Forecasting water consumption is essential for resource management and future planning, especially in water-stressed regions, where daily water deliveries are heavily affected by accurate predictions of demand. Available forecasting methods typically require manual parameter tuning and external factor adjustment for each service area to obtain accurate forecasts. These limitations reduce the scalability of the system to a large number of service areas and require specialists who are not always accessible. This research provides an automated forecasting system based on the Prophet algorithm. The proposed system does not require manual parameter tuning or feature engineering. Parameters are tuned for each service area based on its specific characteristics, and the most influential external factors are automatically selected from a comprehensive set of potential factors for each area. We evaluated the system across 29 diverse service areas in Nablus, Palestine, managing complex consumption patterns from 1,350 to 42,260 m /month (e.g., urban, rural, industrial). Our results show that 93.1% of the areas achieve operational success (MAPE < 10%) with a mean MAPE of 7.3% and median of 6.5% (IQR = 2.4%), representing a 34.2% improvement compared to baseline Prophet models. The system mean execution is 22.7 ± 8.2 seconds per area (median: 19.7 seconds; IQR: 16.1 28.3 seconds), resulting in a total runtime of 659 seconds for the full 12-month walk-forward validation. Execution was parallelized across a 6-core (12-thread) Intel i7 processor. These findings demonstrate readiness for nationwide deployment across 550+ regions in Palestine and establish the foundation for the first national-scale automated water forecasting system in the Middle East, providing a transferable framework for efficient resource allocation in water-stressed regions in other places.

Title

Scientific Reports

Publisher

Springer Nature

Publisher Country

United Kingdom

Indexing

Thomson Reuters

Impact Factor

3.9

Publication Type

Online only

Volume

--

Year

2026

Pages

on line