Chen et al. (2026) Optimal Short-Time Rainfall Time Series for Adapting Rainwater Harvesting System to Climate Change in Urban Runoff Management

Identification

• Journal: Water Resources Management
• Year: 2026
• Date: 2026-01-01
• Authors: Weilun Chen, Zhonghui Liu, Wei Ou, Pengxuan Wang, Xindong Wei, Weijun Gao
• DOI: 10.1007/s11269-025-04418-w

Research Groups

• School of International Education, Jilin Jianzhu University, Changchun, China
• School of Environment and Spatial Informatics, China University of Mining & Technology, Xuzhou, China
• Key Laboratory of Songliao Aquatic Environment, Ministry of Education, Jilin Jianzhu University, Changchun, China
• Faculty of Environmental Engineering, The University of Kitakyushu, Kitakyushu, Japan

Short Summary

This study develops a Bayesian network to identify optimal short-term rainfall time series for designing climate-adaptive rainwater harvesting systems (RWHs) in urban areas, revealing significant spatiotemporal differences in optimal series characteristics across 14 Japanese cities under various climate change scenarios. The framework provides a robust method for RWH planning by inferring optimal rainfall conditions from statistical indicators.

Objective

• To examine the influence of varying lengths of short-term rainfall series on rainwater harvesting system (RWH) performance under climate change scenarios across 14 Japanese cities.
• To develop a Bayesian network to capture the probabilistic features of these impacts and identify optimal rainfall conditions for RWH design under climate change.

Study Configuration

• Spatial Scale: 14 urban locations across Japan, representing different Köppen-Gieger climate zones (Cfa, Dfa, Dfb). A benchmark residential building (three occupants, 100 m² roof area) was assumed in each city.
• Temporal Scale:
  • Historical rainfall data: 1961–2022 (62 years).
  • Baseline period for downscaling: 1961–1990 (30 years).
  • Historical RWH performance evaluation: 1993–2022.
  • Future projections for RWH lifecycle: 2023–2042 (20 years).
  • Short-term rainfall series lengths: 1, 2, ..., 30 years (rolling series).

Methodology and Data

• Models used:
  • Continuous simulation model for RWH performance ("Yield after spillage" criterion).
  • Statistical downscaling: Long Ashton Research Station Weather Generator (LARS-WG6) for Global Climate Model (GCM) data.
  • Ensemble prediction: Averaging outputs from 14 GCMs.
  • Core analytical model: Bayesian network (trained using hill-climbing algorithm, Bayesian information criterion (BIC) for scoring, Maximum Likelihood Estimation (MLE) for parameter learning).
• Data sources:
  • Daily rainfall data: Japan Meteorological Agency (1961–2022).
  • Predicted rainfall data: 14 GCMs from Coupled Model Intercomparison Project phase 6 (CMIP6) for Shared Socioeconomic Pathways (SSP-126, SSP-245, SSP-370, SSP-585).
  • Building and water demand parameters: Ministry of Internal Affairs and Communications in Japan, Ministry of Land, Infrastructure, Transport and Tourism.

Main Results

• Rainfall series longer than 16 years are generally unsuitable for RWH design in cool temperate regions (northern Japan).
• Optimal rainfall series length varies spatially: 1–30 years in southern temperate regions, and 2–16 years in northern cool-temperate regions.
• Optimal rainfall characteristics exhibit a clear north-south contrast:
  • Inland cities: Prioritize series with higher wet-period frequencies.
  • Coastal areas: Prioritize series with lower wet-period frequencies.
  • Northern regions: Optimal series are characterized by longer dry periods, greater annual rainfall, and higher seasonal indices.
  • Southern regions: Optimal series exhibit the reverse pattern (shorter dry periods, lower annual rainfall, lower seasonal indices).
• The influence of climate change emission scenarios on optimal series distribution is minor, but higher non-potable water demand significantly broadens the range of similarity indices.
• The developed Bayesian network reliably infers optimal short-term rainfall time series from statistical indicators, achieving Area Under Curve (AUC) values ranging from 0.88289 to 0.98327, confirming its high predictive accuracy.
• The Bayesian network can also infer local meteorological conditions (wet periods, dry periods, seasonal indices, annual rainfall) from short-term series, addressing data gaps in regions lacking long-term observations.

Contributions

• Introduces a novel Bayesian network-based framework for identifying optimal short-term rainfall time series for climate-adaptive RWH design, overcoming limitations of outdated long-term data and short-term data non-stationarity.
• Quantifies the significant spatiotemporal variability of optimal rainfall series lengths and characteristics across diverse climatic zones in Japan.
• Demonstrates that specific rainfall characteristics (wet/dry periods, annual rainfall, seasonal index) of short-term series, not just length, are crucial for accurate RWH assessment under climate change.
• Provides a robust probabilistic inference tool that can extrapolate local meteorological conditions from short-term data, which is particularly valuable for data-scarce regions.
• Offers practical, location-specific guidance for RWH planning by linking optimal series characteristics to regional climate and building water demand.

Funding

• Natural Science Foundation of Jilin Provincial Department of Science and Technology (YDZJ202601ZYTS123)
• Outstanding Youth Foundation of the Education Department of Jilin Province (No. JJKH20240378KJ)
• Outstanding Youth Foundation of the Education Department of Jilin Province (No. JJKH20261694KJ)

Citation

@article{Chen2026Optimal,
  author = {Chen, Weilun and Liu, Zhonghui and Ou, Wei and Wang, Pengxuan and Wei, Xindong and Gao, Weijun},
  title = {Optimal Short-Time Rainfall Time Series for Adapting Rainwater Harvesting System to Climate Change in Urban Runoff Management},
  journal = {Water Resources Management},
  year = {2026},
  doi = {10.1007/s11269-025-04418-w},
  url = {https://doi.org/10.1007/s11269-025-04418-w}
}