Abstract
Summary Water is essential to life, and tracking trends in the withdrawal of water is paramount if we aspire to become more sustainable. Traditional statistical indicators, such as mean and standard deviation, are useful to track these trends, but they can sometimes fail to capture relevant and nontrivial properties. In this work, we first highlight the limits of these traditional statistical tools and we then offer a new network approach to study water withdrawals. Public supply water withdrawal data from the U.S. Geological Survey (USGS) for the years 1985, 1990, 1995, and 2005 were used in gallons of water per capita per day for all U.S. counties. Essentially, a network is formed between counties when they have withdrawal values within a certain range, ± ξ, of one another. A giant cluster rapidly emerges, containing more than 80% of the nodes for a ξ of 1%. The counties with the highest number of connections are associated with the mode of distribution, and we found multimodal patterns for earlier years. Moreover, the average shortest‐path length can be seen as the spread of a distribution. Overall, beyond a possible process of homogenization, water withdrawal patterns do not seem to have evolved much from 1985 to 2005, and no spatial correlation was detected. Though the methodology is yet to be formalized, it manages to give meaningful insights to supplement traditional statistical analyses.