Optimization and evaluation of the ANTHRO-BGC model for winter crops in Europe
Highlights
► The BIOME-BGC model was updated for agricultural ecosystems (named as ANTHRO-BGC). ► Eddy flux measurements were used as constraints to optimize ANTHRO-BGC model using optimization algorithm. ► The spatially generalized ecophysiological parameters of ANTHRO-BGC model were identified for different crops. ► The performance of optimized ANTHRO-BGC model was similar to other ecosystem models. ► Sensitivity and uncertainty analysis were adopted to explore reliable parameters for ANTHRO-BGC.
Introduction
Climate change and agriculture are interrelated processes, both of which take place on global scale. Weather and climate characteristics such as temperature, precipitation, carbon dioxide and water availability as well as the tropospheric carbon dioxide concentration impact the plant growth directly and also indirectly via the influence of insects, plant diseases, and weeds on agricultural production (Hatfield et al., 2008, Iglesias et al., 2009, Gornall et al., 2010). The prediction of responses of plant growth towards the changing climate is prone to uncertainties and it is sometimes difficult to separate climate related effects from other, e.g. management related factors (c.f. Hatfield et al., 2008). However, simple indicators such as maximum temperature, minimum temperature and precipitation can explain 30% or more of the year-to-year variations in global average yields for the six most grown crop species (Lobell and Field, 2007) and are also known to explain the variations of wheat yields on the continental scale in Australian (Nicholls, 1997). In addition, the increase in the atmospheric CO2 to 550 ppm might on average increase C3 crop yields by 10–20% and C4 crop yields by 0–10% (Long et al., 2004, Ainsworth and Long, 2005). These conditions determine the carrying capacity of the biosphere to produce sufficient food for the human population and domesticated animals (Vandermeer et al., 1998). Assessments of the effects of global climate changes on agriculture might help to properly anticipate and adapt farming to maximize agricultural production (Monzon et al., 2007). The role of agricultural ecosystems for the global climate with respect to greenhouse gas emission needs clarification (Bondeau et al., 2007). Numerical models are good supplement tools to simulate the interactivity and feedbacks between climate and agricultural ecosystems (Osborne et al., 2007).
Numerous process-based crop models have been developed during the last few decades and were applied from the plant and plot level up to regional scales as follows: CERES (Tubiello et al., 1995, Hasegawa et al., 2000), WOFOST (van Diepen et al., 1989), STICS (Brisson et al., 1998, Hebert et al., 2005), Sirius model (Jamieson et al., 1998, Lawless et al., 2005), ARCWHEAT model (Travis et al., 1988, Lawless et al., 2005). These models represented the detailed processes of crop growth. The lack of detailed soil data, management information, and crop parameters, however, hindered the application of such detailed models on large spatial scale.
Several ecosystem models such as LPJmL (Bondeau et al., 2007), Agro-IBIS (Kucharik and Brye, 2003), ORCHIDEE-STICS (Smith et al., 2010b) and SiBcrop (Lokupitiya et al., 2009) were recently extended to simulate the carbon and water exchange between atmosphere and agricultural ecosystems on large scales. Parameters used in these models were either measured in experiments or taken from literature and they have been applied to explore the contribution of agricultural ecosystems to climate change (Bondeau et al., 2007, Lokupitiya et al., 2009) and to estimate yields of crops (Kucharik, 2003, Kucharik and Brye, 2003).
The BIOME-BGC model has been widely used to study natural ecosystems with emphases on parameterization and sensitivity analyses for forest ecosystems and grasslands (White et al., 2000, Tatarinov and Cienciala, 2006, Chiesi et al., 2007, Haszpra et al., 2011). These parameterization and sensitivity analyses were based on literature review. Trusilova et al. (2008) used a Bayesian parameter estimation technique to improve the predictive ability of the BIOME-BGC model for forests using eddy flux measurements as constraints. There are only two studies that we are aware of, which used the BIOME-BGC model to simulate water and carbon exchanges between the atmosphere and agricultural ecosystems. Wang et al. (2005) simulated water and carbon fluxes over crops in China. A recent study updated the BIOME-BGC model (Agro-BGC) to investigate the carbon, nitrogen and water balance of agricultural ecosystems and grasslands (Di Vittorio et al., 2010). The five unavailable parameters for switch grass (C4 perennial grass) were mathematical optimized using observational data from multiple sites (Di Vittorio et al., 2010).
