observ-
ing stations are not generally available, but gridded
near-surface wind components are available from the
NCEP–NCAR reanalysis (Kalnay et al. 1996) on a
6-hourly basis since 1948. These near-surface winds are
constrained by the generally reliable mean sea level
pressure observations. Cloud observations at tempera-
ture observing stations are also generally unavailable,
and the NCEP–NCAR reanalysis cloudiness data are
highly model dependent (Kalnay et al. 1996).
Therefore, daily land surface air temperatures since
the mid–twentieth century from a selection of stations
worldwide have been analyzed separately for windy
and calm conditions, as defined by the NCEP–NCAR
reanalysis, as well as for the full sample, without strati-
fying by cloud amount. Section 2 presents the data and
analytical techniques used. The results in section 3
cover both daytime and nighttime temperatures for the
globe and a selection of large regions and are a sub-
stantial extension of the summary of the results pub-
lished by Parker (2004). Section 4 discusses some addi-
tional comparisons and section 5 draws conclusions.
2. Data and methods
Daily station maximum (T
max
) and minimum (T
min
)
temperature data for 1948 onward were obtained for
the stations in Fig. 1a from the sources indicated. Figure
1a shows the benefits of improved availability of data
from the Global Climate Observing System (GCOS),
but major gaps remain in the Tropics (Mason et al.
2003). Smoothed (Jones et al. 1999) daily climatological
averages of T
max
and T
min
, for 1961–90 where possible
(86% of stations), were created for each station, and
anomalies calculated.
Daily average near-surface wind components were
obtained from the NCEP–NCAR reanalysis (Kalnay et
al. 1996) through the National Oceanic and Atmo-
spheric Administration–Cooperative Institute for Re-
search in Environmental Sciences (NOAA–CIRES)
Climate Diagnostics Center (in Boulder, Colorado)
Web site (see online at http://www.cdc.noaa.gov/cdc/
data.ncep.reanalysis.html). The wind components were
converted to scalar speeds, which were used to classify
15 JUNE 2006 P A R K E R 2883

each date at stations worldwide as “windy”or “calm.”
Even at night, when the coupling between the surface
layers and the free atmosphere is most likely to be
weakened, these daily average reanalysis wind speeds
are generally sufficiently correlated with the limited
available observing-station wind speeds for the pur-
poses of this study (appendix C). Quantiles of the sta-
tistical distributions of wind speed were estimated for
each day of the year, using gamma distributions (Hor-
ton et al. 2001). Each day’s wind speed was then clas-
sified as calm (terce 1 unless otherwise stated) or windy
(terce 3). Temperature anomalies were then matched
to the appropriate date in order to assign them to speed
classes. For stations between 140°E and the date line,
T
min
(which most frequently occurs in the early morn-
ing) was matched with the previous day’s speed. For
FIG. 1. (a) Sources of station data. Data were obtained from the GCOS surface network (GSN)
archive at NOAA’s National Climatic Data Center (Asheville, NC; dots), the European Climate As-
sessment (ECA) Web site (online at http://eca.knmi.nl/; Klein Tank et al. 2002; upward-pointing tri-
angles), the APN dataset (Manton et al. 2001; downward pointing triangles), and several national
sources (Parker et al. 1992; Parker and Alexander 2001; Laursen 2002; Razuvaev et al. 1993; Vincent et
al. 2002; squares). (b) Stations used in the main and subsidiary analyses as described in the legend and
main text.
2884 JOURNAL OF CLIMATE VOLUME 19
Fig 1 live 4/C

stations between 120°W and the date line, T
max
was
matched with the following day’s speed. Where coun-
tries ascribed T
max
to the date of the morning when the
instrument was read, rather than the date on which
T
max
occurred (appendix A), this was allowed for. An-
nual, half-year (October–March, April–September)
and seasonal (December–February, etc.) temperature
anomalies for windy days and for calm days were then
compared for each station.
Figure 2 shows relative warming on calm nights at
Fairbanks, Alaska, which is known to be affected by
urban warming (Magee et al. 1999). Of the 290 stations
studied, only 13 showed significant relative warming of
calm nights (appendix B).
The annual and seasonal anomalies of T
max
and T
min
for each station in Fig. 1b marked with a dot (with or
without a surrounding triangle and/or square) were
gridded on 5°latitude ⫻5°longitude resolution for
windy, calm, and all conditions. The 19 stations marked
“⫹”were not used, generally because they shared a
grid box with another station with a longer or more
complete record. Only one of these showed urban
warming (appendix B), implying minimal potential im-
pact on any worldwide urban warming signal. In the
gridding process, anomalies for all conditions (windy
and calm) were omitted if there were 25 (5) or fewer
applicable days’data in the year or season. Maps of the
gridded anomalies of T
min
for 1975 (Fig. 3) illustrate the
spatial coherence of the fields, and the tendency for the
windy (calm) nights to be relatively warmer (colder),
especially in the extratropics. Coverage was at least 200
grid boxes in 1957–99. Global and regional averages for
each year and season in the period 1950–2000 were
calculated from the gridded data, with weighting by the
cosine of latitude.
It is conceded that this method relies on the assump-
tion that there are no systematic changes in atmo-
spheric circulation during the period of analysis, other
than those expressed as a change in calm and windy
frequencies. Section 3 shows that this assumption does
not hold in western Eurasia in winter. However, owing
to the wide longitudinal distribution of stations in the
extratropical Northern Hemisphere (Fig. 1), these ef-
fects are likely to cancel on a global scale; furthermore,
section 3 finds no such circulation-induced effects in
summer.
To assess the sensitivity of the analysis technique to
differences between the NCEP–NCAR reanalysis daily
average winds and the limited available observations of
nighttime wind at the stations, the analysis was re-
peated on 26 stations in North America and Siberia
(triangles in Fig. 1b) using routine observations of si-
multaneous, instantaneous nighttime temperature
(T
night
) and wind. The nighttime observing hour was
kept constant and as close to dawn as possible to coin-
cide with the most usual time of T
min
. The data were
obtained from the NOAA Integrated Surface Hourly
Dataset, which holds data received in operational
“SYNOP”messages through the Global Telecommuni-
cation System of the World Weather Watch. These sta-
tion nighttime wind data were also compared directly
with the NCEP–NCAR reanalysis daily average winds;
results are summarized in appendix C.
