A marked pulse in annual gross ecosystem productivity detected by E-C measurements in Indiana: is it real or due to a bug?

We have collected eddy-covariance (E-C) based fluxes of CO₂, water vapor, and energy balance components at the AmeriFlux site in the Morgan-Monroe State Forest (MMSF) in Indiana (USA) since 1998. For compilations of annual estimates of net ecosystem production (NEP), and its components, gross ecosystem production (GEP) and ecosystem respiration (RE), we have used standard methods for quality control and gap-filling. In the years 1999-2003 we obtained values of NEP around 330 g C m^-2 (± 10 %), contributed by GEP of 1270 g C m^-2 (± 2 %) and RE of 940 g C m^-2 (± 4 %). Small variations were mostly due to RE and can be linked to climatic controls. It thus came as a surprise that the NEP estimates for 2004 showed an increase of nearly 220 g C m^-2 (contributed by a large jump in GEP and a smaller reduction in RE) over the average of the preceding years. NEP results for 2005 are still large, but less than in 2004. Interestingly, water vapor fluxes for 2004 show no detectable difference to the preceeding years. In this paper we present a detailed uncertainty analysis, explore the structure and distribution of this pulse in NEP in relation to various potential causes and come to a surprising conclusion.