Comment on hess-2021-126 [peer_review]

2021 unpublished
The widespread negative correlation between the atmospheric vapor pressure deficit and soil moisture lends strong support to the complementary relationship (CR) of evapotranspiration. While it has showed outstanding performance in predicting actual evapotranspiration (ETa) over land surfaces, the calibration-free CR formulation has not been tested in the Australian continent dominantly under (semi-)arid climates. In this work, we comparatively evaluated its predictive performance with seven
more » ... ance with seven physical, machine-learning, and land surface models for the continent at a 0.5°×0. 5° grid resolution. 15 Results showed that the calibration-free CR that forces a single parameter to everywhere produced considerable biases when comparing to water-balance ETa (ETwb). The CR method was unlikely to outperform the other physical, machine-learning, and land surface models, overrating ETa in (semi-)humid coastal areas for 2002-2012 while underestimating in arid inland locations. By calibrating the parameter against water-balance ETa independent of the simulation period, the CR method became able to outperform the other models in reproducing the spatial variation of the mean annual ETwb and the interannual variation 20 of the continental means of ETwb. However, interannual the grid-scale variability and trends were captured unacceptably even after the calibration. The calibrated parameters for the CR method were significantly correlated with the mean net radiation, temperature, and wind speed, implying that (multi-)decadal climatic variability could diversify the optimal parameters for the CR method. The other physical, machine-learning, and land surface models provided a consistent indication with the prior global-scale assessments. We also argued that at least some surface information is necessary for the CR method to describe 25 long-term hydrologic cycles at the grid scale. Introduction Terrestrial evapotranspiration (ETa) links water, energy, and carbon exchanges between lands and the atmosphere. On the global scale, more than 60% of land precipitation (P) returns to the atmosphere through plants' vascular systems and soil pores, consuming over 70% of surface net radiation for the phase change of water (Trenberth et al.the warming atmosphere, the upward latent heat flux has received growing attention, because it can control surface water availability, plants productivity, and ecosystem sustainability (Pareek et al., 2020; Kyatengerwa et al., 2020; Jasechko, 2018; Swann et al., 2016). Interacting with the atmosphere, changes in ETa could substantially increase heatwave risks (Miralles et al., 2014a; Mueller and Seneviratne, 2012) and alter precipitation patterns (Koster et al., 2004). However, expensive costs and operational difficulties make ETa observation networks (e.g., the FLUXNET; 35 Baldocchi et al., 2001) subject to limited spatial extent, short data lengths, and questionable data quality. Hence, modeling approaches are inevitable for a regional-or a global-scale ETa analysis, and usually based on physical theories (e.g., Zhang et al., 2016), machine-learning techniques (e.g., Jung et al., 2011), and conceptual land surface schemes (e.g., Haverd et al., 2018) that inherently have numerous error sources. For instance, Jung et al. (2019) produced the global mean ETa from 557 to 668 mm a -1 for 2001-2005 with physical and machine learning models. The previous global mean ETa estimates have varied even 40 in a larger range of 417-650 mm a -1 for a part or the whole period of 1982-2011 (Pan et al., 2020). The large discrepancies in the global means imply that modeling prescriptions for ETa have diverse uncertainty sources, such as forcing errors, ill-posed parameterizations, structural deficiencies, and insufficient training, and thus necessitate intercomparison studies to assess the associated limitations and uncertainties (e.g., Pan et al., 2020). When it depends on a physical equation, such as the Penman-Monteith (Monteith, 1965) or the Priestley-Taylor 45 (Priestley and Taylor, 1972) equations, an ETa model assumes typically that ETa under water deficiency is proportional to the atmospheric evaporative demand (ETp). In the Global Land Evaporation Amsterdam Model (GLEAM; Martens et al., 2017), for example, the Priestley-Taylor equation is multiplied by a stress module to predict ETa under water limited conditions. Similar approaches are easily found with physical and land surface models that adjust the surface roughness length of the Penman-Monteith equation to represent water stress (e.g., Zhang et al., 2016; Pan et al., 2015) . Note that adjusting the surface 50 roughness length is mathematically equivalent to multiplying a coefficient to ETp (Seneviratne et al., 2010). Though it has reliably predicted ETa at multiple scales (e.g., Martens et al., 2017; Fisher et al., 2008), the proportionality