Seleção de grades para mapeamento de plantas espontâneas

Abstract

The adoption of precision agriculture in weed control constitutes an alternative to traditional agriculture, which causes environmental impacts and unnecessary costs when your control is deficient. The maps of spontaneous plants represent the spatial variability of weeds that compete with crops of cultivated fields. The location survey of these plants allows optimizing process control in differentiated application of inputs. The different infestations mapping and location are difficult tasks to be performed on a large scale and require appropriate methodologies. In this context, the present study aimed to evaluate the degree of similarity presented by thematic maps generated by different sampling weed grids in three commercial farming areas, comprising a total area of 23.45 hectares. A counting of monocots and dicots in the 2012/2013 and the 2013/2014 harvests was done before planting soybeans in the town of Realeza, in the west of the state of Paraná, in the fallow period done after a harvest of wheat of both years. A regular grid of 10 x 10 m for invasive sampling was done, which was used as a reference, being made a counting at 1 m² of each sample point, a total of 2344 samples each year, which was compared to regular grids of 30 m and 50 m, generated from deleting data from the standard grid. For soil samples a regular grid of 50 x 50 m was established, comprising 70 composite soil samples, which were submitted to physical and chemical analysis. For the generation of thematic maps, the inverse interpolation of the squared distance, in comparison of maps generated from each grid with the reference map, using the kappa statistic to assess the loss in quality of the maps while reducing the number of sample points. It was observed a poor spatial correlation of the physical and chemical attributes analyzed, with the type of plant wide and narrow leaves; the loss of quality of the maps were lower in 2013 compared to 2012, when reduced the density of sample points. The 30 x 30 m grids described in a satisfactory way the data of wide leaf infestation, and the 50 x 50 m grids described appropriately the infestations of spontaneous narrow leaves, when compared to the standard grid (10 x 10 m).