Advances in Time Series Prediction and Earth Sciences

The field of time series prediction and earth sciences is rapidly evolving, with a focus on developing innovative models and frameworks that can accurately predict complex phenomena. Recent research has emphasized the importance of integrating knowledge from multiple disciplines, such as atmospheric science, computer science, and agricultural science, to improve the accuracy and reliability of predictions. One notable trend is the increasing use of foundation models, which have shown promising results in predicting crop yields, soil greenhouse gas fluxes, and other environmental variables. These models have the potential to simplify the prediction process and reduce the need for extensive feature engineering. Another area of research is the development of hierarchical architectures for time series forecasting, which can capture patterns and relationships at multiple temporal scales. These models have been shown to outperform traditional methods and have significant implications for decision-making in various applications. Noteworthy papers in this area include those that propose novel frameworks for integrating time series data with natural language, such as the ITFormer model, and those that develop new benchmarks for evaluating the performance of machine learning models on tabular data, such as the TabArena system. Overall, the field is moving towards more integrated and interdisciplinary approaches, with a focus on developing models and frameworks that can capture complex relationships and patterns in time series data.

Sources

KG-FGNN: Knowledge-guided GNN Foundation Model for Fertilisation-oriented Soil GHG Flux Prediction

AutoHFormer: Efficient Hierarchical Autoregressive Transformer for Time Series Prediction

Artificial Intelligence for Atmospheric Sciences: A Research Roadmap

TabArena: A Living Benchmark for Machine Learning on Tabular Data

AI-based Approach in Early Warning Systems: Focus on Emergency Communication Ecosystem and Citizen Participation in Nordic Countries

From Rows to Yields: How Foundation Models for Tabular Data Simplify Crop Yield Prediction

Finetuning a Weather Foundation Model with Lightweight Decoders for Unseen Physical Processes

Hierarchical Time Series Forecasting Via Latent Mean Encoding

ITFormer: Bridging Time Series and Natural Language for Multi-Modal QA with Large-Scale Multitask Dataset

SEED: A Structural Encoder for Embedding-Driven Decoding in Time Series Prediction with LLMs

Built with on top of