When scientists want to understand the fate, quality, and future of Earth’s freshwater resources, they often turn to studying watersheds–areas that drain rain or snow into water bodies such as rivers and lakes. Investigating how these environments, a major source of freshwater in the United States, are affected by land development and climate change can help inform water management and establish a water-secure future. However, because watersheds are often so large and vary in characteristics like soil type and landforms within just a few feet, fully grasping and predicting how they function can be challenging. 

EESA scientists Fabio Ciulla and Charuleka Varadharajan recently developed a new approach to grouping watersheds that can help combat these difficulties in a recently published highlight paper in the journal Hydrology and Earth System Sciences

The concept behind grouping watersheds is to use data about their traits, such as weather, plants, soil, and surrounding human activity (i.e., is it near a city or a farm?), to categorize watersheds based on their similarities, similar to how pollsters group people into different categories to understand how they would vote. By grouping watersheds, scientists are able to identify patterns in water quality and availability that can help them better understand what these water resources might look like in the future should these characteristics change. But traditional computing methods are less suitable for large datasets that contain this type of information. 

Ciulla and Varadharajan developed a new way to sort this data using a network analysis, which identifies similarities and connections within datasets, to group over 9,000 watersheds across the United States based on hundreds of traits.

With more comprehensive information about how traits like weather, temperature, and land-use affect our access to this critical resource, this approach can advance how we understand and predict water availability throughout these environments. And more, because this method can help to categorize any large dataset, the new approach can be applied to fields well beyond watershed science.