METEOROLOGICAL APPLICATIONS

Meteorol. Appl. 22: 867–872 (2015)

Published online 24 November 2015 in Wiley Online Library
(wileyonlinelibrary.com) DOI: 10.1002/met.1527

The basics of calibration procedure and estimation of

uncertainty budget for meteorological temperature sensors
Agnieszka Grykałowska,* Aleksandra Kowal and Anna Szmyrka-Grzebyk
Instytut Niskich Temperatur i Badań Strukturalnych, Laboratory of Temperature Standard, Wrocław, Poland

ABSTRACT: The primary aim of this study was to develop the basic principles for calibration of meteorological temperature
sensors. The sensors commonly used in the meteorological community were calibrated using a comparison method in the
temperature range from −40 to +50 ∘ C. Measurements were carried out using a typical liquid thermostatic bath and a climatic
chamber. The experiments allowed the identification of characteristics for each calibration system and the definition of
metrological parameters for sensors used in weather stations. Components of uncertainty budget were identified and estimated,
and the total uncertainty was calculated. The calibration procedure allowed to reach the uncertainty of calibration in a liquid–air
equipment U < 0.10 ∘ C and in a climatic chamber: 0.27 ∘ C ≤ U ≤ 0.99 ∘ C. Presented outcomes were obtained as a result of the
project MeteoMet – Metrology for Meteorology.

KEY WORDS calibration; uncertainty budget; temperature sensor

Received 29 September 2014; Revised 27 April 2015; Accepted 17 May 2015

1. Introduction of meteorological temperature sensors and to assess their metro-

logical parameters.
The success of any meteorological and climate observation
Tests and calibrations of temperature sensors were divided into
depends on the availability of reliable data. This requirement led
the following phases:
to the establishment of active co-operation between two orga-
nizations – the World Meteorological Organization (WMO), 1 characterization of the parameters of measurement systems
which monitors climatic conditions on the Earth, and the used in the experiments, i.e. specially designed thermostat for
International Bureau of Weights and Measures (BIPM), which maintaining suitable thermal parameters (a liquid–air equip-
is responsible for the development and assurance of precise ment and a climatic chamber);
measurements. On the basis of current data, the WMO and the 2 calibration carried out using the liquid–air equipment;
BIPM recognized the urgent need for a joint action (WMO 3 study of the humidity impact on thermometer readings in the
and BIPM, 2010) to improve and standardize the methods climatic chamber;
used to measure meteorological parameters. In April 2010, the 4 calibration in the climatic chamber;
International Committee for Weights and Measures (CIPM) 5 comparative in situ measurements in a Stevenson screen.
document – Mutual Recognition Arrangement (BIPM, 1999),
developed by the BIPM to ensure worldwide measurement
traceability, was signed by the WMO.
In May 2010, the Consultative Committee for Thermome- 2. Estimation of correction and uncertainty of the
try of the CIPM recommended strong co-operation between calibration
National Metrology Institutes and meteorological institutes at The purpose of any physical measurement is to determine the
local, national and international levels. The project MeteoMet value of a measurand. A result of an experiment is only an
entitled ‘Metrology for pressure, temperature, humidity and air- approximation of the measurand and therefore should be given
speed in the atmosphere. Metrology for Meteorology’ was the with an estimate of the approximation. The EA-4/02 M:2013
response to these recommendations. document, Evaluation of the Uncertainty of Measurement In
The structure of the project reflects two key aspects: scientific Calibration (EA-4/02 M, 2013), provides basic recommendations
innovation and practical traceability for end users. It includes on how to present a measured value. This method should also be
development and tests of novel instruments as well as improve- applied in meteorology. Presently, only a part of meteorological
ment in calibration procedures and facilities for ground-based results is disseminated with the associated uncertainty.
observations, in situ calibrations and best practice dissemination. For every measurement, an equation should be created in the
This study presents part of the results of the project MeteoMet. Its form: ( )
primary goal was to develop the basic principles for calibration y = f x1 , x2 , … , xN

where y is an estimate of an output quantity that depends on a

number input estimates xi for values of input quantities.
* Correspondence: A. Grykałowska, Instytut Niskich Temperatur i
The thermometer calibration process consists of determining
Badań Strukturalnych, ul. Okólna 2, 50-422 Wrocław, Poland. E-mail:
[email protected]
the correction and standard uncertainty to be applied to a cali-
brated thermometer (BIPM, 2008). Therefore, the input values xi

© 2015 Royal Meteorological Society

868 A. Grykałowska et al.

have been corrected or the necessary corrections have been intro- automatically (using program MINI 32) after completed mea-
duced as separate input quantities. The correction is equal to the surements. In this study, sensors connected to converter ADAM
negative of the systematic error: are named a ‘thermometer’ with the symbol indicated in Table 1.

