Water use trend analysis: A non-parametric method for the environmental Kuznets curve detection

Understanding of the trend in water use with possible relation to economic development is essential for forecasting future water demand and developing response strategies. The environmental Kuznets curve hypothesis asserts that pollution emissions and natural resource use often follow an inverted-U shape. This paper develops a piecewise Mann-Kendall method to address inverted-U shape detection based on a non-parametric approach. P-values of the piecewise Mann-Kendall method statistic are identified using the exhaustion method, approximation method or analytic method, depending on the data size. Tests of the piecewise Mann-Kendall method on synthetic data indicate that the developed method provides reasonable type I error and power. The method is successfully applied to pattern detection in real water use. Water uses in a few countries and provinces in China present the environmental Kuznets curve. Unlike the traditional parametric quadratic regression method which is sensitive to data skewness, this non-parametric method offers consistent environmental Kuznets curve detection regardless of the original or logarithmic form of data being used.