Gap-filling strategies for annual VOC flux data sets

Up to now the limited knowledge about the exchange of volatile organic compounds (VOCs) between the biosphere and the atmosphere is one of the factors which hinders more accurate climate predictions. Complete long-term flux data sets of several VOCs to quantify the annual exchange and validate recent VOC models are basically not available. In combination with long-term VOC flux measurements the application of gap-filling routines is inevitable in order to replace missing data and make an important step towards a better understanding of the VOC ecosystem-atmosphere exchange on longer time scales. We performed VOC flux measurements above a mountain meadow in Austria during two complete growing seasons (from snowmelt in spring to snow reestablishment in late autumn) and used this data set to test the performance of four different gap-filling routines, mean diurnal variation (MDV), mean gliding window (MGW), look up tables (LUT) and linear interpolation (LIP), in terms of their ability to replace missing flux data in order to obtain reliable VOC sums. According to our findings the MDV routine was outstanding with regard to the minimization of the gap-filling error for both years and all quantified VOCs. The other gap-filling routines, which performed gap-filling on 24 h average values, introduced considerably larger uncertainties. The error which was introduced by the application of the different filling routines increased linearly with the number of data gaps. Although average VOC fluxes measured during the winter period (complete snow coverage) were close to zero, these were highly variable and the filling of the winter period resulted in considerably higher uncertainties compared to the application of gap-filling during the measurement period. The annual patterns of the overall cumulative fluxes for the quantified VOCs showed a completely different behavior in 2009, which was an exceptional year due to the occurrence of a severe hailstorm, compared to 2011. Methanol was the compound which contributed with 381.5 mgCm^-2 and 449.9 mgCm^-2 most to the cumulative VOC carbon emissions in 2009 and 2011, respectively. In contrast to methanol emissions, however, considerable amounts of monoterpenes (-327.3 mgCm^-2) were deposited to the mountain meadow in consequence to the hailstorm in 2009. Other quantified VOCs had considerably lower influences on the annual patterns.