Sie sind auf Seite 1von 7

Quarterly Journal of the Royal Meteorological Society

Q. J. R. Meteorol. Soc. 136: 253259, January 2010 Part A

Natural ventilation effects on temperatures within Stevenson


screens
R. G. Harrison*
Department of Meteorology, University of Reading, Reading, UK
*Correspondence to: R. G. Harrison, Department of Meteorology, University of Reading, PO Box 243, Earley Gate,
Reading, RG6 6BB, UK. E-mail: r.g.harrison@reading.ac.uk

Thermometer screen properties are poorly characterised at low wind speeds.


Temperatures from a large thermometer screen have been compared with those
from an automatically shaded open-air fine-wire resistance thermometer. For the
majority of 5-minute average measurements obtained between July 2008 and 2009,
the screen and fine-wire temperatures agreed closely, with a median difference
<0.05 C. At low wind speeds however, larger temperature differences occurred.
When calm (wind speed at 2 metres, u2 , 0.1 m s1 ), the difference between
screen and open-air temperatures varied from 0.25 C to +0.87 C. At night with
u2 < 0.5 m s1 , this difference was 0.14 C to 0.39 C, and, rarely, up to 0.68 C
to 1.38 C. At the minimum in the daily temperature cycle, the semi-urban site at
Reading had u2 < 1 m s1 for 52% of the observations 19972008, u2 < 0.5 m s1
for 34% and calm conditions for 20%. Consequently uncertainties in the minimum
temperature measurements may arise from poor ventilation, which can propagate
through calculations to daily average temperatures. In comparison with the daily
minimum temperature, the 0900 UTC synoptic temperature measurement has
c 2010 Royal
a much lower abundance (5%) of calm conditions. Copyright 
Meteorological Society
Key Words:

air temperature measurement; meteorological instrumentation

Received 18 September 2009; Accepted 12 October 2009; Published online in Wiley InterScience 19 January 2010
Citation: Harrison RG. 2010. Natural ventilation effects on temperatures within Stevenson screens. Q. J. R.
Meteorol. Soc. 136: 253259. DOI:10.1002/qj.537

1.

Introduction

Accurate measurement of air temperature is well known


to require sheltering of the temperature sensor from
direct solar radiation, whilst retaining adequate ventilation
(Strangeways, 2003). The naturally ventilated doublelouvred Stevenson screen is a long-established design of
thermometer shelter (e.g. Bilham, 1937), but even with
such screens, their internal temperature may differ from
air temperature by extremes of 0.5 C to 2.5 C (WMO,
1983). Differences between air temperature and screen
temperature generally arise from solar heating, long-wave
radiation exchange and, possibly, evaporative cooling after
rain. All of these factors are influenced by ventilation.
Without adequate ventilation, the response time of a screen
is increased (Bryant, 1968; Fitzpatrick, 2007), causing its
c 2010 Royal Meteorological Society
Copyright 

thermometers to respond more slowly to external air


temperature variations. In low ventilation conditions solar
radiation will warm the screen, and, even nocturnally, the
conditions may be unsatisfactory for thermal equilibrium.
Because of these variations in the behaviour of naturally
ventilated screens, the comparability between measurements
from otherwise identical screens at different sites has
been questioned (Painter, 1977). The incidence of poor
ventilation conditions for thermometer screens is considered
further here.
1.1.

Exposure effects on air temperature measurements

Sensitivities to small exposure effects in air temperature


measurements became clear in tests of emerging thermometer screen technologies during the late nineteenth century.

