Raw remotely sensed satellite data have to be processed and mapped into a s
tandard projection in order to produce a multitemporal data set which can t
hen be used for regional or global Earth science studies. However, traditio
nal methods of processing remotely sensed satellite data have inherent limi
tations because they are based on a fixed processing chain. Different users
may need the data in different forms with possibly different processing st
eps; hence, additional transformations may have to be applied to the proces
sed data, resulting in potentially significant errors. Ln this paper, we de
scribe a software system, Kronos, for the generation of custom-tailored pro
ducts from the Advanced Very High Resolution Radiometer (AVHRR) sensor. it
allows the generation of a rich set of products that can be easily specifie
d through a simple interface by scientists wishing to carry out Earth syste
m modeling or analysis. Kronos is based on a flexible methodology and consi
sts of four major components: ingest and preprocessing, indexing and storag
e, a search and processing engine, and a Java interface. After geo-location
and calibration, every pixel is indexed and stored using a combination of
data structures. Following the users' queries, data are selectively retriev
ed and secondary processing such as atmospheric correction, compositing, an
d projection are performed as specified. The processing is divided into two
stages, the first of which involves the geo-location and calibration of th
e remotely sensed data and, hence, results in no loss of information. The s
econd stage involves the retrieval of the appropriate data subsets and the
application of the secondary processing specified by the user. This scheme
allows the indexing and the storage of data from different sensors without
any loss of information and, therefore, allows assimilation of data from mu
ltiple sensors. User specified processing can be applied later as needed.