Naiad: Novel Agentic Intelligent Autonomous System for Inland Water Monitoring#

来源: https://arxiv.org/abs/2601.05256

摘要: NAIAD是一个基于大语言模型(LLM)的智能水质监测系统,为内陆水域监测提供全面解决方案。该系统利用LLM和外部分析工具,通过地球观测(EO)数据进行水质分析。系统的核心特点包括:单一提示界面,可将自然语言查询转化为可操作的见解;基于检索增强生成(RAG)的知识获取;LLM推理能力;外部工具编排;计算图执行;以及智能反馈机制。NAIAD集成了多种工具,包括天气数据分析、Sentinel-2卫星图像处理、遥感指数计算(如NDCI)、叶绿素a估算等。系统性能评估显示在专门的基准测试中,正确性和相关性分别达到77%和85%以上。系统采用Gemma 3(27B)和Qwen 2.5(14B)作为LLM

关键词: 大语言模型, 水质监测, 检索增强生成, 地球观测, 智能系统

Naiad: Novel Agentic Intelligent Autonomous System for Inland Water Monitoring#

NAIAD is an intelligent water quality monitoring system based on Large Language Models (LLMs), providing a comprehensive solution for inland water monitoring. The system leverages LLMs and external analytical tools to perform water quality analysis through Earth Observation (EO) data. Core features of the system include: a single-prompt interface that transforms natural language queries into actionable insights; knowledge acquisition based on Retrieval-Augmented Generation (RAG); LLM reasoning capabilities; external tool orchestration; computational graph execution; and intelligent feedback mechanisms. NAIAD integrates various tools, including weather data analysis, Sentinel-2 satellite image processing, remote sensing index calculations (such as NDCI), and chlorophyll-a estimation. System performance evaluation shows accuracy and relevance rates of over 77% and 85% respectively in specialized benchmark tests. The system employs Gemma 3 (27B) and Qwen 2.5 (14B) as its LLMs.