Skip to content

Class 1 · 100 m

Class 2 · 30 m

Class 3 · 10 m

Class 4 · 3 m

24 m · class 3

0 min

60 min

120 min

180 min

-10 °C

0 °C

10 °C

20 °C

30 °C

Logger · 1-min air temperature

76 flagged

SI kelvin · ITS-90

National metrology institute

Accredited laboratory

Field reference

Station sensor

U 0.040 °C

U 0.057 °C

U 0.101 °C

1 · Observed variable

2 · Purpose of observation

3 · Station/platform

4 · Environment

5 · Instruments and methods

6 · Sampling

7 · Data processing and reporting

8 · Data quality

9 · Ownership and data policy

10 · Contact

0 km

50 km

100 km

150 km

200 km

5 km

10 km

15 km

0.5° · 1.5 km

P

S

30 s · 100 sps

0 dbar

1,000 dbar

2,000 dbar

Ascent · 5.5 h

Drift · 10 day cycle

Class 1 · 100 m; Class 2 · 30 m; Class 3 · 10 m; Class 4 · 3 m

Sensors · Explainer

Interactive

A number is only as good as its chain.

An observation carries where it was made, how it was checked, what it traces back to and who can vouch for it. Seven steps from a weather station’s mast to a float two kilometres down.

Standards
WMO-No. 8 · No. 1192 · GUM
Networks
Weather · radar · seismic · ocean
Chapters
7

01 · Siting

WMO-No. 8

The site sets the ceiling on quality.

WMO-No. 8 grades a site from class 1 to 5 for each variable. For air temperature it is the distance to heat sources such as buildings, car parks and water: class 1 beyond 100 m, then 30, 10 and 3 m. For wind it is distance in obstacle heights: 30, 10, 5 and 2.5 times. Classes 3 to 5 add up to 1, 2 and 5 °C of uncertainty to temperature.

Variable

Siting class
Class 3
Distance for the next class up
30 m for class 2
Added uncertainty, temperature
up to 1 °C

02 · Quality control

Flags

Every value leaves the logger with a flag.

The sensor feeds a logger that averages one-minute values; automatic checks flag each one before it travels. The range check rejects air temperatures outside −80 to +60 °C, the step check doubts a jump of more than 3 °C in a minute, and the persistence check doubts an hour that varies by less than 0.1 °C. Inject a fault and watch the flags.

Faults

Values in three hours
180
Flagged erroneous, range
0
Flagged suspect, step
1
Flagged suspect, persistence
75
Passed as good
104 · 57.8 %

03 · Traceability

GUM

An unbroken chain back to the kelvin.

Each calibration compares an instrument with a better one and adds its own uncertainty. Following the GUM, independent standard uncertainties combine as uc = √Σu², and the expanded uncertainty U = 2uc covers about 95 %. A display that reads in 0.1 °C steps alone contributes 0.1/√12. WMO-No. 8 asks for 0.1 K on air temperature.

Display resolution

Expanded uncertainty down the chain, k = 2

Lab reference

0.040 °C

+ bath

0.057 °C

Station sensor

0.101 °C

WMO-No. 8 target

0.100 °C

04 · WIGOS metadata

WMO-No. 1192

Metadata makes a number reusable.

The WIGOS Metadata Standard describes an observation in ten categories, and a WIGOS station identifier names the station worldwide: series, issuer, issue number and local identifier, as 0-20000-0-71xxx for a station with a WMO number. Fill the categories this station reports.

Categories reported

Issuer of identifier

Categories complete
6 of 10 · 60 %
WIGOS station identifier
0-20000-0-…
Missing
4, 6, 8, 9

05 · Radar volume scan

VCP

Fourteen tilts, and the Earth curves away.

A weather radar sweeps a stack of elevation angles every few minutes. Under standard refraction the beam rides an Earth of 4/3 its radius: h = √(r² + (kea)² + 2rkea sin θ) − kea. Far from the radar, the lowest tilt overshoots the weather near the ground and the beam widens as r·θ. Heights are stretched tenfold.

Radar

Beam centre above the radar
1.46 km
Beam bottom to top
0.00 – 3.03 km
Beam width across
3,142 m
Lowest tilt’s centre
1.46 km at 0.5°

06 · Seismic vault

GNSS time

Two arrivals give the distance; the clock keeps it honest.

The P wave outruns the S wave, about 6 against 3.5 km/s in the crust, so the gap between them grows with distance: S − P = R(1/Vs − 1/Vp), about 8.4 km per second of gap. A GNSS receiver on the vault stamps every sample; any clock error moves the event by Vp × error.

Clock error

Samples per second

P · S arrival
10.14 s · 17.38 s
S − P
7.24 s → 60.8 km
Sample interval · as distance
10 ms · 60 m
Clock error as distance
6 mm

07 · Ocean profile

Argo

A float that profiles two kilometres of ocean.

A profiling float changes its buoyancy with an oil bladder: it sinks to drift at 1,000 dbar, drops to its profile depth, then rises sampling temperature, salinity and pressure, and sends the profile by satellite before it sinks again. A decibar of pressure is close to a metre of depth.

Sample every

Cycle

Ascent time
5.5 h
Samples per profile
1,001
Profiles per year
36.5
Years at 200 cycles
5.5
Building the instruments…
Sensors · ExplainerInteractive

A number is only as good as its chain.

An observation carries where it was made, how it was checked, what it traces back to and who can vouch for it. Seven steps from a weather station’s mast to a float two kilometres down.

