Prediction of tropospheric parameters

We provide predictions of zenith wet delay (ZWD) based on our newly developed ML model, called ZWDX. ZWDX is the successor of the model presented in the publication "Global, spatially explicit modelling of zenith wet delay with XGBoost" by Crocetti et al. (2024). ZWDX is, as its predecessor, based on the XGBoost algorithm but optimized for spatial and temporal ZWD predictions. The details of the methodology of ZWDX will be made available as soon as our paper is published. In the meantime, please cite https://doi.org/10.1007/s00190-024-01829-2 if you use our data.

We provide a ZWD API, the ML models, and precalculated ZWD predicitons on a 0.25 degree latitude, longitude grid. The data may only be used for research purposes. Commercial use is not permitted.

If you have questions or encounter any problems, contact the main developer of the model:
Laura Crocetti (lcrocetti@ethz.ch)

ZWD API

The ZWD API is hosted at https://test-zwd.space01.phys.ethz.ch/. Check out this site for further information on how to obtain ZWD predictions using various programming languages.

ML models

The trained ML models can be found under products/Troposphere/ZWD/model.

Precalculated ZWD prediction grids

Prediction grids of ZWD are available under Products.
The gridded ZWD predictions based on ERA5 are provided in the era5 folder.
The gridded ZWD predictions based on forecasts are provided in the forecast folder.
More information about the upload strategy can be found below.

Latest processing covered: 2024-11-17 12:00:00 till 2024-11-21 06:00:00
Predictions based on ERA5 data: 2023-08-27 00:00:00 till 2024-11-11 23:00:00
Predictions based on forecast data: 2024-11-12 00:00:00 till 2024-11-21 06:00:00

Upload strategy

The upload strategy is illustrated below

  • In red, the ERA5 availability is indicated
  • Other colors represent different forecast data
  • New forecast become available twice per day, at 00:00 and 12:00 UTC
  • As soon as a new forecast is available, it gets processed
  • However, downloading and processing might take few hours (approximately 10-12 hours)
  • ZWD predictions are always based on the newest available forecast data
  • The ERA5 data is processed in daily batches, as soon as the data becomes available (latency is around 7 days)
  • In the figure below, cells that are crossed represent newly generated ZWD predictions



Example:

  • As soon as the forecast data for 08.01. 00:00 gets available, ZWD predictions for 08.01. 00:00 and the following 90 hours are generated. They are stored at the forecast folder, replacing previous predictions.
  • As soon as the forecast data for 08.01. 12:00 gets available, ZWD predictions for 08.01. 12:00 and the following 90 hours are generated. They are stored at the forecast folder, replacing previous predictions.
  • As soon as the ERA5 data for 03.01 gets available, ZWD predictions for 03.01. 00:00 - 23:00 are generated. They are stored at the era5 folder. The forecast predicitons covering the same period of time are removed from the forecast folder.

Reference

Publications

[1] Crocetti, L., Schartner, M., Zus, F., Zhang, W., Moeller, G., Navarro, V., See, L., Schindler, K., & Soja, B. "Global, spatially explicit modelling of zenith wet delay with XGBoost." JGeod 98, 23 (2024). https://doi.org/10.1007/s00190-024-01829-2
[2] Crocetti, L., Schartner, M., Wareyka-Glaner, M.F., Schindler, K., & Soja, B. "ZWDX: A Global Zenith Wet Delay Forecasting Model using XGBoost." Submitted to Earth, Planets (2024, under review).

Presentations

[1] Crocetti, L., Schartner, M., Schindler, K., & Soja, B. "Forecasting of tropospheric parameters using meteorological data and machine learning." EGU General Assembly 2023, Vienna, Austria, 2328 Apr 2023, EGU23-3453, 2023. https://doi.org/10.5194/egusphere-egu23-3453
[2] Crocetti, L., Soja, B., Kłopotek, G., Awadaljeed, M., Rothacher, M., See, L., Weinacker, R., Sturn, T., McCallum, I., and Navarro, V.: Machine learning and meteorological data for spatio-temporal prediction of tropospheric parameters, EGU General Assembly 2022, Vienna, Austria, 23–27 May 2022, EGU22-4531, 2022. https://doi.org/10.5194/egusphere-egu22-4531