Anthropogenic influence on multidecadal changes in reconstructed global evapotranspiration
H. Douville1*, A. Ribes1, B. Decharme1, R. Alkama1 and J. Sheffield2
Global warming is expected to intensify the global hydrological cycle1, with an increase of both evapotranspiration (EVT) and precipitation.Yet, the magnitude and spatial distribution of this global and annual mean response remains highly uncertain2. Better constraining land EVT in twenty-first-century climate scenarios is critical for predicting changes in surface climate, including heatwaves3 and droughts4, evaluating impacts on ecosystems and water resources5, and designing adaptation policies. Continental scale EVT changes may already be underway6,7, but have never been attributed to anthropogenic emissions of greenhouse gases and sulphate aerosols. Here we provide global gridded estimates of annual EVT and demonstrate that the latitudinal and decadal differentiation of recent EVT variations cannot be understood without invoking the anthropogenic radiative forcings. In the mid-latitudes, the emerging picture of enhanced EVT confirms the end of the dimming decades8 and highlights the possible threat posed by increasing drought frequency to managing water resources and achieving food security in a changing climate.
Detection is the process of demonstrating that an observed change cannot be explained by internal climate variability. Attribution of a change to anthropogenic influence requires the additional demonstration that the detected change is consistent with the change simulated in response to a combination of external forcings, including anthropogenic changes in the composition of the atmosphere, and not consistent with alternative explanations. This implies that all important forcing mechanisms (anthropogenic greenhouse gases and aerosols, but also solar radiation and volcanism) must be considered. With the increasing confidence that recent global warming is very likely caused by human activities, detection and attribution (D&A) studies have gradually moved to climate variables more relevant for understanding climate change impacts, such as on the water cycle. Some success has been obtained at detecting human-caused changes in zonal mean precipitation9. D&A of continental scale changes in the other components of the land-surface water budget, however, remains a challenge given the limited instrumental record and the strong spatiotemporal variability of hydrological variables10.
As far as EVT is concerned, only relatively few monitoring sites
operate around the world and the period of record is quite short11. Two recent studies6,7 have used such in situ measurements for tuning global empirical EVT schemes based on remote sensing and standard meteorological data. They agreed on a global increase
in annual mean EVT by about 7mm per year per decade from 1982 to the late 1990s. These results were compared with EVT outputs of process-oriented land-surface models and were found to be relatively robust6. The 19822008 period is, however, too shortfor a formal D&A. Moreover, one study6 suggested that the increase in global EVT could have ceased after 1998, thereby highlighting the need to account for multidecadal variability rather than only linear trends in D&A algorithms.
Here we use two offline global hydrological simulations, respectively from the Interaction SoilBiosphereAtmosphere (ISBA; ref. 12) and Variable Infiltration Capacity (VIC; ref. 13) models, as pseudo-observations to detect and attribute changes in land EVT over the 19512005 period. The offline mode consists of driving the models with a hybrid atmospheric forcing that merges subdaily meteorological reanalyses and monthly means of both in situ and satellite observations14. Our offline simulations are based on two different land-surface schemes (see Supplementary Informations S1 and S2) and two different precipitation forcings (see Supplementary Information S3). As a consequence, they provide four parallel EVT estimates. Not surprisingly, they show some discrepancies on interannual to decadal timescales, which are mainly owing to the different physics and parameters between ISBA and VIC, but share common features on longer timescales.
How reliable are these global annual mean EVT reconstructions? ISBA and VIC belong to a generation of land-surface models in which subgrid variability of hydrological processes has been accounted for and carefully evaluated on the basin scale against observed river discharges (see Supplementary Informations S1 and S2). This explains why ISBA and VIC global annual mean EVT (1.18 and 1:08mmd
