Streamflow statistics and flow duration curves (FDC) (Stedinger, 1993; Yuan, 2013; Zhang, 2017) are commonly used in water resources planning and management practices (Smakhtin, 2001; Chen et al., 2007; Bangash et al., 2012; Zhang and Balay, 2014; Zhang and Kroll, 2007a). For a streamflow gage where reliable long-term continuous streamflow records are available, the needed streamflow statistics could be accurately estimated via a frequency analysis, providing sound basis for water resources planning and management (Zhang and Kroll, 2007b). However, water resources projects or water uses are more often located at ungaged sites where streamflow is not monitored and thus, longterm streamflow records could not be obtained. Water resources managers are confronted with the task of estimating streamflow for ungaged sites to best manage water resources. For prediction in ungaged basins (PUB), the use of a reference gage (also known as an index gage, donor gage, or base station) is generally required, which is assumed to be hydrologically similar to the ungaged site of interest under the premise of similar geologic, topographic, and climatic settings (Sivapalan, 2003; Sivapalan et al., 2003). A wide range of methods existed to estimate streamflow statistics for ungaged sites, such as regional regression (Pandey and Nguyen, 1999; Thomas and Benson, 1970; Vogel et al., 1999; Stagnitta et al., 2018), rainfallrunoff model (Liu and Gupta, 2007; Zhang et al., 2008; Wagener et al., 2009; Mas-Pla et al., 2012; Liu et al., 2015), baseflow correlation (Stedinger and Thomas, 1985; Zhang and Kroll, 2007b), drainage area ratio (Hirsch, 1979), and climate adjustment method (Laaha and Blöschl, 2005). These methods considered three groups of information in terms of choosing a reference gage: spatial distance, basin characteristics, and streamflow correlation (Laaha and Blöschl, 2005). The spatial distance methods include: (1) the use of spatially contiguous hydrologic regions which developed hydrologic homogenous regions and selected the reference gage from the same hydrologic regions where the ungaged site was located (Laaha and Blöschl, 2006); (2) spatial proximity which selected the nearest gage as the reference gage (Archfield and Vogel, 2010; Stedinger, 1993); and (3) the use of a nested gage which employed the immediate downstream or upstream gage as the reference gage (Laaha and Blöschl, 2005). The basin characteristics methods measured similarity between the drainage areas of the potential reference gages and ungaged sites using climatic, land use/land cover, soil, morphologic and geologic characteristics and chose the most similar gage as the reference gage (Merz and Blöschl, 2004; Nathan and McMahon, 1990). The streamflow correlation method required shortterm streamflow records or nominal measurements of streamflow at the ungaged site to select a reference gage. When streamflow correlation could be reliably established, the gage with the highest correlation will be used as the reference gage (Robson and Reed, 1999; Stedinger and Thomas, 1985; Yuan, 2013; Zhang and Kroll, 2007b)