With the model-data fusion methods, eddy covariance measurements of CO2 and H2O fluxes have been used to improve the performance of several process-based ecosystem models (Knorr and Kattge, 2005, Sacks et al., 2006, Mo et al., 2008, Williams et al., 2009). The data provided by eddy covariance measurements can help to better understand the processes driving the carbon and water exchange between atmosphere and biosphere. The multiple-year eddy covariance measurements provided good constraints on photosynthetic parameters (Braswell et al., 2005). It, however, provides relatively poor constraints on parameters related to soil decomposition that vary at considerably longer time scales than canopy photosynthesis and transpiration in forest ecosystems (Braswell et al., 2005). Four parameters of the model at maximum could be determined independently from the eddy covariance measurements, and the mean of the maximum potential electron transport rate of all leaves within the canopy was best determined by the measurements of net CO2 flux (Wang et al., 2001). Only recently, eddy covariance measurements data for agricultural ecosystems became available in Europe (Moors et al., 2010). This provides the possibility to develop and optimize the model parameters for these ecosystems using eddy flux measurements as constraints.
The aim of this study was to identify the spatially generalized ecophysiological parameters of the updated BIOME-BGC (ANTHRO-BGC) model for western and central European croplands by using eddy flux measurements and mathematical methods. The following questions were addressed: (1) Which parameter of ANTHRO-BGC model has relatively most important effects on model simulation? (2) Which observed variable or combination of them provide the best constraint on the parameters of ANTHRO-BGC model? (3) How many eddy flux measurements samples from different sites are enough to constrain the parameters of large scale ecosystem model? (4) Is the range of the parameters derived from literature reliable? (5) Is the estimation of optimized ANTHRO-BGC model on multiple sites and years trustable?
In our study, the global sensitivity analysis (SA) was used to detect the most sensitive parameters of the ANTHRO-BGC. The eddy flux measurements gross primary productivity (GPP), net ecosystem exchange (NEE) and evapotranspiration (ET) were used as constraint to optimize the parameters of ANTHRO-BGC model by using a mathematical optimization algorithm. Uncertainty analysis was adopted to explore the level of confidence of the estimates of optimized ANTHRO-BGC model on multiple site-year.
Section snippets
Model testing sites
The sensitivity analysis, parameter optimization and uncertainty analysis of the ANTHRO-BGC model was based on the daily observation data from eddy covariance measurements of the croplands in Europe (http://gaia.agraria.unitus.it/DATABASE/carboeuropeip/mustlogin.aspx) (Table 1, Fig. 1). Crop rotation and fertilizer management were applied on most of these sites during the carbon flux observation period (Moors et al., 2010, Osborne et al., 2010). Annual mean air temperature and precipitation of
Sensitivity analysis
The model had similar responses to the change of the target parameters for wheat, barley, and rape (Table 3). The model was sensitive to the five parameters: canopy light extinction coefficient (LT_ET), fraction of leaf nitrogen in Rubisco (FLRN), carbon allocation ratio of new fine root to new leaf (FR_to_LF), carbon allocation ratio of new fruits to new leaf (FT_to_LF) and canopy specific leaf area (SP_LAI).
A higher LT_ET indicates that the plant has a higher light absorbing ability (
Conclusions
In this study the most important parameters of the updated ANTHRO-BGC model were detected through SA. The spatially generalized ecophysiological parameters of the ANTHRO-BGC model were identified for wheat, barley, and rape. The simulation ability of the ANTHRO-BGC model was significantly improved in comparison with BIOME-BGC because a new validated phenology model for crops was implemented and the optimal value of the most import parameters of the ANTHRO-BGC model were identified using a
Financial support
A Ph.D. scholarship is provided to SM by the Max-Planck Society (MPG) and the Chinese Academy of Sciences (CAS) through a joint doctoral program and Leibniz-Centre for Agricultural Landscape Research (ZALF). AG acknowledges financial support by the German Research Council (DFG).
