Effects of different gap filling methods and land surface energy balance closure on annual net ecosystem exchange in a semiarid area of China

Based on eddy covariance measurements over two kinds of land surfaces (a degraded grassland and a maize cropland) in a semiarid area of China in 2005 and 2008, the effects of different gap filling methods, energy balance closure and friction velocity threshold (u*) on annual net ecosystem exchange (NEE) were analyzed. Six gap filling methods, including mean diurnal variation (MDV), marginal distribution sampling (MDS), and nonlinear regressions method, were investigated by generating secondary datasets with four different artificial gap lengths (ranging in length from single half-hours to 12 consecutive days). The MDS generally showed a good overall performance especially for long gaps, with an annual sum bias error less than 5 g C m−2 yr−1. There was a large positive annual sum bias error for nonlinear regressions, indicating an overestimate on net ecosystem respiration. The offset in the annual sum NEE for four nonlinear regressions was from 8.0 to 30.8 g C m−2 yr−1. As soil water content was a limiting factor in the semiarid area, the nonlinear regressions considering both soil temperature and soil water content as controlling variables had a better performance than others. The performance of MDV was better in daytime than in nighttime, with an annual sum bias error falling between −2.6 and −13.4 g C m−2 yr−1. Overall, the accuracy of the gap filling method was dependent on the type of the land surface, gap length, and the time of day when the data gap occurred. The energy balance ratio for the two ecosystems was nearly 80%. Turbulent intensity had a large impact on energy balance ratio. Low energy balance ratio was observed under low friction velocity during the night. When there was a large fetch distance in a wind direction, a low energy balance ratio was caused by mismatch of the footprints between the available energy and turbulent fluxes. The effect of energy balance correction on CO2 flux was evaluated by assuming the imbalance caused by the underestimation of sensible heat flux and latent heat flux. The results showed an average increase of 10 g C m−2 yr−1 for annual NEE in both ecosystems with an energy balance correction. On the other hand, the u* threshold also have a large impact on annual sum NEE. Net carbon emission increased 37.5 g C m−2 yr−1 as u* threshold increased from 0.1 to 0.2 m s−1, indicating a large impact of imposing u* threshold on net ecosystem carbon exchange.