254

R. G. Harrison

The Glaisher stand, an early thermometer support frame


(Glaisher, 1868), provided a rotatable shade, allowing free
ventilation to thermometers. Comparison of Glaisher stand
and Stevenson screen temperature data at Camden Square,
London for over forty years (Margary, 1924), showed that
daily maximum Stevenson screen temperatures were lower
than Glaisher stand maxima by 1.3 F (0.7 C) in July,
but that they exceeded the Glaisher stand maxima by 0.5 F
(0.3 C) in December. For daily minima, Stevenson temperatures were consistently warmer than Glaisher temperatures
by 1.0 F (0.56 C). These and similar discrepancies in other
comparative studies arose from the different radiation environments of the thermometers (Laing, 1977). In a Glaisher
stand the thermometers receive radiation from the sky,
ground and shading board, but the radiation exchange of a
Stevenson screen thermometer is intended to occur mostly
within the screen, which emits at close to air temperature.
Screenair temperature differences represent systematic
errors in air temperature measurements, which, unlike
random errors, are not reduced by averaging. However if the
systematic effects only occur rarely, their influence on longterm average temperatures will be small, despite individual
screen temperature measurements poorly representing the
instantaneous air temperature.
1.2.

Context for modern temperature measurements

Because of the consistency required from temperature data


for climate studies, transitions between screen designs
are important to manage (Parker, 1994). For example,
temperatures obtained within modern plastic screens differ
from those within similar wooden screens by less than 0.1 C,
and less than 0.25 C during extremes (Perry et al., 2007).
Whilst such studies very effectively compare one screen
design with another for the purpose of data continuity
at an individual site, they do not address site-dependent
differences between air temperature (Tair ) and screen
temperature (Tscrn ), such as from poor ventilation.
Understanding screen behaviour in extreme conditions of
insolation or low ventilation generally requires the unique
local factors or design aspects to be identified experimentally.
For example, Keil (1996) found that poor maintenance
of the screen led to increases in measured summer
temperatures. Studies of naturally ventilated screens (e.g.
Painter, 1977; Huband et al., 1984; Andersson and Mattison,
1991; Anderson and Baumgartner, 1998; Nakamura and
Mahrt, 2005) are unfortunately relatively rare compared
with climate studies using data from screens. Using an

aspirated psychrometer to determine Tair in three years of


measurements at Kew Observatory, Painter (1977) found
Tair and Tscrn differences for a large thermometer screen of
up to 2 C, which suggests a worst-case temperature excess
in the United Kingdom of 2.5 C (HMSO, 1981). These
differences varied with wind speed and direction, and were
attributed to response time and solar radiation effects.
1.3.

Overview of instrumentation for screen investigations

Effects of poor ventilation on screenair temperature


differences are investigated further here, using instruments
at the University of Reading Atmospheric Observatory, an
semi-urban climatological site. The Observatory currently
operates three thermometer screens: two large screens and
one small screen. Of the two large screens, one contains the
climatological thermometers (maximum, minimum and
dry bulb) read by an observer at 0900 UTC daily, and the
other a platinum resistance thermometer (PRT), recorded
automatically. The remaining small (Stevenson) screen
contains another PRT. Each of the screen PRTs is connected
to a precision resistance-to-voltage converter, from which
the derived voltage is recorded by a logging system.
Periodically, these PRTs are calibrated against a laboratory
standard PRT (Automatic Systems model F250) within a
stirred water bath over the range 0 C to 40 C. Radiation
measurements are also made at the Observatory site. A
Kipp and Zonen solar tracker carries instruments for shortwave radiation measurements, including a pyranometer for
the global solar irradiance (Sg ), and a pyrheliometer for
the direct beam irradiance at normal incidence (Sb ). A
pyrradiometer is used to measure the net radiation (Rn ).
All the radiometer thermopile voltages are amplified to volt
level signals by low-noise amplifiers (Harrison and Knight,
2006), recorded synchronously with the thermometer data
at 1 Hz, from which 5-minute averages are calculated.
Table I summarises the subset of Observatory measurements
considered further for this study,
2.
2.1.

Methodology
Principle

This experimental investigation of screen properties has


compared temperature measurements made within and
outside a large thermometer screen containing a standard
PRT. For this, a fine-wire resistance thermometer was first

Table I. Instrumentation at the Reading Atmospheric Observatory.

