Sugarcane mapping in Paraná State Brazil using MODIS EVI images.
Abstract Sugarcane cultivated in Brazil deserves attention because it makes the Country the world's largest producer of sugar and ethanol. The aim of this work was to develop and evaluate a methodology for sugarcane mapping in Paraná State, Brazil using temporal series of the MODIS EVI, for 201...
| Authors: | , , , |
|---|---|
| Format: | article |
| Status: | Published version |
| Publication Date: | 2020 |
| Country: | Brasil |
| Institution: | Empresa Brasileira de Pesquisa Agropecuária (Embrapa) |
| Repository: | Repositório Institucional da EMBRAPA (Repository Open Access to Scientific Information from EMBRAPA - Alice) |
| Language: | English |
| OAI Identifier: | oai:www.alice.cnptia.embrapa.br:doc/1123618 |
| Online Access: | http://www.alice.cnptia.embrapa.br/alice/handle/doc/1123618 https://doi.org/10.23953/cloud.ijarsg.451 |
| Access Level: | Open access |
| Keyword: | Índice de vegetação Mapeamento de cana-de-açúcar Annual agriculture Timeseries Cana de Açúcar Agricultura Sensoriamento Remoto Agriculture Sugarcane Time series analysis Vegetation index Remote sensing |
| Summary: | Abstract Sugarcane cultivated in Brazil deserves attention because it makes the Country the world's largest producer of sugar and ethanol. The aim of this work was to develop and evaluate a methodology for sugarcane mapping in Paraná State, Brazil using temporal series of the MODIS EVI, for 2010/2011 to 2013/2014 crop seasons. The methodology included supervised classification Fuzzy ARTMAP, taking as input variables such as terms of harmonics amplitude and phase, and phenological metrics of culture. Area estimates indicated a moderate and strong correlation (rs), ranging from 0.62 to 0.71 comparing with IBGE official data and from 0.79 to 0.87 with the Canasat data. To assess mapping accuracy, Canasat vector maps were used as reference to build the confusion matrix. The method developed based on Fuzzy ARTMAP proved efficient to map and estimate the acreage of sugarcane in the State of Paraná, due to digital processing techniques used in homogeneous samples, selection of phenological seasonal metrics, and decomposition of images in accordance with harmonics and supervised training. These together minimized the neural network forecast errors. Results indicate that the methodology is appropriate for sugarcane mapping. |
|---|