Development of hybrid stochastic-mechanistic models for the partitioning and extrapolation of land surface to atmosphere water, energy and carbon fluxes
Final Report Abstract
Using support vector regression, we have related the optimized parameters of a light use efficiency model to climatic and biophysical site characteristics and evaluated them with a leave-one-out-cross-validation. Comparing the extrapolated parameters with the optimized ones at each site showed a good correlation with weighted R^2-values between 0.6 and 0.94. The fluxes modeled with these parameters correlated very well with the fluxes originating from the calibration model results (R^2-values > 0.91) but show biases of maximal 5.5 % difference per year with respect to the annual cumulative sum. The better correlation of the fluxes than that of the parameters is a consequence of the fact that a large part of Fo-dynamics is explained by PAR or APAR itself. The comparison of fluxes resulting from the model runs with the extrapolated parameters with the measured fσ-values revealed R^2-values in the range of 0.43 and 0.95 what is only marginally lower than the coefficients of determination of the calibration performance in the most cases. Due to observed biases between the modeled and measured data sets, however, the normalized RMSE-values differ much more. Nonetheless, the effective % differences of the cumulative fσ-sums per year do not exceed 5.5 %. This study showed the applied scheme to be suitable to extrapolate the optimized parameters of a light use efficiency model. The variations of the model parameters between the study sites could be recaptured in a cross-validation with reasonable precision in most cases. The resulting time series of carbon uptake modeled with the extrapolated parameters yielded in good correlations similar to the original model with the parameters optimized. However, a bias was in some cases introduced leading to deviations of the annual sum of assimilated carbon of 5.5 %. Amongst the attributes selected, those related to seasonality characteristics dominated. It is clear that - due to its high adaption capacities - SVR could also perform well with similar sets of features, but it is also obvious from the out-coming that SVR cannot extrapolate the parameters with arbitrarily attributes. This fact undermines the argument SVR is not suitable for such a task because it somehow can perform with any variable combination via the high dimensional feature space. The not arbitrarily appearing attribute selection also attests a certain biophysical meaning to the parameters of the light use efficiency model to some degree and not a purely empirical manner although the lag parameter has to be reconsidered. The exercise challenging the SVR model with an increase of the SVR parameter responsible for the generalization shows a certain capability of the method in this regard. But dearly, the number of sites is at its minimum in this study. So this study arouses curiosity about performing this exercise with a larger data set like the harmonized FLUXNET La Thuille synthesis data set with currently 253 eddy covariance measurement sites which is principally available for selected researchers.
Publications
- 2006. A semi-parametric gap-filling model for eddy covariance CO2 flux time series data. Global Change Biology 12(9): 1707-1716
Stauch, V.J., Jarvis, A. J.
(See online at https://doi.org/10.1111/j.1365-2486.2006.01227.x) - 2007. Comprehensive comparison of gap-filling techniques for eddy covariance net carbon fluxes. Agricultural and Forest Meteorology, 147(3-4): 209-232
Moffat, A. M., Papale, D., Reichstein, M., Hollinger, D. Y., Richardson, A. D., Barr, A. G., Beckstein, C., Braswell, B. H., Churkina, G., Desai, A. R., Falge, E., Govec, J. H., Heimann, M., Dafeng H., Jarvis, A. J., Kattge, J., Noormetsl, A., Stauch, V. J.
(See online at https://doi.org/10.1016/j.agrformet.2007.08.011) - 2008. Cross-site evaluation of eddy covariance GPP and RE decomposition techniques. Agricultural and Forest Meteorology, 148(1): 821-838
Desai, A. R., Richardson, A. D., Moffat, A. M., Kattge, J., Hollinger, D. Y., Barr, A., Falge, E., Noormets, A., Papale, D., Reichstein, M., Stauch, V. J.
(See online at https://dx.doi.org/10.1016/j.agrformet.2007.11.012) - 2008. Estimation of net carbon exchange using eddy covariance CO2 flux observations and a stochastic model. Journal of Geophysical Research 113(D03101)
Stauch, V. J., Jarvis, A. J., Schulz. K.
(See online at https://doi.org/10.1029/2007JD008603) - 2008. Statistical properties of random CO2 flux measurement uncertainty inferred from model residuals. Agricultural and Forest Meteorology, 148(1): 38-50
Richardson, A. D., Mahecha, M. D., Falge, E., Kattge, J., Moffat, A. M., Papale, D., Reichstein, M., Stauch, V.J., Braswell, B. H., Churkina, G., Kruijt, B., Hollinger, D. Y.
(See online at https://doi.org/10.1016/j.agrformet.2007.09.001) - 2010. Identification of a general light use efficiency model for gross primary production. Biogeosciences Discussions, 7: 7673-7726
Horn, J. E., Schulz, K.
(See online at https://doi.org/10.5194/bg-8-999-2011) - 2010. Post-processing analysis of MODIS LAI subsets. Journal of Applied Remote Sensing, 4(043557)
Horn, J. E., Schulz, K.
(See online at https://doi.org/10.1117/1.3524265)