Quantity

Description

Tscrn

screen temperature

Topen

open-air temperature

u2
Sg
Tmax , Tmin

wind speed at 2 m
global solar radiation
daily maximum and minimum screen temperature
2 m wind speed (5-minute average) measured at
Tmax and Tmin

u2 (Tmax ), u2 (Tmin )

c 2010 Royal Meteorological Society


Copyright 

Sensor
Class A cylindrical Pt100 platinum resistance
thermometer in large thermometer screen
25 m diameter platinum wire resistance thermometer, shaded by solar tracker
Vector Instruments A100M cup anemometer
Kipp and Zonen CM11 solarimeter
Class A cylindrical Pt100 platinum resistance
thermometer in small (Stevenson) screen
Vector Instruments A100M cup anemometer
and small (Stevenson) screen
Q. J. R. Meteorol. Soc. 136: 253259 (2010)

Natural Ventilation Effects in Stevenson Screens

Figure 1. Schematic of apparatus used for screen ventilation investigation.


A fine-wire thermometer is mounted on a solar tracker (left), with a
white shading disk preventing solar radiation from reaching the wire
thermometer. The temperature so-measured is Topen . The fine-wire
platinum resistance thermometer was calibrated before and after its
exposure on the solar tracker in a large thermometer screen, containing
a standard cylindrical platinum resistance sensor (Tscrn ). (The greyed
instruments on the solar tracker are, from the top down, the fine-wire
platinum thermometer, the pyrradiometer and pyrheliometer; the position
in the screen of the fine-wire platinum thermometer during calibration is
also shown in grey).

calibrated within the screen, and then exposed outside


the screen to assess differences between it and the screen
temperature. Shading of the thermometer outside the screen
was achieved using a white painted disk, moved by the
mechanical solar tracker (Figure 1). In terms of shading, this
arrangement resembles the Glaisher screen approach, but
its other deficiencies are minimised by the small shading at
0.5 m from the fine-wire sensor, and the fine-wire sensor
having little area to exchange radiation.
2.2.

Air temperature sensor

The resistance sensor used was a fine-wire platinum


resistance thermometer (fwPRT) of the Reading design,
which has an exponential time response of 40 ms (Harrison
and Pedder, 2001). It consists of about 0.5 m of 25 m
diameter platinum wire, wound in a zigzag fashion across a
36 mm by 28 mm rectangular fibreglass former. The fwPRT
resistance is measured using a precision resistance-tovoltage converter, employing a four-wire Kelvin connection
to correct for errors arising from connection resistances
(Harrison and Rogers, 2006). The resistance measurement
electronics uses a particularly small excitation current
(50 A), to ensure negligible thermometer self-heating.
Amplification raises the thermometer signal voltage to
nominally 100 mV/ C, and the final sensor voltage is sampled
by the logging systems 12-bit analogue-to-digital converter,
giving 0.05 C resolution.
Preliminary experimental tests were made with a pair
of fwPRT sensors, using 5 min average measurements.
Firstly, the radiation error from an unscreened fwPRT
was investigated, using short-wave radiation. For this, the
sensors were unshaded and mounted closely above each
other, at 1.2 m above the surface. Initially both sensors
c 2010 Royal Meteorological Society
Copyright 

255

were edge-on to direct sunshine, but later one of the


sensors was adjusted to be broadside-on to the sun at
local noon. In the edge-on configuration, there was no
differential response to solar radiation. With one sensor
solar broadside-on however, the temperature difference
between the fwPRT sensors was proportional to the
incoming solar radiation. For 2-metre wind speeds (u2 )
greater than 1 m s1 , the sensitivity to solar radiation
was 0.07 C/100 W m2 . Secondly, any orientation effect
was assessed, using nocturnal measurements on consecutive
nights, for sensors similarly aligned and non-aligned. No
significant difference in the measured temperatures between
aligned and non-aligned sensors was found for u2 > 1 m s1 .
Finally, the effect of wind on the sensors was considered,
for two sensors identically aligned edge-on to the sun. For
very low wind conditions (u2 < 0.25 m s1 ), after removing
the median temperature difference, the interquartile range
spanned 0.08 C to 0.06 C. Under moderate ventilation
(u2 > 1.0 m s1 ) the interquartile range was reduced to
0.01 C to 0.02 C.
In summary, these experiments show that an unshaded
fwPRT will have a maximum radiation error of up to 0.6 C
in moderately ventilated (u2 > 1.0 m s1 ) sunny summer
conditions (Sg = 850 W m2 ), and that the errors resulting
from ventilation variations are likely to be less than 0.1 C.
2.3.

