John Nash Upper Air Technology Centre Met Office , UK

1 00/XXXX © Crown copyright

Introduction to performance of modern radiosondes based on WMO

Radiosonde Comparison results, Temperature

John Nash

Upper Air Technology Centre

Met Office , UK

Expert team on Upper Air Systems Comparison, Agenda item 3.1


Progress in radiosondes

Since the mid 1990’s, there have been a major changes in the technology applied to radiosonde design and production by European manufacturers. Next generation European radiosondes will be built using types of electronics and production techniques implemented in mobile phones.

This progress with modern radiosondes has led to a bigger gap in performance compared to the national radiosonde that have not really been modernised yet, i.e. those in China and India, and to some extent Russia.

The changes have been partly driven by the development of GPS radiosondes which require much more processing power on the radiosonde than was ever feasible before. The radiosondes will have more working Memory than the Met Office ground system in 1989.


Modern GPS radiosonde with narrow bandwidth , stable frequency, frequency tuneable before flight


One type of radiosonde(China) used widely in a national network for many years,

but now being replaced , sample from 1989


Test procedures The results shown in the following slides are

derived from WMO Comparison tests or tests of an equivalent standard

Test1(UK) 1984,2(USA) 1985, 3(KAZAKH) 1989, PREF(UK)1992, 4 (JAPAN01993, RH95(USA) 1995, BRAZ GPS (BRAZIL) 2001

In these tests from 3 to 6 radiosondes were flown together suspended from one balloon. An individual data set will include a minimum of 15 successful test flights and some will have more than twenty flights at a given time of day.

Vaisala(Finland) and Sippican (VIZ, USA) radiosonde participated in all tests and are used to link the results together.


Test procedures Comparisons between radiosondes are based on

simultaneous samples of temperature and pressure/height. Thus, differences in the temperature sensors are examined independent of the differences between the radiosondes introduced by pressure/height differences.

Temperatures at night are referenced to the measurements from 3 thermistor radiosondes (available in PREF, and Phase 4) and the equivalent references derived from the link radiosondes ( e.g. the UK MK3 links Phase I to PREF, RS80 + Sippican links Phase 4 to RH95 and BRAZ)

Linking daytime to night time measurements cannot be achieved so accurately, so daytime performance was directly measured against 3 thermistor in PREF and Phase 4, and the other tests are linked by examining day-night differences to indicate the most plausible daytime estimate.


Results on accuracy of best radiosonde temperature measurements

Accuracy depends on temperature sensor error and also the error in the height(pressure) assigned to the temperature

Temperature sensor errors are smaller at night, as long as sensor coating has low emissivity in the infrared (e.g. Vaisala RS80 , RS90, RS92)

Solar heating introduces significant systematic errors, difficult to correct, at pressures lower than 100 hPa

Random errors in temperature are less than 0.2 K at night and less than 0.3 K in daytime in the troposphere and lower stratosphere


Results from WMO Radiosonde Comparison demonstrating the range of systematic errors

in RS80 temperature sensor from 1984 to 2003

Temperature differences of Vaisala RS80 [link radiosonde] at night from the working reference ,

WMO Radiosonde Comparisons + PREFRS

0

5

10

15

20

25

30

35

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2

Temperature difference [K]

Ge

op

ote

nti

al h

eig

ht

[k

m]

RS80(I and 2)night

RS80(PREF)night

RS80(3) night

RS80(4)night

RS80 (RH95) night

RS80 (BRAZ)

Around 1989-91, one of the two calibration facilities was faulty in

the factory giving an additional positive offset at low temperatures for some batches of radiosondes

Increase in error with height

result of wrong software correction at

low pressures used extensively 1985-???

Software correction at low pressures much

smaller in recent software, 1995- 2003

[ Met Office systems ,1990 -2003]

NIGHT


This shows the typical ranges of measurement errors to be found with recent Vaisala radiosonde systems at night

Temperature differences of Vaisala RS80 [link radiosonde] at night from the working reference ,

WMO Radiosonde Comparisonssince 1993 + PREFRS

0

5

10

15

20

25

30

35

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2


Ge

op

ote

nti

al h

eig

ht

[k

m]

RS80(PREF)night

RS90(BRAZ)

RS80(4)night

RS80 (RH95) night

RS80 (BRAZ)

Effect of typical height error of RS80 on reportedtemperature over UK

NIGHT


Effect of cloud on a daytime temperature measurement

The next slide shows a detailed comparison from a PREFRS flight through thick medium and upper cloud, using the average of two three thermistor radiosonde measurements as the reference ( These should be accurate to better than 0.2 °C at these heights )

The Vaisala RS80 ground system has a software correction to compensate solar heating, based on normal conditions. In this flight the backscattering above the cloud is much higher than usual and the software correction was too small by about 0.6 °C


Infrared cooling of VIZ and AIR white rod thermistors increases to a larger extent than solarheating increases onemerging from the cloud,

RS80 temperature correction at 200 hPa is -0.6 deg C but is too low given the high solar albedo of this thick cloud

Reference is average of measurements from VIZ and AIR 3 thermistor radiosondes , PREFRS 1992


Temperature differences of Vaisala RS80 [link radiosonde] from the working reference , day time ,


0

5

10

15

20

25

30

35

-1.5 -1 -0.5 0 0.5 1 1.5


ge

op

ote

nti

al h

eig

ht

[k

m]

RS80(I )day

RS80(2)day

RS80(PREF)day

RS80(3)day

RS80(4)day

Phase 4 Low albedo

over open seaPREFRS,

High albedo over thick cirrus

Results from WMO Radiosonde Comparison in the day .The most reliable estimates are for Phase 4 and PREFRS, with

the large difference between the two the result of the correction algorithm not representing two very different

conditions.

