Dynamic Scaling of the Generalized Complementary Relationship Improves Long-term Tendency Estimates in Land Evaporation

Jozsef Szilagyi,Richard Crago,Ning Ma

doi:10.1007/s00376-020-0079-6

Abstract

Most large-scale evapotranspiration (ET) estimation methods require detailed information of land use, land cover, and/or soil type on top of various atmospheric measurements. The complementary relationship of evaporation (CR) takes advantage of the inherent dynamic feedback mechanisms found in the soil-vegetation-atmosphere interface for its estimation of ET rates without the need of such biogeophysical data. ET estimates over the conterminous United States by a new, globally calibrated, static scaling (GCR-stat) of the generalized complementary relationship (GCR) of evaporation were compared to similar estimates of an existing, calibration-free version (GCR-dyn) of the GCR that employs a temporally varying dynamic scaling. Simplified annual water balances of 327 medium and 18 large watersheds served as ground-truth ET values. With long-term monthly mean forcing, GCR-stat (also utilizing precipitation measurements) outperforms GCR-dyn as the latter cannot fully take advantage of its dynamic scaling with such data of reduced temporal variability. However, in a continuous monthly simulation, GCR-dyn is on a par with GCR-stat, and especially excels in reproducing long-term tendencies in annual catchment ET rates even though it does not require precipitation information. The same GCR-dyn estimates were also compared to similar estimates of eight other popular ET products and they generally outperform all of them. For this reason, a dynamic scaling of the GCR is recommended over a static one for modeling long-term behavior of terrestrial ET.

Highlights

While the globally distributed eddy-covariance flux towers have contributed significantly to our knowledge of ET across a wide range of ecosystems [see a recent review by Baldocchi (2020)], the spatiotemporal variation of global ET and its response to the changing climate remains highly uncertain (Mueller et al, 2011; Liu et al, 2016) because the estimation of long-term, spatially resolved ET is yet laden by difficulties in parameterizing the biophysical processes that control ET in the current land surface models (LSMs) (Ukkola et al, 2016; Ma et al, 2017) and remote sensing algorithms (Vinukollu et al, 2011; Velpuri et al, 2013)
Most LSMs within NLDAS-2 still utilize the NOAA normalized difference vegetation index data developed by Gutman and Ignatov (1998) on a five-year-mean monthly basis without any interannual variation as input (Xia et al, 2012), failing to reasonably capture the impact of vegetation changes on ET
Reduced model performance of generalized complementary relationship (GCR)-stat is apparent in the long-term linear tendencies of the basin-averaged annual evaporation values (Fig. 4) by being less effective than GCR-dyn in reproducing the observed linear trends in the water-balance data

Summary

Introduction

Land surface evapotranspiration (ET) is a central component in the Earth’s energy, water, and carbon cycles (WangDYNAMIC SCALING IMPROVES EVAPORATION ESTIMATES VOLUME 37While the globally distributed eddy-covariance flux towers have contributed significantly to our knowledge of ET across a wide range of ecosystems [see a recent review by Baldocchi (2020)], the spatiotemporal variation of global ET and its response to the changing climate remains highly uncertain (Mueller et al, 2011; Liu et al, 2016) because the estimation of long-term, spatially resolved ET is yet laden by difficulties in parameterizing the biophysical processes (e.g., root water uptake, stomatal resistance and its response to CO2 concentration changes) that control ET in the current land surface models (LSMs) (Ukkola et al, 2016; Ma et al, 2017) and remote sensing algorithms (Vinukollu et al, 2011; Velpuri et al, 2013).

Results

Conclusion