Δtcor = tw − tx 3.2. Measurement systems

where tw is a value of temperature measured with the reference The measurements were carried out in a liquid–air equipment
thermometer, and tx is temperature recorded by the calibrated and in a climatic chamber. In both cases, the same platinum resis-
thermometer. It is worth mentioning that the error is an idealized tance thermometer (PRT), calibrated at fixed points of the Inter-
concept and can never be known exactly. All that can be known national Temperature Scale of 1990 (Preston-Thomas, 1990) and
is its estimated value. Even the systematic error (compensated traceable according to the Polish national temperature standard,
by correction, as opposed to random error) cannot be known was used as a reference thermometer. A resistance bridge with an
perfectly and the compensation cannot be complete (BIPM, internal reference resistor measured the thermometer resistance.
2008). The resolution of the bridge was 0.0001 ∘ C and the measurement
The second valid quantity determined during the calibration is uncertainty was U = 7.0 · 10−5 Ω.
the uncertainty of a measurement (at calibration circumstances).
Based on the measurement equation, a standard uncertainty of 3.2.1. Liquid–air equipment
the output estimate is formulated:
( ) ( ) ( ) Liquid–air equipment was composed of:
u2 (y) = c21 u2 x1 + c22 u2 x2 + .... + c2N u2 xN
1 a liquid thermostatic bath ensuring temperature stability of
where u(xi ) are standard uncertainties of the input estimates xi , ±0.0060 ∘ C;
and ci are the sensitivity co-efficients associated with xi . 2 a cylindrical air chamber made of copper, mounted in the cen-
The uncertainty analysis, commonly called the uncertainty tral axis of the bath, preventing the metrological sensors from
budget of measurement, should include a list of all sources of direct contact with the liquid medium. The temperature stabil-
uncertainty together with the associated standard uncertainties ity obtained in the chamber was equal to ±0.0015 ∘ C. The tem-
and the methods of evaluating them. The EA-4/02 M:2013 docu- perature gradients in the workspace of cylinder (𝜙 = 5 cm, high
ment (EA-4/02 M, 2013) recommends to present the data relevant l = 5 cm) did not exceed 0.007 ∘ C cm−1 . The gradients were
to this analysis in the form of a table. The data shall contain an determined using calibrated resistance thermometers during
expanded uncertainty of measurement U, obtained by multiply- additional tests at various temperatures covering an entire cal-
ing the standard uncertainty u(y) by a coverage factor k: ibration range.

