Transfer Learning for Dead Fuel Moisture Prediction Using Time-Warping Recurrent Neural Networks

2026-05-08

AI Summary

This study addresses the challenge of sparse observational data for fuel moisture content across different time scales (e.g., 1-hour, 100-hour, and 1000-hour fuels), which hinders direct modeling and prediction. To overcome this limitation, the authors propose a transfer learning approach based on a time-warping mechanism that adapts an LSTM dynamics model pretrained on 10-hour fuel data to other time scales. This method enables cross-fuel time-scale alignment and facilitates the transfer of dynamic features, thereby enhancing prediction accuracy for fuel moisture content in data-scarce categories. The effectiveness and superiority of the proposed approach are validated using field-collected datasets from Oklahoma, demonstrating significant improvements in predictive performance under sparse data conditions.

Abstract

This paper proposes a time-warping transfer learning method, a technique for temporally rescaling the learned dynamics of a recurrent neural network (RNN) with a Long Short-Term Memory (LSTM) layer to enable task transfer across fuel moisture classes. Fuel moisture content (FMC) is divided into idealized classes based on characteristic lag time. Large quantities of real-time data are available for 10h fuels from sensors on weather stations, but observations of other fuel classes are sparse in space and time. We use transfer learning to adapt an RNN pretrained on 10h FMC to predict FMC for 1h, 100h, and 1000h fuels. We validate this method using data from a landmark field study conducted in Oklahoma that was used to calibrate the state-of-the-art Nelson fuel moisture model.

Problem

• dead fuel moisture
• transfer learning
• fuel moisture prediction
• temporal rescaling
• sparse observations

Innovation

• time-warping
• transfer learning
• LSTM
• fuel moisture prediction
• recurrent neural networks