Open AccessErrors in Modeling Carbon Turnover Induced by Temporal Temperature Aggregation
Author(s) -
Weihermüller L.,
Huisman J.A.,
Graf A.,
Herbst M.,
Vereecken H.
Publication year - 2011
Publication title -
vadose zone journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.036
H-Index - 81
ISSN - 1539-1663
DOI - 10.2136/vzj2009.0157
Subject(s) - environmental science , amplitude , range (aeronautics) , diurnal temperature variation , lead (geology) , discretization , approximation error , atmospheric sciences , observational error , mathematics , statistics , climatology , materials science , physics , geology , mathematical analysis , quantum mechanics , geomorphology , composite material
Modeling of C turnover is a common tool for the prediction of C stocks and CO 2 efflux. It is well recognized that the choice of the input data (e.g., C pool sizes, hydraulic parameters, atmospheric boundary conditions) determines the outcome of these prediction. Temperature is known to be one of the most important driving factors and it varies in a range of temporal scales. Typically, the time discretization of most models is flexible and can range from minutes to months. However, the implications of variable time discretization for predicted soil C turnover are seldom discussed. In this study, we demonstrated that averaging of input temperature data will lead to changes in predicted C turnover in terms of daily amplitude and the impact of extreme temperatures. The results indicate that averaging from hourly to daily or monthly temperatures will lead to relative errors >4% yr −1 for cumulative CO 2 efflux. Instantaneous CO 2 fluxes are even more affected, where daily and monthly averaging will lead to estimation errors exceeding 20%. We also show that a constant or daily variable temperature amplitude for rescaling daily average temperature did not decrease the error when using daily or monthly mean temperature instead of hourly data. Therefore, instantaneous fluxes are only accurately predicted when hourly temperature input is used. For long‐term modeling (e.g., years to centuries), the relative error in cumulative efflux, and therefore in C stocks loss, is reasonably low (∼4–5% annual error) but will accumulate with time again.