FIG. 2. Annually averaged T
min
(°C, relative to 1961–90) at
Fairbanks, AK, on calm relative to windy nights. The trend cor-
relation is 0.46 and the restricted maximum likelihood (Diggle et
al. 1999) linear trend of 0.34°C decade
⫺1
is significant at the 1%
level.
FIG. 3. Annual anomalies of T
min
(°C, relative to 1961–90) on
(top) calm and (bottom) windy nights in 1975.
15 JUNE 2006 P A R K E R 2885
Fig 3 live 4/C

Although the NCEP–NCAR reanalysis cloudiness
data are highly model dependent (Kalnay et al. 1996),
an analysis of T
min
was carried out using daily average
NCEP–NCAR reanalysis relative humidity at 850 hPa
(RH
850
) as a proxy for low-level cloudiness over 25
stations (squares in Fig. 1b). There was no systematic
tendency for “clear”nights to warm relative to
“cloudy”nights, which would be a symptom of urban
warming. However, at about a quarter of the stations,
the time series of station T
min
for clear minus cloudy
nights showed discontinuities that were not evident in
the analysis with respect to the NCEP–NCAR reanaly-
sis surface winds. Also, the analysis failed to detect the
known urban warming at Fairbanks, Alaska (Magee et
al. 1999). Therefore, the 850-hPa relative humidities
from the NCEP–NCAR reanalysis appear to not be a
reliable classification tool for low cloudiness. Details of
the analysis are in appendix D, where it is also shown
that observed nighttime cloud cover data, while posi-
tively correlated with RH
850
, may be of insufficient
quality to verify it.
3. Results
The main impact of any urban warming is expected
to be on T
min
on calm nights (Johnson et al. 1991).
However, for 1950–2000, the trends of global annual
average T
min
for windy, calm, and all conditions were
virtually identical at 0.20°⫾0.06°C decade
⫺1
(Fig. 4a,b
and Table 1). When the criterion for calm was changed
to the lightest decile of wind strength, the global trend
in T
min
remained 0.20°⫾0.06°C decade
⫺1
(Fig. 4c). So
the overall analysis appears not to show an urban
warming signal and to be robust to the criterion for
calm.
In the supplementary analysis using simultaneous, in-
stantaneous nighttime temperature, T
night
, and wind at
26 stations (section 2), windy and calm nights warmed
at the same rate: 0.20°C decade
⫺1
over the period 1950–
2000. Over this period, T
min
classified by the NCEP–
NCAR reanalysis winds over these 26 stations warmed
by 0.29°C (0.21°C) decade
⫺1
for windy (calm) nights
but the differences between windy and calm, and be-
tween T
min
and T
night
trends, were not statistically sig-
nificant. One reason for differences is that for some
stations, T
night
data but not T
min
data were available for
1999–2000. Trends over 1950–98 averaged over the 26
stations were 0.19°C decade
⫺1
for T
night
for both windy
and calm nights, and 0.26°C (0.23°C) decade
⫺1
for
windy (calm) for T
min
.
The global annual result conceals a relative warming
of windy nights—the opposite of what would be ex-
pected from urbanization—in Europe (35°–70°N,
15°W–40°E) in autumn and winter (Table 1). This re-
sulted in an annual trend of 0.10°⫾0.04°C decade
⫺1
on
windy nights relative to calm nights. There were also
weak relative trends in the same sense in the Arctic
(⬎60°N) and Australasia (10°N–60°S, 90°E–180°; Table
1), and in the contiguous United States (0.02°⫾0.05°C
decade
⫺1
on an annual basis), supporting the reported
FIG. 4. (a) Annual global anomalies of T
max
and (b) T
min
on
(dashed) windy and (dotted) calm days or nights. The linear trend
fits, and the ⫾2
␴
error ranges given in the text were estimated by
restricted maximum likelihood (Diggle et al. 1999), taking into
account autocorrelation in the residuals. (c) Annual global
anomalies of T
min
on (dotted) calm (terce 1) nights and (dash–
dot) very calm (decile 1) nights.
2886 JOURNAL OF CLIMATE VOLUME 19

lack of urban warming there (Peterson 2003; Peterson
and Owen 2005). Conversely, in the Tropics (20°N–
20°S), windy nights cooled (though insignificantly) rela-
tive to calm nights on an annual average (Table 1 and
Fig. 5). The European trends in winter are explained
below in terms of atmospheric circulation changes. In
the extratropical Northern Hemisphere north of 20°N
(NH20; Fig. 6c,d and Table 1), there was no significant
change of T
min
on windy nights relative to calm nights
in summer, when atmospheric circulation changes are
less influential. Any urban warming signal should be
most evident in summer, when urban heat islands are
stronger owing to greater storage of solar heat in urban
structures.
For 1950–2000, the linear trends of global annual av-
erage T
max
were 0.13°⫾0.05°C (0.10°⫾0.06°C) de-
cade
⫺1
for windy (calm) conditions and 0.12°⫾0.05°C
decade
⫺1
for all data. The difference between the
windy-day trend and the calm-day trend is reflected in
Fig. 4a. The annual trends of T
max
for windy minus calm
are positive—not an urbanization signal—and statisti-
cally significant for the globe, NH20, the Arctic, Eu-
rope, Asia (20°–90°N, 40°E–180°), and North America
(20°–90°N, 50°W–180°), but not for the Tropics or Aus-
tralasia (Table 2). As with T
min
, the relative trends
mainly arose from the winter season and to a lesser
extent autumn in Eurasia (Table 2). For NH20 (Fig.
