A role for meta‐analysis in hydrology

Research in hydrology has evolved from a rich history of site‐based (usually catchment‐based) field data collection. These studies typically have a small sample size—that is, one study hillslope or catchment and usually no replication. Although different to many other scientific fields where replication is sacrosanct, such field‐based discovery science has driven much of what we know about infiltration, run‐off processes, and ecohydrological dynamics. But although field work at single sites continues to unearth new forms of hydrological behaviour—and this is much needed to understand better “how catchments work”—process inference by one‐off field experiments has significant limitations. Although some of these limitations can be mediated by replication and stronger adherence to the scientific method and formal hypothesis testing (Hooper, 2001; Pfister & Kirchner, 2017), many limitations remain. One tool that has not yet been fully exploited in hydrology is meta‐ analysis. We are certainly not the first to advocate for meta‐analysis in hydrology, and we note important previous work in large‐scale erosion and sediment delivery (Corbel, 1964; Milliman & Syvitski, 1992) and forest hydrology (Farley, Jobbagy, & Jackson, 2005). Nevertheless, uptake has been limited generally. Muchmore common is the traditional “review paper” where an author identifies studies on a particular topic, summarizes their findings, and reports a conclusion in narrative form. But while useful, reviews lack transparency and are subjective (Borenstein, Hedges, Higgins, & Rothstein, 2010). This is something of a double whammy in a field such as hydrology where the field studies often reviewed are themselves one‐off and exploratory in nature! Consequently, a review paper author may assign implicitly more weight or credence to one study over another based on criteria that are not always articulated. Although such narrative reviews do serve a useful role as a stock‐taking for the state‐of‐knowledge about a particular research question (with many recent examples), they are not objective and they are not quantitative.