Experimental procedure

Each fine-wire thermometer was exposed on the solar tracker


shading system for approximately a week at a time, with its
temperature, the adjacent large-screen temperatures, u2 , Sg
and Rn recorded synchronously. Before and after exposure,
the fwPRT was calibrated in the large screen. The fwPRT
temperatures and screen PRT temperatures on these two
calibration days were fitted by linear regression, providing
two fwPRT calibrations. Both these calibrations were applied
to derive the fwPRT temperature during its exposure, and
the two temperature series obtained were averaged. This
procedure both matches the exposed fine-wire sensor to the
temperature of the adjacent screen, and, because of the use
of one calibration before and another after the exposure,
corrects for drift occurring during exposure.
3.

Results

Measurements were obtained between July 2008 and July


2009, with the days sampled during the year shown in
Figure 2(a). (Values were discarded if the screen and
fwPRT temperatures differed by more than 3 C, usually
due to gross equipment failure such as breakage of the finewire thermometer, water ingress into signal connections, or
poor adjustment of the thermometer shade.) Temperatures
calculated for the fwPRT using the two calibrations available
for each weeks exposure had a median difference of
0 C, with interquartile range 0.05 C to 0.05 C. This
estimates the temporal drift of the sensor during its weekly
exposure, which was clearly smaller than the ventilation
errors.
3.1.

Overview of temperature differences

Figure 2(b) shows a histogram of the difference Tdiff between


screen temperature (Tscrn ) and fwPRT temperature (Topen ),
with positive values indicating that the screen temperature
Q. J. R. Meteorol. Soc. 136: 253259 (2010)

256

R. G. Harrison

Figure 2. (a) Time series of measured screen temperature from 210 days of 5-minute averages obtained between day 200 of 2008 and day 195 of 2009.
(b) Histogram of measured screen to open-air temperature difference Tdiff , (Tdiff = Tscrn Topen ). (c) Tdiff against Tscrn . (d) Tdiff against 2-metre wind
speed (u2 ), for the same period. (c) and (d) include contours showing the proportion of observations (10%, 25%, 50%) within regions of the scatterplot.

was the warmer. The median Tdiff is 0.03 C, with 5% and


95% quantiles 0.41 C and 0.41 C. Clearly, very many of
the screen and fwPRT temperature measurements therefore
fall well within 0.5 C of each other; however, substantial
differences can still rarely occur. Figure 2(c) plots Tdiff
against temperature for the entire dataset. In general, there
is a small increase of Tdiff with temperature, although
the absolute magnitude of the difference decreases with
increasing temperature. For example, at Tscrn = 4.6 C
the median Tdiff = 0.28 C rising to Tdiff = 0.19 C at
Tscrn = 28.9 C. In Figure 2(d), Tdiff is compared with
u2 measured close to the large screen. This shows that
the larger magnitude Tdiff values occur at low wind
speeds, which also suggests that the variation of Tdiff
with Tscrn may arise from ventilation changes with
temperature.
3.2.

