Register now After registration you will be able to apply for this opportunity online.
Detection of Atmospheric Rivers using GNSS Zenith Wet Delay
This project aims to develop a machine learning framework for detecting landfalling atmospheric rivers by using GNSS-derived zenith wet delays. The goal is to explore the feasibility and effectiveness of GNSS data for atmospheric river detection, potentially providing an additional, valuable tool for monitoring and prediction.
Atmospheric rivers are long, narrow bands in the atmosphere composed of condensed water vapour. While many atmospheric rivers are relatively weak systems, stronger atmospheric rivers can transport significant amounts of moisture, which - often combined with higher wind speeds - can lead to extreme rainfall or snowfall. These strong atmospheric rivers can cause catastrophic damage through flooding and landslides. The Western United States in particular is frequently affected by strong landfalling atmospheric rivers (Figure 1).
Given their significance, atmospheric rivers are heavily investigated. Most detection techniques rely on satellite retrievals of atmospheric moisture, such as integrated water vapour (IWV) and/or integrated water vapour transport (IVT), derived from reanalysis datasets or weather models.
GNSS zenith wet delays (ZWDs) offer a complementary approach to detecting atmospheric rivers by providing high-resolution observations of atmospheric water vapour. ZWD is part of the delay experienced by GNSS signals as they pass through the atmosphere. ZWD can also be transformed to IWV. Despite its potential, few studies have investigated the use of GNSS ZWD data for atmospheric river detection.
Atmospheric rivers are long, narrow bands in the atmosphere composed of condensed water vapour. While many atmospheric rivers are relatively weak systems, stronger atmospheric rivers can transport significant amounts of moisture, which - often combined with higher wind speeds - can lead to extreme rainfall or snowfall. These strong atmospheric rivers can cause catastrophic damage through flooding and landslides. The Western United States in particular is frequently affected by strong landfalling atmospheric rivers (Figure 1). Given their significance, atmospheric rivers are heavily investigated. Most detection techniques rely on satellite retrievals of atmospheric moisture, such as integrated water vapour (IWV) and/or integrated water vapour transport (IVT), derived from reanalysis datasets or weather models. GNSS zenith wet delays (ZWDs) offer a complementary approach to detecting atmospheric rivers by providing high-resolution observations of atmospheric water vapour. ZWD is part of the delay experienced by GNSS signals as they pass through the atmosphere. ZWD can also be transformed to IWV. Despite its potential, few studies have investigated the use of GNSS ZWD data for atmospheric river detection.
The goal of this work is to assess the feasibility of detecting landfalling atmospheric rivers using GNSS-derived ZWDs. The project will focus on the development and implementation of a machine learning framework capable of detecting atmospheric river events based on ZWD data.
The study area will be western North America, a region frequently affected by these phenomena. Therefore, the catalog of landfalling atmospheric rivers along the western coast of North America from https://weclima.ucsd.edu/data-products/ will be used to train and validate the detection model. The ZWDs are globally available based on the ZWDX model (Crocetti et al. 2024). Additionally, data from the ERA5 reanalysis might be used to incorporate meteorological variables, such as wind speed. The student will set up a machine learning framework to detect atmospheric rivers.
The goal of this work is to assess the feasibility of detecting landfalling atmospheric rivers using GNSS-derived ZWDs. The project will focus on the development and implementation of a machine learning framework capable of detecting atmospheric river events based on ZWD data. The study area will be western North America, a region frequently affected by these phenomena. Therefore, the catalog of landfalling atmospheric rivers along the western coast of North America from https://weclima.ucsd.edu/data-products/ will be used to train and validate the detection model. The ZWDs are globally available based on the ZWDX model (Crocetti et al. 2024). Additionally, data from the ERA5 reanalysis might be used to incorporate meteorological variables, such as wind speed. The student will set up a machine learning framework to detect atmospheric rivers.
Laura Crocetti (lcrocetti@ethz.ch)
Prof. Dr. Benedikt Soja (soja@ethz.ch)
Laura Crocetti (lcrocetti@ethz.ch) Prof. Dr. Benedikt Soja (soja@ethz.ch)