Acknowledgements
We thank Bernard Heinesch, Corinna Rebmann, Ceschia Eric, Christian Bernhofer, Enzo Magliulo, Eric Larmanou, Laffineur Quentin, Marc Aubinet, Nina Buchmann, Olivier Zurfluh, Pierre Béziat, Pierre Cellier, Paul di Tommasi, Werner Eugster, Werner Kutsch, and Thomas Grünwald for sharing their measurement data. We thank two anonymous reviewers for constructive comments. We thank Xenia Specka for discussing the sensitivity and uncertainty analysis.
References (87)
- et al.
Effects of soil water deficit at different growth stages on rice growth and yield under upland conditions. 2. Phenology, biomass production and yield
Field Crops Res.
(1996) - et al.
Application of BIOME-BGC to simulate Mediterranean forest processes
Ecol. Model.
(2007) - et al.
Impact of weather on yield components of winter rye over 30 years
Agric. Forest Meteorol.
(2000) - et al.
Development and optimization of an Agro-BGC ecosystem model for C4 perennial grasses
Ecol. Model.
(2010) - et al.
The REFLEX project: comparing different algorithms and implementations for the inversion of a terrestrial ecosystem model against eddy covariance data
Agric. Forest Meteorol.
(2009) - et al.
Testing CERES model predictions of crop growth and N dynamics, in cropping systems with leguminous green manures in a Mediterranean climate
Field Crops Res.
(2000) - et al.
Spatiotemporal image correlation spectroscopy (STICS) theory, verification, and application to protein velocity mapping in living CHO cells
Biophys. J.
(2005) - et al.
Carbon budget of mature no-till ecosystem in North Central Region of the United States
Agric. Forest Meteorol.
(2005) - et al.
Hydrological model uncertainty assessment in southern Africa
J. Hydrol.
(2010) - et al.
Parameter estimation and uncertainty analysis of a large-scale crop model for paddy rice: application of a Bayesian approach
Agric. Forest Meteorol.
(2009)
Interaction of temperature and daylength on leaf appearance rate in wheat and barley
Agric. Forest Meteorol.
Making sense of wheat development: a critique of methodology
Field Crops Res.
Effects of temperature and photoperiod on phenology as a guide to the selection of annual legume cover and green manure crops for hillside farming systems
Field Crops Res.
Maximum conductances for evaporation from global vegetation types
Agric. Forest Meteorol.
A wheat canopy model linking leaf area and phenology
Eur. J. Agron.
Growing degree-days: one equation, two interpretations
Agric. Forest Meteorol.
Multiple sources of predictive uncertainty in modeled estimates of net ecosystem CO2 exchange
Ecol. Model.
Modelling management strategies for wheat–soybean double crops in the south-eastern Pampas
Field Crops Res.
Variability in carbon exchange of European croplands
Agric. Ecosyst. Environ.
River flow forecasting through conceptual models. Part I—A discussion of principles
J. Hydrol.
Key questions and uncertainties associated with the assessment of the cropland greenhouse gas balance
Agric. Ecosyst. Environ.
A general model of forest ecosystem processes for regional applications. I. Hydrologic balance, canopy gas exchange and primary production processes
Ecol. Model.
SPATULA links daytime temperature and plant growth rate
Curr. Biol.
Measurements necessary for assessing the net ecosystem carbon budget of croplands
Agric. Ecosyst. Environ.
Application of BIOME-BGC model to managed forests: 1. Sensitivity analysis
For. Ecol. Manage.
Modeling and measuring the effects of disturbance history and climate on carbon and water budgets in evergreen needleleaf forests
Agric. Forest Meteorol.
Modelling the timing of the early development of winter wheat
Agric. Forest Meteorol.
Interactions of CO2, temperature and management practices: simulations with a modified version of CERES-Wheat
Agric. Syst.
Definition and calculation of uncertainty in ecological process models
Ecol. Model.
Global change and multi-species agroecosystems: concepts and issues
Agric. Ecosyst. Environ.
Annual carbon dioxide exchange in irrigated and rainfed maize-based agroecosystems
Agric. Forest Meteorol.
Simulation of water and carbon fluxes using BIOME-BGC model over crops in China
Agric. Forest Meteorol.
A review of applications of model-data fusion to studies of terrestrial carbon fluxes at different scales
Agric. Forest Meteorol.
