Abstract: Regional agricultural yield assessments allowing for weather effect quantifications are a valuable basis for deriving scenarios of climate change effects and developing adaptation strategies. Assessing weather effects by statistical methods is a classical approach, but for obtaining robust results many details deserve attention and require individual decisions as is demonstrated in this paper. We evaluated regression models for annual yield changes of winter wheat and silage maize in more than 300 German counties and revised them to increase their predictive power. A major effort of this study was, however, aggregating separately estimated time series models (STSM) into panel data models (PDM) based on cluster analyses. The cluster analyses were based on the per-county estimates of STSM parameters. The original STSM formulations (adopted from a parallel study) contained also the non-meteorological input variables acreage and fertilizer price. The models were revised to use only weather variables as estimation basis. These consisted of time aggregates of radiation, precipitation, temperature, and potential evapotranspiration. Altering the input variables generally increased the predictive power of the models as did their clustering into PDM. For each crop, five alternative clusterings were produced by three different methods, and similarities between their spatial structures seem to confirm the existence of objective clusters about common model parameters. Observed smooth transitions of STSM parameter values in space suggest, however, spatial autocorrelation effects that could also be modeled explicitly. Both clustering and autocorrelation approaches can effectively reduce the noise in parameter estimation through targeted aggregation of input data. (C) 2015 Elsevier B.V. All rights reserved.

Abstract: • Systematic analysis of the influence of spatial soil data resolution on simulated regional yields and total growing season evapotranspiration. • The responses of four crop models of different complexity are compared. • Differences between models are larger than the effect of the chosen spatial soil data resolution. • Low influence of soil data resolution due to: high precipitation amount, methods for calculating water retention and method of data aggregation. The spatial variability of soil properties is an important driver of yield variability at both field and regional scale. Thus, when using crop growth simulation models, the choice of spatial resolution of soil input data might be key in order to accurately reproduce observed yield variability. In this study we used four crop models (SIMPLACE<LINTUL-SLIM>, DSSAT-CSM, EPIC and DAISY) differing in the detail of modeling above-ground biomass and yield as well as of modeling soil water dynamics, water uptake and drought effects on plants to simulate winter wheat in two (agro-climatologically and geo-morphologically) contrasting regions of the federal state of North-Rhine-Westphalia (Germany) for the period from 1995 to 2008. Three spatial resolutions of soil input data were taken into consideration, corresponding to the following map scales: 1:50 000, 1:300 000 and 1:1 000 000. The four crop models were run for water-limited production conditions and model results were evaluated in the form of frequency distributions, depicted by bean-plots. In both regions, soil data aggregation had very small influence on the shape and range of frequency distributions of simulated yield and simulated total growing season evapotranspiration for all models. Further analysis revealed that the small influence of spatial resolution of soil input data might be related to: (a) the high precipitation amount in the region which partly masked differences in soil characteristics for water holding capacity, (b) the loss of variability in hydraulic soil properties due to the methods applied to calculate water retention properties of the used soil profiles, and (c) the method of soil data aggregation. No characteristic “fingerprint” between sites, years and resolutions could be found for any of the models. Our results support earlier recommendation to evaluate model results on the basis of frequency distributions since these offer quick and better insight into the distribution of simulation results as compared to summary statistics only. Finally, our results support conclusions from other studies about the usefulness of considering a multi-model approach to quantify the uncertainty in simulated yields introduced by the crop growth simulation approach when exploring the effects of scaling for regional yield impact assessments.