Model-data fusion approach to quantify evapotranspiration and net ecosystem exchange across the sagebrush ecosystem at different temporal resolutions

作者：Mitra Bhaskar^*; Mackay D Scott; Pendall Elise; Ewers Brent E; Kwon Hyojung; Cleary Meagan B; Naithani Kusum J

来源：Ecohydrology, 2018, 11(5): UNSP e1957.

DOI：10.1002/eco.1957

• ecosystem respiration (R-e)
• gross primary production (GPP)
• Markov chain Monte Carlo
• model-data fusion
• net ecosystem exchange (NEE)

Evapotranspiration (ET) and net ecosystem exchange (NEE) are driven by both high and slow frequency scalar fluxes. Quantifying the variation of these two processes at different timescales remains a challenge. Bridging this knowledge gap is crucial in order to improve insights of the impact of biotic and abiotic factors modulating these fluxes as well as for accurate estimation of gross primary productivity (GPP) and ecosystem respiration (R-e). This issue was addressed using a model-data fusion approach within a Bayesian framework by running the model against ET and NEE observations at three different time steps: subdaily (30min), daily (1day), and intermediate (7days). The model was tested against eddy covariance data collected for a 2-month period (June and July) from a sagebrush-steppe ecosystem in the United States. The 95% credible interval (CI) of fast processes such as transpiration and photosynthesis reduced by more than 90% compared with its a priori range when model was run at 30-min time step. The reduction in CI of the same parameters varied between 30% and 70% when the model was run at 1- or 7-day time step. The 95% CI of slow process such as root respiration reduced by 89% and 73% when model was run at 7-day and 30-min time step, respectively. We found strong confidence in predicting ET and NEE at subdaily timescale, whereas uncertainty increased with increase in temporal resolution. GPP and R-e varied strongly as the system transitioned from a traditionally wet (June) to a dry (July) month.

• 出版日期2018-7