Gap filling strategies and error in estimating annual soil respiration

Soil respiration (R_soil) is one of the largest CO₂ fluxes in the global carbon (C) cycle. Estimation of annual R_soil requires extrapolation of survey measurements or gap filling of automated records to produce a complete time series. Although many gap filling methodologies have been employed, there is no standardized procedure for producing defensible estimates of annual R_soil. Here, we test the reliability of nine different gap filling techniques by inserting artificial gaps into 20 automated R_soil records and comparing gap filling R_soil estimates of each technique to measured values. We show that although the most commonly used techniques do not, on average, produce large systematic biases, gap filling accuracy may be significantly improved through application of the most reliable methods. All methods performed best at lower gap fractions and had relatively high, systematic errors for simulated survey measurements. Overall, the most accurate technique estimated R_soil based on the soil temperature dependence of R_soil by assuming constant temperature sensitivity and linearly interpolating reference respiration (R_soil at 10 °C) across gaps. The linear interpolation method was the second best-performing method. In contrast, estimating R_soil based on a single annual R_soil - T_soil relationship, which is currently the most commonly used technique, was among the most poorly-performing methods. Thus, our analysis demonstrates that gap filling accuracy may be improved substantially without sacrificing computational simplicity. Improved and standardized techniques for estimation of annual R_soil will be valuable for understanding the role of R_soil in the global C cycle.