University of Twente Student Theses


Yield assessment at field level using satellite measurements and semi-empirical crop modelling: A case study in Gutland area, Germany

Sithole, Vhusomuzi B. (2009) Yield assessment at field level using satellite measurements and semi-empirical crop modelling: A case study in Gutland area, Germany.

[img] PDF
Abstract:Crop models seek to quantify relationships between weather, crop growth and effects of managements to allow yield prediction, diagnosis, management decision and environmental assessment. Remote sensing provide alternative methods of measuring crop variables such as Leaf Area Index (LAI) and canopy cover which when used together with meteorological data improve crop models’ accuracy. Winter rape is essential for oil seed extraction which supports a number of other industries and families in European Union (EU) countries such as Germany. Most EU countries experience near overcast weather conditions which makes acquisition of active remote sensing data problematic due to cloud cover. The year 2003 was generally warmer and cloud free in EU hence this study focussed yield assessment for winter rape using the 2003 growing season. The study assesses if a combination of SPOT data and Monteith’s model allows an accuracy improvement compared to an empirical model in estimating crop yield within fields of winter rape in Gutland, Germany. This was done with a special focus on the optimisation of LUE within the model. Three fields of winter rape were measured for yield and global solar radiation. This data was used together with fAPAR and climate efficiency to optimise LUE and then compare accuracy levels of Monteith’s model and empirical model. In empirical modelling, measured yield was related to ANDVI using regression analysis. Field 1 data was used for calibration while fields 2 and 3 data was used for validation. Results show that Montheith’s model with pixel optimised LUE values does improve yield prediction accuracy and the accuracy levels are higher than using empirical model, or than Montheith’s model with a constant from literature or than Monteith’s model with a constant field optimised LUE value. Average RMSE were 0.64, 1.18, 0.66 and 0.41 ton/ha for empirical modelling, semi-empirical with LUE from literature, semi-empirical with field-optimised LUE and semi-empirical with pixel optimised LUE respectively. Average relative RMSE were 28, 36, 28 and 13 percent of average measured yield for the respective models. Pixel optimised LUE values were investigated for consistency with crop condition classification from an aerial photograph for the same winter rape growing season. Future work should consider validating pixel-optimised LUE values of winter rape using measured values.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page