Selecting model scenarios of real hydrodynamic forcings on mesotidal and macrotidal estuaries influenced by river discharges using K-means clustering Javier F. B  arcena * , Paula Camus, Andr  es García, C  esar  Alvarez Environmental Hydraulics Institute “IH Cantabria”, Universidad de Cantabria, Isabel Torres N  15, Parque Científico y Tecnol ogico de Cantabria, 39011, Santander, Spain article info Article history: Received 12 August 2014 Received in revised form 27 November 2014 Accepted 2 February 2015 Available online Keywords: K-means (KMA) Tidal-river estuaries River flow (Q) Astronomical tide (A) Model efficiency (CE) Modelling scenarios Suances estuary (SE) abstract The long-term (>30 years) simulation of 3D estuarine hydrodynamics with high-resolution meshgrids is still a challenge in numerical modeling because of the large data set of results and the computational cost requirements. Meso and macrotidal estuaries are governed by tidal action and could be influenced by river. The complexity of their behavior, suggest data mining methods may be particularly effective in selecting short-term series from a long-term series to identify the major modes of forcing variability. This study uses K-means clustering for two aims: explaining the variability of astronomical tides and river flows, and selecting scenarios of real forcings to obtain the mean behavior with a dimensional reduction. The application to the Suances estuary has highlighted the ability to classify long-term series in small number of groups. Before conducting any simulation, the proposal also determines the minimum and optimal number of groups to consider the combined effect of both forcings. © 2015 Elsevier Ltd. All rights reserved. 1. Introduction The modelling of hydrodynamic and transport processes in es- tuaries is a key component in order to estimate and understand water renewal or stratification patterns, calculate the physical and chemical dilution fluctuations of wastewaters, specify the alloca- tion and sizing of mixing zones or optimize the design of a field campaign. Meso and macrotidal estuaries are mainly governed by astro- nomical tides , whereas other forcing such as meteorological tides, river inflows, wind, waves, evaporation or precipitation are signif- icant on a site-specific basis. In this study, we are focused on es- tuaries where the astronomical tide and river discharges are the most important forcings to understand the mean behavior of the estuarine hydrodynamics. These type of estuaries are worldwide spreading so a methodology to understand their mean behavior could help researchers, technicians and/or regulators to manage them more efficiently. Among others, estuaries falling in this category could be Suances (B arcena et al., 2012a); Huelva (S amano et al., 2012); Urdaibai (García et al., 2010a); Mandovi (Vijith and Shetye, 2012); Mondego (Kenov et al., 2012); Hudson (Warner et al., 2005); Alafia(Chen, 2007); Tanshui (Liu et al., 2002); Columbia (Chawla et al., 2008), Yaquina (Frick et al., 2007); Ribble (Kashekipour et al., 2001); Haihe (Bai et al., 2003); or Humber (Edwards and Winn, 2006). Thus, an accurate classification of these two forcings, taking into account their environmental variability, is of high value to re- searchers as a source of condensed information and as a selection of model scenarios to identify the major modes of forcing variability and conduct numerical simulations. The effect of higher order moments or extreme values is out of the scope because there are specific methodologies to understand such specific cases (maximum annual values, peak over threshold …). In the last decades, long-term time series of the river flow and the astronomical tide from numerical models have been improving the knowledge of the temporal distribution of these two variables (see, for instance, Singh and Woolhiser, 2002; García et al., 2008; Andersen, 1994; or Pawlowicz et al., 2002). These databases are especially important at locations where instrumental data are not * Corresponding author. Tel.: þ34 942201616x1323; fax: þ34 942266361. E-mail addresses: barcenajf@unican.es (J.F. B arcena), camusp@unican.es (P. Camus), garciagan@unican.es (A. García), alvarezc@unican.es (C.  Alvarez). Contents lists available at ScienceDirect Environmental Modelling & Software journal homepage: www.elsevier.com/locate/envsoft http://dx.doi.org/10.1016/j.envsoft.2015.02.007 1364-8152/© 2015 Elsevier Ltd. All rights reserved. Environmental Modelling & Software 68 (2015) 70e82