Submitted to: USDA Greenhouse Gas Symposium
Publication Type: Abstract Only
Publication Acceptance Date: 3/24/2005
Publication Date: 3/25/2005
Citation: Parkin, T.B., Kaspar, T.C. 2005. Filling data gaps in soil respiration measurements using autocorrelation. USDA Greenhouse Gas Symposium. http://soilcarboncenter.k-state.edu/conference/Abstract_Pages.htm Interpretive Summary:
Technical Abstract: Field respiration measurements are commonly performed using chambers placed on the soil surface at periodic intervals. Calculation of cumulative carbon dioxide (CO2) flux over time is then estimated by linear interpolation between measurement points. Because soil CO2 fluxes often exhibit pulses following rainfall events or other pertubations (i.e. tillage), measurements at infrequent intervals may fail to adequately characterize the temporal flux dynamics. If this occurs biased estimates of cumulative CO2 loss may be obtained. This paper explores the use of autocorrelation analysis to improve interpolation between measurement points, and thus, improved estimates of cumulative CO2 flux from soil respiration. An automated chambers was used to measure soil CO2 fluxes at hourly intervals from a fallow soil from April 16 through Sept. 5, 2001. All the hourly measurements were then used to compute cumulative CO2 flux from the site. This value was used as the best estimate of cumulative CO2 flux. Two interpolation techniques (linear interpolation and autocorrelation analysis) were then tested with regard to how well they provided estimates of cumulative CO2 flux relative to the best estimate. In this analysis the population of hourly chamber fluxes was subsampled by selecting individual hourly flux measurments at intervals ranging from 1 d to 20 d. The two interpolation techniques were then applied and a cumulative flux calculated. We observed that there was no difference in the two interpolation techniques when sampling interval was 4 d or less. However, as sampling interval was increased beyond 4 d the variance associated with estimates obtained by linear interpolation increased, whereas the variance associated with estimates obtained by autocorrelation were substantially less and remained relatively constant. Additional evaluations are being conducted to refine the autocorrelation technique.