Ventilation effects in cloudy and clear skies

Wind speed effects on screen temperature were accordingly


investigated more closely for short series of daily
measurements made under cloudy and clear skies. Figure 3
shows a collection of synchronous measurements made
under cloudy conditions in early summer (days 178 to
183 of 2009), during which the air temperatures steadily
increased from day to day. The variability apparent in the
daytime net radiation (Rn ) and global solar irradiation
(Sg ) traces arose from the cloud present. Periods of calm
c 2010 Royal Meteorological Society
Copyright 

conditions are occasionally evident in the u2 data, notably


during the early hours of days 180 and 181, when Tdiff
became positive, i.e. the screen temperature was warmer
than that of the open-air thermometer. Figure 4 shows
similar data for mostly clear sky conditions on days 76 to
80, as evident from smooth Sg and Rn traces compared
with Figure 3. Sustained calm periods occur during the
early part of day 76, and from sunset on day 77 to
after sunrise on day 78. At sunset on day 77 the screen
temperature was warmer than the open-air temperature,
but this situation reversed and the screen temperature
became colder at sunrise on day 78. For most of the
rest of the data, u2 > 1 m s1 and Tdiff is close to
zero, but, during the daytime temperature maxima, screen
temperatures became slightly warmer than the open-air
temperatures.
To remove the complexity caused by solar heating effects
during the daytime measurements, only the nocturnal
measurements of Figures 3 and 4 are considered further.
Figure 5(a) shows a comparison between Tdiff and u2 for
the nocturnal data of Figure 3. Using a locally weighted
lowess fit (Cleveland, 1981), Tdiff increases as u2 decreases.
Figure 5(b) divides the data to form box plots; these show
that, for the 122 values of u2 < 0.25 m s1 , Tdiff has a
median of 0.39 C, with extreme values from 0.5 C to
1.4 C. For u2 between 0.5 m s1 and 1 m s1 , Tdiff is not
significantly different from zero. Figure 5(c) and (d) show
the same analysis for the data of Figure 4. In this case for the
Q. J. R. Meteorol. Soc. 136: 253259 (2010)

15

257

179

180
181
day in 2009

182

10
5

Discussion

The analysis presented in Figure 5 illustrates that, in low


ventilation nocturnal conditions (u2 < 0.5 m s1 ), the openair thermometer Topen and the screen thermometer Tscrn
differ, by (median values) Tdiff = 0.14 C to 0.39 C, and
at most, Tdiff = 0.68 C to 1.38 C. Clearly therefore, an
uncertainty in Tscrn depends on the local ventilation, both
in magnitude and sign.
In part these differences can be understood in terms of
long-wave radiation exchange. The net radiation traces on
Figures 3 and 4 show differences in the steady values during the cloudy and clear nights. On clear nights, the Rn is
generally more negative than for the cloudy nights, i.e. the
upwards long-wave radiation is greatest on clear radiation
nights. A radiative loss to clear skies will cool the screen
on calm nights, in addition to the extended time response
of the screen to local air temperature from associated low
wind speed conditions (Bryant, 1968; Painter, 1977; HMSO,
1981). Radiative cooling of the screen in clear skies (Figure 4)
can therefore lead to the screen becoming cooler than its surrounding air, if there is insufficient ventilation. Conversely,
in cloudy conditions, less radiative cooling of the screen will
occur, therefore the long lag time at low ventilation becomes
dominant in determining the screen response. This factor
alone could cause decreasing temperature fluctuations to be
underestimated. For example, with a long lag time, the screen
temperature will not track the decreases in air temperature

u/

(Rn & Sg) /

500
0

Wm2

179 values of u2 < 0.25 m s1 , the extreme Tdiff values are


0.7 C and +0.7 C, with a median Tdiff of 0.14 C. Again,
for 0.5 m s1 u2 1 m s1 , Tdiff is not significantly
different from zero.

c 2010 Royal Meteorological Society


Copyright 

ms1

0
1
1

183

Figure 3. Screen thermometer measurements obtained in cloudy


conditions. Time series (plotted as day of year) showing: (upper trace)
screen temperature (Tscrn , thick line) and open-air temperature (Topen ,
thin line); (second trace) 2-metre wind speed (u2 ); (third trace) global
solar irradiance (Sg , larger values), net radiation (Rn , smaller and
negative values); and (lower trace) screenair temperature difference
(Tdiff = Tscrn Topen )( C). This figure is available in colour online at
www.interscience.wiley.com/journal/qj

4.

