Abstract: In crop insurance, the accuracy with which the insurer quantifies the actual risk is highly dependent on the availability on actual yield data. Crop models might be valuable tools to generate data on expected yields for risk assessment when no historical records are available. However, selecting a crop model for a specific objective, location and implementation scale is a difficult task. A look inside the different crop and soil modules to understand how outputs are obtained might facilitate model choice. The objectives of this paper were (i) to assess the usefulness of crop models to be used within a crop insurance analysis and design and (ii) to select the most suitable crop model for drought risk assessment in semi-arid regions in Spain. For that purpose first, a pre-selection of crop models simulating wheat yield under rainfed growing conditions at the field scale was made, and second, four selected models (Aquacrop, CERES-Wheat, CropSyst and WOFOST) were compared in terms of modelling approaches, process descriptions and model outputs. Outputs of the four models for the simulation of winter wheat growth are comparable when water is not limiting, but differences are larger when simulating yields under rainfed conditions. These differences in rainfed yields are mainly related to the dissimilar simulated soil water availability and the assumed linkages with dry matter formation. We concluded that for the simulation of winter wheat growth at field scale in such semi-arid conditions, CERES-Wheat and CropSyst are preferred. WOFOST is a satisfactory compromise between data availability and complexity when detail data on soil is limited. Aquacrop integrates physiological processes in some representative parameters, thus diminishing the number of input parameters, what is seen as an advantage when observed data is scarce. However, the high sensitivity of this model to low water availability limits its use in the region considered. Contrary to the use of ensembles of crop models, we endorse that efforts be concentrated on selecting or rebuilding a model that includes approaches that better describe the agronomic conditions of the regions in which they will be applied. The use of such complex methodologies as crop models is associated with numerous sources of uncertainty, although these models are the best tools available to get insight in these complex agronomic systems. (C) 2015 Elsevier B.V. All rights reserved.

Abstract: Highlights • Regularization techniques for regression outperformed the classical regression techniques in predicting crop yields. • Different regression techniques with similar prediction accuracy showed different responses of major climatic variables to crop yield. • The regression models showed some responses of crop yield to climatic conditions that is mostly absent in process based crop models. Abstract Statistical regression models represent alternatives to process-based dynamic models for predicting the response of crop yields to variation in climatic conditions. Regression models can be used to quantify the effect of change in temperature and precipitation on yields. However, it is difficult to identify the most relevant input variables that should be included in regression models due to the high number of candidate variables and to their correlations. This paper compares several regression techniques for modeling response of winter oilseed rape yield to a high number of correlated input variables. Several statistical regression methods were fitted to a dataset including 689 observations of winter oilseed rape yield from replicated field experiments conducted in 239 sites in Denmark, covering nearly all regions of the country from 1992 to 2013. Regression methods were compared by cross-validation. The regression methods leading to the most accurate yield predictions were Lasso and Elastic Net, and the least accurate methods were ordinary least squares and stepwise regression. Partial least squares and ridge regression methods gave intermediate results. The estimated relative yield change for a +1°C temperature increase during flowering was estimated to range between 0 and +6 %, depending on choice of regression method. Precipitation was found to have an adverse effect on yield during autumn and winter. It was estimated that an increase in precipitation of +1 mm/day would result in a relative yield change ranging from 0 to −4 %. Soil type was also important for crop yields with lower yields on sandy soils compared to loamy soils. Later sowing was found to result in increased crop yield. The estimated effect of climate on yield was highly sensitive to the chosen regression method. Regression models showing similar performance led in some cases to different conclusions with respect to effect of temperature and precipitation. Hence, it is recommended to apply an ensemble of regression models, in order to account for the sensitivity of the data driven models for projecting crop yield under climate change.

Abstract: Heat and drought stress can reduce crop yields considerably which is increasingly assessed with crop models for larger areas. Applying these models originally developed for the field scale at large spatial extent typically implies the use of input data with coarse resolution. Little is known about the effect of data resolution on the simulated impact of extreme events like heat and drought on crops. Hence, in this study the effect of input and output data aggregation on simulated heat and drought stress and their impact on yield of winter wheat is systematically analyzed. The crop model SIMPLACE was applied for the period 1980-2011 across Germany at a resolution of 1 km x 1 km. Weather and soil input data and model output data were then aggregated to 10 km x 10 km, 25 km x 25 km, 50 km x 50 km and 100 km x 100 km resolution to analyze the aggregation effect on heat and drought stress and crop yield. We found that aggregation of model input and output data barely influenced the mean and median of heat and drought stress reduction factors and crop yields simulated across Germany. However, data aggregation resulted in less spatial variability of model results and a reduced severity of simulated stress events, particularly for regions with high heterogeneity in weather and soil conditions. Comparisons of simulations at coarse resolution with those at high resolution showed distinct patterns of positive and negative deviations which compensated each other so that aggregation effects for large regions were small for mean or median yields. Therefore, modelling at a resolution of 100 km x 100 km was sufficient to determine mean wheat yield as affected by heat and drought stress for Germany. Further research is required to clarify whether the results can be generalized across crop models differing in structure and detail. Attention should also be given to better understand the effect of data resolution on interactions between heat and drought impacts. (C) 2015 Elsevier B.V. All rights reserved.