Understanding the DayCent model: Calibration, sensitivity, and identifiability through inverse modeling

Abstract

a b s t r a c tThe ability of biogeochemical ecosystem models to represent agro-ecosystems depends on their correct integration with field observations. We report simultaneous calibration of 67 DayCent model parameters using multiple observation types through inverse modeling using the PEST parameter estimation software. Parameter estimation reduced the total sum of weighted squared residuals by 56% and improved model fit to crop productivity, soil carbon, volumetric soil water content, soil temperature, N 2 O, and soil NO 3 À compared to the default simulation. Inverse modeling substantially reduced predictive model error relative to the default model for all model predictions, except for soil NO 3 À and NH 4 þ . Post-processing analyses provided insights into parametereobservation relationships based on parameter correlations, sensitivity and identifiability. Inverse modeling tools are shown to be a powerful way to systematize and accelerate the process of biogeochemical model interrogation, improving our understanding of model function and the underlying ecosystem biogeochemical processes that they represent.

Underestimation of N2O emissions in a comparison of the DayCent, DNDC, and EPIC models

“…Using DayCent to simulate N 2 O emissions, soil temperature, and soil water from a corn–soybean rotation in Iowa, Necpálová et al. () reported similar values in calibration and validation for model index of agreement. Model performance in our simulations may be influenced by the agricultural contexts and sample sizes included in our data set.…”

“…Whereas Necpálová et al. () calibrated DayCent to a single crop rotation with a total sample size of n = 24 for each variable, we calibrated multiple crop rotations (Arlington) and multiple fertilizer treatments (Marshfield) simultaneously using sample sizes of n = 365 and 516 for each variable in calibration and validation, respectively. By calibrating with large sample sizes over multiple treatments, we intended to calibrate to static system properties (e.g., soil texture) and limit the influence of individual data points, which we hoped would improve model performance in validation.…”

“…Previously identified sensitive parameters (Necpálová et al. , Rafique et al. ), nitrification and denitrification parameters, soil physical properties, and manure nutrient content were included in the parameterization schema (Appendix : Tables S1 and S2).…”

Evaluating four nitrous oxide emission algorithms in response to N rate on an irrigated corn field

“…Recent DAYCENT model simulations of 6 cropland and 3 grassland sites in the United Kingdom showed that N 2 O emissions were sensitive to soil pH and clay content, particularly when the initial values for these model inputs were low (Fitton et al, 2014). Regarding internal model parameters, Necp alov a et al (2015) found that N 2 O emissions were sensitive to parameters controlling soil heat flux, the temperature effect on soil organic matter decomposition rates, and crop growth rates. Vogeler et al (2013) compared APSIM and DNDC models and found that soil temperature had a larger effect on nitrification in APSIM model, while soil water content had more effect on nitrification in the DNDC model.…”

“…Many process-based simulation models have been developed for understanding agricultural ecosystem carbon and nitrogen biochemical cycle and the response of N 2 O emissions to different agricultural managements (Chen et al, 2008;Pattey et al, 2007;Perlman et al, 2013;Necp alov a et al, 2015;Qin et al, 2013). Most commonly used models are APSIM (McCown et al, 1996), DNDC (Li et al, 1992), DAYCENT Del Grosso et al, 2000), Ecosys (Grant and Pattey, 1999), ExpertN (Engel and Priesack, 1993), FASSET (Olesen et al, 2002), NASA-Ames version of the CASA (Carnegie-Ames-Stanford approach) model (Potter et al, 1997), NOE (Henault et al, 2005), RZWQM and WNMM (Li et al, 2007).…”