Optimizing duration of incubation experiments for understanding soil carbon decomposition

Abstract

Laboratory incubation is a commonly used method to measure the decomposition of soil organic carbon (SOC). While incubation experiments are conducted across a wide range of durations that may vary from hours to years, no method is available to determine an optimal duration of the incubation experiment so that SOC decomposition can be best understood. Here we presented a novel approach to determine the optimal duration called OPtimal Incubation Duration (OPID). The OPID approach quantifies information gained from an ongoing incubation experiment and determines the time point when SOC decomposition rates can be well quantified. Statistically, the OPID approach is based on a progressive data assimilation algorithm that iteratively assimilates data from an ongoing incubation experiment into a three-pool first-order SOC decomposition model. Using a published incubation data set under different temperatures as a case study, we first generated synthetic daily data, and then fed the data into the three-pool model iteratively to observe the changes of model performance. We found that the accuracy of model projections increased with incubation period and exhibited a trade-off between initial model performance and the time towards accurate projection among different temperatures of incubation. The optimal incubation duration was 347, 212, and 126 days under incubation temperatures of 15 °C, 25 °C and 35 °C, respectively. Comparing the parameters with which from the synthetic daily data, if the incubation period was shorter than the optimal durations, then the decomposition rate of the fast-turnover pool was underestimated and those of the slow pools were overestimated. Sensitivity analysis indicated that optimal incubation duration was negatively correlated with proportion of slow-turnover carbon pools, turnover rates, and initial carbon content, respectively. Our study suggested that long-term incubation experiments are necessary for capturing the dynamics of slow-turnover carbon pools. However, the additional data may not be helpful for model performance if the incubation duration is longer than the optimum. Our study provides a tool for soil scientists to design more effective incubation experiments.DA: data assimilation; PDA: progressive data assimilation; Cum CO2: cumulative CO2 emission from soil; opt: optimal duration;