Multi-model ensemble prediction of terrestrial evapotranspiration across north China using Bayesian model averaging

Abstract

Using high‐quality dataset from 12 flux towers in north China, the performance of four evapotranspiration (ET) models and the multi‐model ensemble approaches including the simple averaging (SA) and Bayesian model average (BMA) were systematically evaluated in this study. The four models were the single‐layer Penman–Monteith (P–M) model, the two‐layer Shuttleworthe–Wallace (S–W) model, the advection–aridity (A–A) model, and a modified Priestley–Taylor (PT‐JPL). Based on the mean value of Taylor skill (S) and the regression slope between measured and simulated ET values across all sites, the order of overall performance of the individual models from the best to the worst were: S–W (0.88, 0.87), PT‐JPL (0.80, 1.17), P–M (0.63, 1.73) and A–A (0.60, 1.68) [statistics stated as (Taylor skill, regression slope)]. Here, all models used the same values of parameters, LAI and fractional vegetation cover as well as the forcing meteorological data. Thus, the differences in model performance were mainly attributed to errors in model structure. To the ensemble approach, the BMA method has the advantage of generating more skillful and reliable predictions than the SA scheme. However, successful implementation of BMA requires accurate estimates of its parameters, and some degradation in performance were observed when the BMA parameters generated from the training period were used for the validation period. Thus, it is necessary to explore the seasonal variations of the BMA parameters according the different growth stages. Finally, the optimal conditional density function of half‐hourly ET approximated well by the double‐exponential distribution. Copyright © 2016 John Wiley & Sons, Ltd.

Parameter sensitivity analysis and optimization for a satellite‐based evapotranspiration model across multiple sites using Moderate Resolution Imaging Spectroradiometer and flux data

“…Additionally, to evaluate the final performance of the model, we used a Taylor diagram [ Taylor , ] since this method is especially useful in testing multiple aspects of complex models [ International Panel on Climate Change , ]. Generally, the Taylor diagram characterizes a single point to indicate three different statistical relationships between the “test” field (simulation) and the “truth” field (observation) (correlation, ratio of the standard deviations, and root‐mean‐square difference of the patterns) [ Zhu et al ., ]. The statistics of each point can be scored using where S is the model skill metric bound by zero and unity (unity indicates agreement with observations), σ m is the standard deviation of the simulation, and σ o is the standard deviation of the observation.…”

“…Among them, the Priestly‐Taylor Jet Propulsion Laboratory (PT‐JPL) model proposed by Fisher et al . [] has been widely used to estimate ET because of its minimal requirements for ground‐based measurements and its good performance [ Feng et al ., ; Michel et al ., ; Zhu et al ., ]. For example, Ershadi et al .…”

A hierarchical Bayesian approach for multi‐site optimization of a satellite‐based evapotranspiration model

“…R 2 ranges between 0 and 1, with higher values indicating a good simulation result; the NSE values range from −∞ to 1, with NSE = 1 being the optimal value (Moriasi et al, ). The calculation of the statistical measures can be found in Zhu et al ().…”

“…Among these methods, modelling provides a powerful tool and is becoming more and more popular (Shugart, ). The Priestly–Taylor Jet Propulsion Laboratory (PT‐JPL) model (Fisher, Tu, & Baldocchi, ), which has a process‐based structure to partition total ET into its different components, is physically sound and rigorous, and has been widely used in previous studies due to its minimal requirements for ground‐based measurements and its good performance (Feng et al, ; Michel et al, ; Zhang et al, ; Zhu et al, ).…”

“…First, the PT‐JPL model is highly complex with some ecophysiological parameters that may vary with the environmental conditions, plant functional types (PFTs), and other factors (Zhang et al, ). The widely used method to estimate parameters is to calibrate the mechanistic process model independently for each site through simple Bayesian (SB) analysis (i.e., Samanta, Mackay, Clayton, Kruger, & Ewers, ; Zhu et al, , ; Zhang et al, ). However, the SB method aggregated all uncertainties into a single term (residual error), which makes it be sensitive to measurement noise (Clark, ).…”