ICDSUPL1-P014

Volume: 1, 2022
1^st International PhD Student’s Conference at the University of Life Sciences in Lublin, Poland: ENVIRONMENT  – PLANT  – ANIMAL  – PRODUCT

Abstract number: P014

DOI: https://doi.org/10.24326/ICDSUPL1.P014

Published online: 26 April 2022

ICDSUPL, 1, P014 (2022)

A case study on potato yield prediction with remote sensing methods

Renata Leszczyńska¹*, Stanisław Samborski¹, Dariusz Gozdowski²

¹Faculty of Agriculture and Ecology; Department of Agronomy; Warsaw University of Life Sciences; Nowoursynowska 159, 02-776 Warsaw, Poland

² Department of Biometry, Warsaw University of Life Sciences, Nowoursynowska 159, 02-776 Warsaw, Poland

^{* Corresponding author: renata_leszczynska@sggw.edu.pl}

Abstract

Free access to images from Sentinel-2 satellites and getting more cost-effective platforms such as unmanned aerial vehicles (UAVs) significantly increase the opportunities to obtain timely information on the variability of crop canopy. A critical issue supporting agrotechnical and marketing decision making in farms is the prediction of potato yield before harvest. However, remote sensing is not often used when referring to single potato fields. The objective of this research was to quantify the relationship between potato yield and vegetation indices (VIs) derived from satellite and UAV imagery and estimate yield prediction error. Two potato fields cropped with a medium early cv. Innovator located in Pomerania voivodeship were used for this study, field 1 in Damno (16.9 ha) and field 2 in Wiślinka (10.1 ha). Cloudless Sentinel-2 (European Space Agency, https://eos.com/landviewer), images were obtained nine times for field 1 and 13 times for field 2, from 15 to 139 days after planting (DAP). Normalized Difference Vegetation Index (NDVI) was derived from these images using the QGIS raster calculator. UAV images were collected four times using YUNEEC H520 (Yuneec Holding Ltd. Company, Hong Kong, China) attached with an RGB camera and one time using DJI, Phantom 4 Multispectral (DJI, Shenzhen, China) from 46 to 125 DAP. The UAVs were flown during ±2 h solar noon (10 AM to 2 PM) at an altitude of 80 m. Images acquired for each site were stitched together using the Pix4Dfields software (Pix4D, Lausanne, Switzerland) to generate qualitative, high-resolution (4 cm/pixel) orthomosaic imageries. Respectively, NDVI was generated from DJI imageries and Normalized Green-Red Difference Index (NGRDI) from Yuneec imageries using the QGIS raster calculator. The potato was manually harvested at the termination of the crop over an area of 3 m2, respectively, at thirty and seventeen sampling points on field one and field 2.  The highest correlations between yield and VIs were achieved at mid to end of August at the end of tuber bulking/beginning of vines yellowing: – satellite Sentinel-2: (R2=0.38-0.53), – UAV: (R2=0.62-0.69). The Mean Absolute Error of yield prediction for VI derived from imagery was for satellite Sentinel-2: 3.54-3.60 (t∙ha-1) in Damno and 2.54-2.97 (t∙ha^-1) in Wiślinka, and for UAVs: 2.85-3.56 (t∙ha-1) in Damno and 2.31-2.85 (t∙ha-1) in Wiślinka.

How to cite

R. Leszczyńska, S. Samborski, D. Gozdowski, 2022. A case study on potato yield prediction with remote sensing methods. In: 1st International PhD Student’s Conference at the University of Life Sciences in Lublin, Poland: Environment – Plant – Animal – Product. https://doi.org/10.24326/ICDSUPL1/P014