Improving Short-Term Urban Water Demand Forecasts with Reforecast Analog Ensembles: Fid-Bau Portal

Improving Short-Term Urban Water Demand Forecasts with Reforecast Analog Ensembles

Tian, Di / Martinez, Christopher J. / Asefa, Tirusew

Urban water demand forecasting is key to municipal water supply management. Short-term urban water demands are influenced by weather conditions. Thus, short-term urban water demand forecasting could be improved by using accurate weather forecasting information. This study explores the potential of using an analog approach with a newly developed retrospective weather forecast (reforecast) of a numerical weather prediction (NWP) for improving short-term urban water demand forecasting. The analog method derives an analog ensemble forecast resampled from observed data (analogs) based on the reforecast of a NWP: the Global Ensemble Forecast System (GEFS). The probabilistic and ensemble mean forecasts generated from analogs of weekly total rainfall (WeekRain), number of rainy days in one week (RainDays), number of consecutive rainy days in one week (CosRainDays), number of hot days in one week (HotDays), and daily mean temperature of the first seven lead days (T) from the reforecast were evaluated using in situ observations. The analog ensemble forecasts were used to drive seven water demand forecasting models based on autoregressive integrated moving average with exogenous inputs (ARIMAX) to make water demand forecasts in the Tampa Bay region of Florida. The GEFS-based analog forecast generally showed moderately high skill for WeekRain, RainDays, CosRainDays, and T but no skill for HotDays. The water demand forecasts driven by analog forecasts mostly showed higher skill than the original ARIMAX forecasts. Besides improving forecast accuracy, the analog-driven water demand forecasts accounted for the uncertainty of the weather forecasts, allowing for the assessment of demand forecast uncertainty. These results indicated that NWP-based analogs showed promising features for advancing the accuracy of short-term urban water demand forecasts.