A Systematic Review of TinyML-Enabled IoT Architectures for Water Pipeline Leak Detection

Water loss through pipeline leakage is a critical challenge in both urban and rural regions, particularly in developing countries such as Tanzania. The degradation of infrastructure, lack of monitoring systems, and limited access to real-time data contribute significantly to this persistent issue. Recent technological advancements in Tiny Machine Learning (TinyML) and Internet of Things (IoT) have opened promising pathways for addressing this problem through real-time, low-power, and embedded leak detection mechanisms. This systematic literature review aims to explore the current state of TinyML-enabled IoT systems specifically designed for water pipeline leak detection, evaluating their architecture, sensor technologies, machine learning models, and real-world deployment strategies. Utilizing the PICO framework and tools such as Zotero and Excel. Following PRISMA 2020 guidelines, we conducted a structured search and screening process across major databases including Google Scholar (89 articles),  ScienceDirect (75), and PubMed (2), focusing on literature published between 2021 and 2025. The findings reveal major trends in machine learning methodologies, geographic gaps in deployments (notably in Africa), and the lack of energy-autonomous solutions suitable for underground application. We conclude with targeted recommendations for future research that aligns with real-world deployment in Tanzanian contexts.