Jasechko et al. reply

Abstract

replying to A. M. J. Coenders-Gerrits et al. Nature506,http://dx.doi.org/10.1038/nature12925(2014)

In their Comment, Coenders-Gerrits et al.¹ suggest that our conclusion that transpiration dominates the terrestrial water cycle² is biased by unrepresentative input data and optimistic uncertainty ranges related to runoff, interception and the isotopic compositions of transpired and evaporated moisture. We clearly presented the uncertainties applied in our Monte-Carlo sensitivity analysis, we reported percentile ranges of results rather than standard deviations to best communicate the nonlinear nature of the isotopic evaporation model, and we highlighted that the uncertainty in our calculation remains large, particularly in humid catchments (for example, figure 2 in our paper²).

Main

A global runoff flux of 37,288 km³ per year was obtained from a previous compilation of discharge measurements³, and has been independently reproduced within 1,300 km³ per year (36,055 km³ per year) using satellite measurements⁴. The larger uncertainty reported by Coenders-Gerrits et al.¹ (± 5,100) is obtained using a higher runoff value from a schematic in a review⁵, which also included direct groundwater discharge to oceans, making this number 10% higher than it should be. Because the uncertainty estimate recommended by Coenders-Gerrits et al.¹ also includes water fluxes not included in our runoff parameter, we stand by the original values used, which seem to us to be the best approximation of this uncertainty^3,4.

Global interception, derived from satellite measurements^6,7 (7,500 km³ per year, as used in our paper²), global climate model outputs⁸ (11,900 km³ per year, shown in figure 6b of ref. 4) and stand-level measurements⁹ (median value of 17.9% of precipitation, or 19,700 km³ per year), all fall into a range from 7,500 km³ per year to 19,700 km³ per year. We note that the high end-member applied by ref. 1 is derived from stand-level measurements and was not intended to be used to estimate a global interception flux. Its use in this manner¹ introduces bias towards regions where interception is expected to be important, because interception measurements are not often reported in areas free of canopy cover.

The Comment¹ also suggests that the uncertainties we applied to the deuterium excess (d = δ²H – 8δ¹⁸O) of transpired moisture (d_T ) and evaporate (d_E ) were understated². First, Coenders-Gerrits et al.¹ propose a deuterium excess uncertainty of ±19‰ for transpired moisture, quoting the individual uncertainties in our paper in the δ¹⁸O and δ²H models. This uncertainty is too large because they do not recognize that δ¹⁸O and δ²H values covary¹⁰, resulting in a well-constrained range of deuterium excess values for groundwater tapped by plant roots. The uncertainty we applied² for d_T (±3‰) is validated by a database of stable oxygen and hydrogen isotope values for shallow and deep groundwater across the continental USA (the Water Quality Portal, http://www.waterqualitydata.us). Samples from depths of <4.6 m—the global average rooting depth¹¹—show a deuterium excess of 7.8 ± 4.1‰ (n = 1,021; 25th to 75th percentile range of data set), found to be nearly identical in deuterium excess to that of samples collected from depths greater than 4.6 m: d = 7.9 ± 3.1‰ (n = 24,309). These ranges are consistent with our estimate² of d_T of 8 ±3‰. Coenders-Gerrits et al.¹ suggest that the deuterium excess value ascribed to evaporating moisture should be higher than ±30‰. Our uncertainty of ±30‰ was derived from the range of deuterium excess calculations for 73 lakes used in our study. We used percentile ranges so that the high sensitivity (and associated high uncertainty) would be carried through to our final, global scale calculation of transpiration and evaporation fluxes.

Finally, Coenders-Gerrits et al.¹ suggest that we neglect uncertainty in water-use efficiency (WUE), despite our use of the standard error of regression that resulted in a global WUE of 3.4 ± 0.9 mmol CO₂ per mol H₂O (ref. 2). CO₂ fertilization does indeed affect WUE¹², as we acknowledged², and continued research into WUE will allow us to monitor and map these changes as atmospheric CO₂ increases. We strongly disagree with Coenders-Gerrits et al.¹ that gross primary production models “fail at the regional and local scale,” given the success of the FLUXNET project in constraining this vital component of the carbon cycle (123 ± 8 Gt of C per year reported¹³). Finally, WUE has in fact been measured at scales that are larger than the leaf level (see Supplementary Infomation refs 105, 114, 116 and 118 in ref. 2).

We do not agree with the magnitude of the increase in uncertainty in ref. 1, nor with the unidirectional changes to input parameters that produce lower transpiration fluxes. However, we agree with the final point of the Comment¹: that further method development and the use of stable O and H isotopes in hydrology can help to reduce uncertainty in terrestrial evaporation and transpiration.