Burnt areas as a result of wildfires can be readily detected from high resolution aerial photographs or satellite imagery of the zone that includes the wildfire. Moderate resolution remote sensing data, as provided by MODIS, can also be used to detect active or past wildfires, most usually from daily records of a suitable combination of reflectance bands. The objective of the present work was to test some simple algorithms and variations for automatic blind detection of burnt areas from MODIS biweekly vegetation indices time series data. MODIS derived NDVI 250m time series data for the Valencia region, Southeast Spain, were subjected to a two-steps process for the detection of candidate burnt areas, and the results compared with the record of wildfires with affected area greater than 100 hectares. For each pixel and date in the data series, a model was fitted to both the previous and posterior time series data. Discrepancies or jumps between the pre- and post- models exceeding a certain threshold were used as seeds to define cluster of pixels, the candidate burnt areas, with similarities between pixels either from their extreme discrepancy dates or from their parameters in the fitted models. Results using a simple combination of a constant fitted model and pixel similarity from jump dates were in good agreement with the perimeters of the actual burnt areas. A computationally efficient implementation of the method was developed using a digital filter type approach.