01 · SitingWMO-No. 8

The site sets the ceiling on quality.

WMO-No. 8 grades a site from class 1 to 5 for each variable. For air temperature it is the distance to heat sources such as buildings, car parks and water: class 1 beyond 100 m, then 30, 10 and 3 m. For wind it is distance in obstacle heights: 30, 10, 5 and 2.5 times. Classes 3 to 5 add up to 1, 2 and 5 °C of uncertainty to temperature.

Variable
Siting class
–
Distance for the next class up
–
Added uncertainty, temperature
–
02 · Quality controlFlags

Every value leaves the logger with a flag.

The sensor feeds a logger that averages one-minute values; automatic checks flag each one before it travels. The range check rejects air temperatures outside −80 to +60 °C, the step check doubts a jump of more than 3 °C in a minute, and the persistence check doubts an hour that varies by less than 0.1 °C. Inject a fault and watch the flags.

Faults
Values in three hours
180
Flagged erroneous, range
–
Flagged suspect, step
–
Flagged suspect, persistence
–
Passed as good
–
03 · TraceabilityGUM

An unbroken chain back to the kelvin.

Each calibration compares an instrument with a better one and adds its own uncertainty. Following the GUM, independent standard uncertainties combine as uc = √Σu², and the expanded uncertainty U = 2uc covers about 95 %. A display that reads in 0.1 °C steps alone contributes 0.1/√12. WMO-No. 8 asks for 0.1 K on air temperature.

Display resolution
Expanded uncertainty down the chain, k = 2
04 · WIGOS metadataWMO-No. 1192

Metadata makes a number reusable.

The WIGOS Metadata Standard describes an observation in ten categories, and a WIGOS station identifier names the station worldwide: series, issuer, issue number and local identifier, as 0-20000-0-71xxx for a station with a WMO number. Fill the categories this station reports.

Categories reported
Issuer of identifier
Categories complete
–
WIGOS station identifier
–
Missing
–
05 · Radar volume scanVCP

Tilt by tilt, the Earth curves away.

A weather radar sweeps a stack of elevation angles every few minutes. Under standard refraction the beam rides an Earth of 4/3 its radius: h = √(r² + (kea)² + 2rkea sin θ) − kea. Far from the radar, the lowest tilt overshoots the weather near the ground and the beam widens as r·θ. Heights are stretched tenfold.

Radar
Beam centre above the radar
–
Beam bottom to top
–
Beam width across
–
Lowest tilt’s centre
–
06 · Seismic vaultGNSS time

The gap between arrivals gives the distance; the clock keeps it honest.

The P wave outruns the S wave, about 6 against 3.5 km/s in the crust, so the gap between them grows with distance: S − P = R(1/Vs − 1/Vp), about 8.4 km per second of gap. A GNSS receiver on the vault stamps every sample; any clock error moves the event by Vp × error.

Clock error
Samples per second
P · S arrival
–
S − P
–
Sample interval · as distance
–
Clock error as distance
–
07 · Ocean profileArgo

A float that profiles two kilometres of ocean.

A profiling float changes its buoyancy with an oil bladder: it sinks to drift at 1,000 dbar, drops to its profile depth, then rises sampling temperature, salinity and pressure, and sends the profile by satellite before it sinks again. A decibar of pressure is close to a metre of depth.

Sample every
Cycle
Ascent time
–
Samples per profile
–
Profiles per year
–
Years at 200 cycles
–
08 · Scale

One station, then a network whose values compare.

The smallest unit is one station, sited to a known class. A site places several instruments around shared power and a clock; a cluster shares a calibration bench and a travelling reference; a region runs one network with one metadata practice; the global layer is many operators’ networks made comparable. The WMO Integrated Global Observing System is the framework that lets observations from different owners be used together.

Sources WMO-No. 1160, Manual on the WMO Integrated Global Observing System

09 · Exchange

From the station to every user who needs it.

An observation is checked, encoded in a standard format and published once. In the WMO Information System 2.0, centres publish notifications through message brokers using MQTT, users subscribe and download over the web, and a global discovery catalogue describes every dataset. A value that fails quality control is flagged or held back, so a subscriber receives data together with its quality, not raw readings alone.

Sources WMO-No. 1060, Manual on the WMO Information System, including WIS 2.0

10 · Maintenance

Calibration visits, planned from the data.

Every sensor drifts. A network keeps them honest with routine field checks against a travelling reference, recalibration on a schedule and replacement in rotation. Comparing each station with its neighbours flags the one whose readings wander, so the next visit goes where it is needed rather than on a fixed round. The WMO guide sets out inspection and calibration practice for each kind of instrument.

Sources WMO-No. 8, Guide to Instruments and Methods of Observation

11 · Lifecycle

A record that survives new instruments.

A station lives through a site survey, installation, calibration, years of operation and, often, a move or a new sensor. Each change can put a step into a long record. The climate monitoring principles ask that changes be assessed before they are made and that old and new systems overlap for a while; homogenisation methods then detect and adjust the breaks that remain.

Sources GCOS Climate Monitoring Principles; WMO-No. 1245, Guidelines on Homogenization

12 · Buyers’ questions

What to ask of an observing network.

Which siting class does each station meet, and where is that written down? Can every value be traced to a calibration record? Is the metadata complete enough for someone else to use the data years from now? Under which data policy is it shared? The WMO unified data policy separates core data, exchanged free and without restriction, from recommended data that may carry conditions.

Sources WMO Unified Data Policy, Resolution 1 of the Extraordinary Session of the World Meteorological Congress