Bellocchi, G., & Ma, S. (2014). Results of uncalibrated grassland model runs (Vol. 3).
Abstract: This deliverable focuses on the some illustrative results obtained with the grassland models selected (D-L2.1.1) to simulate biomass and flux data from grassland sites in Europe and peri-Mediterranean regions (D-L2.1.1 and D-L2.1.2). This is a blind exercise, carried out without model calibration. The complete set of results will include simulations from calibrated models. The results shown are illustrative of the methodology adopted for grassland model intercomparison in MACSUR. The insights gained from this ongoing study are relevant for some crop and vegetation models, which in some cases proved comparable to grassland-specific models to simulate biomass data from managed grasslands. The results reported here cannot be considered conclusive. Additional results will be published as they become available together with calibration results, as well as the comprehensive evaluation of models with fuzzy logic-based indicators. No Label
|
Rötter, R. P., & Semenov, M. A. (2014). Development of methods for the probabilistic assessment of climate change impacts on crop production (Vol. 3).
Abstract: Various attempts have been made to determine the relative importance of uncertainties in climate change impact assessments stemming from climate projections and crop models, respectively, and to analyse yield outputs probabilistically. For example, in the ENSEMBLES project, probabilistic climate projections (Harris et al. 2010) have been applied in conjunction with impact response surfaces (IRS), constructed by using impact models, to estimate the future likelihood (risk) of exceeding critical thresholds of crop yield impact (see, Fronzek et al., 2011, for an explanation of the method). In this task, we aimed to further develop and operationalize these methods and testing them in different case study regions in Europe. The method combines results of a sensitivity analysis of (one or more) impact model(s) with probabilistic projections of future temperature and precipitation (Fronzek et al., 2011). Such an overlay is one way of portraying probabilistic estimates of future impacts. By further accounting for the uncertainties in crop and biophysical parameters (using perturbed parameter approaches), the outcome represents an ensemble of impact risk estimates, encapsulating both climate and crop model uncertainties. No Label
|
Wallach, D., & Rivington, M. (2014). A framework for assessing the uncertainty in crop model predictions (Vol. 3).
Abstract: It is of major importance in modeling to understand and quantify the uncertainty in model predictions, both in order to know how much confidence to have in those predictions, and as a first step toward model improvement. Here we show that there are basically three different approaches to evaluating uncertainty, and we explain the advantages and drawbacks of each. This is a necessary first step toward developing protocols for evaluation of uncertainty and so obtaining a clearer picture of the reliability of crop models. No Label
|
Kersebaum, K., & C,. (2014). Results of uncalibrated model runs available (ROTATIONEFFECTS) (Vol. 3).
Abstract: The study ROTATIONEFFECT aims to compare the output of different models simulating field data sets with multi-year crop rotations including different treatments. Data sets for 5 locations in Europe were distributed to 19 interested modeller groups comprising a total of 201 crop growth seasons. In a first step only minimal information for calibration were provided to the modellers. In total 14 modelling teams sent their “uncalibrated” results as single-year calculations and/or calculations of rotation depending on the capability of the model. 7-10 models were capable to run the rotations as continuous runs. Up to 12 models provided single year simulations of at least one crop. Comparing results of models which provided both single year and continuous runs, show a little lower root mean square error for the continuous rotations runs. Cereal crop yields were generally better simulated than tuber/beet yields. Additionally, the models’ response to various treatments (irrigation/rainfed, nitrogen level, CO2 level, residue management/ tillage, catch crops) were compared to observed differences. First indicators of model performance have been developed and presented at international conferences. No Label
|
Bellocchi, G., Rivington, M., & Acutis, M. (2014). Protocol for model evaluation (Vol. 3).
Abstract: This deliverable focuses on the development of methods for model evaluation in order to have unambiguous indications derived from the use of several evaluation metrics. The information about model quality is aggregated into a single indicator using a fuzzy expert system that can be applied to a wide range of model estimates where suitable test data are available. This is a cross-cutting activity between CropM (C1.4) and LiveM (L2.2). No Label
|