1R. Melchiori, 2N. Bergmann, 1S. Albarenque, 2A. Campos, 1A. Kemerer, 2F. Romagnoli, 1J. Pautasso
1. National Institute of Agricultural Technology INTA EEA Paraná, Argentina
2. Taguay S.A.
The use of drone technologies is growing in agriculture, and research groups are working worldwide in order to use them, however scarce practical application in agriculture has been developed up today. The aim of this research was to determine the performance of drone images to prescribe in season Nitrogen (N) fertilization rates for wheat. A drone based nitrogen rate calculator was developed using a vegetation index obtained from high resolution images. In order to obtain a practical use a virtual reference was used to estimate the N response index using the 95% percentile value of the vegetation index of the field.
The flight was performed at crop stage of first visible node (GS 31, Zadoks scale), with a DJI Matrice 100 quadcopter equipped with a Zenmuse X3 BGNIR modified camera. The survey was made at 120 m height and 70% of frontal and lateral overlap. The algorithm was applied at pixel level, and image resampling procedures were applied to customize the prescription size to a better match with the pneumatic fertilizer applicator path-wide (14 m).
A field experiment was performed in a spring wheat in Don Alfredo´s farm located in Diamante (Entre Rios, Argentina) (32° 12.560'S; 60° 33.145'W). Wheat field was cultivated according to common local farmer practices. The evaluated treatments were: i) Farmer practice (FM); ii) Regional recommendation model (RRM); and iii) drone based recommendation (D-NCR). The treatments were applied in 20 m wide and 400 m long strips with three replications. Wheat canopy showed low variability of vegetation index at first node stage, in consequence variable N prescriptions were in a narrowed range.
Fertilizations rates for the evaluated treatment were 46 kg N ha-1 (at sowing) for FP, 46 (at sowing) + 62 (at GS31) kg N ha-1 for RRM, and ranged from 12.8 to 20.2 kg N ha-1 (at GS31) for D-NCR. Grain yield was, in average, 3927, 3906 and 4035 kg ha-1 for FM, RRM, and D-NCR, respectively. In average, on the season N fertilization D-NCR improved grain yield (+108 kg ha-1), reaching a NUE of 6.6 kg grain per kg of N. These results showed that D-NCR method adequately predict a low N response, accordingly prescribe low N rates improving grain yield with a NUE higher than historical grain/fertilizer price ratio.