6a,b), T
max
warmed by 0.29°⫾0.11°C (0.15°⫾0.13°C)
decade
⫺1
for windy (calm) conditions in winter but by
0.07°⫾0.08°C (0.08°⫾0.07°C) decade
⫺1
for windy
TABLE 1. Trends in average T
min
,°C decade
⫺1
, 1950–2000. The linear trend fits and the ⫾2
␴
error ranges were estimated by restricted
maximum likelihood (Diggle et al. 1999), taking into account autocorrelation in the residuals. Trends for “windy minus calm”and
“windy minus all”are only given where significant: 5% (italic) or 1% (bold).
Region Season All (2
␴
) Windy (2
␴
) Calm (2
␴
)
Windy
minus calm
Windy
minus all
Globe Year 0.20 (0.06) 0.20 (0.06) 0.20 (0.05)
Arctic (60°–90°N) Year 0.25 (0.09) 0.27 (0.09) 0.22 (0.09) 0.05
Europe (35°–70°N, 15°W–40°E) Year 0.17 (0.14) 0.22 (0.13) 0.13 (0.14) 0.10 0.06
Dec–Feb 0.27 (0.27) 0.37 (0.23) 0.21 (0.28) 0.19 0.11
Mar–May 0.17 (0.14) 0.17 (0.13) 0.17 (0.14)
Jun–Aug 0.12 (0.10) 0.14 (0.12) 0.10 (0.10)
Sep–Nov 0.08 (0.14) 0.16 (0.16) 0.03 (0.14) 0.13 0.08
North America (20°–90°N, 50°W–180°) Year 0.25 (0.13) 0.24 (0.12) 0.23 (0.13)
Dec–Feb 0.28 (0.24) 0.25 (0.23) 0.26 (0.23)
Mar–May 0.35 (0.17) 0.33 (0.16) 0.33 (0.18)
Jun–Aug 0.20 (0.07) 0.19 (0.06) 0.19 (0.08)
Sep–Nov 0.15 (0.10) 0.15 (0.12) 0.13 (0.10)
Asia (20°–90°N, 40°E–180°) Year 0.30 (0.07) 0.32 (0.07) 0.28 (0.07)
Dec–Feb 0.49 (0.15) 0.55 (0.17) 0.44 (0.15) 0.10
Mar–May 0.34 (0.12) 0.33 (0.11) 0.33 (0.12)
Jun–Aug 0.14 (0.07) 0.16 (0.05) 0.15 (0.06) 0.02
Sep–Nov 0.23 (0.08) 0.25 (0.10) 0.22 (0.09)
NH north of 20°N Year 0.24 (0.08) 0.26 (0.08) 0.21 (0.08) 0.04 0.02
Dec–Feb 0.34 (0.14) 0.38 (0.14) 0.30 (0.13) 0.09 0.04
Jun–Aug 0.16 (0.08) 0.17 (0.08) 0.16 (0.07)
Tropics (20°N–20°S) Year 0.18 (0.04) 0.17 (0.03) 0.18 (0.05)
Australasia (10°–60°S, 90°E–180°) Year 0.11 (0.04) 0.12 (0.05) 0.10 (0.04) 0.02 0.01
Dec–Feb 0.13 (0.06) 0.13 (0.05) 0.11 (0.05)
Mar–May 0.09 (0.06) 0.12 (0.06) 0.06 (0.07) 0.05 0.03
Jun–Aug 0.09 (0.07) 0.09 (0.08) 0.08 (0.07)
Sep–Nov 0.13 (0.05) 0.15 (0.06) 0.11 (0.05) 0.04 0.02
FIG. 5. Same as in Fig. 4a, but for tropical (20°N–20°S)
anomalies of T
min
on (dashed) windy and (dotted) calm nights.
15 JUNE 2006 P A R K E R 2887

(calm) conditions in summer. An influence of atmo-
spheric circulation changes in winter is suggested. The
observed tendency of an increased positive phase of the
North Atlantic Oscillation (Folland et al. 2001b) im-
plies that the windier days in western Eurasia tended
toward increased warm advection from the west or
southwest (Hurrell and van Loon 1997), yielding
greater warming in windy conditions. This finding im-
plies that the comparison of trends of T
max
or T
min
between windy and calm conditions may not always be
able to detect small urban warming trends at individual
stations or on subcontinental scales in winter in middle
and high latitudes. However, in other seasons and on
hemispheric and global scales, atmospheric circulation
effects are likely to be small and/or cancelled out, al-
lowing any urban warming signal to be detected.
The observed decline in diurnal temperature range
T
range
(Folland et al. 2001b; Easterling et al. 1997) is
insignificantly weakened (from ⫺0.08°to ⫺0.07°⫾
0.02°C decade
⫺1
) by subsampling on windy days (Table
3). The overall decline of T
range
is significant and con-
sistent with reported increases in cloudiness (Folland et
al. 2001b; Dai et al. 1997).
4. Additional comparisons
Because a small (though widespread) sample was
used, the robustness of the results was tested by com-
paring global trends for 1950–93 with published all-
conditions trends based on a much larger sample (East-
erling et al. 1997). For that period, the global annual
trends of T
min
in the present study were 0.16°⫾0.07°C
decade
⫺1
for all conditions and 0.16°⫾0.07°C (0.17°⫾
0.07°C) decade
⫺1
for windy (calm) conditions. Global
annual trends of T
max
for 1950–93 were 0.08°⫾0.06°C
decade
⫺1
(full sample) and 0.09°⫾0.06°C (0.06°⫾
0.07°C) decade
⫺1
for windy (calm) conditions. These
are close to the annual average of the seasonal global
all-conditions trends for ⬎5000 nonurban stations for
1950–93 (Easterling et al. 1997): 0.17°C decade
⫺1
for
FIG. 6. (a) Anomalies of T
max
for winter (December–February) and (b) summer (June–August) in the NH north of 20°N on (dashed)
windy and (dotted) calm days. The linear trend fits and the ⫾2
␴
error ranges given in the text were estimated by restricted maximum
likelihood (Diggle et al. 1999), taking into account autocorrelation in the residuals. T
max
is, as expected, lower on windy than on calm
days in summer (b), but higher on windy than on calm days in winter (a), because persistent near-surface inversions are limited to
calm weather. (c), (d) Same as in (a), (b), but for T
min
.T
min
is, as expected, higher on windy than on calm nights in both winter and
summer.
