Use of neural network to predict ﬂash ﬂood and attendant water qualities of a mountainous stream on Oahu, Hawaii G.B. Sahoo a , C. Ray a, * , E.H. De Carlo b a Department of Civil and Environmental Engineering, University of Hawaii at Manoa, Water Resources Research Center, 2540 Dole Street, 383 Holmes Hall, Honolulu, HI 96822, United States b Department of Oceanography, University of Hawaii at Manoa, Honolulu, HI 96822, United States Received 15 November 2005; received in revised form 26 November 2005; accepted 26 November 2005 Summary Frequent ﬂash ﬂoods of Hawaii streams pose continuous threats to the coastal environ- ment because the streams respond rapidly to high runoff and huge transport quantities of sedi- ments, to which are sorbed nutrients, heavy metals, and persistent hydrophobic organic compounds. High-frequency stream ﬂow and water quality estimation are essential to correctly assess water quality variations and pollutant loads during ﬂash ﬂoods, because stream ﬂow and tur- bidity in Hawaii can change by a factor of 60 and 30, respectively, in 15 min. This study shows the application of artiﬁcial neural networks (ANNs) to assess ﬂash ﬂoods and their attendant water quality parameters using measured data of a Hawaii stream. The paper illustrates that ANNs predict stream ﬂow with a correlation coefﬁcient (R) greater than 0.99 and turbidity and speciﬁc conduc- tance with R-values greater than 0.80. Although the R-values for the estimation of dissolved oxy- gen, pH, and water temperature were somewhat low, most of the estimated stream water quality values (turbidity, speciﬁc conductance, dissolved oxygen, pH, and water temperature) were within the limits of ±30% deviations of the 1:1 line. The R-value for the estimation of stream water qual- ities could have been signiﬁcantly improved if high resolution (at 15 min or lower measurement fre- quency), noise-free, and continuous data were available for a longer period of time. The paper demonstrates that the upstream water quality parameters depend on weather forces and land use of the watershed and the downstream water quality parameters additionally inﬂuenced by oce- anic tides. Stream stage is found to be an important input parameter for stream ﬂow prediction using ANN; however, the predictive performance of ANN for the estimation of stream ﬂow is improved if weather data, rainfall, and evapotranspiration are included in the input data set.  c 2005 Elsevier B.V. All rights reserved. KEYWORDS Flash ﬂood; Surface hydrology; Water quality; Non-point source pollution; Watershed; Coastal oceanography; Artiﬁcial neural network 0022-1694/$ - see front matter  c 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.jhydrol.2005.11.059 * Corresponding author. Tel.: +1 808 956 9652; fax: +1 808 956 5014. E-mail address: cray@hawaii.edu (C. Ray). Journal of Hydrology (2006) 327, 525– 538 available at www.sciencedirect.com journal homepage: www.elsevier.com/locate/jhydrol