Tdiff / degC

500
0
1
178

(Rn & Sg) / Wm2

u / ms1

(Tscrn & Topen) / degC

10
1

20

(Tscrn & Topen) / degC


Tdiff / degC

30

Natural Ventilation Effects in Stevenson Screens

76

77

78
day in 2009

79

80

Figure 4. Screen thermometer measurements obtained in clear conditions.


(Time series as for Figure 3). This figure is available in colour online at
www.interscience.wiley.com/journal/qj

sufficiently rapidly to reach brief minima in air temperature.


Hence the screen temperature would remain warmer than
the air temperature at the temperature minimum.
The combination of both radiative and ventilation
effects causing uncertainty in screen temperatures would
seem to preclude a straightforward correction being
made, particularly if little more than basic climatological
measurements are available. For synoptic meteorology, the
screen temperature measurements made at the daily minima
and maxima and at 0900 UTC have particular importance.
The prevalence of poor ventilation conditions for these
synoptic observation times (and therefore the extent to
which ventilation effects may occur) is now considered
further, using wind speed measurements made over a long
period at the Reading site.
4.1.

Prevalence of low wind speeds

Wind speeds at 2 metres have been continuously measured at


the Observatory site from 1997 to 2008, using, throughout,
the same cup anemometer instrument, recording at 0.1 m s1
resolution. (This anemometer mast was adjacent to the large
screen in the present experiments.) Figure 6 shows the
distribution of 5-minute average u2 at different points in
the day. Using temperatures measured simultaneously in
the small thermometer screen over the same period, u2
values corresponding to the screens daily minima and
maxima (Tmin and Tmax ) have been identified. Figure 6(a)
and (b) show histograms of u2 obtained at Tmin and Tmax
respectively, and Figure 6(c) shows the u2 histogram for the
0900 UTC synoptic time. For comparison, Figure 6(d) shows
the histogram of the daily mean u2 . Table II summarises the
properties of the distributions. At Tmin , about one-half of
u2 measurements lie below 1 m s1 , and about one-third lie
below 0.5 m s1 . In contrast, at Tmax , or indeed at 0900 UTC,
the proportion of low wind speeds is much smaller, generally
no more than 10%. For this site, the daily mean wind speed
Q. J. R. Meteorol. Soc. 136: 253259 (2010)

1.5

2.0

0.125

0.375
u2 (m/s)

1.25

0.125

0.375
u2 (m/s)

1.25

1.0
0.5
0.0
0.5

0.0

0.5

Tdiff(degC)

1.0

(d)

0.5

Tdiff(degC)

0.5
0.5

1.0
u2 (m/s)

1.5

0.5

1.5

0.0

(c)

0.0

Tdiff(degC)

0.5
0.0
0.5

Tdiff(degC)

1.0

(b)

1.0

(a)

1.5

R. G. Harrison

1.5

258

0.0

0.5

1.0
u2 (m/s)

1.5

2.0

Figure 5. Nocturnal measurements of screen to open-air temperature difference (Tdiff ) against 2 m wind speed u2 , for (a) days 178 to 182 of 2009 (from
Figure 3), and (c) days 76 to 79 of 2009 (from Figure 4). A locally-weighted robust fit (lowess) line has been added in each case. (b) and (d) show notched
box plots formed by dividing the data in (a) and (c) respectively at the dotted vertical lines. Note that u2 was recorded throughout at 0.1 m s1 resolution.
(Thick bar in box centre marks the median, with the 95% confidence range shown by the notch. Upper and lower box edges are at the upper and lower
quartiles, with whiskers extending to 1.5 times the interquartile range. Box width is proportional to the square root of the number of values contained).

Table II. Properties of measured wind speed distributions (19972008).

