Method
Our investigation was structured by process class (Anderson and McDonnell, 2005), for example “evapotranspiration”, “overland flow”, or “groundwater flow”. McMillan (2020) provided a list of papers describing processes in experimental watersheds. We searched literature on perceptual models, including recent discussions of perceptual model use and scope (Beven and Chappell, 2021; Wagener et al., 2021), perceptual models of well-known watersheds such as Panola (Aulenbach et al., 2021), Maimai (McGlynn et al., 2002) and the Attert Basin (Wrede et al., 2015), and perceptual models of specific processes such as infiltration (Beven, 2004). Many papers do not explicitly refer to perceptual models but equivalently describe runoff generation processes. We surveyed these papers with particular attention to studies from a wide range of climate and landscape types. These included arid (Ries et al., 2017), humid (Dunne and Black, 1970; Hewlett and Hibbert, 1967), cold region (Peters et al., 1995; Pomeroy et al., 1999; Quinton and Marsh, 1999; Rango, 1993), forested (Bonell, 1993; Jones, 2000), and karst watersheds (Hartmann et al., 2013).
From each paper, we extracted all names or short phrases describing runoff generation processes. Where available, we referred to previous process classifications such as the typology of groundwater–surface water interaction by Dahl et al. (2007), and processes in earth system models (Clark et al., 2015; Fan et al., 2019). We noted alternative names for each process, although it was sometimes difficult to ascertain minor differences in meaning between terms; see further comment in the Discussion section below.
To integrate the new taxonomy with previous classification systems, we specified a functional class for each process. Bracken et al. (2013) divided hydrological function into structural knowledge - to do with stores - and process-based knowledge - to do with fluxes. Wagener’s classification (2007) adds partitioning andrelease. We added a class Complex process for emergent behaviors, and subcategories of each class were added as needed (Table 1).
The process taxonomy was collated into spreadsheet format that tracked the hierarchical structure by assigning a ‘parent’ to each process name (see Supplemental Information). The spreadsheet was processed inR using the collapsibleTree package to create an interactive tree diagram (Khan et al., 2018). Processes were shown as nodes that expand and collapse, and were coloured by functional class. Tool tips displayed alternative names and the identifier.