2888 JOURNAL OF CLIMATE VOLUME 19

TABLE 2. Same as in Table 1, but for T
max
.
Region Season All (2
␴
) Windy (2
␴
) Calm (2
␴
)
Windy
minus calm
Windy
minus all
Globe Year 0.12 (0.05) 0.13 (0.05) 0.10 (0.06) 0.02
Arctic (60°–90°N) Year 0.15 (0.09) 0.17 (0.09) 0.11 (0.09) 0.07 0.03
Europe (35°–70°N, 15°W–40°E) Year 0.11 (0.12) 0.17 (0.11) 0.06 (0.13) 0.11 0.06
Dec–Feb 0.25 (0.25) 0.36 (0.18) 0.14 (0.25) 0.24 0.12
Mar–May 0.12 (0.16) 0.12 (0.15) 0.07 (0.17)
Jun–Aug 0.05 (0.09) 0.05 (0.12) 0.02 (0.09)
Sep–Nov 0.03 (0.10) 0.08 (0.13) ⫺0.01 (0.10) 0.09 0.05
North America (20°–90°N, 50°W–180°) Year 0.15 (0.11) 0.18 (0.10) 0.13 (0.12) 0.05 0.03
Dec–Feb 0.18 (0.19) 0.19 (0.18) 0.16 (0.19)
Mar–May 0.27 (0.14) 0.31 (0.15) 0.23 (0.17)
Jun–Aug 0.12 (0.06) 0.11 (0.07) 0.12 (0.07)
Sep–Nov 0.02 (0.13) 0.06 (0.10) ⫺0.02 (0.15) 0.09 0.05
Asia (20°–90°N, 40°E–180°) Year 0.17 (0.08) 0.19 (0.08) 0.13 (0.08) 0.06
Dec–Feb 0.32 (0.17) 0.37 (0.18) 0.21 (0.16) 0.16
Mar–May 0.22 (0.12) 0.21 (0.12) 0.20 (0.12)
Jun–Aug 0.07 (0.08) 0.04 (0.09) 0.06 (0.07) ⫺0.03
Sep–Nov 0.11 (0.11) 0.13 (0.10) 0.08 (0.10)
NH north of 20°N Year 0.14 (0.08) 0.17 (0.08) 0.10 (0.08) 0.07 0.02
Dec–Feb 0.24 (0.13) 0.29 (0.11) 0.15 (0.13) 0.14 0.05
Jun–Aug 0.09 (0.07) 0.07 (0.08) 0.08 (0.07) ⫺0.02
Tropics (20°N–20°S) Year 0.08 (0.05) 0.08 (0.05) 0.07 (0.05)
Australasia (10°–60°S, 90°E–180°) Year 0.10 (0.04) 0.11 (0.05) 0.10 (0.04)
Dec–Feb 0.10 (0.05) 0.08 (0.08) 0.11 (0.06)
Mar–May 0.09 (0.05) 0.11 (0.06) 0.07 (0.06)
Jun–Aug 0.12 (0.06) 0.12 (0.07) 0.12 (0.07)
Sep–Nov 0.10 (0.07) 0.14 (0.08) 0.06 (0.07) 0.08 0.04
TABLE 3. Same as in Table 1, but for the diurnal temperature range.
Region Season All (2
␴
) Windy (2
␴
) Calm (2
␴
)
Windy
minus calm
Windy
minus all
Globe Year ⫺0.08 (0.02) ⫺0.07 (0.02) ⫺0.10 (0.02) 0.02
Arctic (60°–90°N) Year ⫺0.11 (0.03) ⫺0.09 (0.04) ⫺0.11 (0.02) 0.01
Europe (35°–70°N, 15°W–40°E) Year ⫺0.05 (0.03) ⫺0.05 (0.03) ⫺0.06 (0.03)
Dec–Feb ⫺0.01 (0.04) 0.00 (0.04) ⫺0.06 (0.06) 0.06
Mar–May ⫺0.05 (0.05) ⫺0.05 (0.05) ⫺0.10 (0.07)
Jun–Aug ⫺0.07 (0.05) ⫺0.09 (0.05) ⫺0.08 (0.05)
Sep–Nov ⫺0.04 (0.05) ⫺0.07 (0.05) ⫺0.03 (0.05) ⫺0.04
North America (20°–90°N, 50°W–180°) Year ⫺0.10 (0.08) ⫺0.06 (0.07) ⫺0.11 (0.07) 0.04 0.03
Dec–Feb ⫺0.10 (0.09) ⫺0.05 (0.09) ⫺0.09 (0.07) 0.04
Mar–May ⫺0.08 (0.09) ⫺0.02 (0.08) ⫺0.11 (0.09) 0.08 0.05
Jun–Aug ⫺0.07 (0.04) ⫺0.08 (0.03) ⫺0.07 (0.03)
Sep–Nov ⫺0.14 (0.08) ⫺0.09 (0.10) ⫺0.16 (0.09) 0.05
Asia (20°–90°N, 40°E–180°) Year ⫺0.12 (0.03) ⫺0.13 (0.03) ⫺0.15 (0.04)
Dec–Feb ⫺0.17 (0.07) ⫺0.17 (0.06) ⫺0.24 (0.08) 0.06 0.03
Mar–May ⫺0.12 (0.03) ⫺0.12 (0.04) ⫺0.14 (0.04)
Jun–Aug ⫺0.08 (0.03) ⫺0.13 (0.03) ⫺0.09 (0.03) ⫺0.05
Sep–Nov ⫺0.13 (0.05) ⫺0.12 (0.05) ⫺0.14 (0.06)
NH north of 20°N Year ⫺0.09 (0.03) ⫺0.08 (0.05) ⫺0.12 (0.03) 0.03
Dec–Feb ⫺0.10 (0.04) ⫺0.09 (0.04) ⫺0.15 (0.05) 0.05
Jun–Aug ⫺0.08 (0.03) ⫺0.10 (0.03) ⫺0.09 (0.02)
Tropics (20°N–20°S) Year ⫺0.09 (0.01) ⫺0.09 (0.03) ⫺0.10 (0.02)
Australasia (10°–60°S, 90°E–180°) Year ⫺0.01 (0.04) ⫺0.01 (0.04) ⫺0.00 (0.04)
Dec–Feb ⫺0.03 (0.04) ⫺0.05 (0.06) ⫺0.01 (0.05) ⫺0.05
Mar–May 0.01 (0.07) ⫺0.01 (0.07) 0.01 (0.07)
Jun–Aug 0.03 (0.07) 0.03 (0.07) 0.05 (0.07)
Sep–Nov ⫺0.03 (0.06) ⫺0.02 (0.06) ⫺0.05 (0.07)
15 JUNE 2006 P A R K E R 2889

T
min
and 0.07°C decade
⫺1
for T
max
. The trend in global
mean temperature in 1950–2000 in the sample used in
the present study, 0.16°⫾0.05°C, is also not signifi-
cantly different from that in the full Jones and Moberg
(2003) dataset, 0.15°⫾0.05°C (Fig. 7). The robustness
of the small sample (up to 270 stations) arises because
the spatial coherence of surface temperature variations
and trends yields only about 20 spatial degrees of free-
dom for annual averages, and fewer on decadal time
scales (Jones et al. 1997). The interannual variability of
global temperature anomalies in the sample used in this
study slightly exceeds that in Jones and Moberg (2003;
Fig. 7) because of its sparser coverage and greater rela-
tive concentration of stations over the Northern Hemi-
sphere continents.
