The objective of this work was to develop and evaluate a method for estimating soybean crop area on a regional scale and to calculate the statistical error associated with the estimate. The method (Geosafras), which combines statistical sampling techniques with characteristics of images obtained by orbital remote sensing, was applied to obtain an objective sampling estimation for soybean crop area in the 2005/2006 harvest season in the state of Rio Grande do Sul (RS), Brazil. Soybean‑producing municipalities in RS were distributed into ten strata according to preexisting data on the area cultivated with the crop. The number of municipalities selected in each stratum followed Neyman's allocation rule. In each selected municipality, points corresponding to the pixels of images were randomized and classified as "soybean" or "non‑soybean" after site visitation. From the data of 3,000 points in the 30 selected municipalities across the ten strata, soybean crop area in RS was estimated, totaling 4,069,887 ha, with a coefficient of variation (CV) of 3.4%. This estimate was consistent with official data. The stratified objective sampling method, supported by remote sensing, allows for the estimation of the area cultivated with soybean in the state of Rio Grande do Sul and is able to quantify the error associated with the calculated estimate.
Glycine max; sampling error; agricultural statistics; stratification; satellite image; crop forecasting
