Determining yield-limiting factors using optical satellite data
MetadataShow full item record
The constantly growing population and food demand require us to use agricultural land as efficiently as possible. The high costs of fertilisers and plant protection products together with their impact on the environment motivate farmers to seek and implement new technologies for maximising the yield while reducing the costs of production. In order to achieve that, farmers need to address Yield-Limiting Factors (YLFs). This is where remote sensing comes in handy - instead of using ground-based techniques for the detection and monitoring of yield-limiting factors that are costly and time-consuming, satellite imagery is offering a cheap and fast alternative. In this research, the focus is put on yield-limiting factors for corn crops. The test fields are located in the central part of Ukraine, namely in the Cherkasy province. Two cornfields were selected for analysis. Whereas field 6 was used as the main research field, field 10 was used for validation purposes to compare the results obtained from both fields. The size of the fields is 227 ha and 181 ha respectively, making this research important for large-scale farming. Eight satellite images were selected, one for every month in the growing season (March to October 2017). Those images were visually and statistically analysed for correlation with the final yield and several yield-limiting factors. Near Infrared (NIR), Normalized Difference Vegetation Index (NDVI) and Atmospherically Resistant Vegetation Index (ARVI) were used to test the possibility of the detection of yield variability using remote sensing data. NIR was determined to be the most correlated with the final yield (with the maximum R2 of 0.72). NDVI was the second-best tool for the detection of yield variability with the maximum R2 reaching 0.698. ARVI did not show any particularly outstanding results. The average amount of explained variation in NIR is around 28%; it is then followed by NDVI of 19% and ARVI of 14%. The results show that September is the best moment for the detection of yield variability, the highest correlation can be found at that time (R2 = 0.72). The beginning of the growing season can also produce a high correlation considering that it is only bare soil (R2 is 22% on average). Organic Matter (OM) content was deemed as the YLF that has the greatest potential for detection using satellite imagery in field 6 (R2 of 0.368). The reason for that is the big variations in OM content in field 6. Soil Potential of Hydrogen (pH) and Electrical Conductivity (EC) showed smaller correlation (R2 = 0.038 and R2 = 0.051, respectively), while elevation needed a different approach before the correlation could be found (R2 = 0.312). During the validation of the results, it was found that a lot of them are field dependent and cannot be reproduced on other fields/farms. Especially when it comes to yield-limiting factors, no consistent results were found; hence there is no specific method which is the most optimal for yield-limiting factors detection. The same applies to the best moment for the detection of yield-limiting factors. For different yield-limiting factors different VI/spectral bands showed the highest results, and it also varied per field. The same goes for the temporal aspects of detection. However, both fields showed that September was found to be the month of the highest correlation between yield variability and optical satellite imagery.