Abstract:

[Background] Cryosphere research faces core challenges such as inconsistent spatiotemporal references and weak semantic correlations in multi-source heterogeneous data, which severely hinder the in-depth mining of data value. [Objective] This study proposes a spatiotemporal cell-based framework for multimodal data fusion and intelligent management, aiming to enhance the operability and knowledge discovery efficiency of massive scientific data through standardized spatiotemporal unit construction and semantic enhancement technologies. [Methods] First, based on unified spatiotemporal references, observation areas are divided into regularized spatiotemporal cells with preset resolutions, ensuring spatiotemporal consistency of meteorological, remote sensing, and other multimodal data through standardized processing. Second, a cell feature calculation model is designed to mine data variation patterns and interaction relationships within and across cells using statistical algorithms and association analysis. Furthermore, large language models and semantic embedding technologies are integrated to establish a “precise condition + semantic reasoning” dual-mode retrieval mechanism, supporting structured queries and natural language-driven deep correlation mining. [Conclusions] The framework effectively addresses multi-source data fusion and weak semantic correlation issues, significantly improving the intelligent management level of cryosphere scientific data. Researchers can dynamically explore data evolution patterns through interactive interfaces, supporting scientific discovery and decision-making, thereby providing efficient tool support for climate system research and sustainable development.

Key words: spatio-temporal cells, data fusion, feature correlation analysis, semantic reasoning