This paper proposes a new approach of change detection that reduces seasonality in time series by using Photosynthetic Vegetation Time Series (PVTS) from satellite images. With this approach, each pixel value represents at the subpixel level a fraction of the photosynthetic forest's activity. Our hypothesis is based on an assumption that photosynthetic vegetation fractions will remain constant until a disturbing agent (natural or anthropic) occurs. Using Landsat data, we compared our approach with the Carnegie Landsat Systems Analysis-Lite (CLASlite) approach and with the national reports of the Ministry of the Environment of Perú (MINAM). After reducing seasonal variations in Landsat data, we detected deforestation events with a new detection method. Our approach (which was called PVts-β) of detection is a simple method that does not model the seasonality and it only requires as inputs: i) the average and standard deviation of the time series of a pixel and ii) a threshold magnitude (β) that was calibrated to detect deforestation events in tropical forests. For the PVts-β approach, the results of calibration show that deforestation was optimally detected for β = (5,6), higher or lower than this range, the biases favor to false detections and favor the omission of deforestation too. On the other hand, the overall accuracy for the PVts-β approach was 91.1%, with an omission and commission of 8.3% and 0.5% respectively, while for CLASlite the overall accuracy was 79.2%, with an omission and commission of 20.8% and 0.0% respectively. The differences in the overall accuracy between the PVts-β and CLASlite approach were significant, being atmospheric noise a main problem which CLASlite usually does not work optimally. The strength of our PVts-β approach is the early detection at the subpixel level of deforestation events that, added to our new method of change detection explain the little omission obtained in the results. Therefore, the PVts-β approach -that we propose here- provides the opportunity to monitoring deforestation events in tropical forests at sub-annual scales using Landsat data, and it can be used for near-real-time change detection monitoring without a doubt.