Remote sensing and simulation modelling as tools for improving nitrogen efficiency for winter oilseed rape (Brassica napus L.)
Winter oilseed rape (OSR) became more and more important during the last decades due to its profitability, the potential to grow OSR as a renewable resource for biofuel production and its beneficial value as preceding crop. Its high fertilizer requirements and its low nitrogen efficiency enhance the risk of N losses from soil-crop system to the environment by leaching during winter rainfall. Site specific fertilization is one approach to improve N efficiency for OSR. Therefore, high resoluted spatial and time concerning information about current crop parameters like biomass and N uptake as well as information about yield potential is necessary. Two approaches were developed to realize these requirements. First, site specific differences of crop development should become measurable in realtime and for a large scale. Second, time courses of biomass production and N uptake should be comprehended by model simulation. Subsequently, appropriate fertilizer algorithms should be implemented for the derivation of fertilizer recommendations. Furthermore, the applicability of these approaches on field scale must be investigated in the course of additional studies. The first objective of this thesis was the establishment of a non-destructive, remote sensing based method to achieve information on site specific variability of above ground biomass and crop uptake of OSR during the growth period by measuring canopy reflectance in order to save intensive destructive measurements in the fields. A systematic analysis of hyperspectral reflectance measurements by testing all possible waveband combinations enabled to derive new vegetation indices. These indices were successfully used to estimate crop canopy parameters like the green area index (GAI), shoot dry matter (DMshoot) and nitrogen in the shoot (Nshoot) during the vegetation period and their prediction power exceeded commonly used indices in terms of r2 and RMSE. Particularly, combinations of near-infrared wavebands 740 nm and 780 nm served best to estimate GAI and DMshoot and Nshoot. Regarding calibration and validation, the indices showed stable results. Using these indices enables to identify small differences between crop canopy parameters in realtime and for large areas. These information can be used to initialize and parameterize a crop growth model for OSR. A further objective of this thesis was the development of a growth model to predict plant parameters like biomass and canopy N in the time course, to derive the amount of fertilizer demand. The presented crop growth model follows a dynamic, empirical approach for predicting dry matter production based on the concept of light use efficiency (LUE) for vegetative growth from emergence to flowering. Data gave evidence that LUE before winter was not constant but depended on radiation intensity. Therefore, it was calculated according to a linear regression function for this specific period. Furthermore, the model calculates dry matter partitioning into leaf and stem fraction according to an allometric approach. For the calculation of nitrogen uptake, nitrogen concentration had to be determined. For leaves, it was set to be constant, whereas stem nitrogen concentration depended on stem dry matter according to an exponential dilution curve. Nitrogen uptake for the shoot was obtained by adding leaf and stem nitrogen uptake. Additionally, the model calculated dry matter losses during winter, which were calculated according to temperatures below 0 °C. Taking these losses into account, enabled for a continuously prediction of dry matter from emergence to flowering without interruption. The variable LUE approach before winter improved dry matter predictions compared to calculations with an optimized, constant LUE. The model could reproduce the dynamic of dry matter losses during winter. Only few data sets were available for calibration and validation, so that parameterization was not robust regarding standard errors for optimized parameters. Prediction for validation data set revealed high differences concerning quality, indicating that additional data set are needed for improving parameterization and prediction of the crop growth model. Because crop growth differs for different developmental stages, another object of this study was the development of a phenological model. Knowledge about time and duration of growth stages can also be used for a growth specific crop management like N fertilization. The presented phenological model based on BRASNAP-PH, which was developed for north-west Europe. Because BRASNAP-PH predicts only four main developmental stages, modifications were conducted, to predict highly resoluted developmental stages according to BBCH. Vernalization and photoperiod were the most important factors besides temperature to estimate leaf development, whereas temperature was determinant during all phases of development. The modified model was calibrated and validated on a large data set throughout Germany. Optimized parameter values and corresponding standard errors for calibration and cross calibration gave advice for a robust parameterization. Validation and cross validation let to a root mean square error (RMSE) of 3.11 and 3.25 BBCH stages, respectively, for the whole developmental phase. The presented results of this thesis can contribute to improve N efficiency for OSR in terms of a site specific fertilization implementing fertilizer algorithms to the growth model. Therefore, an accurate, adjusted, site specific N fertilization recommendation should be derived from information of current crop status and of yield potential. Since the site specific fertilization deals only with one aspect for improving N efficiency for OSR, other approaches must be developed and investigated. Especially approaches concerning production management and crop rotations reveal promising for improvement.