Title : Spatio-temporal water storage solutions for climate whiplash: Downscaling, AI-enhanced detection, and cloud-orchestrated management for flood prevention and water security
Abstract:
Climate whiplash is becoming a serious challenge for water managers because extreme rainfall and drought now occur in faster and more disruptive cycles. This paper presents a practical framework for responding to that problem through precipitation forecasting, artificial intelligence, and smart water infrastructure. Using Merced County, California as the main test case, the study examines how future climate conditions may increase flood risk and how excess stormwater can be captured and stored for later use during dry periods.
The research begins with a historical analysis of precipitation from 2000 to 2024 and compares it with downscaled climate projections for 2026 to 2050 and 2051 to 2099. The results show a strong rise in extreme precipitation events, with flood risk increasing by about 150 percent in the near future and about 250 percent in the long term compared with the historical baseline. These projections suggest that current drainage systems and flood control infrastructure may not be enough to manage future storm intensity and frequency.
To move from forecasting to action, the study uses a regional convolutional neural network (RCNN) to scan satellite images and identify land areas that could be used for water storage. The model classifies terrain into useful categories such as lakes, reservoir areas, farms, and suburbs, and then helps locate sites that are suitable for capture and retention of floodwater. Candidate locations were checked against elevation data and terrain maps to confirm whether water could be safely stored in those areas. This process reduced the time needed to assess potential storage sites and created a faster way to support climate adaptation planning.
Based on the spatial analysis, the paper introduces the concept of Engineered Wetlands. These are hybrid water storage systems that combine natural landscape features with small dams, barriers, and flow control structures. In the Merced County case study, one proposed engineered wetland could store up to 2.5 million cubic meters of surface water. The design is intended to capture a large share of excess water during intense storms while also supporting water availability during dry months. In this way, the same system can help reduce flood damage and improve water security.
The study also proposes an Internet of Things and cloud-based monitoring system for real time water management. Sensors can track water levels and flow conditions, while cloud platforms can support remote control, data analysis, and automated response. This integrated approach offers a scalable and timely solution for regions facing climate whiplash. The paper shows that AI driven environmental intelligence can support both prediction and adaptation, providing a human centered and practical path toward more resilient water management.
