Estimating cropland carbon fluxes: A process-based model evaluation at a Swiss crop-rotation site

Abstract

In support of climate change mitigation strategies and food security, there is an ever-increasing requirement to understand the key drivers of cropland yields and carbon (C) budgets. Process-based crop models, driven by meteorological observations, can provide realistic simulations and diagnosis of the variability in crop C budgets. However, the field-scale observations required to validate the models, particularly across multiple growing seasons and crop types, are seldom available.This research evaluates the Soil-Plant-Atmosphere Crop (SPA-Crop) model for estimating the C fluxes of multiple crop types and seasons at the Swiss long-term flux site Oensingen (CH-OE2). SPA-Crop is a detailed parametric model, simulating ecosystem photosynthesis and water balance at fine temporal (half-hourly time-steps) and vertical scales (multiple canopy and soil layers). We initialised the model at the point-scale using management information and meteorological observations recorded during ten crop seasons from 2004 to 2015. SPA-Crop has been developed for simulating winter cereals (wheat and barley), we further evaluated the model performance for these crops at CH-OE2 and extended the simulation to include a crop-specific calibration for winter rapeseed and pea crops. Model estimates of net ecosystem exchange (NEE) of carbon dioxide (CO2), ecosystem respiration (Reco), and gross primary production (GPP) were evaluated based on gap-filled and partitioned eddy covariance (EC) fluxes. Further comparisons were made between modelled and observed yields.The daily SPA-Crop estimates had a high agreement with the EC measurements across the different crop types with a mean R2 of the NEE of 0.74 (wheat), 0.73 (barley), 0.71 (rapeseed) and 0.69 (pea). The modelled yield estimates had biases ranging from −22% (barley, measured-model difference −70 gC m–2; 0.31 t ha−1) to +56% (rapeseed, measured-model difference −85 gC m–2; −0.38 t ha−1), with the largest bias of wheat being only −13% (measured-model difference 34 gC m–2; 0.15 t ha−1) across four wheat seasons. We recommend future SPA-Crop developments – particularly with regards to improving the precision of Reco estimates. Our research demonstrates that SPA-Crop is a reliable tool for simulating the C budget and yield across the four crop types: winter wheat, winter barley, winter rapeseed, and peas.