U = k · u (y) Both sensors, the reference PRT and the tested one, were placed
in the air chamber along its vertical axis. The entire length
The assigned expanded uncertainty corresponds to a coverage of the sensors was inside the chamber together with a pair of
probability of ∼95%. In case of a normal distribution of the electrical wires. Active elements of both sensors were placed at
measurand, the standard coverage factor k = 2 shall be used. the same height within a short distance (1–2 cm) from each other.
The numerical value of U shall be given up to, at most, two Sensors were protected from thermal contact between them and
significant figures (ILAC, 2013). In reporting the final results, it the chamber wall (Figure 1).
may sometimes be appropriate to round up uncertainties rather The system included a computer for data acquisition. A pro-
than to the nearest digit (BIPM, 2008). The numerical value of gram using LabView package allowed to set up the characteristic
the measurement result in the final statement shall be rounded co-efficients (specified by the manufacturer) and the sampling
up to the least significant figure in the value of the expanded time for each sensor.
uncertainty assigned to the measurement result (ILAC, 2013). Properly made measurement required stable thermal conditions
in the measurement system. The liquid–air equipment achieved
the stable state after 1.5–2.5 h, because heat exchange in the
3. Experiment in the laboratory conditions air chamber was caused by convection only, without forced
air flow.
3.1. Selection of tested sensors
The system required a preliminary observation to check
Appropriate preparation of measurements needs an adequate whether the tested thermometers generate any heat just after
selection of sensors for tests and calibration. In this study, the turning on, disrupting the thermal conditions in the chamber.
selected sensors are commonly used by the meteorological ser- Such an effect was observed for the Minikin thermometers,
vices and represent not only different models but also different which affected the time of achieving thermal equilibrium in the
operating times. One of these sensors was new, and the remain- system, and thus the duration of the calibration process.
ing had been previously used by a meteorological service. Table 1
3.2.2. Climatic chamber
summarizes the basic information about the selected sensors. It
contains the measurement uncertainties – at room temperature The climatic chamber used in the experiments allowed the tem-
and at extreme temperatures of scope – provided by the manu- perature value to be changed from −40 to +50 ∘ C and the relative
facturers. Operating time means the working period in the mete- humidity from 20 to 90% above 10 ∘ C. The temperature stability
orological service. in the chamber was ±0.050 ∘ C.
During measurements, the sensors were connected to an exter- During the tests, the difference between the reference and
nal analog digital converter Advantech ADAM-4017. Minikin tested sensor positions in the vertical plane was 3 cm, and in the
thermometers were an exception as their construction allows to horizontal plane it was determined by the thickness of the sensors
collect data independently, they consist of both a sensor and (maximum 5 cm). Only the part of the wires needed to mount the
a logger. The Minikin temperature registration was performed converter was pulled outside the chamber.

© 2015 Royal Meteorological Society Meteorol. Appl. 22: 867–872 (2015)

 The basics of calibration procedure of temperature sensors 869

Table 1. Selected meteorological sensors.

Model Producer Uncertainty at Uncertainty at Operating time Symbol

room temperature (∘ C) extreme temperatures of scope (∘ C)

HMP 155 Vaisala 0.15 0.3 ∘ C at –50 ∘ C New HMP 155

0.35 ∘ C at +50 ∘ C
MP 103 A Rotronic 0.3 No data 10 years RT
HMP 45 AC Vaisala 0.2 0.5 ∘ C at –40 ∘ C 5 years HMP 45A
0.35 ∘ C at +50 ∘ C
HMP 45 AC Vaisala 0.2 0.5 ∘ C at –40 ∘ C A few months HMP 45B
0.35 ∘ C at +50 ∘ C
Minikin T EMS Brno 0.2 No data 4 years M5
Minikin TH EMS Brno 0.2 No data 4 years M9

60

40

20

tx (°C)
0
–60 –40 –20 0 20 40 60
–20

–40

–60
tw (°C)

Figure 2. A typical linear characteristic tx = f (tw ) of meteorological

thermometers.

3.4. Calibration results

The obtained temperature characteristics of all the

meteorological sensors were approximately linear with the
linear regression co-efficient R2 ≈ 1 (Figure 2). In reality,
the characteristics of the thermometers are not exactly linear.
Figure 3 depicts the obtained results in the form Δt = f (tw ), where
Δt = (tx – tw ), and highlights the deviations from the linearity.
All values of Δt obtained during calibration in the climatic
chamber agree (within the estimated uncertainty) with those
obtained in the liquid–air equipment.

3.5. Analysis of the results

The components of the uncertainty budget u(xi ) (the standard

Figure 1. Scheme of the liquid–air equipment. uncertainties of input estimates xi ) are presented in Table 2. The
division into three parts (i.e. (1) components derived from the
reference thermometer, (2) components derived from the mea-
3.3. Measurement procedure surement system and (3) components derived from the meteo-
The temperature measurements of the reference PRT and the rological thermometer) was made in order to facilitate further
tested meteorological thermometers were carried out simultane- analysis. The contribution of these groups of components was
ously, i.e. the measurements were made at the same time and at analysed in detail.
the same time intervals (1–3 s). Two series of the same num- The component related to the homogeneity of temperature in
ber of measurements were performed for each temperature (at the area of calibrated sensor u(𝛿tx hom ) was omitted due to its
least 1200 measurements for the sensors connected to the ADAM negligible value (the active element is small in size).
converter and at least 250 measurements for the Minikin ther-
3.5.1. Uncertainty budget of calibration in the liquid–air
mometers in each series).
equipment
The measurements performed in the liquid–air equipment
were made at every 10 ∘ C in the temperature range between −40 The components associated with the reference thermometer
and + 50 ∘ C (for the HMP 155 thermometer, the calibration range showed the smallest impact to the total calibration uncertainty
was from −50 to +50 ∘ C). The tests in the climatic chamber were u (k = 1) – between 0.001 and 0.005 ∘ C. The value of the uncer-
performed at temperatures of about −37, 0, 10, 20 and 50 ∘ C for tainty derived from the measurement system used was in the
three different values of relative humidity (20, 50 and 90%) at range of 0.010–0.015 ∘ C. The uncertainty derived from the cali-
temperatures ≥ 10 ∘ C. brated thermometer was larger and varied from 0.015 to 0.050 ∘ C.

