Estimasi Kandungan Hara Daun Kelapa Sawit (Elaeis guineensis Jacq.) Menggunakan Citra Multispektral Berbasis Unmanned Aerial Vehicle

Abstract

Adequate nutrition is one of the important factors in its contribution to oil palm productivity. Usually, Usually, fertilization is needed to obtain sufficient nutrients based on the analysis of soil and leaf nutrients. Measuring leaf nutrient content conventionally lacks flexibility, is impractical, is labour-intensive, and takes time and is expensive, so remote sensing can be an alternative to addressing this problem. The research uses an Unmanned Aerial Vehicle (UAV) equipped with multispectral sensors that have the advantages of high spatial resolution, low cost, easier operating systems, can fly low, and can be operated at any time The independent variable used reflection value extraction for each band and 10 vegetation index transformations obtained from 3 multispectral camera bands (red, green, and near-infrared), while the dependent variable used analysis of laboratory leaf samples. The data is used to determine the best predictive model, independent variable, acquisition time, and leaf nutrient mapping in oil palm plantations. The best prediction model of nutrients N, P, K, Ca, and Mg use polynomial regression of double order 4 with R2 values in the succession of 0.986; 0.975; 0.981; 0,970; and 0.968, as well as Adjusted (Adj) R2 values successively 0.861; 0.761; 0,812; 0,710; and 0.690. The NDVI and GNDVI vegetation index combined with the single band NIR become the best predictor variables in the predictive model. The best time for image data acquisition is daytime with a MAPE value ratio of 13.03%, the lowest than morning and afternoon acquisitions for all predictive models. The output prediction is the nutritional value of the leaves of each nutrient for each plant, presented spatially to provide accuracy and ease of use.