Circumstances of wind
speed measurements

At daily minimum temperature Tmin


At daily maximum temperature Tmax
At 0900 UTC synoptic observation time
All daily (5 min average) values

of Proportion
of Proportion
of
Median 2 m Proportion
wind
speed measurements with measurements with measurements with
u2 0.1
ms1 u2 0.5 ms1 (%) u2 1 ms1 (%)
(m s1 )
(calm) (%)
1.0
2.6
1.9
2.1

does not provide a good indicator of wind speeds at Tmin as


has previously been assumed for urban sites (Parker, 2004);
the wind speed at 0900 UTC follows the mean daily wind
speed more closely.
Further evidence in support of the 0900 UTC wind speed
distribution determined exists in subjective wind speed
estimates made at 0900 UTC daily by a Meteorological
Observer. In these records, which began in 1968 (yielding
14 976 values), the mean daily rate of 0900 UTC observations
having wind speeds <1 m s1 was 16%, but with appreciable
variations between years. As might be expected, calms
were more common in years having long anticyclonic
periods, particularly those generating light easterly winds.
Furthermore, years with the lowest annual variability in
pressure had the greatest abundance of calms.
c 2010 Royal Meteorological Society
Copyright 

20
4
5
3

4.2.

34
6
9
6

52
12
19
16

Effect on daily averages

A ventilation effect on Tmin can also affect the daily average


temperature. This is because the daily average temperature
Tavg is often (DallAmico and Hornsteiner, 2006) calculated
from the mean of the daily extremes i.e. as the bimean, Tavg = 12 (Tmin + Tmax ). Tmax and Tmin are readily
found using maximum and minimum thermometers at
climatological sites, read once daily. Where this method
of calculating the daily mean is used, an uncertainty in
Tmin because of ventilation effects will therefore propagate
through to affect Tavg . A particularly poor measurement
of minimum air temperature will contribute uncertainty
to the daily average temperature, with a weighting of one
half. (There may also be a ventilation effect on Tmax , but,
as the Tmax wind speed distribution indicates calms are
Q. J. R. Meteorol. Soc. 136: 253259 (2010)

Noel Fitzpatrick undertook the first study using this


apparatus in summer 2007 and made many important
improvements. Andrew Lomas, Rosemarie Wilson and
Michael Stroud provided technical support.

50
0

400

150

Acknowledgements

800

(b)

Frequency

(a)

8 10

u2(Tmin)

8 10

200
0

100

150

Frequency

300

250

(d)

0 50

Frequency

u2(Tmax)

(c)

u2(9UT)

8 10

10

u2mean

Figure 6. Histogram of 5 min average wind speed at 2 metres measured


from 1997 to 2008, at the time of daily (a) screen minimum temperature
Tmin , (b) screen maximum temperature Tmax , and (c) 0900 UTC synoptic
observation. (d) Histogram of 2-metre mean daily wind speeds (u2mean ).

less common for Tmax than for Tmin , this has not been
investigated further.)
5.

259

Frequency

Natural Ventilation Effects in Stevenson Screens

Conclusions

For the majority of the measurements considered here, the


screen temperature and open-air fine-wire thermometer
temperature agree closely, to a median difference <0.05 C.
However, larger temperature differences do occasionally
occur, often at low (u2 < 1 m s1 ) wind speeds, for which
thermometer screen properties are not reliably characterised.
Low wind speeds are particularly common around the minimum of the diurnal cycle in temperature, and therefore the
daily minimum temperature is likely to be the most badly
affected of the maximum, minimum and 0900 UTC temperature measurements. This is unfortunate, as the daily minimum temperature is also often used in the bi-mean method
to calculate the daily average temperature, in combination
with the daily maximum temperature. Such a procedure
to determine the daily average will carry forward the daily
minimum measurement vulnerable to ventilation errors.
The results obtained here are specific to the measurement
site at Reading. Whilst they will not be representative of
all sites globally, they are probably unlikely to be entirely
atypical of semi-urban sites in the United Kingdom, in terms
of the diurnal cycle of thermometer screen ventilation, or
in the year-to-year variability in calm periods arising from
different circulation patterns. Consequently the 0900 UTC
measurement may also provide a temperature observation at
other sites which is less affected by poor ventilation than the
daily minimum, although the overall prevalence of natural
ventilation effects on screen temperature measurements
seems unlikely to remain constant with time.

