Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Personalized Federated Learning for improving radar based precipitation nowcasting on heterogeneous areas

Judith Sáinz-Pardo Díaz, María Castrillo, Juraj Bartok, Ignacio Heredia Cachá, Irina Malkin Ondík, Ivan Martynovskyi, Khadijeh Alibabaei, Lisana Berberi, Valentin Kozlov, Álvaro López García

TL;DR

The paper addresses privacy and data-sharing constraints in radar-based precipitation nowcasting by applying a novel adaptive personalized federated learning (adapFL) framework to distributed VIL radar images. The data are split into four zones to simulate independent data owners, with adapFL combining a global FL phase and subsequent local adaptation to each zone. Results show adapFL consistently outperforms the COTREC baseline and both independent and standard FL models across zones, and central-area generalization is enhanced through adaptation. This work demonstrates the feasibility and value of privacy-preserving, heterogeneity-aware learning for meteorological nowcasting, with clear implications for multi-radar deployments and cross-institution collaboration.

Abstract

The increasing generation of data in different areas of life, such as the environment, highlights the need to explore new techniques for processing and exploiting data for useful purposes. In this context, artificial intelligence techniques, especially through deep learning models, are key tools to be used on the large amount of data that can be obtained, for example, from weather radars. In many cases, the information collected by these radars is not open, or belongs to different institutions, thus needing to deal with the distributed nature of this data. In this work, the applicability of a personalized federated learning architecture, which has been called adapFL, on distributed weather radar images is addressed. To this end, given a single available radar covering 400 km in diameter, the captured images are divided in such a way that they are disjointly distributed into four different federated clients. The results obtained with adapFL are analyzed in each zone, as well as in a central area covering part of the surface of each of the previously distributed areas. The ultimate goal of this work is to study the generalization capability of this type of learning technique for its extrapolation to use cases in which a representative number of radars is available, whose data can not be centralized due to technical, legal or administrative concerns. The results of this preliminary study indicate that the performance obtained in each zone with the adapFL approach allows improving the results of the federated learning approach, the individual deep learning models and the classical Continuity Tracking Radar Echoes by Correlation approach.

Personalized Federated Learning for improving radar based precipitation nowcasting on heterogeneous areas

TL;DR

The paper addresses privacy and data-sharing constraints in radar-based precipitation nowcasting by applying a novel adaptive personalized federated learning (adapFL) framework to distributed VIL radar images. The data are split into four zones to simulate independent data owners, with adapFL combining a global FL phase and subsequent local adaptation to each zone. Results show adapFL consistently outperforms the COTREC baseline and both independent and standard FL models across zones, and central-area generalization is enhanced through adaptation. This work demonstrates the feasibility and value of privacy-preserving, heterogeneity-aware learning for meteorological nowcasting, with clear implications for multi-radar deployments and cross-institution collaboration.

Abstract

The increasing generation of data in different areas of life, such as the environment, highlights the need to explore new techniques for processing and exploiting data for useful purposes. In this context, artificial intelligence techniques, especially through deep learning models, are key tools to be used on the large amount of data that can be obtained, for example, from weather radars. In many cases, the information collected by these radars is not open, or belongs to different institutions, thus needing to deal with the distributed nature of this data. In this work, the applicability of a personalized federated learning architecture, which has been called adapFL, on distributed weather radar images is addressed. To this end, given a single available radar covering 400 km in diameter, the captured images are divided in such a way that they are disjointly distributed into four different federated clients. The results obtained with adapFL are analyzed in each zone, as well as in a central area covering part of the surface of each of the previously distributed areas. The ultimate goal of this work is to study the generalization capability of this type of learning technique for its extrapolation to use cases in which a representative number of radars is available, whose data can not be centralized due to technical, legal or administrative concerns. The results of this preliminary study indicate that the performance obtained in each zone with the adapFL approach allows improving the results of the federated learning approach, the individual deep learning models and the classical Continuity Tracking Radar Echoes by Correlation approach.
Paper Structure (19 sections, 3 equations, 16 figures, 8 tables, 1 algorithm)

This paper contains 19 sections, 3 equations, 16 figures, 8 tables, 1 algorithm.

Table of Contents

  1. Introduction and motivation
  2. State of the art
  3. Deep learning techniques for meteorology
  4. Federated Learning and Personalized Federated Learning
  5. Data
  6. Methodology
  7. Classical models
  8. Deep Learning and Federated Learning models
  9. Convolutional neural architecture
  10. Computing resources
  11. Results and analysis
  12. Central area
  13. Discussion
  14. Conclusions
  15. Data distribution
  16. ...and 4 more sections

Figures (16)

  • Figure 1: Example of the radar images under study after reducing to 100$\times$100 resolution with the information on vertically integrated liquid.
  • Figure 2: Example of an image where the four quadrants into which it has been divided to create the four zones are shown in blue.
  • Figure 3: Average of all the radar images available in April, May, June and July 2016 by zone once processed by area.
  • Figure 4: Schema of the three learning paradigms implemented: individual, federated and adaptive federated learning.
  • Figure 5: Example of an image where the four quadrants into which it has been divided to create the four zones are shown in blue and the central area is marked in red.
  • ...and 11 more figures