Currently, the crop forecast in Brazil is made subjectively by means of questionnaires to producers and technicians of the main agricultural regions, by employees of the institutions responsible for estimating crop. Thus, the levels of accuracy are highly dependent on interviewers and respondents involved in the survey.
Although there is the knowledge of the potential use of satellite images, in Brazil, there is not a solid model, based on remote sensing and statistics, which can provide information about agricultural production in an accurate, objective and serve parameter for policy-making in the agricultural sector. The most promising way for the establishment of a viable crop forecast, or at least to estimate the area with the major crops, is through the use of satellite images, with the support of GIS. The introduction of remote sensing as a tool in agriculture was the culmination of the development of agricultural survey data. Remote sensing is a potential tool for acquiring information relating to coffee due to her character multispectral, repeatability in time and space, and relatively low cost compared to other methods.
Research has shown that the cultivation of coffee, although with variations in spectral behavior due to factors such as spacing, age, time of year, can be identified and mapped on satellite images with reasonable accuracy, since there is additional information about the culture, whether subjective surveys or technicians working at the place of interest. Another factor that can influence the spectral behavior of the cultivation of coffee is the topography that significantly influences the spectral response of targets in a specific area.
Still in the context of using mapping images for coffee plantations, these data can be obtained in two ways: by means of visual interpretation and classification algorithms, which in turn can be performed through the process pixel by pixel, or by extraction of homogeneous regions. Thus, this article aims mapping coffee plantations in Paraná State, in Brazil, by use of geotechnology, in order to support the survey of coffee crops.
Material and methods
The State of Paraná has a 93.250 hectares of coffee plantations, according to information from the fourth estimate of coffee held by the National Supply Company – (Conab), in 2010, been highlighted as the sixth largest producer in the Brazilian national ranking.
The climate, according to Köppen classification, is the CFA type, characterized by a subtropical climate with hot summers and infrequent frosts. Average temperatures are below 18 ºC in winter and above 22 ºC in the summer time, season that concentrates the highest rainfalls. However, there is no dry season in winter. In this region, rainfall occurring between 1300 mm to 1500 mm per year.
To update the mapping of crops were selected TM sensor images – Landsat 5 in a digital format from the National Institute for Space Research (INPE). We obtained 10 scenes, which overlie the entire State of Paraná, dated between July and September 2010, corresponding to the period without interference from clouds.
In the period from October to March, there is the seed filling stage, with the greater vigour and with maximum photosynthetic activity, ideal for acquiring satellite images for mapping purposes. However, in this period other crops are cultivate as well, which contributes to a higher probability of mapping error due to the spectral similarity between different plant species. For this reason, it was important to use images obtained during the dry season, between the months of June to September, because at this time of year the spectral contrast of coffee and other targets is very highlighted, even though it was made the harvest and caused detachment of leaves.
The acquired images were of medium spatial resolution (30 m pixel), however, through restoration techniques could improve the quality of degraded image and also obtain an image with better effective spatial resolution (20 m pixel).
After restoration, the georeferencing was done frame by frame through the ENVI software, which was adopted Geocover Landsat 5 TM images as the basis and was admitted an error RMS less than 0.5.
Once created the database with Landsat TM-5 images, the mapping of coffee zones was done through visual interpretation of the crops and through delimitation of areas by polygons through software ArGIS in each Landsat scene. In terms of technology and procedures, the development of a method of image analysis was highlighted, which was consisted in to combine the use of a computer package of geographic information system (ArcGIS) and visual mapping with the aid of pictures of medium and high resolution available in Google Earth. Thus, Google Earth images served accessory for the interpretation of ground truth, since these enable greater expansion of scale for the identification of coffee zones.
After mapping, it was collected georeferenced points in coffee plantations for a statistical validation of the use of geo mapping. Thus, the validation consisted basically in the field gathering points equivalent to coffee plots, using a GPS navigation and a browser – GPS TrackMaker® – for comparison and validation with mapping based on satellite images.
Results and discussion
During the interpretation of satellite images phase, it was observed that the spectral behavior of coffee, when properly conducted, in Paraná State, expressed a reddish hue, which varied depending on the leaf area index of the crop.The variation in hue of spectral response of coffee is related to leaf area index of the crop, which was influenced by planting system (traditional, folded or thickened), age and detachment of leaves during harvest, handling of pruning, intercropping with other cultures and use of windbreaks.
In the field validation, performed in the first mapping, in 2008, were evaluated different crops coffee management as a radical prune, which is held clearcutting coffee; alternating prune, when clearcutting is conducted in alternating lines in culture; the framing or radical pruning of the plant; cultivation of culture shaded, it means in conjunction with woody species; and the growth of new areas, and also interference relief and soil class, both with great variability in the State. Even with this diversity in management and landscape, coffee can be identified in satellite images, however with some difficulty due to the large amplitude of the spectral response of the coffee culture in the State of Paraná.
In Paraná, the use of geotechnology identified 70.790 ha of coffee, which represented 76% of the total area that was released by Conab. This difference can be explained, first, due to the methodology adopted, where the geoprocessing tool is characterized as an objective method of obtaining data, it means, the field is identified promptly while the estimates made by Conab start from subjective methodologies. Therefore, it is natural to estimate distinct areas. Coffee plantations aged up to a year and a half were not noticeable in the Landsat images, because it doesn’t express a typical spectral behavior of coffee. The various managements, cropping systems and possible interference of relief may also have contributed to the identification of areas cultivated with coffee, for influence on their spectral characteristics.
Another difficulty highlighted is the confusion between the spectral pattern of coffee with other crops, especially with the orange, lemon, eucalyptus and banana. In some cases, areas of native forest and field areas near small coffee zones, around three hectares, also may have caused some confusion.
Due to the difficulties mentioned above, the field validation is necessary and the use of geo-technologies should be considered as a complementary and not exclusive of estimated crop. In Figure 1, you can display the result of the coffee mapping by use of geotechnology in Paraná in 2010. On the map, there is a concentration of the coffee plantations in the northeastern portion of the State, which, as noted in the validation field, climatic conditions and the type of clayey soil promote the development of culture at lower costs.
The accuracy or reliability of the coffee mapping, in the state of Paraná was 78%, as the validation performed on the first mapping in 2008, after correlation mapping performed on satellite images and field verification.
Figure 1: Mapping of the coffee crop in Parana State by using geothecnology.
Coffee culture has spectral response rather complex because of the different farming systems and management. The use of geotechnology was considered viable for mapping the coffee plantations in the State of Paraná. It was mapped an area of 70.790 ha of coffee in Paraná.