Globally partitioning the simultaneous impacts of climate-induced and human-induced changes on catchment streamflow: A review and meta-analysis

Abstract

Streamflow (Q) change is simultaneously driven by climate-induced changes and various human activities. To understand their respective roles, a large number of partitioning studies were conducted. However, each of the scattered case studies may only provide local-to-regional understanding of the contributions from climate-induced changes and human-induced changes to streamflow variations over various study periods. To address if it is possible to reveal the relative contribution of climate-induced changes (dQ_CIC) and human-induced changes (dQ_HIC) from the spatially distributed case studies, and then to predict regional to global spatial pattern of streamflow changes due to climate-induced changes, we reviewed 103 studies providing access to 1321 catchment cases, and performed a meta-analysis which characterized how that population varied with the number of catchment cases per study and the length of study period, and analyzed if and to what extent the bias of that population was influenced by various study factors. Possible catchment streamflow change, relative to the 1960s−2000s, was projected for the 2040s–2060s. Our meta-analysis suggested that, over the past decades, total streamflow change (i.e. dQ_Tot), on average, was −11.9 mm.a−1, and streamflow changes due to dQ_CIC and dQ_HIC were + 3.6 ± 48.1 mm.a−1 and −15.5 ± 44.4 mm.a−1, respectively. We also found that several study factors including catchment size, partitioning method, and climate type (mainly mean annual precipitation) were not important sources of bias for the meta-analysis, as they each only explained a small portion of the variability in dQ_CIC or dQ_HIC. Although study periods of the case studies and climate oscillation influenced the magnitude and direction of dQ_CIC (and dQ_HIC), dQ_CIC showed a strongly temporal invariant dependence on dP, with a ratio of 0.48 for the fitted linear line (p < 0.05). The mean changing rate of dQ_CIC (i.e. dQ_CIC/dt) over the past decades was +0.07 ± 4.13 mm.a−2. Our study heightened the imperative to perform analysis with respect to the spatial variation of dQ response to climate perturbations and/or variation across various regions, and to disentangle hydrological contribution of interactions/feedbacks between climate-induced change and land-use/land-cover changes from the other anthropogenic changes in future partitioning studies for a better understanding of dQ response to climate-induced change.