Different types of changes are investigated using remotely sensed data: (i) abrupt changes, such as the changes caused by a fire or flooding, and (ii) gradual changes such as urban growth. Besides these kinds of changes, remote sensing community differentiates between transitional changes and conditional changes. Transitional changes refer to a major change of land surface such as conversion of forest to pasture or the expansion of mangroves into the surrounding water. Conditional changes refer to the change in condition at the surface such as water stress in an agricultural field, forest degradation caused by pest.
In the past, many remote sensing studies used two images to detect different types of changes such as deforestation, land cover change or change in the health or condition of the vegetation (e.g. pest infestation). Meanwhile, satellite image time series are used to assess the change. Time series analysis allows for monitoring more subtle changes and for providing temporal patterns of change. In this way, the timing of changes and drivers of change can be easily identified.
Different methods are being used in change detection studies. There are studies that analyze individual images available in the investigated time series to map the target class/phenomena/events at the time when images were collected and to identify the changes: e.g. mapping the mangroves extent on an year basis and measuring it to identify changes. Alternative studies search for breaks in time series for detecting changes. The breaks are used to segment the time series into before and after changes periods which are further classified using one of the existing supervised or unsupervised classification methods (K-means, fuzzy k-means, Random Forest, Support Vector Machine etc.).