PARENT SESSION
Poster Session # 13: Biogeochemistry, Photosynthesis, and Respiration.
Tuesday, August 5 Presentation from 5:00 PM to 6:30 PM. SITCC Exhibit Hall B.
Modeling net ecosystem exchange using auto-regression model with exogenous inputs.
Xu, Tao*,1, Hui, Dafeng2, Luo, Yiqi2, White, Luther*,1, 1 Department of Mathematics,601 Elm Avenue ,University of Oklahoma, ,Norman, ,Oklahoma, ,USA2 Department of Botany and Microbiology,University of Oklahoma,770 Van Vleet Oval, ,Norman, ,Oklahoma, ,USA
ABSTRACT- Improving the prediction accuracy of net ecosystem exchange (NEE) in terrestrial ecosystems is an important task in ecosystem modeling. Considering that the present time NEE not only depends on the current and historical climatic variables, but also correlates with its own past history, we proposed an approach by using a time series analysis model. The model can be best described by the ARX (auto-regression with exogenous inputs) model in the form of A(q)y(t)=B(q)u(t)+e(t), where y(t) is output NEE, u(t) is multiple-input including photosynthetically active radiation, air temperature, relative humidity, vapor pressure deficit etc. and e(t) is model error. Weekly averaged NEE and climatic variables from year 1992 to year 1999 in Harvard Forest were used in analysis. Compared to multivariate linear regression model, ARX model greatly improved the NEE estimation. The variation in NEE explained by the ARX model increased on the average to 65% from 52.3% given by multiple regression when the auto-regressor had time delay from 1 to 20, and that was increased to more than 71% on the average if the auto-regressor had time delay ranging from 45 to 55. This indicates that the NEE pattern of the present season, while being influenced by the current environmental factors, is either closely auto-correlated with its most recent history or, at another extreme, closely auto-correlated with the patterns of the same season of the previous year. While the former has an obvious explanation, the later strongly suggests that it is possible to use historical patterns of NEE of a specific site, coupled with the current/predicted environmental factors, to generate more accurate current/predicted NEE at that specific site.
Key words: Time Series, ARX Model, NEE Modeling