Evaluation of supervised classification capability of Landsat-8 and Sentinel-2A Satellite images in determining type and area of Pistachio Cultivars

Subject Areas : Agriculture, rangeland, watershed and forestry

Hadi  Zare khormizi ¹ (PhD Student of Range Management, Faculty of Natural Resources, University of Tehran, Karaj, Iran)
Hamid Reza  Ghafarian Malamiri ² (Assist. Prof. Department of Geography, Faculty of Humanities & Social Sciences, Yazd University, Iran)
Morad  Mortaz ³ (MSc. Graduated of Horticulture and Agronomy, Department of Plant Science, University of California, Davis, California)

Received: 2019-12-21 Accepted : 2020-03-07 Published : 2020-03-20

Keywords: Pistachio cultivars, remote sensing, Maximum likelihood algorithm, Kappa coefficient, Supervised classification,

Abstract :

Remote sensing technique is one of the most effective tools for monitoring, studying and determining the cultivation area of agricultural and horticultural crops, especially on a large scale. Planners, managers, and farmers, with knowledge of the type and extent of crop cultivation, can adopt appropriate management and enforcement policies. The purpose of the present study was to evaluate the supervised classification ability to classify Landsat 8 and Sentinel-2A multi-band satellite imagery in determining the cultivated area and type of four varieties of Pistachio namely such as; Akbari, Kalle Ghuchi, Ahmad Aghaei and Fandooki in an orchard in the Yazd province. In the present study, the accuracy of four classification algorithms, namely: Parallelepiped classification, Minimum distance, Mahalanobis distance and Maximum likelihood, as well as the optimum time in the separation of pistachio cultivars, were investigated. According to the classification results of a Landsat-8 image, on June 12, 2018, the Maximum likelihood algorithm with a final accuracy and Kappa coefficient of 76.8% and 0.67% and Parallelepiped classification algorithm with the final and Kappa coefficients of 64.7 and 0.47, were of highest and lowest accuracy among others, respectively. Also, according to the results, the best time for the separation of Pistachio cultivars was in late June. The Kappa coefficient of maximum likelihood classification algorithm on June 22, July 23, August 24 and September 25 of 2018 were 0.67, 0.64, 0.63 and 0.63, respectively. The final accuracy and Kappa coefficient of maximum likelihood classification algorithm on the Sentinel-2A Satellite images on 12 June  2018, were 80% and 0.71, respectively. By applying the median filter with a 3×3 dimensional kernel window size on the classified image, the final accuracy and Kappa coefficient was increased to 82.6% and 0.75, respectively. The final accuracy and Kappa coefficient of classification and separation of Pistachio cultivars in Sentinel-2A images were higher than in Landsat-8 images. Overall, based on our results, the remote sensing classification techniques, as well as multi-spectral satellite imagery, are suitable for agricultural and horticultural mapping.