A multisite analysis of temporal random errors in soil CO₂ efflux

Abstract

An important component of the terrestrial carbon balance is the efflux of CO 2 from soils to the atmosphere, which is strongly influenced by changes in soil moisture and temperature. Continuous measurements of soil CO 2 efflux are available around the world, and there is a need to develop and improve analyses to better quantify the precision of the measurements. We focused on random errors in measurements, which are caused by unknown and unpredictable changes such as fluctuating environmental conditions. We used the CO 2 gradient flux method with two different algorithms to study the temporal variation of soil CO 2 efflux and associated random errors at four different ecosystems with wide ranges in mean annual temperature, soil moisture, and soil CO 2 efflux. Our results show that random errors were better explained by a double-exponential distribution, had a mean value close to zero, were nonheteroscedastic, and were independent of soil moisture conditions. Random errors increased with the magnitude of soil CO 2 efflux and scale isometrically (scaling exponent ≈ 1) within and across all sites, with a single relation common to all data. This isometric scaling is unaffected by ecosystem type, soil moisture conditions, and soil CO 2 efflux range (maximum and minimum values within an ecosystem). These results suggest larger uncertainty under extreme events that increase soil CO 2 efflux rates. The accumulated annual uncertainty due to random errors varied between ±0.38 and ±2.39%. These results provide insights on the scalability of the sensitivity of soil CO 2 efflux to changing weather conditions across ecosystems.

Large Uncertainty in Estimating pCO2 From Carbonate Equilibria in Lakes

“…Although the tLocat distribution provided a better fit over Gaussian distribution in ALK and DIC observation within M ALK to H ALK groups, neither of these two distributions nor any other probability density functions sufficiently characterized error distribution in some ALK groups (i.e., pH in M ALK ; Figure S1 and Table ). Our results agree with CO 2 ‐related studies showing non‐Gaussian distributions of random errors (Ciais et al, ; Cueva et al, ; Richardson & Hollinger, ) and imply that errors derived from normal distribution will underestimate both small and large random errors. Furthermore, the variety of error distributions limits the use of statistical and modeling techniques (i.e., assuming normal distribution) in characterizing random parameter uncertainties and propagating them onto p CO 2 derived from carbonate equilibria.…”

The value of soil respiration measurements for interpreting and modeling terrestrial carbon cycling

“…Chamber system measurement error 30 % Pumpanen et al 2004Random error 20 % of flux magnitude (heteroskedastic) Lavoie et al 2014, Savage et al (2008), Cueva et al (2015) Wind advection 233 % in unvented chambers Bain et al (2005) Missed subniveal fluxes 35-28 % Monson et al (2006) Soil collar damage to fine roots 15 % Heinemeyer et al 2011Scaling up to tower footprint (spatial variability)…”

“…Hashimoto 2012), instantaneous R S measures from autochambers and survey campaigns have rarely been synthesized (Bahn et al 2010;Vargas et al 2010a;Lavoie et al 2014;Cueva et al 2015), and are not easily available to modelers because they are not in organized data repositories (see Vargas and Leon 2015 for an exception). These small but geographically widespread datasets are generally in the hands of individual investigators, and are part of what Dietze et al (2013) refers to as the 'long tail' of data.…”

Soil CO2concentrations and efflux dynamics of a tree island in the Pantanal wetland

“…Random errors in the values of F 0 are expected in cases of large changes in environmental conditions [Cueva et al, 2015], as seen during the rapid saturation and drying of the soil in the February-March 2014 period, which we believe affected our model's steady state assumption. For every 30 min estimate of F 0 , we have assumed a constant CO 2 flux between our two selected depths (as F 0.05 and F 0.20 in equation (9)), which might not hold all the time.…”

“…The largest systematic errors in F 0 came from the uncertainty in soil CO 2 concentrations of ± (3000 ppm + 2% reading), especially for soil CO 2 values closer to atmospheric CO 2 concentrations. Similar sensors with smaller concentration measurement ranges (e.g., 0-10,000 ppm) have been used widely in research using the gradient methods [Cueva et al, 2015;Vargas et al, 2010;Baldocchi et al, 2006;Jassal et al, 2005;Tang et al, 2003] with an accuracy of ± (20 ppm + 2% reading). Future research in the Pantanal should consider additional sensors in the topsoil able to measure soil CO 2 concentration at similar ranges and with greater accuracy.…”