Kasper Stener Hintzs
A thesis for the degree of Doctor of Philosophy defended august 2019.
The PhD School of Science, Faculty of Science, PICE, Niels Bohr Institute, University of Copenhagen
Professor Eigil Kaas (NBI)
Dr Henrik Vedel (DMI)
On The Usage of Crowdsourced Data in Numerical Weather Prediction
Observations are a vital part of Numerical Weather Prediction (NWP) to give accurate forecasts of future weather. As the spatial resolution of NWP models increases, so does the need for more observations for use in data assimilation. Installing new professional meteorological observing equipment is costly and expensive to maintain. Crowdsourced data is a new potential data source for NWP that have emerged in recent years. Crowdsourced data is an overall term covering reports from users and data from equipment owned and operated by the public. Such data are often less accurate than traditional meteorological observations, but there are much more available. This PhD project studies the potential use of crowdsourced data in NWP. The main focus is on handheld wind measurements from smartphones, pressure observations from smartphones and on Personal Weather Stations (PWS). It is shown that handheld wind Measurements can, in some cases, be more representative than traditional wind observations and a method of estimating the surface roughness length from a handheld measurement is presented. Software for collecting Smartphone Pressure Observations (SPOs) has been developed and have in one year successfully collected more than 60 million observations from Denmark. Also, a scheme for quality control of the SPOs has been developed.
SPOs and pressure observations from PWS have been assimilated into the DMI HARMONIE NWP model using 3-Dimensional Variational (3D-Var) data assimilation. It is shown that SPOs can contribute positively to NWP, but it is also concluded that the full potential has not yet been reached. Other assimilation techniques than 3D-Var are likely to be more suitable for assimilating crowdsourced data. Observations from PWS are likely to be useful for NWP and observation based nowcasting, but a quality monitoring system needs to be developed before this can be applied. The Developments and findings in this PhD project constitute fundamental contributions to advance the work on utilising crowdsourced data as it is shown that such data is of a quality that can contribute to improving weather forecasts from NWP models.