WATERSHED CLASSIFICATION USING CANONICAL CORRESPONDENCE ANALYSIS AND CLUSTERING TECHNIQUES: A CAUTIONARY NOTE¹

Abstract

Watershed classification using multivariate techniques requires the incorporation of continuous datasets representing controlling environmental variables. Often, out of convenience and availability rather than importance to the structure of the system being modeled, the environmental data used originate from a variety of sources and scales. To demonstrate the importance of appropriate environmental data selection, classifications of six‐digit hydrologic units (1:24,000) across selected geographic areas within the Interior Columbia River Basin were produced. Canonical correspondence analysis was used to select and test environmental variables important in predicting Rosgen stream types and valley bottom classes. Then, hierarchical agglomerative clustering was used to group (classify) watersheds based on these variables. Statistically significant results were derived from the use of organized classification data with presumed predictive relationships to watershed properties, and a random distribution of environmental variables from the same datasets provided similar results. The results contained herein demonstrate that these analysis techniques do not necessarily select meaningful variables from a broad spectrum of data and that significant results are easily generated from randomly associated data. It is suggested that classifications produced using these multivariate techniques, especially when using multi‐scale data or data of unknown significance, are subject to invalid inferences and should be used with caution.

Comparative hydrology across AmeriFlux sites: The variable roles of climate, vegetation, and groundwater

“…The results presented here are suggestive of components of a catchment classification scheme based on emergent process controls on the properties of the hydrological filter of catchments. Such a process‐based approach would complement existing, largely statistically driven approaches to catchment classification [ McDonnell and Woods , 2004; Detenbeck et al , 2000; Caratti et al , 2004] and has strong analogies with recent approaches based on catchment “functionality” [ Sawicz et al , 2010], although to date, such approaches have focused on the features of streamflow. As illustrated in Figure 9, the AmeriFlux sites considered here appear to broadly cluster across aridity gradients, with secondary controls arising on the basis of rooting depth (which determines the potential for water table dynamics to interact with ET at a site) and, in energy‐limited sites, the suite of controls that determine the behavior of early season ET.…”

Las seis reglas de la regionalización en hidrología

“…Antes de generar grupos homogéneos es necesario verificar la consistencia de las variables que van a utilizarse (Ebisemiju, 1979;Nathan and McMahon, 1990;Caratti et al, 2005;Campos-Aranda 2017). En general todo análisis exploratorio de datos es recomendado (Tukey, 1977).…”

River Classification as a Geographic Tool in the Age of Big Data and Global Change

“…Although data mining and machine learning are sometime derided as black‐box techniques, they can lead to new knowledge discoveries when used by experts (Miller ). Moreover, it is necessary to use caution when classifications are based on multivariate datasets with different sources and scales (Caratti and others ). Nevertheless, there is great potential in accomplishing process‐based classification through bottom‐up approaches, which benefit greatly from the availability of big data and the development of machine‐learning algorithms that can reveal the patterns in the data and leave it up to human experts to understand the processes behind those patterns.…”