1. Crowdsourced data for quality control and nowcasting - early studies by Met Norway
Mareile Wolff1 with input from Cristian Lussana2 and Thomas Nipen3
Norwegian Meteorological Service
1 observation quality and data processing
2 climate services
3 center for development of weather forecast

2. Netatmo – low cost weather sensors

3. MET Norway
Access to live datastream and archive (2013-current)

4. MET Norway
Precipitation
Number of stations

5. Netatmo’s temperature
Outside temperature sensor (and «shield») has to be protected from rain and needs to be sited within the line of sight of the indoor sensor (to guarantee their connection)
close to a window and under a roof.
Comparison of MET Norway’s AWS Oslo-Blindern and Netatmo
Better agreement during winter months

6. Netatmo’s temperature
«Summer» problems:
Higher and later daily maximum mostly because of unsufficient radiation shielding
Higher daily minimum temperature during night because of changed long wave radiation budget due to siting.

7. Quality control of Netatmo temperatures
Altitude check
± 5 STDEV of neighbouring altitudes
Buddy check
± 5 STDEV of neighbouring observations
Forecast ensemble check
± 5 ensemble STDEV of ensemble mean
Spatial consistency check*
Close to the cross-validated field 1 2 3 4 z T
* Lussana C. et al.,2010 – Q.J.R. Meteorol. Soc. 136:

8. Example Feb 20, UTC

9. Use of Netatmo temperatures
Challenge:
The 2.5 km model does not resolve
the actual topography
2 m
Model-topography
yr.no distributes temperature forecasts which may be more than 3 ºC off the actual values, even if some postprocessing during inversion situations is already in place
Real topography

10. Use of Netatmo temperatures
Model
Model + Netatmo
(°C)

11. MET Norway’s observation network
Varying impact
MET Norway’s observation network
Netatmo
30 km
1 km
Cities
Countryside / cabins
No infra-structure
30 km
15 km

12. Use of Netatmo temperatures
Measureable effect: Reducing amount of large errors in wintertime from 15% to 5%

13. Use of Netatmo temperatures
Netatmo Nowcast for Hakadal, a town just outside Oslo, based on 13 Netatmo stations located in and near the town: great improvements of January temperatures

14. Use of Netatmo temperatures
… not so easy to deal with the «summer» problems, larger outliers are easily detected, but smaller biases (probably affecting all sensors) might not get caught by QC: working on further improvements

15. Netatmo Precipiation
…Netatmo sensors don’t measure snow
-2°C
+5°C

16. Netatmo Wind
…underestimates by a factor of ~3.5. Correlation not too bad

17. Netatmo humidity
Overestimates by 2%. Correlation of 0.98.

18. Other external and crowdsourced data @MET Norway
MET Norway is collecting and quality controlling data from the road authorities-network (some hundreds of stations)
We also have an agreement to collect car data from the road- authorities service cars, currently working on establishing data transfer
Running tests for the use of stations in Bergen area (school network)
Discussing agreement with Volvo to collect temperature data of Volvo- cars in Norway
Other external weather stations which are associated to our network (and quality controlled)
Precipitation stations of municipalitys/citys (control of sewage system and flood managing)
Hydropower companies (no free data, but can be used for assimilation and verification of models)
Agricultural meteorological service

19. Summary Quite effective quality control for temperature data in place
Use of temperature observations reduces amount of large errors significantly and generally improves nowcasts for individual locations during winter months
Use of summer data more complicated due to large influence of radiation (no shielding and altered long wave radiation budget)
Precipitation data only usable in summer (no snow) needs more sophisticated QC (combination with radar)
Wind and humidity are severely biased, but still have high correlations
An increasing number and variety of crowdsourced data are available and advanced quality control techniques are under development

20. What do we get at what costs?
Data collection
IT (and other) Resources
Number of stations
Diversity of sensors
Metadata availa-bility
Comment
Contract with individual providers (Netatmo)
low
(only one data format to deal with; no need to motivate contributors)
high
Lowest IT costs, but highly dependent on provider, high risk for inhomogenities, individual station owners may not be aware of that they provide data
Station owners provide data (i.e. WOW)
(collection service and database to be maintained and kept attractive)
moderate
Best possibility to get metadata, lower amount of stations, but higher diversity of sensors
Citizens report weather impact (visual observations)
(App development and database to be maintained, motivation work to achieve critical amont of observations)
Possibility for standardized information (but not too complicated),
Data mining
highest
(under constant adaption, several memberships to be maintained)
partly
High amount of stations and diversity, needs lots of IT resources

21. THANK YOU !