Open AccessUse of remote sensing data obtained from UAVs to assess the biomass productivity of Silphium perfoliatum
Author(s) -
Tamara Myslyva,
Branislava Sheliuta,
P. P. Nadtochy,
Alesia Kutsayeva
Publication year - 2021
Publication title -
vescì nacyânalʹnaj akadèmìì navuk belarusì. seryâ agrarnyh navuk
Language(s) - English
Resource type - Journals
eISSN - 1817-7239
pISSN - 1817-7204
DOI - 10.29235/1817-7204-2021-59-2-186-197
Subject(s) - biomass (ecology) , environmental science , productivity , remote sensing , raster graphics , sampling (signal processing) , vegetation (pathology) , coefficient of determination , pixel , mathematics , statistics , computer science , geology , medicine , filter (signal processing) , pathology , artificial intelligence , economics , computer vision , macroeconomics , oceanography
Agromonitoring is one of the most important sources of obtaining up-to-date and timely information about the state of agricultural crops. It is possible to speed up and reduce the cost of its implementation process using remote sensing data (RSD) obtained with the help of unmanned aerial vehicles (UAVs). Possibility of using ultra-high-resolution remote sensing to determine productivity of Silphium perfoliatum biomass has been evaluated using Phantom-4ProV 2.0 UAV. The shooting was carried out in RGB mode, the shooting height was 50 m, the spatial resolution was 2.5 cm. Based on the results of the survey, a height map and orthomosaic were created, which were later used to assess productivity of plants. To obtain the plant height values, the difference between the vegetation cover heights obtained from the surface model raster and the minimum height determined within the raster has been calculated. The actual height of plants measured in the field was compared with the data obtained using the UAV, and after the biomass productivity calculated from the actual and predicted heights was determined. The determination coefficient for equation of paired linear regression between the actual and predicted values of productivity made 0.97, and the value of the average approximation error was 3.3 %. To verify the results obtained, 60 samples of biomass were taken in the field within the study area, with the length of the plants determined using a tape measure, and the sampling sites coordinated using GPS positioning. 13 vegetation indices have been determined using pixel-based calibrated orthomosaic and normalized RGB channels, four of which (ExG, VARI, WI, and EXGR) showed to be suitable for creating a predictive model of multiple linear regression, which allows estimating and predicting the productivity of Silphium perfoliatum biomass during stemming phase with an error not exceeding 2 %. The results of the study can be useful both in development of prediction methods and in the direct prediction of Silphium perfoliatum biomass and other forage crops productivity, in particular Helianthus annuus and Helianthus tuberosus.