Satellite image time series analysis plays an important role in different domains including vegetation dynamics monitoring, estimating crop yields, discriminating between different land cover classes, exploring human-nature interactions, monitoring land cover change, assessing environmental threats, or evaluating ecosystems-climate feedbacks or urbanization.
Time series analysis requires high quality time series which are reconstructed by removing any source of contamination such as clouds, cloud shadows, or scan-line corrector (SLC) gaps of the Enhanced Thematic Mapper plus sensor (ETM+) on Landsat 7. Removed pixels are usually filled in with data predicted from a different date (temporal interpolation), nearby pixels (spatial interpolation) or from both (spatiotemporal interpolation). Different methods are available for screening and masking out clouds and shadows in satellite images including mono-temporal methods such as Function of mask (Fmask), or multitemporal mask (e.g. Tmask algorithm). Fmask is used by the United States Geological Survey (USGS) to produce a cloud mask layer of Landsat images. European Space Agency (ESA) is using Sen2cor processor to produce Level 2A Sentinel-2 data with a shadow and cloud shadow mask. All images used in the time series have to be co-registered, i.e. they align as closely as possible.
Time series analysis is used to (1) investigate various surface properties such as evapotranspiration, land surface temperature, (2) map the cover of the Earth surface (e.g. land cover mapping, crop mapping etc.), (3) detect different type of changes such as abrupt changes (fire event) or gradual changes (urbanization), and (4) study the trends.
To map surface features from satellite image time series, numerous studies make use of the vegetation phenology extracted from a spectral-temporal trajectory of a given spectral vegetation index such as the normalized difference vegetation index (NDVI) or enhanced vegetation index (EVI). Several metrics can be used to characterized vegetation phenology: metrics of greenness and metrics of time. The metrics of greenness include the minimum and maximum spectral vegetation indices, their difference or amplitude, seasonally averaged greenness etc. The metrics of time include start and end of the growing season, duration or length of the growing season or the timing of maximum greenness. Changes, on the other hand, are identified either by investigating two images acquired at two different points in time or by identifying breaks in a dense (annual or multi-annual) satellite image time series.