The carbon balance of European croplands: a cross-site comparison of simulation models
Agric. Ecosyst. Environ.
What have we learned from 15 years of free-air CO2 enrichment (FACE)? A meta-analytic review of the responses of photosynthesis, canopy properties and plant production to rising CO2
New Phytol.
Assessing the eddy covariance technique for evaluating carbon dioxide exchange rates of ecosystems: past, present and future
Glob. Change Biol.
Modelling the role of agriculture for the 20th century global terrestrial carbon balance
Glob. Change Biol.
Estimating diurnal to annual ecosystem parameters by synthesis of a carbon flux model with eddy covariance net ecosystem exchange observations
Glob. Change Biol.
STICS: a generic model for the simulation of crops and their water and nitrogen balances. I. Theory and parameterization applied to wheat and corn
Agronomie
Simple scaling of photosynthesis from leaves to canopies without the errors of big-leaf models
Plant Cell Environ.
Canopy structure and vertical patterns of photosynthesis and related leaf traits in a deciduous forest
Oecologia
A biochemical model of photosynthetic CO2 assimilation in leaves of species
Planta
Implications of climate change for agricultural productivity in the early twenty-first century
Phil. Trans. R. Soc. B: Biol. Sci.
Cited by (12)
Conditional interval reduction method: A possible new direction for the optimization of process based models
2022, Environmental Modelling and SoftwareCitation Excerpt :Here we refer to these intervals and optimum parameter set as GLUE-based, constrained intervals. Optimized model performance was evaluated using the square of the linear correlation coefficient (R2), bias (systematic error), root mean square error (RMSE) and Nash-Sutcliffe modelling efficiency (ME) (Ma et al., 2011; Sándor et al., 2016). The performance indicators were calculated based on the observed and the simulated maize yield time series.
Simulating the net ecosystem CO<inf>2</inf> exchange and its components over winter wheat cultivation sites across a large climate gradient in Europe using the ORCHIDEE-STICS generic model
2016, Agriculture, Ecosystems and EnvironmentCitation Excerpt :Recent efforts have led to a better representation of croplands within some GTEMs. This is the case for the ORCHIDEE-STICS model (de Noblet-Ducoudre et al., 2004; Gervois et al., 2004), the LPJmL model (Bondeau et al., 2007), the SIBcrop model (Lokupitiya et al., 2009), the Jules-sucros model (Van den Hoof et al., 2011), the ANTHRO-BGC model (Ma et al., 2011), the CESM1 model (Levis et al., 2012) or the CLM model (Bilionis et al., 2015). New model developments for croplands need to be evaluated especially against site data like eddy-flux measurements.
Exploring the relationships between climatic variables and climate-induced yield of spring maize in Northeast China
2015, Agriculture, Ecosystems and EnvironmentCitation Excerpt :These process-based models are typically developed and tested using experimental trials and thus offer the distinct advantage of leveraging decades of research on crop physiology and reproduction, agronomy, and soil science, among other disciplines (Lobell and Burke, 2010). For example, the ANTHRO-BGC model (ANTHROpogenic BioGeoChemistry) (Ma et al., 2011), dynamic crop yield model WOFOST (Ceglar et al., 2011), STICS model (Corre-Hellou et al., 2009), APSIM model (Hammer et al., 2009), DNDC model (Han et al., 2014) and the CERES model (Timsina and Humphreys, 2006). Yet these models also require extensive data input on cultivars, management, and soil conditions that are unavailable in many parts of the world.
Effect of spatial heterogeneity on the validation of remote sensing based GPP estimations
2013, Agricultural and Forest MeteorologyDevelopment of the Biome-BGC model for simulation of managed herbaceous ecosystems
2012, Ecological ModellingCitation Excerpt :Biogeochemical models are under continuous development. Most of the models were already extended to include realistic cropland phenology and management practices (Gervois et al., 2004, 2008; Bondeau et al., 2007; Vuichard et al., 2007a; Lokupitiya et al., 2009; Van den Hoof et al., 2010; Ma et al., 2011b). However, relatively few studies deal with grassland-related developments in spite of the fact that grasslands occupy significant areas (e.g. around 22% in EU-25), and they have the potential to sequester more carbon in the soil than forest soils (Schulze et al., 2009).