© 2015 Royal Meteorological Society Meteorol. Appl. 22: 867–872 (2015)

870 A. Grykałowska et al.

(a) 0.1 (b) 0.4

0.2
0

0
Δt (°C)

Δt (°C)
–0.1
–0.2

–0.2 –0.4

–0.3 –0.6
–60 –40 –20 0 20 40 60 –60 –40 –20 0 20 40 60
tW (°C) tW (°C)

Figure 3. Characteristics Δt = f (tw ) of the chosen meteorological thermometers: (a) HMP 155 and (b) RT.

Table 2. The calibration uncertainty budget components.

Components derived from Components derived from Components derived from the
the reference thermometer the measurement system meteorological thermometer

• u(𝛿twdev ) – standard deviation of tw • u(𝛿tstab ) – temperature stability • u(𝛿tx dev ) – repeatibility of tx results
results • u(𝛿tgradH ) – temperature gradient in the • u(𝛿tx resol ) – resolution
• u(𝛿tw scat ) – scattering of tw value horizontal plane of the air chamber • u(𝛿tx rep ) – reproducibility
• u(𝛿tw std ) – uncertainty of PRT according • u(𝛿tgradV ) – temperature gradient in the • u(𝛿tx hist ) – hysteresis
to its Calibration certificate vertical plane of the air chamber • u(𝛿tx hom ) – homogeneity
• u(𝛿tw drift ) – drift of PRT • u(𝛿tx rad ) – radiation effect (in climatic
• u(𝛿tb std ) – uncertainty of resistance chamber only)
bridge according to its Calibration
certificate
• u(𝛿tb drift ) – drift of resistance bridge

As an example, the HMP 155 uncertainty budget for calibra- the expanded uncertainty. This is particularly evident in the case
tion at about −10 ∘ C is presented in detail in Table 3. The budget of the RT with a parabolic characteristic (Figure 3(b)).
includes the estimates and standard uncertainties of all known The WMO suggests that the expanded uncertainty U obtained
quantities that influenced the calibration results. Drift of a resis- in laboratory conditions during calibration by the comparison
tance meter was set to zero, since it was unknown (it was a new method should be ∼0.1 ∘ C at the 95% confidence level (WMO,
resistance bridge). The last line of the right column shows the 2010). This uncertainty mentioned in the document does not con-
calculated total uncertainty u (k = 1). On the same row (second tain effects due to non-linearity, drift, repeatability and repro-
column from the left), the estimated correction is presented. The ducibility of the meteorological sensor and its transducer. All
table gives an idea about the orders of magnitude of the u(xi ) the results presented above prove the accuracy of the performed
components. The expanded uncertainty in the presented example calibration and the adequacy of the thermometers used for mete-
is U = 0.074 ∘ C (the coverage factor k = 2, which corresponds to orological measurements (Table 5, values U 1 ).
the coverage probability of ∼95%).
The results of the HMP 155’s calibration are presented in 3.5.2. Uncertainty budget of calibration in the climatic
Table 4, which shows the measured values tx and calculated Δtcor chamber
for all calibrations performed in the temperature range from −50 The measurement uncertainty obtained in the climatic chamber is
to +50 ∘ C. The expanded uncertainty, given to one significant much larger than that obtained in the liquid–air equipment. The
digit, did not change for all calibrations at reference tempera- main components affecting this difference are:
tures tw . A similar situation occurred for other meteorological
thermometers with two exceptions: HMP 45B and M5, where the • the radiation effect;
value of the uncertainty strongly depended on the standard devi- • the temperature stability in the climatic chamber;
ation of tx results – u(𝛿tx dev ). Among all the investigated Vaisala • the temperature gradient in the climatic chamber;
sensors (Table 1), the smallest uncertainty values were obtained • the dispersion of the values tx .
for thermometers with the longest history of work.
The measurement results presented in the form of a linear The smallest impact on the calibration uncertainty u (k = 1) in
equation require the addition of a real thermometer characteristic the climatic chamber was derived from the reference thermome-
to the uncertainty budget of a component arising from lineariza- ter (∼0.005 ∘ C). This value has not changed significantly in com-
tion. Table 5 presents the uncertainties obtained for all ther- parison with the measurements in the liquid–air equipment. The
mometers, at temperature ∼20 ∘ C, where U 1 is an uncertainty value of the uncertainty associated with the measurement system
containing components derived from the reference thermometer, ranged between 0.13 and 0.47 ∘ C. The contribution of the uncer-
the measurement system and the calibrated thermometer, and U 2 tainty derived from the calibrated thermometer ranged between
is the uncertainty with an additional component derived from the 0.02 and 0.14 ∘ C. The expanded uncertainty of the measurement
linearization of the characteristic. The results show that the addi- U (k = 2) varied with the temperature and humidity condition and
tion of the linearization component caused significant increase of ranged between 0.27 and 0.99 ∘ C.

