Abstract: A spatial resolution needs to be determined prior to using models to simulate crop yields at a regional scale, but a dilemma exists in compromising between different demands. A fine spatial resolution demands extensive computation load for input data assembly, model runs, and output analysis. A coarse spatial resolution could result in loss of spatial detail in variability. This paper studied the impact of spatial resolution, data aggregation and spatial heterogeneity of weather data on simulations of crop yields, thus providing guidelines for choosing a proper spatial resolution for simulations of crop yields at regional scale. Using a process-based crop model SIMPLACE (LINTUL2) and daily weather data at 1 km resolution we simulated a continuous rainfed winter wheat cropping system at the national scale of Germany. Then we aggregated the weather data to four resolutions from 10 to 100 km, repeated the simulation, compared them with the 1 km results, and correlated the difference with the intra-pixel heterogeneity quantified by an ensemble of four semivariogram models. Aggregation of weather data had small effects over regions with a flat terrain located in northern Germany, but large effects over southern regions with a complex topography. The spatial distribution of yield bias at different spatial resolutions was consistent with the intra-pixel spatial heterogeneity of the terrain and a log-log linear relationship between them was established. By using this relationship we demonstrated the way to optimize the model resolution to minimize both the number of simulation runs and the expected loss of spatial detail in variability due to aggregation effects. We concluded that a high spatial resolution is desired for regions with high spatial environmental heterogeneity, and vice versa. This calls for the development of multi-scale approaches in regional and global crop modeling. The obtained results require substantiation for other production situations, crops, output variables and for different crop models. (C) 2014 Elsevier B.V. All rights reserved.

Abstract: Field-scale crop models are often applied at spatial resolutions coarser than that of the arable field. However, little is known about the response of the models to spatially aggregated climate input data and why these responses can differ across models. Depending on the model, regional yield estimates from large-scale simulations may be biased, compared to simulations with high-resolution input data. We evaluated this so-called aggregation effect for 13 crop models for the region of North Rhine-Westphalia in Germany. The models were supplied with climate data of 1 km resolution and spatial aggregates of up to 100 km resolution raster. The models were used with 2 crops (winter wheat and silage maize) and 3 production situations (potential, water-limited and nitrogen-water-limited growth) to improve the understanding of errors in model simulations related to data aggregation and possible interactions with the model structure. The most important climate variables identified in determining the model-specific input data aggregation on simulated yields were mainly related to changes in radiation (wheat) and temperature (maize). Additionally, aggregation effects were systematic, regardless of the extent of the effect. Climate input data aggregation changed the mean simulated regional yield by up to 0.2 t ha(-1), whereas simulated yields from single years and models differed considerably, depending on the data aggregation. This implies that large-scale crop yield simulations are robust against climate data aggregation. However, large-scale simulations can be systematically biased when being evaluated at higher temporal or spatial resolution depending on the model and its parameterization.

Abstract: Eddy covariance data from four European grassland sites are used to probabilistically invert the CARAIB (CARbon Assimilation In the Biosphere) dynamic vegetation model (DVM) with 10 unknown parameters, using the DREAM((ZS)) (DiffeRential Evolution Adaptive Metropolis) Markov chain Monte Carlo (MCMC) sampler. We focus on comparing model inversions, considering both homoscedastic and heteroscedastic eddy covariance residual errors, with variances either fixed a priori or jointly inferred together with the model parameters. Agreements between measured and simulated data during calibration are comparable with previous studies, with root mean square errors (RMSEs) of simulated daily gross primary productivity (GPP), ecosystem respiration (RECO) and evapotranspiration (ET) ranging from 1.73 to 2.19, 1.04 to 1.56 g C m(-2) day(-1) and 0.50 to 1.28 mm day(-1), respectively. For the calibration period, using a homoscedastic eddy covariance residual error model resulted in a better agreement between measured and modelled data than using a heteroscedastic residual error model. However, a model validation experiment showed that CARAIB models calibrated considering heteroscedastic residual errors perform better. Posterior parameter distributions derived from using a heteroscedastic model of the residuals thus appear to be more robust. This is the case even though the classical linear heteroscedastic error model assumed herein did not fully remove heteroscedasticity of the GPP residuals. Despite the fact that the calibrated model is generally capable of fitting the data within measurement errors, systematic bias in the model simulations are observed. These are likely due to model inadequacies such as shortcomings in the photosynthesis modelling. Besides the residual error treatment, differences between model parameter posterior distributions among the four grassland sites are also investigated. It is shown that the marginal distributions of the specific leaf area and characteristic mortality time parameters can be explained by site-specific ecophysiological characteristics.