Spatial Timeseries for Automated high-Resolution multi-Sensor data fusion (STARS)

Margaret C. Johnson (she/her)
maggie.johnson@jpl.nasa.gov
Principal investigator: lead of data fusion methodological development and Julia code implementations.
NASA Jet Propulsion Laboratory 398L

Gregory H. Halverson (they/them)
gregory.h.halverson@jpl.nasa.gov
Lead developer for data processing pipeline design and development, moving window implementation, and code organization and management.
NASA Jet Propulsion Laboratory 329G

Jouni I. Susiluoto
jouni.i.susiluoto@jpl.nasa.gov
Technical contributor for methodology development, co- developer of Julia code for Kalman filtering recursion.NASA Jet Propulsion Laboratory 398L

Kerry Cawse-Nicholson (she/her)
kerry-anne.cawse-nicholson@jpl.nasa.gov
Concept development and project management. Advised on technical and scientific requirements for application and mission integration.
NASA Jet Propulsion Laboratory 329G

Joshua B. Fisher (he/him)
jbfisher@chapman.edu
Concept development and project management
Chapman University

Glynn C. Hulley (he/him)
glynn.hulley@jpl.nasa.gov
Advised on technical and scientific requirements for application and mission integration.
NASA Jet Propulsion Laboratory 329G

Nimrod Carmon (he/him)
nimrod.carmon@jpl.nasa.gov
Technical contributor for data processing, validation/verification, and hyperspectral resampling
NASA Jet Propulsion Laboratory 398L

STARS is a general data fusion methodology utilizing spatiotemporal statistical models to optimally combine high spatial resolution VSWIR measurements with high temporal resolution measurements from multiple instruments. The methods are highly-scalable, able to fuse <100 m spatial resolution products in near-real time (<24 hrs) on regional to global scales, to facilitate online data processing as well as large-scale reprocessing of mission datasets. The statistical spatiotemporal modeling framework provides with each fused surface reflectance product associated pixel-level uncertainties incorporating any known data source measurement uncertainties, bias characteristics, and degree of historical data missingness.

The specific capabilities offered by STARS are:

1. automatic, high-resolution spatial and temporal gap-filling,
2. a tunable fusion framework allowing the user to choose a level of accuracy vs computational complexity, and
3. quantifiable uncertainties that can be used for downstream product sensitivity/uncertainty assessments and that can be incorporated into higher-order data product quality flags.

STARS is a significant advancement for surface reflectance data fusion and for quantifying (and potentially reducing) the uncertainty associated with satellite-derived inputs in retrievals of science quantities of interest.

The Julia implementation for the STARS data fusion algorithm is inSTARS.jl.

There are several supporting sub-components in generalized Julia packages, including:

• SentinelTiles.jl for geo-referencing Sentinel UTM tiles
• MODLAND.jl for geo-referencing MODIS/VIIRS sinusoidal tiles
• CMR.jl for searching the Common Metadata Repository (CMR)
• HLS.jl for searching and downloading the Harmonized Landsat Sentinel (HLS) dataset

Repositories