Supporting weather forecast with crowdsourced GNSS data in the CAMALIOT project: Application of Machine Learning Technology for GNSS IoT data fusion

Last Updated: 15/12/2022 16:33     Created at: 06/12/2022 09:02

Final Presentation of NAVISP Project EL1-038 BIS now available

On Wednesday 30th November 2022, the CAMALIOT project was presented by ETH Zurich (Institute of Geodesy and Photogrammetry), Switzerland and the International Institute for Applied Systems Analysis, Austria.

GNSS infrastructure has been growing significantly in recent years, in the space segment as well as on ground. Meanwhile, low-cost Internet-of-things (IoT) devices, including smartphones, can facilitate the collection of raw GNSS measurements and are used by billions of people worldwide. Data from this dynamic network of GNSS-capable IoT devices have the potential for scientific exploitation with unprecedented spatio-temporal resolution, however access to IoT data is currently limited and the large-scale data processing required presents a challenge.

The CAMALIOT project developed an Android app and conducted a dedicated crowdsourcing campaign to utilise the concept of citizen science, combining smartphone-based GNSS data with observations acquired by high quality GNSS stations. The crowdsourcing campaign reached over 12k users worldwide and collected over 150 billion GPS observations!

A robust automated framework for GNSS big data processing was developed in the CAMALIOT project. The quality of the smartphone data was assessed, and anomalous data were detected using machine learning (ML). The fusion of smartphone GNSS data, Geodetic GNSS data and data from other environmental sensors was based on ML and the processed data were applied to science use-cases for prediction of weather forecasting and space weather monitoring. One use-case was the improved spatial modelling of tropospheric parameters, such as the zenith wet delay, which is important as the presence of water vapour in the troposphere may have spatial variation down to the km level and can delay GNSS signal propagation. Another use-case was the temporal prediction of the ionosphere to improve total electron content (TEC) maps which have temporal variation between seconds and days, important because the GNSS signal propagation speed varies with electron density.

The project outcomes have extended the capabilities of ESA’s GNSS Science Support Centre (GSSC), which offers GNSS data and processing services for various domains. The data collected in this campaign will be available on the GSSC portal.

This activity was supported by NAVISP Element 1, which is dedicated to European industry technology innovation in the wide PNT sector.

For the slides of the Final Presentation, please click here.