assumption unavoidably requires soil moisture information to quantify the degree of water stress, giving rise to practical difficulties such as data unavailability, computational inefficiency, and delayed data dissemination. Importantly, the assumption of the positive relation between ETa and ETp could be rejected by observational evidence that supports negative correlations between the two 55 (e.g., Brutsaert, 2006; Ramírez et al., 2005; Hobbins et al., 2004) . Han et al. (2014) emphasized that the correlation between ETa and ETp depends mainly on water availability rather than being always positive. The drawbacks of the proportionality assumption can be remedied at least in part by employing the complementary relationship (CR) of evapotranspiration. Bouchet (1963) found that the pan evaporation rate over a small wet patch surrounded by water-limited areas is higher than when the same surroundings are entirely wet. Since the small wet patch hardly influences 60 the overlying atmosphere, ETp over the wet surface is raised by blending with the drier and hotter surroundings. This "oasis effect", by contrast, is negligible in the case that the surrounding areas are entirely wet and large enough to transform the overpassing atmosphere. In other words, even under the same surface radiation and wind speed conditions, ETp responds to changes in regional water availability. Hence, one can predict water-limited ETa by gauging how much ETp is raised from the https://doi.org/10.5194/hess-2021-126 Preprint. Discussion started: 16 March 2021 c Author(s) 2021. CC BY 4.0 License. hypothetical evaporation rate that should occur under the full wetness (referred to as the wet-environment ET; ETw). Since a 65 higher adjustment in ETp indicate a lower water availability and thus ETa, the CR supports inverse correlations between ETa and ETp (i.e., complementarity). In practice, the complementarity allows users to predict ETa with no surface information, because ETp and ETw are all obtainable from meteorological data. among others. While those CR methods have been deemed mere heuristic methods with limited reliability (Shuttleworth et al., 2009), the non-dimensional derivation of Brutsaert (2015) and the following modifications (Szilagyi et al., 2017; Crago et al., 2016) have suggested the generality and definitiveness of the CR principle. The non-dimensional CR formulations have shown outstanding performance in predicting water-limited ETa at local, regional, and global scales (e.g., Brutsaert et al., 2020; Crago and Qualls, 2018; Brutsaert et al., 75 2017), and applications are extended to drought assessments (Kyatengerwa et al., 2020; Kim et al., 2019b) and even used for predicting the crop coefficient under the proportionality assumption (Kim et al., 2019a). Though the non-dimensional CR formulations are still under improvement based on the thermodynamic foundations (Szilagyi, 2021; Qualls and Crago, 2020), they mostly require any ETa observations to identify required parameters. Szilagyi et al. (2017) is the only calibration-free CR method that analytically determines the parameter for ETw with no requirement of 80 ETa data. By transferring the parameter analytically obtained in highly humid locations to the entire region of interest, the calibration-free CR formulation showed superior performance in predicting ETa to typical land surface and machine-learning models in the conterminous United States and China where climates are very diverse (Ma et al., 2019; Ma and Szilagyi, 2019). However, the same approach has not been examined in a continent where only small parts are under humid climates, and thus it is questionable whether the parameter transferring from humid locations is valid. 85 In this work, therefore, we applied the calibration-free CR formulation for the Australian continent where land surfaces are mostly under (semi-)arid climates, and its predictive performance was compared with a bunch of physical, machine-learning, and land surface models. Here, we addressed that the use of a single parameter for the entire continent could lead the CR method to low performance in preserving spatial coherence, interannual variability, and decadal trends of waterbalance ETa. We also provided some perspectives for improving the non-dimensional CR formulations. 90 2 Methodology and data The generalized complementary relationship approach The CR principle explains the feedback response of ETp to regional water deficiency using the three evaporation rates, namely, ETa, ETp, and ETw. Again, ETa is the actual water flux from a homogeneous land surface to the atmosphere, and ETp is the atmospheric capacity to receive water vapor that responds actively to water availability on the surface. ETw is the 95 hypothetical ETa rate that would take place under the same atmospheric conditions but ample water. ETp under regional water
doi:10.5194/hess-2021-126-rc1 fatcat:osuz2usa3vc53ioxzjatdzzxie