5. Conclusions
The analysis would have benefited from more daily
data over Africa and South America (Mason et al. 2003;
Fig. 1 of this paper). A future increase in daily data
availability may be expected through the GCOS Imple-
mentation Plan (GCOS 2004). For detailed urban
warming studies, however, higher-resolution (e.g.,
hourly) data will be required.
Nevertheless, the network of stations used in this
study has been shown to be adequate for the represen-
tation of large-scale air temperature trends over land.
The analysis of T
min
demonstrates that neither urban-
ization nor other local instrumental or thermal effects
have systematically exaggerated the observed global
warming trends in T
min
. The robustness of the analysis
to the criterion for “calm”implies that the estimated
overall trends are insensitive to boundary layer struc-
ture and small-scale advection, and to siting, instrumen-
tation, and observing practices that increasingly influ-
ence temperatures as winds become lighter. Further-
more, even at windy sites (e.g., St. Paul, Aleutian
Islands, in Fig. C1), the calmest terce and especially the
calmest decile will be strongly affected by occasions
with very light winds in passing ridges or blocking an-
ticyclones, and should reveal any urban warming influ-
ence.
The analysis of T
max
supports the findings for T
min
while also showing the influence of regional atmo-
spheric circulation changes.
In view of the error estimates (Tables 1 and 2) and
the different periods of analysis, the results are com-
patible with the IPCC TAR conclusion that urban
warming is responsible for an uncertainty of 0.06°Cin
the global warming in the twentieth century. Neverthe-
less, the reality of global-scale warming is strongly sup-
ported by the overall near equality of temperature
trends on windy days with trends based on all data.
The reality of urban warming on local (Arnfield
2003; Johnson et al. 1991) and small regional scales
(e.g., Zhou et al. 2004) is not questioned by this work;
it is the impact of urban warming on estimates of global
and large regional trends that is shown to be small.
Because changes of siting, instrumentation, or ob-
serving practice are likely to have the greatest impact in
calm, cloudless weather, analyses of individual station
temperature data in such conditions could be used to
supplement or validate station metadata. In turn, im-
proved metadata will allow refined estimates of climatic
changes, especially on local and small regional scales.
When data are averaged over hemispheres and the
globe, these types of heterogeneity are likely to cancel
out to some extent, and the results of the present study
also suggest that they have not affected the estimates of
temperature trends.
Acknowledgments. Data were provided by J. Cap-
pelen (Danish Meteorological Institute), D. Kiktev
(Russian Hydrometeorological Institute), R. Ray (Na-
tional Climatic Data Center), M. J. Salinger [NIWA
New Zealand: Asia–Pacific network (APN) data], L.
Vincent (Environment Canada), and P. Zhai (China
Meteorological Administration). C. K. Folland pro-
vided useful comments. This work was supported by
the U.K. Government Research Program and this pa-
per is U.K. Crown Copyright.
APPENDIX A
National Practices for Ascribing Dates to
Maximum Temperatures
When a 24-h maximum temperature is read in the
morning, some countries record it against the date of
FIG. 7. Annual global anomalies of T
mean
from the (solid) sta-
tions used in this analysis and from the (dashed) Jones and
Moberg (2003) dataset.
2890 JOURNAL OF CLIMATE VOLUME 19

reading the instrument whereas others record it against
the previous day when, it is presumed, the maximum
temperature usually occurred. A further complication
arises where the local date differs from the UTC date.
For example, between 135°E and the date line, at 0900
local time (often adopted as the time for reading instru-
ments), the previous day is still current at the Green-
wich meridian on which UTC is based. Listed below are
national conventions for recording maximum tempera-
ture. This information was used to align the maximum
temperatures with the appropriate daily mean wind
strength from the NCEP–NCAR reanalysis (Kalnay et
al. 1996).
The following information was provided by national
contacts.
Australia: Maxima and minima are ascribed to the
local date of occurrence in Western Australia, the
Indian Ocean Islands, and Antarctica but to the
previous date in central and eastern states (P. Della
Marta 2002, personal communication).
China: Maxima are ascribed to the local day of oc-
currence (P. Zhai 2002, personal communication).
Denmark and Greenland: The maxima are ascribed
to the following date, when the instruments are
read (J. Cappelen 2002, personal communication).
The same is implied by Laursen (2002).
New Zealand and the southwest Pacific islands (APN
dataset): Maxima are ascribed to the local date of
occurrence.
