研究目的
Investigating the improvement of evapotranspiration (ET) monitoring through the use of ensemble model averaging, specifically Bayesian Model Averaging (BMA), over a saltmarsh scrub area in South France using MODIS data.
研究成果
The ensemble model averaging, particularly the weighted average approach, improved ET estimates compared to simple averaging or individual models. The method demonstrated potential for integrating information from various models to reduce epistemic uncertainty in ET estimation.
研究不足
The study considered models strongly related in terms of basic principles, excluding models based on more physical descriptions of turbulent exchanges or transpiration processes. The ranking of models and weighting coefficients could vary with different training datasets.
1：Experimental Design and Method Selection
The study employed Bayesian Model Averaging (BMA) to weight each model according to their performances for deriving an ensemble average of ET. The methodology was applied to MODIS data over a saltmarsh scrub area.
2：Sample Selection and Data Sources
The study area was the Crau-Camargue pilot site in south-eastern France, characterized by highly contrasted wet and dry areas. Ground measurements of energy balance fluxes were acquired over a saltmarsh site using a 3D sonic anemometer, a net radiometer, and characterization of ground heat flux.
3：List of Experimental Equipment and Materials
3D sonic anemometer for measuring sensible heat flux, net radiometer for measuring net radiation, and equipment for ground heat flux characterization.
4：Experimental Procedures and Operational Workflow
Energy balance data were processed, including eddy covariance data and flux estimation uncertainties. Daily evapotranspiration was obtained by integrating latent heat flux data over the day.
5：Data Analysis Methods
The evaluation of the averaged models was performed by considering the weighting coefficient and the model ranking obtained over the training dataset. The performances were evaluated using mean absolute error.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
