The Mekong Basin in Southeast Asia exemplifies these issues with growing irrigation water demand (Pech and Sunada, 2008), greater flood-risk exposure (Osti et al., 2011), and hydropower-induced changes in seasonal river flow and ecology (Arias et al., 2012 and Ziv et al., 2012). Adaptation measures are hampered by Selleck Dasatinib uncertainties in projected
streamflow changes (Kingston et al., 2011). A number of hydrological models have been developed for the Mekong Basin to predict streamflow variability, however their complexity and lack of transparency (Johnston and Kummu, 2012), often limit possible users to modeling experts, instead of the practitioners working closely with populations affected by flow extremes. Additionally, the majority of models have been developed to predict flow along the Mekong mainstem, precluding accurate assessments in headwater catchments where populations are repeatedly exposed to flash floods and/or water resource shortages. Flow duration curves (FDCs) provide an integrated representation of flow variability selleck compound that can be used for water resource planning, storage design and flood risk management
(Castellarin et al., 2013). A period-of-record FDC indicates the percentage of time (duration) a particular value of streamflow is exceeded over a historical period. Similarly, a median annual FDC can reflect the percentage of time a particular value of streamflow is exceeded in a typical or median year
(see Vogel and Fennessey, 1994). Various parametric and nonparametric statistical methods exist to predict an FDC in ungauged catchments and have been applied in many parts of the world (Castellarin et al., 2004). We present a set of new multivariate power-law models to predict FDC percentiles as well as other flow metrics, at any location along the tributaries of the Lower Mekong River (Fig. 1) using easily determined catchment characteristics. Section 2 describes the main steps of the multiple regression analysis. Section 3 presents Mirabegron the data used to empirically develop the models. Section 4 presents the equations of the power-law models, discusses their significance and compares their performance with other case studies. We used a multivariate power-law equation (Eq. (1)), already used in many parts of the world (Vogel et al., 1999 and Castellarin et al., 2004), to estimate the river flow Q from m catchment characteristics Xi (i = 1, …, m). A logarithmic transformation of Eq. (1) results in a log-linear model (Eq. (2)) whose coefficients βi (i = 1, …, m) can be determined by multiple linear regression. equation(1) Q=expβ0⋅X1β1⋅X2β2⋅⋅⋅Xmβm⋅ν equation(2) ln(Q)=β0+β1⋅ln(X1)+β2⋅ln(X2)+⋯+βm⋅ln(Xm)+εln(Q)=β0+β1⋅ln(X1)+β2⋅ln(X2)+⋯+βm⋅ln(Xm)+ε β0 is the intercept term of the model. v (Eq. (1)) and ɛ (Eq. (2)) are the log-normally and normally distributed errors of the models, respectively.