Wallach, D., Nissanka, S. P., Karunaratne, A. S., Weerakoon, W. M. W., Thorburn, P. J., Boote, K. J., et al. (2016). Accounting for both parameter and model structure uncertainty in crop model predictions of phenology: A case study on rice. European Journal of Agronomy, .
Abstract: We consider predictions of the impact of climate warming on rice development times in Sri Lanka. The major emphasis is on the uncertainty of the predictions, and in particular on the estimation of mean squared error of prediction. Three contributions to mean squared error are considered. The first is parameter uncertainty that results from model calibration. To take proper account of the complex data structure, generalized least squares is used to estimate the parameters and the variance-covariance matrix of the parameter estimators. The second contribution is model structure uncertainty, which we estimate using two different models. An ANOVA analysis is used to separate the contributions of parameter and model uncertainty to mean squared error. The third contribution is model error, which is estimated using hindcasts. Mean squared error of prediction of time from emergence to maturity, for baseline +2 °C, is estimated as 108 days2, with model error contributing 86 days2, followed by model structure uncertainty which contributes 15 days2 and parameter uncertainty which contributes 7 days2. We also show how prediction uncertainty is reduced if prediction concerns development time averaged over years, or the difference in development time between baseline and warmer temperatures.
|
Maiorano, A., Martre, P., Asseng, S., Ewert, F., Müller, C., Rötter, R. P., et al. (2016). Crop model improvement reduces the uncertainty of the response to temperature of multi-model ensembles. Field Crops Research, 202, 5–20.
Abstract: To improve climate change impact estimates and to quantify their uncertainty, multi-model ensembles (MMEs) have been suggested. Model improvements can improve the accuracy of simulations and reduce the uncertainty of climate change impact assessments. Furthermore, they can reduce the number of models needed in a MME. Herein, 15 wheat growth models of a larger MME were improved through re-parameterization and/or incorporating or modifying heat stress effects on phenology, leaf growth and senescence, biomass growth, and grain number and size using detailed field experimental data from the USDA Hot Serial Cereal experiment (calibration data set). Simulation results from before and after model improvement were then evaluated with independent field experiments from a CIMMYT world-wide field trial network (evaluation data set). Model improvements decreased the variation (10th to 90th model ensemble percentile range) of grain yields simulated by the MME on average by 39% in the calibration data set and by 26% in the independent evaluation data set for crops grown in mean seasonal temperatures >24 °C. MME mean squared error in simulating grain yield decreased by 37%. A reduction in MME uncertainty range by 27% increased MME prediction skills by 47%. Results suggest that the mean level of variation observed in field experiments and used as a benchmark can be reached with half the number of models in the MME. Improving crop models is therefore important to increase the certainty of model-based impact assessments and allow more practical, i.e. smaller MMEs to be used effectively.
|
Kuhnert, M., Yeluripati, J., Smith, P., Hoffmann, H., van Oijen, M., Constantin, J., et al. (2016). Impact analysis of climate data aggregation at different spatial scales on simulated net primary productivity for croplands. European Journal of Agronomy, 88, 41–52.
Abstract: For spatial crop and agro-systems modelling, there is often a discrepancy between the scale of measured driving data and the target resolution. Spatial data aggregation is often necessary, which can introduce additional uncertainty into the simulation results. Previous studies have shown that climate data aggregation has little effect on simulation of phenological stages, but effects on net primary production (NPP) might still be expected through changing the length of the growing season and the period of grain filling. This study investigates the impact of spatial climate data aggregation on NPP simulation results, applying eleven different models for the same study region (∼34,000 km2), situated in Western Germany. To isolate effects of climate, soil data and management were assumed to be constant over the entire study area and over the entire study period of 29 years. Two crops, winter wheat and silage maize, were tested as monocultures. Compared to the impact of climate data aggregation on yield, the effect on NPP is in a similar range, but is slightly lower, with only small impacts on averages over the entire simulation period and study region. Maximum differences between the five scales in the range of 1–100 km grid cells show changes of 0.4–7.8% and 0.0–4.8% for wheat and maize, respectively, whereas the simulated potential NPP averages of the models show a wide range (1.9–4.2 g C m−2 d−1 and 2.7–6.1 g C m−2 d−1for wheat and maize, respectively). The impact of the spatial aggregation was also tested for shorter time periods, to see if impacts over shorter periods attenuate over longer periods. The results show larger impacts for single years (up to 9.4% for wheat and up to 13.6% for maize). An analysis of extreme weather conditions shows an aggregation effect in vulnerability up to 12.8% and 15.5% between the different resolutions for wheat and maize, respectively. Simulations of NPP averages over larger areas (e.g. regional scale) and longer time periods (several years) are relatively insensitive to climate data.
|
Coles, G. D., Wratten, S. D., & Porter, J. R. (2016). Food and nutritional security requires adequate protein as well as energy, delivered from whole-year crop production. PeerJ, 4, 17.
Abstract: Human food security requires the production of sufficient quantities of both high-quality protein and dietary energy. In a series of case-studies from New Zealand, we show that while production of food ingredients from crops on arable land can meet human dietary energy requirements effectively, requirements for high-quality protein are met more efficiently by animal production from such land. We present a model that can be used to assess dietary energy and quality-corrected protein production from various crop and crop/animal production systems, and demonstrate its utility. We extend our analysis with an accompanying economic analysis of commercially available pre-prepared or simply-cooked foods that can be produced from our case-study crop and animal products. We calculate the per-person, per-day cost of both quality-corrected protein and dietary energy as provided in the processed foods. We conclude that mixed dairy/cropping systems provide the greatest quantity of high quality protein per unit price to the consumer, have the highest food energy production and can support the dietary requirements of the highest number of people, when assessed as all-year-round production systems. Global food and nutritional security will largely be an outcome of national or regional agroeconomies addressing their town food needs. We hope that lour model will be used for similar analyses of food production systems in other countries, agroecological zones and economies.
|
Wallach, D., Thorburn, P., Asseng, S., Challinor, A. J., Ewert, F., Jones, J. W., et al. (2016). Estimating model prediction error: Should you treat predictions as fixed or random. Env. Model. Softw., 84, 529–539.
Abstract: Crop models are important tools for impact assessment of climate change, as well as for exploring management options under current climate. It is essential to evaluate the uncertainty associated with predictions of these models. We compare two criteria of prediction error; MSEPfixed, which evaluates mean squared error of prediction for a model with fixed structure, parameters and inputs, and MSEPuncertain(X), which evaluates mean squared error averaged over the distributions of model structure, inputs and parameters. Comparison of model outputs with data can be used to estimate the former. The latter has a squared bias term, which can be estimated using hindcasts, and a model variance term, which can be estimated from a simulation experiment. The separate contributions to MSEPuncertain(X) can be estimated using a random effects ANOVA. It is argued that MSEPuncertain(X) is the more informative uncertainty criterion, because it is specific to each prediction situation.
|