c 2010 Royal Meteorological Society


Copyright 

References
Andersson T, Mattison I. 1991. A field test of thermometer screens.
Swedish Meteorological and Hydrological Institute, SMHI Report
RMK No. 62, 41 pp.
Anderson SP, Baumgartner MF. 1998. Radiative heating errors in
naturally ventilated air temperature measurements made from buoys.
J. Atmos. Oceanic Technol. 15: 157173.
Bilham EG. 1937. A screen for sheathed thermometers. Q. J. R. Meteorol.
Soc. 63: 309322.
Bryant D. 1968. An investigation into the response of thermometer
screens The effect of wind speed on the lag time. Meteorol. Mag. 97:
183186, 256.
Cleveland WS. 1981. LOWESS: A program for smoothing scatterplots by
robust locally weighted regression. The American Statistician 35: 54.
DallAmico M, Hornsteiner M. 2006. A simple method for estimating
daily and monthly mean temperatures from daily minima and
maxima. Int. J. Climatol. 26: 19291936.
Fitzpatrick N. 2007. Uncertainty in Stevenson screen temperature
measurements. MSc dissertation, University of Reading.
Glaisher J. 1868. Description of thermometer stand. Symonss Monthly
Meteorol. Mag. 33: 155157.
Harrison RG, Knight JR. 2006. Thermopile radiometer signal
conditioning for surface atmospheric radiation measurements. Rev.
Sci. Inst. 77: 116105, DOI:10.1063/1.2370752.
Harrison RG, Pedder MA. 2001. Fine wire thermometer for air
temperature measurement. Rev. Sci. Inst. 72: 15391541.
Harrison RG, Rogers GW. 2006. Fine wire resistance thermometer
amplifier for atmospheric measurements. Rev. Sci. Inst. 77: 116112,
DOI:10.1063/1.2400013.
HMSO. 1981. Handbook of meteorological instruments, Volume 2
(Measurement of temperature). Meteorological Office, 2nd edition, Her
Majestys Stationery Office.
Huband NDS, King SC, Huxley MW, Butler DR. 1984. The performance
of a thermometer screen on an automatic weather station. Agric. Forest
Meteorol. 33: 249258.
Keil M. 1996. Temperature measurement in Stevenson screens. MSc
dissertation, University of Reading.
Laing J. 1977. Maximum summer temperatures recorded in Glaisher
stands and Stevenson screens. Meteorol. Mag. 106: 220228.
Margary ID. 1924. Glaisher stand versus Stevenson screen: A comparison
of forty years observations of maximum and minimum temperature
as recorded in both screens at Camden Square, London. Q. J. R.
Meteorol. Soc. 50: 209226.
Nakamura R, Mahrt L. 2005. Air temperature measurement errors in
naturally ventilated radiation shields. J. Atmos. Oceanic Technol. 22:
10461058.
Painter HE. 1977. An analysis of the differences between dry-bulb
temperatures obtained from an aspirated psychrometer and those from a
naturally ventilated large thermometer screen at Kew Observatory.
Meteorological Office O.P. Memo. 36 (unpublished, National
Meteorological Library, Exeter).
Parker DE. 1994. Effects of changing exposure of thermometers at land
stations. Int. J. Climatol. 14: 131.
Parker DE. 2004. Climate: Large-scale warming is not urban. Nature
432: 290.
Perry MC, Prior MJ, Parker DE. 2007. An assessment of the suitability
of a plastic thermometer screen for climatic data collection. Int. J.
Climatol. 27: 267276.
Strangeways I. 2003. Measuring the natural environment, 2nd edition.
Cambridge University Press.
WMO. 1983. Guide to meteorological instruments and methods of
observation. WMO-No. 8. World Meteorological Organisation:
Geneva.

Q. J. R. Meteorol. Soc. 136: 253259 (2010)

Das könnte Ihnen auch gefallen