Abstract:Due to recent climate changes, we have seen more frequent and severe wildfires in the United States. Predicting wildfires is critical for natural disaster prevention and mitigation. Advances in technologies in data processing and communication enabled us to access remote sensing data. With the remote sensing data, valuable spatiotemporal statistical models can be created and used for resource management practices. This paper proposes a distributed learning framework that shares local data collected in ten locations in the western USA throughout the local agents. The local agents aim to predict wildfire grid maps one, two, three, and four weeks in advance while online processing the remote sensing data stream. The proposed model has distinct features that address the characteristic need in prediction evaluations, including dynamic online estimation and time-series modeling. Local fire event triggers are not isolated between locations, and there are confounding factors when local data is analyzed due to incomplete state observations. Compared to existing approaches that do not account for incomplete state observation within wildfire time-series data, on average, we can achieve higher prediction performance.

What problem does this paper attempt to address?

### Problems the paper attempts to solve This paper aims to solve the problem of increasingly frequent and severe wildfire prediction in the western United States. Specifically, the article proposes a distributed learning framework to achieve multi - time - point predictions of wildfire grid maps in the next 1 week, 2 weeks, 3 weeks, and 4 weeks by sharing local remote - sensing data from 10 different locations. #### Problem background In recent years, due to the impact of climate change, the frequency and severity of wildfires in the United States have increased significantly. Accurate prediction of wildfires is crucial for natural disaster prevention and mitigation. With the progress of data processing and communication technologies, researchers can use remote - sensing data to create valuable spatio - temporal statistical models for resource management practices. However, existing methods have limitations in dealing with incomplete state observations, which leads to a decline in prediction performance. #### Main challenges 1. **Incomplete state observations**: Sensors can only indirectly observe part of the environmental state, rather than the complete state of the entire system. 2. **Spatio - temporal correlations**: Wildfire events have temporal and spatial correlations, and existing methods are difficult to effectively handle such complex correlations. 3. **Data sharing and optimization**: How to improve prediction performance through data sharing while maintaining communication resource constraints is also an important challenge. #### Solutions To solve the above problems, this paper proposes a distributed learning framework that combines system identification and a multi - agent data - sharing mechanism. The main contributions include: 1. **Introducing system identification tasks**: To mitigate the impact of incomplete state observations, the author adds system identification as an auxiliary task to the prediction task to better estimate the observation - generation model. 2. **Data - sharing mechanism**: Improve prediction performance by sharing training data among multiple local agents while keeping the use of communication resources within constraints. 3. **Dynamic computing network**: Use a dynamic computing network including convolutional layers, gated recurrent units (GRU), and deconvolutional layers for real - time multi - step prediction. 4. **Reinforcement learning optimization**: Use reinforcement learning to optimize the data - sample - exchange strategy to further improve the generalization ability and prediction accuracy of the model. Through these methods, this paper has achieved better prediction performance in dealing with incomplete state observations and spatio - temporal correlations in wildfire prediction.

Multi-time Predictions of Wildfire Grid Map using Remote Sensing Local Data

Next Day Wildfire Spread: A Machine Learning Data Set to Predict Wildfire Spreading from Remote-Sensing Data

Super Real-Time Forecast of Wildland Fire Spread by A Dual-Model Deep Learning Method

Wildfire Detection Through Deep Learning Based on Himawari-8 Satellites Platform

Wildfire spreading prediction using multimodal data and deep neural network approach

Enhancing Wildfire Forecasting Through Multisource Spatio-Temporal Data, Deep Learning, Ensemble Models and Transfer Learning

A Multi-Modal Wildfire Prediction and Personalized Early-Warning System Based on a Novel Machine Learning Framework

Estimating fire Duration using regression methods

Wildfire Risk Prediction: A Review

TeleViT: Teleconnection-driven Transformers Improve Subseasonal to Seasonal Wildfire Forecasting

Using Echo State Networks to Inform Physical Models for Fire Front Propagation

Spatio-Temporal Knowledge Graph Based Forest Fire Prediction with Multi Source Heterogeneous Data

Seasonal Fire Prediction using Spatio-Temporal Deep Neural Networks

An interpretable wildfire spreading model for real-time predictions

A Deep Learning Framework: Predicting Fire Radiative Power From the Combination of Polar-Orbiting and Geostationary Satellite Data During Wildfire Spread

Wildfire Prediction to Inform Fire Management: Statistical Science Challenges

Wildfire Impact Analysis and Spread Dynamics Estimation on Satellite Images Using Deep Learning

Deep Learning Approaches for Wildland Fires Using Satellite Remote Sensing Data: Detection, Mapping, and Prediction

Analyzing Multispectral Satellite Imagery of South American Wildfires Using Deep Learning

Dynamic prediction of global monthly burned area with hybrid deep neural networks