A Mechanistic Model to Fill the Data Gaps in Water Vapor Flux

Due to system/sensor malfunctions or certain natural circumstances such as thunderstorms, data gaps are unavoidable problems in long term measurements for all FLUXNET research sites, which will result in difficulties in estimating annual net ecosystem exchange (NEE), sensible heat flux (H) and latent heat flux (LE), and bias in the relationships among them as well. To fill the gaps in long term measurements, various gap-filling strategies had been developed and applied at different sites, for instance, statistical methods, mechanistic models and neural networks. Here a mechanistic model based on the re-arranged Penman- Monteith equation was developed with the aim to fill the gaps in hourly water vapor flux measured at Morgan Monroe State Forest site (MMSF), Indiana. This mechanistic model includes two sub-models, one is for the aerodynamic conductance corrected by atmospheric stability, and the other one is for the canopy conductance. The stability functions derived from the data at MMSF during the study period are found to be similar to the classic Kansas results, but with more scatter, possibly because measurements at MMSF were taken in the roughness sublayer instead of inertial sublayer. The canopy conductances for MMSF were found to be dependent on both meteorological and physiological factors. Among the controlling factors examined, the best agreement was found between canopy conductance and the assimilation rate of CO2, therefore the sub model to compute canopy conductance is constructed with the assimilation rate as predictor. In addition, Stewart's parametric model is found to perform well for the estimation of canopy conductance of our forest. Overall, by comparing the modeled with measured values, this mechanistic model from a re-arranged Penman-Monteith equation performs well with R-square value of 0.827, but the plot becomes scattered as LE increases.