Evaluation of the Maizsim Model Under Irrigated and Dryland Conditions.

Crop models can be used as aids in interpreting experimental results, as agronomic research tools, or even as agronomic grower tools. By using them as research tools users are able to try different modifications in the agronomic system and see the consequences of those changes in terms of yield, biomass formation and senescence among other variables. A drawback to the use of crop simulation models is that they often have to be calibrated to use them in a new location or even with a new set of environmental variables. This process requires a comprehensive and specific data set before being capable to forecast precise crop responses. MAIZSIM is a soil-plant-atmosphere model constructed by coupling a process-based maize simulation model (MAIZSIM) and a 2D finite element-based soil model (2DSOIL). MAIZSIM has been developed to require a minimal amount of calibration by incorporating non-linear temperature dependencies of basic processes that are not dependent on growing degree days, and using a biochemical model of C4 photosynthesis rather than radiation use efficiency. The objective of this study was to test MAIZSIM against observed data from irrigated and dryland sites in Kansas, U.S.A. The simulation results are also compared against those from CERES-MAIZE that had been calibrated in a previous study. The observed data used for this analysis were acquired from ten field trials conducted in five different locations across Kansas between 2005 and 2006. MAIZSIM was able to predict leaf area index values with high accuracy in most of the cases (d>0.85), while in two particular trials it was very imprecise (d<0.17). Similar results were observed with leaf dry matter. Moreover, the simulation showed correct values of leaf number, presenting d values over 0.95 in most of the cases. The model also showed high accuracy simulating phenological stages of silking and tasseling (d=0.97; RMSE=3.7). The predicted values of shoot dry matter were generally close to the observed ones during the beginning and middle of the growing season. The model tended to over-predict stem dry matter, especially at the end of the season (RMSE>730). Overall, MAIZSIM was able to properly simulate ear dry matter during the season but not the final values, therefore simulated yield values were not as close to the measured ones as expected (d=0.79; EF=0.35). In conclusion, MAIZSIM is capable of predicting phenological stages, leaf area index, leaf dry matter, leaf number and shoot dry matter with reasonable accuracy, while it needs further development to improve the simulation processes of stem and ear dry matter, and yield.