© 2015 Royal Meteorological Society Meteorol. Appl. 22: 867–872 (2015)

 The basics of calibration procedure of temperature sensors 871

Table 3. The uncertainty budget of the HMP 155 calibration.

Quantity, Estimate, Standard uncertainty, Sensitivity Contribution to the total standard uncertainty,
xi (∘ C) xi (∘ C) u(xi ) (∘ C) co-efficient, ci ci u(xi ) (∘ C)

tw –10.0157
𝛿tw dev 0 2.58 × 10 –04 1 2.58 × 10 –04
𝛿tw scat 0 4.55 × 10 –04 1 4.55 × 10 –04
𝛿tw std 0 3.15 × 10 –04 1 3.15 × 10 –04
𝛿tw drift 0 1.07 × 10 –03 1 1.07 × 10 –03
𝛿tb std 0.00015 3.50 × 10 –04 1 3.50 × 10 –04
𝛿tb drift 0 0.00 × 10+00 1 0.00 × 10+00
𝛿tstab 0 1.03 × 10 –04 1 –1.03 × 10 –04
𝛿tgradH 0 9.67 × 10 –03 1 9.67 × 10 –03
𝛿tgradV 0 7.74 × 10 –03 1 7.74 × 10 –03
tx –10.1008
𝛿tx dev 0 3.22 × 10 –02 1 3.22 × 10 –02
𝛿tx resol 0 4.04 × 10 –03 1 4.04 × 10 –03
𝛿tx rep 0 1.04 × 10 –02 1 1.04 × 10 –02
𝛿tx hist 0 9.51 × 10 –03 1 9.51 × 10 –03
Δtcor 0.0853 0.037

Table 4. Results of the HMP 155 thermometer calibration.

HMP 155
Reference temperature Measured temperature Correction Uncertainty (k = 2)
tw (∘ C) tx (∘ C) Δtcorr (∘ C) U (∘ C)

50.26 50.21 0.05 0.08

40.26 40.26 0.00 0.08
30.26 30.23 0.03 0.08
23.06 23.05 0.01 0.08
10.26 10.26 0.00 0.08
5.27 5.23 0.04 0.08
0.02 −0.06 0.08 0.08
−10.02 −10.10 0.09 0.08
−25.00 −25.18 0.18 0.08
−39.99 −40.07 0.08 0.08
−49.97 −50.06 0.09 0.08

