Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Localized Weather Prediction Using Kolmogorov-Arnold Network-Based Models and Deep RNNs

Ange-Clement Akazan, Verlon Roel Mbingui, Gnankan Landry Regis N'guessan, Issa Karambal

TL;DR

This study benchmarks deep recurrent neural networks such as LSTM, GRU, BiLSTM, BiLSTM, BiGRU, and Kolmogorov-Arnold-based models for daily forecasting of temperature, precipitation, and pressure in two tropical cities and introduces two customized variants of $ \texttt{TKAN}$ that replace its original $\texttt{SiLU}$ activation function with $ GeLU and MiSH, respectively.

Abstract

Weather forecasting is crucial for managing risks and economic planning, particularly in tropical Africa, where extreme events severely impact livelihoods. Yet, existing forecasting methods often struggle with the region's complex, non-linear weather patterns. This study benchmarks deep recurrent neural networks such as $\texttt{LSTM, GRU, BiLSTM, BiGRU}$, and Kolmogorov-Arnold-based models $(\texttt{KAN} and \texttt{TKAN})$ for daily forecasting of temperature, precipitation, and pressure in two tropical cities: Abidjan, Cote d'Ivoire (Ivory Coast) and Kigali (Rwanda). We further introduce two customized variants of $ \texttt{TKAN}$ that replace its original $\texttt{SiLU}$ activation function with $ \texttt{GeLU}$ and \texttt{MiSH}, respectively. Using station-level meteorological data spanning from 2010 to 2024, we evaluate all the models on standard regression metrics. $\texttt{KAN}$ achieves temperature prediction ($R^2=0.9986$ in Abidjan, $0.9998$ in Kigali, $\texttt{MSE} < 0.0014~^\circ C ^2$), while $\texttt{TKAN}$ variants minimize absolute errors for precipitation forecasting in low-rainfall regimes. The customized $\texttt{TKAN}$ models demonstrate improvements over the standard $\texttt{TKAN}$ across both datasets. Classical \texttt{RNNs} remain highly competitive for atmospheric pressure ($R^2 \approx 0.83{-}0.86$), outperforming $\texttt{KAN}$-based models in this task. These results highlight the potential of spline-based neural architectures for efficient and data-efficient forecasting.

Localized Weather Prediction Using Kolmogorov-Arnold Network-Based Models and Deep RNNs

TL;DR

This study benchmarks deep recurrent neural networks such as LSTM, GRU, BiLSTM, BiLSTM, BiGRU, and Kolmogorov-Arnold-based models for daily forecasting of temperature, precipitation, and pressure in two tropical cities and introduces two customized variants of that replace its original activation function with $ GeLU and MiSH, respectively.

Abstract

Weather forecasting is crucial for managing risks and economic planning, particularly in tropical Africa, where extreme events severely impact livelihoods. Yet, existing forecasting methods often struggle with the region's complex, non-linear weather patterns. This study benchmarks deep recurrent neural networks such as , and Kolmogorov-Arnold-based models for daily forecasting of temperature, precipitation, and pressure in two tropical cities: Abidjan, Cote d'Ivoire (Ivory Coast) and Kigali (Rwanda). We further introduce two customized variants of that replace its original activation function with and \texttt{MiSH}, respectively. Using station-level meteorological data spanning from 2010 to 2024, we evaluate all the models on standard regression metrics. achieves temperature prediction ( in Abidjan, in Kigali, ), while variants minimize absolute errors for precipitation forecasting in low-rainfall regimes. The customized models demonstrate improvements over the standard across both datasets. Classical \texttt{RNNs} remain highly competitive for atmospheric pressure (), outperforming -based models in this task. These results highlight the potential of spline-based neural architectures for efficient and data-efficient forecasting.

Paper Structure

This paper contains 17 sections, 13 equations, 12 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Materials and Methods
  4. Study Area and Data
  5. Study Area
  6. Data
  7. Data Information.
  8. Data Set Manipulation.
  9. Methods
  10. Long Short Term Memory (LSTM)
  11. Gated recurrent units (GRU)
  12. Bidirectional Recurrent Neural Network-Based Models
  13. Ensemble Model
  14. Kolmogorov Arnold Network (KAN)
  15. Temporal Kolmogorov-Arnold Networks
  16. ...and 2 more sections

Figures (12)

  • Figure 1: Kigali City
  • Figure 2: Abidjan City
  • Figure 3: Target Variables Distribution: Row 1 is for Kigali city and row 2 for Abidjan city
  • Figure 4: Comparison of predicted vs. actual temperature values in Abidjan using the best-performing KAN-based model and the top-performing deep RNN model
  • Figure 5: Comparison of predicted vs. actual temperature values in Kigali using the best-performing KAN-based model and the top-performing deep RNN model
  • ...and 7 more figures