Comparison of selected remote sensing sensors for crop yield variability estimation
MetadataNäita täielikku nimetuse kirjet
Currently, spectral indices are very common tool how to describe various characteristics of vegetation. In fact, these are mathematical operations which are calculated using specific bands of electromagnetic spectrum. Nevertheless, remote sensing sensors can differ due to the variations in bandwidth of the particular spectral channels. Therefore, the main aim of this study is to compare selected sensors in terms of their capability to predict crop yield by NDVI utilization. The experiment was performed at two locations (Prague-Ruzyně and Vendolí) in the year 2015 for both locations and in 2007 for Prague-Ruzyně only, when winter barley or spring barley grew on the plots. The cloud-free satellite images were chosen and normalised difference vegetation indices (NDVI) were calculated for each image. Landsat satellite images with moderate spatial resolution (30 m per pixel) were chosen during the crop growth for selected years. The other data sources were commercial satellite images with very high spatial resolution – QuickBird (QB) (0.6 m per pixel) in 2007 and WorldView-2 (WV-2) (2 m per pixel) in 2015 for Prague-Ruzyně location; and SPOT-7 (6 m per pixel) satellite image in 2015 for Vendolí location. GreenSeeker handheld crop sensor (GS) was used for collecting NDVI data for both locations in 2015 only. NDVI calculated at each of images was compared with the yield data. The data sources were compared with each other at the same term of crop growth stage. The results showed that correlation between GS and yield was relatively weak at Ruzyně. Conversely, significant relation was found at Vendolí location. The satellite images showed stronger relation with yield than GS. Landsat satellite images had higher values of correlation coefficient (in 30 m spatial resolution) at Ruzyně in both selected years. However, at Vendolí location, SPOT-7 satellite image has significantly better results compared to Landsat image. It is necessary to do more research to define which sensor measurements are most useful for selected applications in agriculture management.