|
|
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.
|
|
|
|
Nendel, C., Wieland, R., Mirschel, W., Specka, X., Guddat, C., & Kersebaum, K. C. (2013). Simulating regional winter wheat yields using input data of different spatial resolution. Field Crops Research, 145, 67–77.
Abstract: The success of using agro-ecosystem models for the high-resolution simulation of agricultural yields for larger areas is often hampered by a lack of input data. We investigated the effect of different spatially resolved soil and weather data used as input for the MONICA model on its ability to reproduce winter wheat yields in the Federal State of Thuringia, Germany (16,172 km(2)). The combination of one representative soil and one weather station was insufficient to reproduce the observed mean yield of 6.66 +/- 0.87 t ha(-1) for the federal state. Use of a 100 m x 100 m grid of soil and relief information combined with just one representative weather station yielded a good estimator (7.01 +/- 1.47 t ha(-1)). The soil and relief data grid used in combination with weather information from 14 weather stations in a nearest neighbour approach produced even better results (6.60 +/- 1.37 t ha(-1)); the same grid used with 39 additional rain gauges and an interpolation algorithm that included an altitude correction of temperature data slightly overpredicted the observed mean (7.36 +/- 1.17 t ha(-1)). It was concluded that the apparent success of the first two high-resolution approaches over the latter was based on two effects that cancelled each other out: the calibration of MONICA to match high-yield experimental data and the growth-defining and -limiting effect of weather data that is not representative for large parts of the region. At the county and farm level the MONICA model failed to reproduce the 1992-2010 time series of yields, which is partly explained by the fact that many growth-reducing factors were not considered in the model. (C) 2013 Elsevier B.V. All rights reserved.
|
|