Table 5. Comparison of the uncertainty values, U 1 and U 2 of tested calibration in the climatic chamber is less accurate than in the
meteorological thermometers at temperature ∼20 ∘ C. liquid–air equipment, it allows to evaluate the influence of the
parameters as a whole.
U 1 (∘ C) U 2 (∘ C)
3.5.3. Study of humidity influence on meteorological
HMP 155 0.08 0.15 thermometer parameters
RT 0.04 0.72
HMP 45A 0.05 0.09 The temperature sensors in the meteorological equipment used
HMP 45B 0.05 0.10 for this study were integrated with a humidity sensor (the M5
M5 0.08 0.15 thermometer is an exception) and in practice they were never
M9 0.06 0.12 protected from the effects of humidity. It was the main reason
why the meteorological sensors were tested in the climatic
chamber. Our measurements allowed to examine the influence
of humidity on the readings of the meteorological thermometers
It can be seen that the largest contribution to the total uncer-
as well. There was no strict relation of the humidity and the
tainty depended on the measurement system. Reduction of the
recorded temperature. Furthermore, there was no difference in
calibration uncertainty can be achieved by reducing the calibra-
the behaviour of the new thermometers and those previously used
tion area in the chamber (reduced gradients) or by using an addi-
in weather stations for a long time. It is also worth noting that
tional container for the thermometer (Lopardo et al., 2012). the tests in the climatic chamber have not changed the metro-
Large dispersion of the values indicated by the investigated logical characteristics of the thermometers, as confirmed by
thermometer (tx ) results from the fact that the sensors in the subsequent measurements carried out in the liquid thermostatic
climatic chambers are influenced by additional external fac- bath (Grykałowska et al., 2014; Wełna et al., 2014).
tors (e.g. variable air speed, radiation). Commercial climatic
chambers are not able to carry out controlled change of these
4. In situ tests
parameters, which would allow for an accurate analysis of their
impact. Therefore, it is impossible to estimate the contribution In situ tests were performed in a Stevenson screen to check the
of these factors to the uncertainty budget. However, although the accuracy of temperature measurement in typical conditions for

© 2015 Royal Meteorological Society Meteorol. Appl. 22: 867–872 (2015)

872 A. Grykałowska et al.

0.4 in the laboratory conditions (the liquid–air equipment) allowed

Liquid–air
to reach a small uncertainty of the calibration – below 0.1 ∘ C.
In situ
0.2 The measurements carried out in the climatic chamber, which
allowed to examine the influence of the humidity on the meteoro-
logical thermometers, have confirmed a lack of relation between
0
the recorded temperature and humidity. Moreover, the tests per-
Δt (°C)

formed in the climatic chamber have not changed the metrolog-

–0.2 ical characteristics of the thermometers. The uncertainty of the
calibration in the climatic chamber was found to be larger than
–0.4 that obtained in the liquid–air equipment, but still possible to be
estimated. The calibration uncertainty obtained in the laboratory
–0.6 conditions does not reflect the actual measurement uncertainty,
–40 –20 0 20 40 60 which can be expected in operating conditions.
tW (°C)

