Disentangling ecosystem transpiration from evapotranspiration observations employing simplified vegetation-substrate energy balance model
<p>Evapotranspiration (E<sub>ET</sub>) observed by eddy covariance (EC) towers is composed of physical evaporation (E<sub>E</sub>) from wet surfaces and biological transpiration (E<sub>T</sub>), that involves soil moisture uptake by roots and water vapor transfer regulated through the canopy-stomatal conductance (g<sub>C</sub>) during photosynthesis. E<sub>T</sub> plays a dominant role in the global water cycle and represents 80% of the total terrestrial E<sub>ET</sub>. Understanding the magnitude and variability of E<sub>T</sub> are critical to assess the ecophysiological responses of vegetation to drought. While separating E<sub>T</sub> signals from lumped E<sub>ET</sub> observations and/or simulating E<sub>T</sub> by terrestrial systems models is insufficiently constrained owing to the large uncertainties in disentangling g<sub>C</sub> from the aggregated canopy-substrate conductance (g<sub>cS</sub>), evaluating ecosystem E<sub>T</sub> derived through partitioning E<sub>ET</sub> observations (or model simulation) is also challenging due to the absence of any ecosystem-scale measurements of this biotic flux and g<sub>C</sub>. To date, the main methods for partitioning EC-E<sub>ET</sub> observations are largely based on regressing E<sub>ET</sub> with gross photosynthesis (P<sub>g</sub>) and atmospheric vapor pressure deficit (D<sub>A</sub>) observations. However, such methods ignore the essential feedback of the surface energy balance (SEB) and canopy temperature (T<sub>C</sub>) on g<sub>C</sub> and E<sub>T</sub>.</p><p>This study demonstrates partitioning E<sub>ET</sub> observations into E<sub>T</sub> and E<sub>E</sub> [soil evaporation (E<sub>Es</sub>) and interception evaporation (E<sub>Ei</sub>)] through an &#8216;analytical solution&#8217; of g<sub>C</sub>, T<sub>C</sub> and canopy vapor pressures by employing a Shuttleworth-Gurney vegetation-substrate energy balance model with minimal complexity. The model is called TRANSPIRE (Top-down partitioning evapotRANSPIRation modEl), which ingests remote sensing land surface temperature (LST) and leaf area index (L<sub>ai</sub>) information in conjunction with meteorological, sensible heat flux (H) and E<sub>ET</sub> observations from EC tower into the SEB equations for retrieving canopy and soil temperatures (T<sub>S</sub>, T<sub>C</sub>), g<sub>C</sub>, and E<sub>T</sub>.</p><p>E<sub>T</sub> estimates from TRANSPIRE were tested and evaluated with a remote sensing based E<sub>T</sub> estimate from an analytical model (STIC1.2), where lumped E<sub>ET</sub> was partitioned by employing a moisture availability constraints across an aridity gradient in the North Australian Tropical Transect (NATT) by using time-series of 8-day MODIS Terra LST and LAI products in conjunction with EC measurements from 2011 to 2018. Both methods captured the seasonal pattern of E<sub>T</sub>/E<sub>ET</sub> ratio in a very similar way. While E<sub>T</sub> accounted for 60&#177;10% of the annual E<sub>ET</sub> in the tropical savanna, E<sub>T</sub> in the arid mulga contributed 75&#177;12% of the annual E<sub>ET</sub>. Seasonal variation of E<sub>T</sub> was higher in the arid, semi-arid ecosystems (50 - 90%), as compared to the humid tropical ecosystem (10 - 50%). The TRANSPIRE model reasonably captured E<sub>T</sub> variations along with soil moisture and precipitation dynamics in both sparse and homogeneous vegetation and showed the potential of partitioning E<sub>ET</sub> observations for cross-site comparison with a variety of models.</p>