Use of inverse modeling to evaluate CENTURY-predictions for soil carbon sequestration in US rain-fed corn production systems.

Hoyoung Kwon,Carmen M Ugarte,Stephen M Ogle,Stephen A Williams,Michelle M Wander,Lucas C.R Silva

doi:10.1371/journal.pone.0172861

Abstract

We evaluated the accuracy and precision of the CENTURY soil organic matter model for predicting soil organic carbon (SOC) sequestration under rainfed corn-based cropping systems in the US. This was achieved by inversely modeling long-term SOC data obtained from 10 experimental sites where corn, soybean, or wheat were grown with a range of tillage, fertilization, and organic matter additions. Inverse modeling was accomplished using a surrogate model for CENTURY’s SOC dynamics sub-model wherein mass balance and decomposition kinetics equations from CENTURY are coded and solved by using a nonlinear regression routine of a standard statistical software package. With this approach we generated statistics of CENTURY parameters that are associated with the effects of N fertilization and organic amendment on SOC decay, which are not as well quantified as those of tillage, and initial status of SOC. The results showed that the fit between simulated and observed SOC prior to inverse modeling (R2 = 0.41) can be improved to R2 = 0.84 mainly by increasing the rate of SOC decay up to 1.5 fold for the year in which N fertilizer application rates are over 200 kg N ha-1. We also observed positive relationships between C inputs and the rate of SOC decay, indicating that the structure of CENTURY, and therefore model accuracy, could be improved by representing SOC decay as Michaelis-Menten kinetics rather than first-order kinetics. Finally, calibration of initial status of SOC against observed levels allowed us to account for site history, confirming that values should be adjusted to account for soil condition during model initialization. Future research should apply this inverse modeling approach to explore how C input rates and N abundance interact to alter SOC decay rates using C inputs made in various forms over a wider range of rates.

Highlights

The CENTURY soil organic matter (SOM) model [1] is an agro-ecosystem model developed to simulate the dynamics of multiple SOM “compartments” or “pools”, which differ in their size and degree of physical and/or chemical stabilization, under various agronomic practices and soil/climatic conditions (Fig 1)
We evaluated the fit between observed soil organic carbon (SOC) at experimental sites and CENTURY-modeled SOC from the dataset and output files compiled by Ogle et al [11] (Fig 3A), which revealed a less than optimal fit (Fig 3A; R2 = 0.41, P
Drawing on data from multiple sites, we were able to generate robust parameter estimates using climatic factors and C input rates derived from CENTURY outputs simulated by Ogle et al [11] when they were combined with yield-based estimates of C inputs

Summary

Introduction

The CENTURY soil organic matter (SOM) model [1] is an agro-ecosystem model developed to simulate the dynamics of multiple SOM “compartments” or “pools”, which differ in their size and degree of physical and/or chemical stabilization, under various agronomic practices and soil/climatic conditions (Fig 1).The model has been widely used to guide soil-based nutrient management, mitigate agricultural non-point source pollution, and promote soil organic carbon (SOC) sequestration. Accuracy and precision of predictions are generally improved for sitespecific application with model calibration wherein the model’s initial conditions and parameter values are adjusted for a site or region against observed data. CENTURY calibration typically includes adjustment of i) the initial distribution of SOC among three SOM pools (active, slow, and passive) and ii) one or more of the parameters related with decay rates of those pools [6, 7]. This type of calibration usually involves manual adjustments, which result in variable success and uncertainties in model prediction that partly depend on the modeler’s experience [8]

Methods

Results

Conclusion