Satellite images accessed via the Internet have become popular for finding addresses or for seeing the layout and surroundings of a future trip destination. In a professional context, especially in agriculture, these images now allow monitoring of large areas and estimation of crop production. This is done through remote sensing techniques used to process and interpret the photographs and data obtained by satellite-based sensors. Recently, a new type of agricultural monitoring, related to the cultivation of sugarcane, won first place in the Academic category of the fifth edition of the Top Ethanol Award, an initiative of the Sugarcane Industry Union (Unica) in partnership with other sugarcane producer associations and companies, such as Dedini, Basf, Monsanto and Syngenta. The winning team is from the National Institute for Space Research (INPE) in São José dos Campos, São Paulo state. Researcher Marcio Pupin Mello led the study, and developed software to automate the mapping of sugarcane crops throughout the season, using satellite images.
The method, which was published in 2013 in the journal IEEE Transactions on Geoscience and Remote Sensing, was first developed in England in 2000 by Professor Carlos Vieira. Currently at the Federal University of Santa Catarina (UFSC), Viera was finishing his doctorate at the University of Nottingham at the time. In 2007, the method began to be integrated into applications within Canasat, an INPE program to monitor sugar cane plantations in the South Central region of Brazil via satellite.
Launched in 2003, Canasat, in addition to estimating and mapping sugarcane crops, is also employed to identify whether a specific section of a crop was harvested with or without the use of straw burning in the pre-harvest phase. Agricultural-environmental legislation in the state of São Paulo requires the gradual reduction of straw burning and complete elimination of this type of agricultural process by 2031. Anticipating the provisions in the law, the São Paulo State Secretary of the Environment signed a protocol with the sugarcane industry to eliminate this practice throughout most of the state by 2014. Since 2006, the images have been used to evaluate the progress of the gradual reduction of the area burned in the state of São Paulo. From 2006 to 2013, this area was reduced from 65.8% to 16.3%.
The analysis is performed by technicians who examine images one by one to determine the type of sugarcane harvesting: with or without the use of fire. Despite achieving very high levels of accuracy, mapping based on visual interpretation is cumbersome because technicians need to interpret and map each sugarcane harvest region as seen in several images collected over time. The new system does the same work, but with automatic analysis. “It processes the satellite images obtained on different dates throughout the season and automatically detects whether the sugarcane harvest was carried out with or without straw burning,” says Mello, who worked on the development of the new system during his master’s and part of his PhD in the INPE Graduate Program in Remote Sensing.
“From images obtained from U.S. satellites Landsat-5 and Landsat-7, we identified changes in the energy reflected by areas of sugarcane crops,” said Mello, who completed part of his PhD studies at the Institute for Geoinformatics, University of Münster, Germany, and is currently coordinator of remote sensing research at Boeing in Brazil. “We identify the variation in the energy reflected by the sugarcane over time. With this information, we can interpret what is in the field, whether it is straw, or a growing plant, for example, and identify each harvested section throughout the season on the images,” says Professor Bernardo Rudorff, who retired from INPE, where he coordinated the Canasat project, and is now a partner in Agrosatélite, a company in Florianópolis, Santa Catarina state that specializes in remote sensing applied to agriculture.
The new system called STARS (Spectral-Temporal Analysis by Response Surface), may also be useful for monitoring deforestation. Automatic evaluation could be performed by analyzing changes in the spectral patterns of areas of vegetation over time, from forests to exposed soil. “I believe this method could be helpful to supervisory agencies, both for checking sugarcane burning and for environmental monitoring of forests,” says Mello.
At the moment the software is not yet in commercial use, but the algorithms can be accessed on the INPE website at www.dsr.inpe.br/~mello. To improve image processing and make the software operational, in addition to the possibility of exploring other applications, Mello says he is waiting for an incoming graduate student at INPE, where he supervises students, to continue the work. “My degree was in engineering and I implemented the software for my research, but if an expert in software programming were to take on the project I believe he could turn it into a product,” says Mello. Software for automating remote sensing, principally for time analysis of images, is in high demand in the field. “There are more and more satellites with plenty of capacity to obtain images of the earth’s surface, soil data, and agricultural crops. Therefore, we need to increase the automation of analyses,” says Rudorff.
AGUIAR, D. A. et al. Remote sensing images in support of environmental protocol: Monitoring the sugarcane harvest in São Paulo state, Brazil. Remote Sensing. v. 3, n. 12, p. 2682-703. 2011.
MELLO, M.P. et al. Stars: A new method for multitemporal remote sensing. IEEE Transactions on Geoscience and Remote Sensing. v. 51, n. 4, p. 1897-913. abr. 2013.