This systematic review follows PRISMA guidelines to comprehensively examine applications of Large Language Models (LLMs) and Natural Language Processing (NLP) techniques in hydrology from 2020-2025. The review systematically synthesizes evidence across five key application domains: flood forecasting, drought monitoring, emergency response, policy support, and hydrologic modeling. Through systematic search of multiple databases (Web of Science, Scopus, Google Scholar, arXiv) and rigorous screening processes, this review identifies advances, challenges, and open problems in hydrology-AI integration. Expected outcomes include development of best-practice recommendations, identification of reproducibility gaps, and establishment of a research agenda for physics-informed language models in water resources management.