The main requirement for BACI SSV dataset was to provide frequent time series of remote sensing information in different domains of electromagnetic spectrum covering largest possible regions. It was important to have data which allows change detection to be as precise as possible without attribution. The dataset combines layers of optical, thermal infrared and microwave data providing comprehensive set of information.

The process used MODIS reflectance, MODIS land surface temperature and Sentinel-1 VV/VH backscatter. It also employed linear Kernel BRDF models to normalise reflectance to nadir view. i.e.and an inversion of the Kernel models to obtain kernels and then it is easy to calculate reflectance at nadir. In the case of thermal and SAR information the process used identity operator i.e. smoother to fill gaps and estimate uncertainty. This allows minimum loss of information and makes data sets compatible.
The main difference between SSV datasets and conventional way of representing data is availability of information about associated uncertainties. This allows to see the extent to which we can trust specific pixel at specific date/time. Most of the conventional change detection and time series decomposition methods do not take uncertainty into account. This can lead to misinterpretation of data due to atmospheric effects, processing or model errors. The result was smooth continuous time series with associated uncertainties and restored time/space gaps. We exploit temporal regularization which was presented in see {Quaife2010} and {Lewis2012a} in data set documentation). This technique allows filling gaps in the time series of parameters and explicitly characterize the output uncertainties.

Inputs to the BACI SSV are MODIS daily reflectance and LST data, Sentinel 1 backscatter and historical microwave (ENVISAT ASAR). A key innovation of the BACI SSV processing chain is the use of the multitasking facilities of CEMS/JASMIN cluster to process almost 20 years of EO data across domains .