New Zealand and its southwest Pacific islands (not
APN dataset): Data are ascribed to the New
Zealand local date at the time of observation
(A. Harper 2003, personal communication). For
example, Campbell Island (169°E) T
max
and T
min
ascribed to 23 May refer to the 24-h period from
0900 on 22 May to 0900 on 23 May, New Zealand
Time, so both should be matched with winds for 22
May in UTC time. An irregularity is that the
Pitcairn Island (across the date line at 130°W) T
max
ascribed to 23 May refers to the 24-h period from
0600 on 22 May to 0600 on 23 May, New Zealand
Time (from 0900 on 21 May to 0900 on 22 May,
Pitcairn Island Time), so T
max
and T
min
should also
be matched with winds for 22 May in UTC time.
United Kingdom: Maxima are ascribed to the date of
occurrence.
United States: The policy is that observations are
recorded as if they were made when the thermom-
eters were read. Some observers, however, ascribe
the observations to the day before. Whenever this
practice is detected, the data are reassigned to
the date on which the observations were made
(T. Peterson 2002, personal communication). Thus,
T
max
is, in theory, ascribed to the date after it oc-
curs. An analysis (by the author of this paper) of
average maximum temperature anomalies on calm
versus windy days, with and without a 1-day offset,
suggested that T
max
occurred on the day before the
ascribed date as implied by the policy, except at
San Juan, Majuro, and Koror. At these three sta-
tions, T
max
occurred on the ascribed date. Their
data were analyzed accordingly.
The following was deduced from published books of
data.
Japan, Philippines, South Korea, and Thailand:
Maxima are ascribed to the local date of occur-
rence.
The following was deduced from samples of plotted
data on Met Office or NOAA daily weather reports:
Canada, the former Soviet Union, and Ireland:
Maxima are ascribed to the local date of occur-
rence.
The following was deduced from mean sea level pres-
sure patterns from the NCEP–NCAR reanalysis (Kal-
nay et al. 1996).
South Africa (Marion Island): Maxima are ascribed
to the local date of occurrence.
The following was deduced from average maximum
temperature anomalies on calm versus windy days, with
and without a 1-day offset.
Algeria, Austria, France, Germany, Iceland, Ireland,
Israel, Kuwait, Malaysia, Mauritius, Namibia, New
Caledonia, Norway, Poland, Seychelles, Sweden,
Switzerland, and Uruguay: Maxima are ascribed to
the local date of occurrence.
Colombia, Papua New Guinea, Saudi Arabia, and
Syria: The evidence is unclear; assume that
maxima are ascribed to the local date of occur-
rence.
Greece and Mongolia: The maxima are ascribed to
the following date.
APPENDIX B
Stations Showing Urban Warming or Other
Heterogeneities
Of the 290 stations studied, only 13, listed in Table
B1 with their World Meteorological Organization num-
bers, showed significant warming of calm relative to
windy T
min
. However, comparison of calm with windy
T
min
did not always detect the known slight urban
15 JUNE 2006 P A R K E R 2891

warming influences at individual locations, for example,
in central England temperature (Manley 1974; Parker
et al. 1992), because, for instance, the source of the air
on windy days may change systematically (see main
text). Another 21 stations, listed in Table B2, showed
other systematic discontinuities or trends of uncertain
origin in calm relative to windy T
max
or T
min
. Stations in
Tables B1 and B2 marked with an asterisk were not
used in the global and regional analyses because a sta-
tion with a longer or more complete record shared the
same 5°grid box (see main text). In addition, Shanghai,
although not sharing a 5°grid box, was not used as it
showed marked cooling on calm relative to windy days
and nights, possibly indicating a site change.
A discontinuity at Fresno, California, is illustrated in
Fig. B1. Because changes of siting, instrumentation, or
observing practice are likely to have the greatest impact
in calm, cloudless weather (see main text), analyses like
Fig. B1 could be used, along with appropriate statistical
tests, to supplement or validate station metadata.
APPENDIX C
Comparisons between Station Nighttime Winds
and NCEP–NCAR Reanalysis Daily Average
Winds
Station wind speeds for 26 locations in North
America and Siberia (triangles in Fig. 1b) were com-
pared with corresponding NCEP–NCAR reanalysis
daily average winds. The station winds were selected to
be for a constant nighttime hour as close to dawn as
possible to coincide with the most usual time of T
min
.
For 18 of the stations, correlations were in the range
0.6–0.8; the least well correlated station, Phoenix (Ari-
zona), scored 0.2–0.5. At the 14 Russian stations, the
ratio (station wind speed/reanalysis wind speed) aver-
aged 17% less in the period 1979–2000 than previously,
when the reanalysis lacked satellite data. For the 12
North American stations, this ratio fell by 5%. None-
theless, scatterplots show that both before and after
1979, an NCEP–NCAR reanalysis daily average wind
in terce 1 represented an enhanced likelihood of very
light wind being observed at the station, even at poorly
correlated Phoenix (Fig. C1). The ratio of NCEP–
NCAR reanalysis wind speed at the upper limit of terce
1 to that at the upper limit of decile 1 is typically nearly
2 at the 26 stations. At eight tropical stations represent-
ing a wide range of longitudes, this ratio is also typically
nearly 2, except during seasons of persistent maritime
trade winds at three of these stations, when it is 1.25–
1.5. Now the systematic global offset between very calm
(decile 1) and calm (terce 1) T
min
is less than 0.1°C (Fig.
4c). So the small fractional changes in station wind
speed relative to NCEP–NCAR reanalysis wind speed
are unlikely to have caused a global bias in trends of
T
min
in calm conditions approaching ⫺0.1°C over this
50-yr period, that is, ⫺0.02°C decade
⫺1
or one-tenth of
the magnitude of the observed warming signal.
APPENDIX D
Analysis Using Proxy Low-Level Cloudiness
A selection of 25 stations (locations with squares in
Fig. 1b) was used in a comparison of T
min
on dates with
low (bottom terce, assumed to represent clear condi-
tions) and high (top terce, assumed to represent cloudy
TABLE B1. Stations showing significant warming of calm relative
to windy T
min
.
