International journal of applied earth observation and geoinformation2019，Vol.8210.DOI:10.1016/j.jag.2019.05.005

Detailed agricultural land classification in the Brazilian cerrado based on phenological information from dense satellite image time series

Bendini, Hugo do Nascimento Fonseca, Leila Maria Garcia Schwieder, Marcel Korting, Thales Sehn Sanches, Ieda Del Arco Leitao, Pedro J. Hostert, Patrick Rufin, Philippe
International journal of applied earth observation and geoinformation2019，Vol.8210.DOI:10.1016/j.jag.2019.05.005
✕
  • NSTL
  • Elsevier

Detailed agricultural land classification in the Brazilian cerrado based on phenological information from dense satellite image time series

Bendini, Hugo do Nascimento 1Fonseca, Leila Maria Garcia 1Schwieder, Marcel 2Korting, Thales Sehn 1Sanches, Ieda Del Arco 1Leitao, Pedro J. 2Hostert, Patrick 2Rufin, Philippe2
扫码查看

作者信息

  • 1. Natl Inst Space Res INPE, Sao Jose Dos Campos, SP, Brazil
  • 2. Humboldt Univ, Geog Dept, Unter Linden 6, D-10099 Berlin, Germany
  • 折叠

Abstract

The paradox between environmental conservation and economic development is a challenge for Brazil, where there is a complex and dynamic agricultural scenario. This reinforces the need for effective methods for the detailed mapping of agriculture. In this work, we employed land surface phenological metrics derived from dense satellite image time series to classify agricultural land in the Cerrado biome. We used all available Landsat images between April 2013 and April 2017, applying a weighted ensemble of Radial Basis Function (RBF) convolution filters as a kernel smoother to fill data gaps such as cloud cover and Scan Line Corrector (SLC)-off data. Through this approach, we created a dense Enhanced Vegetation Index (EVI) data cube with an 8-day temporal resolution and derived phenometrics for a Random Forest (RF) classification. We used a hierarchical classification with four levels, from land cover to crop rotation classes. Most of the classes showed accuracies higher than 90%. Single crop and Non-commercial crop classes presented lower accuracies. However, we showed that phenometrics derived from dense Landsat-like image time series, in a hierarchical classification scheme, has a great potential for detailed agricultural mapping. The results are promising and show that the method is consistent and robust, being applicable to mapping agricultural land throughout the entire Cerrado.

Key words

Big data/Time-Series mining/Random forest algorithm/Land use and Land cover mapping (LULC)/Multi-Sensor

引用本文复制引用

出版年

2019
International journal of applied earth observation and geoinformation

International journal of applied earth observation and geoinformation

SCI
ISSN：0303-2434
被引量16
参考文献量50
段落导航相关论文