Acknowledgements
Figure 4. The comparison of the M9 thermometer results obtained in the
Stevenson screen and in the liquid–air equipment. This study was performed within the frame of the European
Metrology Research Program (EMRP) joint research project
meteorological measurements. The sensors were hanged verti- ‘METEOMET’. The EMRP is jointly funded by the EMRP
cally during the experiment without pulling and moving them. participating countries within the EURAMET and the European
The distance between the active elements of the sensors – the Union.
tested and the PRT ones – was ∼4 cm. The procedure used The present study was done under the project of Ministry of
allowed to perform measurements without opening the screen Science and Higher Education 2373/GRANT KE/2012/2.
door and did not interfere with the sensors’ location. The resis-
tance bridge was located outside the Stevenson screen. The air References
temperature during the test changed in the range of 15–35 ∘ C.
The tests indicated that the discrepancy of the results can be sig- BIPM. 1999. Mutual recognition arrangement. http://www.bipm.org/en/
cipm-mra/participation/signatories.html (accessed 29 September
nificantly larger (even 10 times) than the expanded uncertainties
2014).
determined in the liquid–air equipment (Figure 4). The criterion BIPM. 2008. JCGM 100:2008 GUM 1995 with minor corrections, evalu-
of the achievable measurement uncertainty (i.e. based on sensor ation of measurement data – guide to the expression of uncertainty in
performance under nominal and recommended exposure that can measurement. http://www.BIPM.org (accessed 29 September 2014).
be achieved in operational practice) (WMO, 2010), U ≤ 0.2 ∘ C, EA-4/02 M. 2013. Evaluation of the uncertainty of measurement in cal-
ibration. September 2013 rev01. http://www.european-accreditation.
was not met. This result indicates that sources of error, which org/publication/ea-4-02-m (accessed 29 September 2014).
are not taken into account during calibration, are quite impor- Grykałowska A, Kowal A, Szmyrka-Grzebyk A. 2014. INTiBS partic-
tant. It should be emphasized that the two types of the sensors, ipation in the MeteoMet project. In Proceedings of Polish Nation-
the PRT and the meteorological ones, were of different construc- wide Seminar: Measurement Problems in Meteorology, 24 June 2014.
Wrocław, Poland.
tion, and therefore they may have responded in a different way ILAC. 2013. ILAC-P14:01/2013 ILAC Policy for uncertainty in cal-
to wind speed, solar radiation, salinity of the environment and ibration. http://ilac.org/publications-and-resources/ilac-documents/
atmospheric pressure. procedural-series (accessed 29 September 2014
Additional contributions to the calibration uncertainty need Lopardo G, Marengo D, Meda A, Merlone A, Moro F, Pennecchi FR,
to be taken into account for in situ measurements. The Guide et al. 2012. Traceability and online publication of weather station mea-
surements of temperature, pressure, and humidity. Int. J. Thermophys.
(WMO, 2010) lists some of these sources: 33: 1633.
Merlone A, Bell S, Benyon R, Bergerud AR, Bertigia F, Boese N,
• the effectiveness of the heat transfer between the thermometer Brunet M, Buee B, del Campo D, Dahl J, Deboli R, Dobre M,
element and the air in the thermometer shelter, which should Ebert V, Emardsson R, Garcia Izquierdo C, Georgin E, Gilabert A,
Grudniewicz E, Heinonen M, Hermandez S, Hogstrom R, Holm-
ensure that the element is at thermal equilibrium with the air;
sten M, Holstein-Rathlou C, Hudoklin D, Jarlemark P, Jahans-
• the effectiveness of the thermometer shelter, which should son J, Kajastie H, Kaykisisli H, Klason P, Knorova R, Lakka A,
ensure that the air in the shelter is at the same temperature Lau P, Lopardo G, Melvad C, Merrison J, Migała K, Mokdad S,
as the air surrounding it; Piccato A, Pitre L, Roggero R, Ruiz S, Saathoff H, Sairanen H,
• the exposure, which should ensure that the shelter is at a tem- Sanna F, Saxholm S, Smorgon D, Sparasci F, Spazzini P, Stein-
metz E, Strnad R, Szmyrka-Grzebyk A, Verge A, Vidal V, Vuiller-
perature, which is representative of the region to be monitored. moz E. 2013. The “MeteoMet” project – Metrology for Meteorol-
ogy: challenges and results. In XII Symposium on Temperature and
The results of our measurements as well as those presented by Thermal Measurements in Industry and Science TEMPMEKO 2013:
other partners of the MeteoMet project (Merlone et al., 2013) 14–18 October 2013. Funchal.
strongly indicate that the development of a reliable uncertainty Preston-Thomas H. 1990. The international temperature scale of 1990.
Metrologia 28: 3–10.
budget for measurements in operating conditions is currently of Wełna A, Wiśniewska B, Kozicki M, Grudniewicz E. 2014. Procedures
great importance. Further joint research will be focused on this of maintaining and calibrating of Automatic Weather Stations. In
issue. Proceedings of Polish Nationwide Seminar: Measurement Problems
in Meteorology, 24 June 2014. Wrocław, Poland.
WMO. 2010. Guide to meteorological instruments and methods of
observation. http://library.wmo.int/pmb_ged/wmo_8_en-2012.pdf
5. Summary (accessed 29 September 2014).
WMO and BIPM. 2010. Report on the WMO-BIPM Workshop on
The results presented in this study showed a good reproducibility measurement challenges for global observation systems for climate
of metrological parameters of the sensors used in meteorology for change monitoring. Traceability, stability and uncertainty. WMO:
temperature measurements. The calibration procedure applied Geneva.

© 2015 Royal Meteorological Society Meteorol. Appl. 22: 867–872 (2015)