Gap filling strategies for defensible annual sums of net ecosystem exchange

Heightened awareness of global change issues within both science and political communities has increased interest in using the global network of eddy covariance flux towers to more fully understand the impacts of natural and anthropogenic phenomena on the global carbon balance. Comparisons of net ecosystem exchange (F-NEE) responses are being made among biome types, phenology patterns, and stress conditions. The comparisons are usually performed on annual sums of F-NEE; however, the average data coverage during a year is only 65%. Therefore, robust and consistent gap filling methods are required. We review several methods of gap filling and apply them to data sets available from the EUROFLUX and AmeriFlux databases. The methods are based on mean diurnal variation (MDV), look-up tables LookUp, and nonlinear regressions (Regr.), and the impact of different gap filling methods on the annual sum of FNEE is investigated. The difference between annual F-NEE filled by MDV compared to F-NEE filled by Regr. ranged from -45 to +200 g C m(-2) per year (MDV-Regr.). Comparing LookUp and Regr. methods resulted in a difference (LookUp-Regr.) ranging from -30 to +150g Cm-2 per year. We also investigated the impact of replacing measurements at night, when turbulent mixing is insufficient. The nighttime correction for low friction velocities (u(*)) shifted annual F-NEE on average by +77 g C m(-2) per year, but in certain cases as much as +185 g C m-2 per year. Our results emphasize the need to standardize gap filling-methods for improving the comparability of flux data products from regional and global flux networks. (C) 2001 Elsevier Science B.V. All rights reserved.