Növénytermelés / Volume 69 / Issue 1 (March 2020) / pp. 73-94
RAGÁN PÉTER – SULYOK DÉNES ZSOLT – CSATÁRI NÁNDOR – VÁNTUS ANDRÁS – HAGYMÁSSY ZOLTÁN – NAGY JÁNOS – HARSÁNYI ENDRE – RÁTONYI TAMÁS
Correlation analysis of data of various origin in a field variety experiment
Unfavourable soil characteristics and extreme weather factors negatively affect the yield of farm crops. Soil and plant analyses are absolutely necessary for site-specific crop production, but traditional, grid-based soil sampling is difficult and expensive. Today, yield zone maps made from satellite images by using some kind of vegetation indexes are becoming increasingly widespread in precision agriculture. We performed our examinations in a field maize experiment on a 56.3 ha plot in Hajdú-Bihar county in Debrecen, Hungary. Soil analysis was performed with a Veris U3 soil scanner in 2018 and again in 2019, after the field was drained using trenches. During our analyses, the plot was mapped on two occasions also with a DJI Phantom 4 Agro UAV equipped with an NGB camera. In 2019, NDVI, GNDVI and bNDVI images were made from 23 Sentinel 2 satellite images free from clouds or other disturbing factors. Using Quantum GIS, a 10×10 m spatial resolution vector GIS and then numerical database were prepared from the vector data of Veris U3 and the raster data of UAV and Sentinel 2. Our aim was to evaluate contact soil analysis and remotely sensed data, as well as to perform a correlation analysis on maize yield. The ECa obtained in 2019 and the three Sentinel 2-based vegetation indexes together showed a moderately strong (r=0.6) correlation with maize yield. The Veris and Sentinel data, supplemented with the bNDVI data of the two UAV flights, a strong correlation (r=0.79) was shown with yield. UAV-based GNDVI distribution, UAV-based bNDVI means, Sentinel 2-based yearly bNDVI, GNDVI and NDVI means, the ECa, IR/R and pH distribution had a strong and significant correlation (r=0.86) with maize yield. The soil scanner and the remotely sensed data showed a 74.6% correlation with maize yield. Based on the performed analyses, it can be concluded that neither ECa, nor the one-year-long remotely sensed data series alone provide enough information on plot heterogeneity.
Keywords: Veris U3 data, UAV- and Sentinel 2-based vegetation indexes, complex GIS database, linear and multilinear regression