Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

KAN-FIF: Spline-Parameterized Lightweight Physics-based Tropical Cyclone Estimation on Meteorological Satellite

Jiakang Shen, Qinghui Chen, Runtong Wang, Chenrui Xu, Jinglin Zhang, Cong Bai, Feng Zhang

TL;DR

KAN-FIF introduces a lightweight, spline-parameterized Kolmogorov-Arnold Network framework for multitask tropical cyclone estimation on resource-limited edge devices. By replacing standard CNN/MLP components with KAN layers, incorporating physics-guided constraints, and fusing temporal and infrared modalities, the approach achieves substantial parameter reductions (~94.8%) and faster inference while improving MSW and RMW accuracy. The method demonstrates robust edge deployment feasibility, achieving $14.41$ ms per-sample latency on an Ascend 310 NPUs and maintaining competitive performance under model compression. This work enables real-time, onboard TC monitoring with potential for broad operational impact in satellite and edge-vision contexts.

Abstract

Tropical cyclones (TC) are among the most destructive natural disasters, causing catastrophic damage to coastal regions through extreme winds, heavy rainfall, and storm surges. Timely monitoring of tropical cyclones is crucial for reducing loss of life and property, yet it is hindered by the computational inefficiency and high parameter counts of existing methods on resource-constrained edge devices. Current physics-guided models suffer from linear feature interactions that fail to capture high-order polynomial relationships between TC attributes, leading to inflated model sizes and hardware incompatibility. To overcome these challenges, this study introduces the Kolmogorov-Arnold Network-based Feature Interaction Framework (KAN-FIF), a lightweight multimodal architecture that integrates MLP and CNN layers with spline-parameterized KAN layers. For Maximum Sustained Wind (MSW) prediction, experiments demonstrate that the KAN-FIF framework achieves a $94.8\%$ reduction in parameters (0.99MB vs 19MB) and $68.7\%$ faster inference per sample (2.3ms vs 7.35ms) compared to baseline model Phy-CoCo, while maintaining superior accuracy with $32.5\%$ lower MAE. The offline deployment experiment of the FY-4 series meteorological satellite processor on the Qingyun-1000 development board achieved a 14.41ms per-sample inference latency with the KAN-FIF framework, demonstrating promising feasibility for operational TC monitoring and extending deployability to edge-device AI applications. The code is released at https://github.com/Jinglin-Zhang/KAN-FIF.

KAN-FIF: Spline-Parameterized Lightweight Physics-based Tropical Cyclone Estimation on Meteorological Satellite

TL;DR

KAN-FIF introduces a lightweight, spline-parameterized Kolmogorov-Arnold Network framework for multitask tropical cyclone estimation on resource-limited edge devices. By replacing standard CNN/MLP components with KAN layers, incorporating physics-guided constraints, and fusing temporal and infrared modalities, the approach achieves substantial parameter reductions (~94.8%) and faster inference while improving MSW and RMW accuracy. The method demonstrates robust edge deployment feasibility, achieving ms per-sample latency on an Ascend 310 NPUs and maintaining competitive performance under model compression. This work enables real-time, onboard TC monitoring with potential for broad operational impact in satellite and edge-vision contexts.

Abstract

Tropical cyclones (TC) are among the most destructive natural disasters, causing catastrophic damage to coastal regions through extreme winds, heavy rainfall, and storm surges. Timely monitoring of tropical cyclones is crucial for reducing loss of life and property, yet it is hindered by the computational inefficiency and high parameter counts of existing methods on resource-constrained edge devices. Current physics-guided models suffer from linear feature interactions that fail to capture high-order polynomial relationships between TC attributes, leading to inflated model sizes and hardware incompatibility. To overcome these challenges, this study introduces the Kolmogorov-Arnold Network-based Feature Interaction Framework (KAN-FIF), a lightweight multimodal architecture that integrates MLP and CNN layers with spline-parameterized KAN layers. For Maximum Sustained Wind (MSW) prediction, experiments demonstrate that the KAN-FIF framework achieves a reduction in parameters (0.99MB vs 19MB) and faster inference per sample (2.3ms vs 7.35ms) compared to baseline model Phy-CoCo, while maintaining superior accuracy with lower MAE. The offline deployment experiment of the FY-4 series meteorological satellite processor on the Qingyun-1000 development board achieved a 14.41ms per-sample inference latency with the KAN-FIF framework, demonstrating promising feasibility for operational TC monitoring and extending deployability to edge-device AI applications. The code is released at https://github.com/Jinglin-Zhang/KAN-FIF.
Paper Structure (30 sections, 11 equations, 4 figures, 7 tables, 1 algorithm)

This paper contains 30 sections, 11 equations, 4 figures, 7 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Tropical Cyclone Estimation
  4. Edge-device inference
  5. Problem Definition
  6. Methods
  7. Kolmogorov-Arnold Networks
  8. KAN-based Shared Feature Extraction
  9. Temporal Feature Extraction with KAN-LSTM
  10. Spatial Feature Extraction with KAN-CNN
  11. Final Shared Representation
  12. KAN-based Task-Specific Feature Extraction
  13. Physics-Guided Constraint Modeling
  14. Multimodal Feature Fusion
  15. Deployment Preparation for TC Estimation
  16. ...and 15 more sections

Figures (4)

  • Figure 1: The comparison of frameworks between the MLP-based and KAN-based models. (a) Conventional MLP-based model (Phycoco) (b) our KAN-based lightweight model
  • Figure 2: The offline deployment verification process of the FY-4 series satellite processor on the Qingyun-1000 development board. (a) Deployment verification on Ascend 310 NPU (b) Edge-device inference process of tropical cyclone estimation on FY-4 series satellite
  • Figure 3: The architecture of the KAN-FIF learning framework: (a) The Task-specific Feature Extraction Module uses KAN layer and center-aware attention to extract the task features of MSW and RMW respectively (b) The Physical Constraint Module is designed to conduct constraints among task-specific features ; (c) The KAN Fusion Decoder fuse the features from(a)(b)(d) and obtain the final output ; (d) The Shared Feature Extraction Module take the multi-channel image and the temporal sequence data as input and obtain the shared feature between TC tasks; (e) the architecture of KAN layers; (f) the architecture of Multiscale Conv
  • Figure 4: Visual comparison between KAN-FIF and Phy-CoCo models under different tropical cyclone structures