Comparison of two summer crop mapping methods in the state of Paraná in 2008

Comparison of two summer crop mapping methods in the state of Paraná in 2008
Daniel Garbellini Duft1; Jerry Adriani Johann2; Jansle Vieira Rocha1; Rubens Augusto Camargo Lamparelli3

1. State University of Campinas - FEAGRI/UNICAMP, P.O. Box: 6011, ZIP Code:13081-970- Campinas, SP, Brasil (danielduft@gmail.com; jansle.rocha@feagri.unicamp.br)
2. State University of West Paraná – UNIOESTE, P.O. Box: 0701, ZIP Code: 85819-110 – Cascavel, PR, Brasil (jerry.johann@hotmail.com)
3. Interdisciplinary Center for Energy Planning – NIPE, Campinas, SP, Brasil (rubens.lamparelli@gmail.com)

Abstract: The accuracy of crop mapping is still very dependent on the quality of processing remote sensing data, so the objective of this paper is to compare the accuracy of two different methodologies for mapping summer crops in the State of Paraná, Brazil. EVI (Enhanced Vegetation Index) images were used for this purpose, following these methodologies: classic mask generation by the difference in EVI this methodology were used in some studies such as (Schroeder et al. 1999; Labus et al. 2002; Jakubauskas et al, 2002;Fontana et al, 2007), and the RGB method, as used by Johann et al (2011).The quality of the methods was measured using higher spatial resolution images (LANDSAT and AWIFS / IRS) as a reference to calculate the Kappa Index (KI) (Congalton, 1991), Overall Accuracy(EG), omission (EO) and commission (EI) errors. With 400 random samples for each mask, an error matrix was drawn up. The results for the classic methodology presented values of: KI=0.78, EO=11% and EI=11%, while the RGB methodology presented a KI=0.895, EO=6.1% EI=5.9%. KI was 12.8% higher for the RGB methodology as EO and EI were 4.9% and 5.1% lower, respectively. The RGB method presented a greater accuracy potential than the conventional method.

Keywords: Mapping, Crops, Paraná, EVI, Kappa Index

AttachmentSize
DuftAccuracy2012.pdf164.38 KB