06700 Geneva-Cointrin* 07650 Marseille 40061 Palmyra
47112 Inchon 59758 Haikou 70261 Fairbanks
71082 Alert 71924 Resolute 71946 Fort Simpson
78526 San Juan 91753 Hihifo 91765 Pago Pago
98653 Surigao
TABLE B2. Stations showing other systematic discontinuities or
trends in calm relative to windy T
max
or T
min
.
04092 Teigarhorn 17062 Istanbul 17240 Isparta
23849 Surgut* 31369 Nikolaev-na-
Amur
36177 Semipalatinsk*
44231 Muren 44259 Choibal 47165 Mokpo
48500 Prachuap 48568 Songkhla 50527 Hailar
52203 Hami 54342 Shenyang 56294 Chengdu
58362 Shanghai 71090 Clyde 72389 Fresno
91348 Ponape 91376 Majuro 91592 Noumea
FIG. B1. Annually averaged T
max
(°C, relative to 1961–90) at
Fresno, CA, on calm relative to windy days. The discontinuity in
the mid-1970s is not evident in the corresponding analysis of T
min
(not shown).
2892 JOURNAL OF CLIMATE VOLUME 19

conditions) daily average relative humidity at 850 hPa
(RH
850
) in the NCEP–NCAR reanalysis (Kalnay et al.
1996). The 850-hPa level was selected because low-level
clouds, being warmer in general, have a greater influ-
ence than higher-level clouds on nighttime surface tem-
peratures. The detection of colder nights by this proxy
for clear skies was often, but not always, stronger than
the detection of colder nights by light winds. There was
no systematic tendency for clear nights to warm relative
to cloudy nights, which would be a symptom of urban
warming. However, at about a quarter of the stations,
the time series of station T
min
for clear minus cloudy
nights showed discontinuities, mainly but not exclu-
sively before 1975, which were not evident in the analy-
sis with respect to the NCEP–NCAR reanalysis surface
winds. These discontinuities may have arisen from het-
erogeneities in the reanalysis resulting from changes in
radiosonde humidity data availability, instrumentation,
or processing techniques (Elliott and Gaffen 1991; El-
liott et al. 1998). Figure D1 is an example for a Siberian
station. The analysis failed to detect the known urban
warming at Fairbanks, Alaska (Magee et al. 1999),
whereas the analysis using near-surface winds did so
(cf. Fig. D2 with Fig. 2), and the relative change of T
min
appeared as a discontinuity in the late 1960s.
At 15 of the stations, sufficient cloud cover observa-
tions, made visually from ground level at the same
nighttime hour as the wind observations used in appen-
dix C, were available to allow comparisons with the
daily average NCEP–NCAR reanalysis RH
850
. Where
available, the coverage of the lowest main cloud deck
(SYNOP code N
h
) was selected, otherwise the total
cloud cover Nwas used. Correlations were typically
0.5–0.7, but 0.2–0.3 in some seasons at one-third of the
stations. Although there was a tendency for RH
850
to
decline relative to observed cloud cover, there was no
clear relationship between changes or discontinuities in
RH
850
/N
h
and changes or discontinuities in T
min
on low
RH
850
nights relative to high RH
850
nights. The ob-
FIG. C1. Comparison of NCEP–NCAR reanalysis near-surface
and station winds at St. Paul (Aleutian Islands), Phoenix (AZ),
and Vytegra (Russia). The plotted lines represent the equality of
wind speeds.
FIG. D1. Annually averaged T
min
(°C, relative to 1961–90) at
Kljuci in eastern Siberia (56°N, 161°E) on nights with low (bottom
terce) 850-hPa relative humidity relative to nights with high (top
terce) relative humidity. The discontinuities in the late 1960s and
early 1990s are not evident in the corresponding analysis of T
min
on calm relative to windy nights (not shown).
15 JUNE 2006 P A R K E R 2893
Fig C1 live 4/C

served cloud cover data may be of insufficient quality to
verify the RH
850
data.
REFERENCES
Arnfield, A. J., 2003: Two decades of urban climate research: A
review of turbulence, exchanges of energy and water, and the
urban heat island. Int. J. Climatol., 23, 1–26.
Dai, A., A. D. DelGenio, and I. Y. Fung, 1997: Clouds, precipi-
tation and temperature range. Nature, 386, 665–666.
Diggle, P. J., K. Y. Liang, and S. L. Zeger, 1999: Analysis of Lon-
gitudinal Data. Clarendon Press, 253 pp.
Easterling, D. R., and Coauthors, 1997: Maximum and minimum
temperature trends for the globe. Science, 277, 364–367.
Elliott, W. P., and D. J. Gaffen, 1991: On the utility of radiosonde
humidity archives for climate studies. Bull. Amer. Meteor.
Soc., 72, 1507–1520.
——, R. J. Ross, and B. Schwartz, 1998: Effects on climate records
of changes in National Weather Service humidity processing
procedures. J. Climate, 11, 2424–2436.
Folland, C. K., and Coauthors, 2001a: Global temperature change
and its uncertainties since 1861. Geophys. Res. Lett., 28, 2621–
2624.
——, and ——, 2001b: Observed climate variability and change.
Climate Change 2001: The Scientific Basis, J. T. Houghton et
al., Eds., Cambridge University Press, 99–181.
GCOS, 2004: Implementation plan for the Global Observing Sys-
tem for Climate in support of the UNFCCC. Global Climate
Observing System GCOS-92, WMO Tech. Doc. 1219, 136 pp.
[Available online at http://www.wmo.int/web/gcos/gcoshome.
html.]
Hansen, J., R. Ruedy, J. Glascoe, and M. Sato, 1999: GISS analy-
sis of surface temperature change. J. Geophys. Res., 104
(D24), 30 997–31 022.
——,——, M. Sato, M. Imhoff, W. Lawrence, D. Easterling, T.
Peterson, and T. Karl, 2001: A closer look at United States
and global surface temperature change. J. Geophys. Res., 106
(D20), 23 947–23 963.
Horton, E. B., C. K. Folland, and D. E. Parker, 2001: The chang-
ing incidence of extremes in worldwide and Central England
temperatures to the end of the twentieth century. Climatic
Change, 50, 267–295.
