In Mato Grosso (Brazil), agricultural practices are evolving rapidly and their monitoring (e.g. sowing dates or double cropping systems) becomes a relevant issue. The objective of this paper is to test the potential of MODIS satellite data to detect vegetation dynamics over agricultural land in Mato Grosso. We especially focus on the estimate of the sowing dates of soybean crops. First, the MODIS MCD12Q2 product, composed of phenology transition dates, was tested at the field scale, but turned out to be unusable due to a large number of missing data and inconsistencies. We then developed an alternative method based on the MOD13Q1 Enhanced Vegetation Indices (EVI) time series. We applied a 3x3 window Savitzky-Golay filter to the EVI time series, and extracted the 2006-2007 growing period. We then calculated the dates at which different EVI values were reached (from 0.1 to 0.9, step 0.1), and correlated these dates to a set of sowing dates observed in the fields over the same period. The best result was obtained with a EVI threshold value of 0.6, with a Pearson coefficient equals to 0.8.