Timesat for processing time-series data from satellite sensors for land surface monitoring

Abstract

The TIMESAT software package has been developed to enable monitoring of dynamic land surface processes using remotely sensed data. The monitoring capability is based on processing of time-series for each image pixel using either of three smoothing methods included in TIMESAT: asymmetric Gaussian fits, doublelogistic fits, and Savitzky-Golay filtering. The methods have different properties and are suitable for a wide range of data with different character and noise properties. The fitting methods can be upper-envelope weighted and can take quality data into account. Based on the fitted functions, growing season parameters are then extracted (beginning, end, amplitude, slope, integral, etc.), and can be merged into images. TIMESAT has been used in a number of application fields: mapping of phenology and phenological variations; ecological disturbances; vegetation classification and characterization; agriculture applications; climate applications; and for improving remote sensing signal quality. Future developments of TIMESAT will include new methods to better handle long gaps in time-series, handling of irregular time sampling, improved smoothing methods, and incorporation of the spatial domain. These modifications will enable use of TIMESAT also for high-resolution data, e.g. data from the planned ESA Sentinel-2 satellite.