Core Insights - The article discusses the development of GeoEvolve, a framework that integrates geographic knowledge into AI to enhance geospatial modeling, allowing AI to autonomously improve algorithms rather than merely assisting researchers [2][4][19]. Research Background - Geospatial modeling is crucial for understanding climate change and promoting sustainable urban development, traditionally relying on expert experience for hypothesis formulation and algorithm design [4]. - Recent advancements in large language models (LLMs) show potential for automated code evolution, but these systems lack an understanding of geography, which can lead to ineffective models [4][11]. GeoEvolve Framework - GeoEvolve is conceptualized as a research team comprising an AI (acting as a PhD student) and a geographic knowledge base (acting as a mentor), ensuring that the evolution of algorithms aligns with spatial theories [5]. - The framework consists of four core modules: 1. Code Evolver (automatically generates and mutates candidate algorithms) 2. Code Analyzer (diagnoses issues and suggests improvements) 3. Geographic Knowledge Retriever (GeoKnowRAG, provides spatial theory and classic methods) 4. Knowledge-Driven Prompt Generator (translates complex geographic knowledge into AI-understandable optimization instructions) [5][8]. Case Study: Automation of Kriging Improvement - Ordinary Kriging, a classic spatial interpolation method, has seen limited structural improvements over time, primarily relying on external combinations with regression models [13]. - GeoEvolve introduces several enhancements to the Kriging model, including: - Adaptive empirical variogram estimation to reduce the impact of outliers [14]. - Multi-start global fitting to avoid local optima [15]. - Adaptive data transformation for better residual distribution [16]. - Experimental results show that GeoEvolve significantly outperforms traditional and other automated models, achieving lower RMSE and MAE across various metal predictions [18]. Conclusion - GeoEvolve demonstrates that AI can autonomously evolve stronger classic models under the guidance of geographic knowledge, suggesting a shift towards fully automated algorithm development in geospatial modeling [19]. - This advancement opens new possibilities for AI applications in geographic science and sustainable development, positioning AI as a collaborative research partner rather than just a tool [20].