DAY


White rod thermistors White rod thermistors manufactured by VIZ Manufacturing

Co in the USA ( later taken over to become Sippican, Inc) were used widely throughout the world on many radiosonde types

In each of the WMO Comparisons there were usually two systems using these sensors, so the results obtained should have been comparable in accuracy with the Vaisala RS80 results.

Errors from these sensors are large at night because although the sensors were white in the visible the sensors were effectively black in the infrared.

In the UK stratospheric temperatures change considerably with season, so this gave the chance to demonstrate that the infrared errors errors were not stable, but depended on how far ambient temperature was from radiative equilibrium at the given height.


Changes in infrared errorof white rod thermstor on passing through cloud, PREFRS Flight 57, dark

[values referenced to AIR + VIZ 3 thermistor radiosondes + Vaisal RS80 and UK Mk3

0

5

10

15

20

25

30

35

40

45

50

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

Temperature error [K]

Tim

e in

to f

lig

ht

[m

inu

tes]

Cloud layer

T= -65 °C

T= -30 °C

Temperature indicated by white rod thermistor at this level is about 0.4°C lower than with no cloud

NIGHT


Results from WMO Radiosonde Comparison show the errors in US white rod thermistors at night are influenced by infrared radiation and are less consistent than for Vaisala radiosondes

Difference of rod thermistor [link radiosonde]at night from the working reference


0

5

10

15

20

25

30

35

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2


Ge

op

ote

nti

al h

eig

ht

[km

]

(I)night

(2)night

(PREF)night

3TH(PREF)

(3)night

(4)night

VIZ(RH95)night

VIZ(BRAZ)night

NIGHT


IR errors at 10 hPa depend on atmospheric temperature and vary with cloud cover, surface temperature,etc . At

very low temperatures the infrared exchange produces a positive bias in temperature. Data are for flights with low

cloud amount. Infrared error in rod thermistor temperature at night at 10 hPa from multiple radiosonde comparisons 1984-1996 in the UK

-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

-80 -70 -60 -50 -40 -30

Temperature [degrees C]

Te

mp

era

ture

err

or

fro

m in

fra

red

he

at

ex

ch

an

ge

[K

]

Radiative equilibrium at -66 °C

10 hPa,NIGHT


The range of errors at 32 hPa is not as large at 10 hPa, with the stratosphere over the UK rarely warm enough

to give large negative temperature errors. Errors have been estimated from days with low cloud amount.

Infrared error in rod thermistor temperature at night at 32 hPafrom multiple radiosonde comparisons 1984-1996 in the UK

-1.5

-1

-0.5

0

0.5

1

1.5

-90 -80 -70 -60 -50 -40


Te

mp

era

ture

err

or

fro

m in

fra

red

he

at

ex

ch

an

ge

[K

]Radiative equilibrium at -62 °C

32 hPa,NIGHT


Infrared error in rod thermistor temperature at night at 200 hPafrom multiple radiosonde comparisons 1984-1996 in the UK

-1

-0.5

0

0.5

1

-70 -60 -50 -40


tem

pe

ratu

re e

rro

r fr

om

infr

are

d h

ea

t e

xc

ha

ng

e [

K]

200 hPa,NIGHT

The range of errors at 200 hPa is relatively low, on days with low cloud amount.

Radiative equilibrium at -56 °C, on average


The errors in US rod thermistors introduced during the day [not usually corrected].

PREFRS and Phase 4 results were the most reliable.

Difference of rod thermistors [link radiosondes] from the working reference, daytime ,

WMO Radiosonde Comparison +PREFRS

0

5

10

15

20

25

30

35

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2


Ge

op

ote

nti

al

he

igh

t [

km

]

Rod(I)day

Rod(2)day

Rod(PREF)day

Rod(3)day

Rod(4)day

DAY

PREFRS, High albedo

over thick cirrus

Phase 4 Low albedo

over open sea


Temperature differences from the working reference at night , WMO Radiosonde Comparison

0

5

10

15

20

25

30

35

-3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1


Ge

op

ote

nti

al h

eig

ht

[km

]

G78C(I)

GM60(2)

China(3)

Japan RS2-80(4)

Japan RS2-91(4)

MOD(BRAZ)

Results from other radiosondes at night [1]. All these radiosondes show signs of infrared cooling at night , either from a white sensor or the black coating on the inside of protective ducts apart from the RS2-91 (Japan) .

NIGHT


Temperature differences from the working reference at night , WMO Radiosonde Comparison

0

5

10

15

20

25

30

35

-2 -1.5 -1 -0.5 0 0.5 1 1.5 2


ge

op

ote

nti

al h

eig

ht

[km

]

UKRS3(1)

India(2)

Russia(3)

Sipp(BRAZ)

RS90(BRAZ)

Results from other radiosondes at night [2]. Indian and Russian radiosondes have some infrared cooling, but the support wires to the sensor are unpainted and relatively thick. So, the infrared error

on the sensor is compensated by thermal conduction from the uncoated wires . UKRS3, new Sippican and RS90 have very small infrared errors

NIGHT


Conclusions Results from WMO Comparisons can be expected to

represent operational results at night for radiosondes where temperature sensors are insensitive to infrared radiation errors.

Linking daytime measurements is more difficult. Hence the reason to reduce susceptibility to daytime heating in next generation radiosonde designs.

The other limitation on relating temperature comparisons to operational data is is where pressure errors are significant . Changes in performance of pressure sensors are poorly documented in many countries.

John Nash Upper Air Technology Centre Met Office , UK

Documents