Hurrell, J. W., and H. van Loon, 1997: Decadal variations in cli-
mate associated with the North Atlantic Oscillation. Climatic
Change, 36, 301–326.
Johnson, G. T., T. R. Oke, T. J. Lyons, D. G. Steyn, I. D. Watson,
and J. A. Voogt, 1991: Simulation of surface urban heat is-
lands under “ideal”conditions at night. Part 1: Theory and
tests against field data. Bound.-Layer Meteor., 56, 275–294.
Jones, P. D., and A. Moberg, 2003: Hemispheric and large-scale
surface air temperature variations: An extensive revision and
an update to 2001. J. Climate, 16, 206–223.
——, P. Ya. Groisman, M. Coughlan, N. Plummer, W. C. Wang,
and T. R. Karl, 1990: Assessment of urbanization effects in
time series of surface air temperature over land. Nature, 347,
169–172.
——, T. J. Osborn, and K. R. Briffa, 1997: Estimating sampling
errors in large-scale temperature averages. J. Climate, 10,
2548–2568.
——, E. B. Horton, C. K. Folland, M. Hulme, D. E. Parker, and
T. A. Basnett, 1999: The use of indices to identify changes in
climate extremes. Climatic Change, 42, 131–149.
Kalnay, E., and M. Cai, 2003: Impact of urbanization and land-use
change on climate. Nature, 423, 528–531.
——, and Coauthors, 1996: The NCEP/NCAR 40-Year Reanaly-
sis Project. Bull. Amer. Meteor. Soc., 77, 437–471.
Karl, T. R., C. N. Williams, P. J. Young, and W. M. Wendland,
1986: A model to estimate the time of observation bias asso-
ciated with monthly mean maximum, minimum and mean
temperatures for the United States. J. Climate Appl. Meteor.,
25, 145–160.
Klein Tank, A. M. G., and Coauthors, 2002: Daily dataset of
20th-century surface air temperature and precipitation series
for the European Climate Assessment. Int. J. Climatol., 22,
1441–1453.
Laursen, E. V., 2002: Observed daily precipitation, maximum
temperature and minimum temperature from Ilulissat and
Tasiilaq, 1873–2000, version 2. Danish Meteorological Insti-
tute Tech. Rep. 02-15, 10 pp.
Lin, X., K. G. Hubbard, E. A. Walter-Shea, and J. R. Brandle,
2001: Some perspectives on recent in situ air temperature
observations: Modeling the microclimate inside the radiation
shields. J. Atmos. Oceanic Technol., 18, 1470–1484.
Magee, N., J. Curtis, and G. Wendler, 1999: The urban heat island
at Fairbanks, Alaska. Theor. Appl. Climatol., 64, 39–47.
Manley, G., 1974: Central England temperatures: Monthly means
1659 to 1973. Quart. J. Roy. Meteor. Soc., 100, 389–405.
Manton, M. J., and Coauthors, 2001: Trends in extreme daily rain-
fall and temperature in southeast Asia and the South Pacific:
1961–1998. Int. J. Climatol., 21, 269–284.
Mason, P. J., and Coauthors, 2003: The second report on the ad-
equacy of the Global Observing Systems for Climate in sup-
port of the UNFCCC. Global Climate Observing System
GCOS-82, WMO Tech. Doc. 1143, 74 pp.
Parker, D. E., 1994: Effects of changing exposure of thermom-
eters at land stations. Int. J. Climatol., 14, 1–31.
——, 2004: Large-scale warming is not urban. Nature, 432, 290.
——, and L. V. Alexander, 2001: Global and regional climate in
2000. Weather, 56, 255–267.
——, T. P. Legg, and C. K. Folland, 1992: A new daily central
England temperature series. Int. J. Climatol., 12, 317–342.
FIG. D2. Annually averaged T
min
(°C, relative to 1961–90) at
Fairbanks, AK, on nights with low (bottom terce) 850-hPa relative
humidity relative to high (top terce) relative humidity. The rela-
tive change of T
min
appears as a discontinuity in the late 1960s and
the trend correlation is only 0.32, as against 0.46 in Fig. 2.
2894 JOURNAL OF CLIMATE VOLUME 19

Peterson, T. C., 2003: Assessment of urban versus rural in situ
surface temperatures in the contiguous United States: No
difference found. J. Climate, 16, 2941–2959.
——, and T. W. Owen, 2005: Urban heat island assessment: Meta-
data are important. J. Climate, 18, 2637–2646.
——, K. P. Gallo, J. Lawrimore, A. Huang, and D. A. McKittrick,
1999: Global rural temperature trends. Geophys. Res. Lett.,
26, 329–332.
Razuvaev, V. N., E. G. Apasova, and R. A. Martuganov, cited
1993: Daily temperature and precipitation data for 223 USSR
stations. ORNL/CDIAC-56, NDP-040, ESD Publication
4194, Carbon Dioxide Information Data Center, Oak Ridge
National Laboratory, Oak Ridge, TN. [Available online at
http://cdiac.ornl.gov/ftp/ndp040/ndp040.txt.]
Simmons, A. J., and Coauthors, 2004: Comparison of trends and
low-frequency variability in CRU, ERA-40, and NCEP/
NCAR analyses of surface air temperature. J. Geophys. Res.,
109, D24115, doi:10.1029/2004JD005306.
Trenberth, K. E., 2004: Rural land-use change and climate. Na-
ture, 427, 213.
Vincent, L. A., X. Zhang, B. R. Bonsal, and W. D. Hogg, 2002:
Homogenization of daily temperatures over Canada. J. Cli-
mate, 15, 1322–1334.
Vose, R. S., T. R. Karl, D. R. Easterling, C. N. Williams, and M. J.
Menne, 2004: Impact of land-use change on climate. Nature,
427, 213–214.
Zhou, L., R. E. Dickinson, Y. Tian, J. Fang, Q. Li, R. K. Kauf-
mann, C. J. Tucker, and R. B. Myneni, 2004: Evidence for a
significant urbanization effect on climate in China. Proc.
Natl. Acad. Sci. USA, 101, 9540–9544.
15 JUNE 2006 P A R K E R 2895