Research Paper: SWdSoil and Water Improved irrigation water demand forecasting using a soft-computing hybrid model Inmaculada Pulido-Calvo*, Juan Carlos Gutie ´rrez-Estrada Dpto. Ciencias Agroforestales, EPS, Campus Universitario de La Ra ´bida, Universidad de Huelva, 21819 Palos de la Frontera (Huelva), Spain article info Article history: Received 5 December 2007 Received in revised form 24 September 2008 Accepted 29 September 2008 Available online 30 November 2008 Recently, Computational Neural Networks (CNNs) and fuzzy inference systems have been successfully applied to time series forecasting. In this study the performance of a hybrid methodology combining feed forward CNN, fuzzy logic and genetic algorithm to forecast one-day ahead daily water demands at irrigation districts considering that only flows in previous days are available for the calibration of the models were analysed. Individual forecasting models were developed using historical time series data from the Fuente Pal- mera irrigation district located in Andalucı´a, southern Spain. These models included univariate autoregressive CNNs trained with the Levenberg–Marquardt algorithm (LM). The individual models forecasting were then corrected via a fuzzy logic approach whose parameters were adjusted using a genetic algorithm in order to improve the forecasting accuracy. For the purpose of comparison, this hybrid methodology was also applied with univariate autoregressive CNN models trained with the Extended-Delta-Bar-Delta algo- rithm (EDBD) and calibrated in a previous study in the same irrigation district. A multicriteria evaluation with several statistics and absolute error measures showed that the hybrid model performed significantly better than univariate and multivariate autore- gressive CNNs. ª 2008 IAgrE. Published by Elsevier Ltd. All rights reserved. 1. Introduction. General scope of the work Information regarding water demand in irrigated areas is basic information for the development and implementation of successful water resource management tools given that irri- gated agriculture is the largest user of water throughout the world, accounting for 87% of consumptive uses (ONU, 1997; Sumpsi et al., 1998). Also, forecasting of water demand is one of the main problems in the design, management and modernisation of water supply and distribution systems. Actually, most pressurised irrigation systems operating on-demand deliver water with the flow rate and pressure required by farm irrigation systems, sprinkling or micro-irri- gation, and respecting the time, duration and frequency decided by the farmers. Therefore, they allow farmers to operate their irrigation systems with a large freedom with respect to other types of delivery schedules. Usually the Cle ´ ment formula (Cle ´ ment, 1966; Cle ´ ment and Galand, 1979) is used to design collective irrigation systems operating on- demand. This approach does not permit to take into consid- eration the variety of flow regimes occurring in a collective irrigation system. So, a risk threshold is accepted, i.e. during the operation of the system, flow rates higher than those assumed at design may occur with low probability due to the seasonal and daily variation in water demand. Consequently, a large spatial and temporal variability of pressure and flow rates available in the hydrants may occur and affect network performance and even crop yield (Pereira, 1999; Lamaddalena * Corresponding author. E-mail addresses: ipulido@uhu.es (I. Pulido-Calvo), juanc@uhu.es (J.C. Gutie ´ rrez-Estrada). Available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/issn/15375110 1537-5110/$ – see front matter ª 2008 IAgrE. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.biosystemseng.2008.09.032 biosystems engineering 102 (2009) 202–218