• meteoblue MultiModel performs similarly well on dewpoint temperature as the reanalysis model ERA5.
• meteoblue MultiModel improves the forecast of dewpoint compared to raw prediction models.
• meteoblue reaches the best accuracy amongst several weather providers for dewpoint temperature.

Dewpoint temperature is an important and commonly used weather variable, especially in use cases where air temperature and relative humidity plays an important role. It is a useful weather variable not only for weather prediction e.g., fog events or cloud development, but also for city climate management and therefore, human health.
In the following, we give a summary of the meteoblue forecast performance on dewpoint temperature by a verification study and a provider comparison.

Spatial analysis on dewpoint temperature

meteoblue predicts the dewpoint temperature with the meteoblue MultiModel. For spatial analysis and comparison, the meteoblue MultiModel, historical data of the ERA5 reanalysis model and forecast data of the prediction model GFS have been verified with measurements from 2017 and 2018 of more than 10'000 meteorological stations worldwide.

Analysis shows that the accuracy of the meteoblue MultiModel with an average mean absolute error (MAE) of 1.8 K is in similar range as the ERA5 model with an MAE of 1.6 K. The spatial distribution of the MAE can be found in the following figures for all named models.

MAE [K] of the weather model GFS (left) and the reanalysis model ERA5 (right) based on more than 9'000 measurements worldwide of the year 2017.

Provider comparison

Additionally, meteoblue conducted a provider comparison of several weather forecast providers and the meteoblue dewpoint temperature forecast. The performances of these forecasts were verified with 475 METAR measurements worldwide of 2021. The hourly 12 - 35h forecast created at 12:00 UTC was stored in our database and analyzed.

meteoblue vs. raw models

The figure below shows the distribution of MBE of the meteoblue forecast and different forecast models as well as the ERA5 reanalysis data. meteoblue forecast performs better than all displayed prediction models, except for the ERA5 reanalysis model.

meteoblue vs. other weather forecast providers

The forecast of different weather forecast providers and the meteoblue forecast have been compared and verified for the period from January 2021 until 31st of November 2021. Final results are summarized in the table below. meteoblue performed the best on dewpoint temperature with an MAE of 1.43 K, while the other providers show a forecast error of more than 1.78 K.

MAE and MBE of the meteoblue forecast and other weather forecast providers validated at 475 METAR stations worldwide for the year 2021.

	Provider	MAE [K]	MBE [K]
1	meteoblue	1.43	-0.21
3	Provider_D	1.88	0.20
2	Provider_C	1.78	-0.32
	Provider_G	2.19	-1.29
	Provider_B	2.10	-1.15
	Provider_H	2.09	-0.90