Abstract

In today’s information-driven geospatial world, an increasing number of Earth-observing satellite sensors are being launched to meet insatiable demands of timely and accurate datasets, for use in understanding the complex Earth landscape. Traditional methods for evaluating radiometric performance of the vast quantity of short lived satellites quickly are coming up short. One traditional method for evaluating sensor radiometric performance is via the use of Pseudo-Invariant Calibration Sites (PICS). PICS have been extensively used for radiometric calibration and temporal stability monitoring of optical satellite sensors, but traditionally the calibration approach only used a small sets of ROIs over established PICS located throughout North Africa. Coupled with cloud and weather, led to a limited number of data points being available for any give location (~1/month). The team at South Dakota State University, has expanded the scope of PICS from small isolated regions, to vast continental scaled regions, through the use of global unsupervised classifications maps of invariant natural targets to create extended PICS (EPICS).

EPICS offer uncertainties between 3-5% temporal stability levels, but at temporal frequencies at or greater then a data point a day. A key advantage of the cluster approach is that large numbers of pixels are aggregated into contiguous homogeneous regions sufficiently distributed across the continent to allow multiple imaging opportunities per day, as opposed to imaging a typical PICS once during the sensor’s revisit period. Initial temporal results from this work using a single cluster called “Cluster 13” will be used to evaluate Landsat 7/8 and Sentinel 2 A/B.

In addition, when information from Cluster 13 is combined with Hyperion hyperspectral results, the site becomes a vast cross calibration site, capable of correcting for spectral band differences. Results using this approach will be used expand the number of cross calibration opportunities between Landsat 8 and Sentinel 2A to ~122 near coincident cross-calibrations pairs over Cluster 13 in a year, as compared to 8 using traditional methods over the same period. The calibration gains of different bands derived from the cluster-based approach are similar to the traditional PICSbased method but with lower uncertainty.

Furthermore, the Extended PICS Absolute Calibration Model (ExPAC) model was developed using Landsat 8 OLI Cluster 13 data and the knowledge of geometry illumination angles; solar zenith, sun azimuth, satellite zenith, and satellite azimuth angles to create BRDF. Cluster 13 hyperspectral profile is used to create a scale factor to match any satellites to Landsat 8 OLI. The validation of ExPAC model with Landsat 8, Landsat 7, Sentinel 2A and 2B shows that the model can predict satellite measurements with accuracy to be within 2% for all spectral bands with random uncertainties better than 3% for all bands except Coastal Aerosol and Blue band with 4%. The ExPAC model can increase temporal resolution for calibration points to be on a daily basis. It can be an alternative independent technique, reliable and less expensive to monitor the performance of satellites; especially the short-lived satellites, thanks to the coverage of Cluster 13 in Northern Africa.

This new expansion of PICS to EPICS, is an exciting advancement for calibration, addressing many of the challenges radiometric calibration on the quantity, and shorter lived small satellites on orbit.

Share

COinS
 
Jun 17th, 4:55 PM

Classification of North Africa for Use as an Extended Pseudo Invariant Calibration Site for Radiometric Calibration and Stability Monitoring of Optical Satellite Sensors

In today’s information-driven geospatial world, an increasing number of Earth-observing satellite sensors are being launched to meet insatiable demands of timely and accurate datasets, for use in understanding the complex Earth landscape. Traditional methods for evaluating radiometric performance of the vast quantity of short lived satellites quickly are coming up short. One traditional method for evaluating sensor radiometric performance is via the use of Pseudo-Invariant Calibration Sites (PICS). PICS have been extensively used for radiometric calibration and temporal stability monitoring of optical satellite sensors, but traditionally the calibration approach only used a small sets of ROIs over established PICS located throughout North Africa. Coupled with cloud and weather, led to a limited number of data points being available for any give location (~1/month). The team at South Dakota State University, has expanded the scope of PICS from small isolated regions, to vast continental scaled regions, through the use of global unsupervised classifications maps of invariant natural targets to create extended PICS (EPICS).

EPICS offer uncertainties between 3-5% temporal stability levels, but at temporal frequencies at or greater then a data point a day. A key advantage of the cluster approach is that large numbers of pixels are aggregated into contiguous homogeneous regions sufficiently distributed across the continent to allow multiple imaging opportunities per day, as opposed to imaging a typical PICS once during the sensor’s revisit period. Initial temporal results from this work using a single cluster called “Cluster 13” will be used to evaluate Landsat 7/8 and Sentinel 2 A/B.

In addition, when information from Cluster 13 is combined with Hyperion hyperspectral results, the site becomes a vast cross calibration site, capable of correcting for spectral band differences. Results using this approach will be used expand the number of cross calibration opportunities between Landsat 8 and Sentinel 2A to ~122 near coincident cross-calibrations pairs over Cluster 13 in a year, as compared to 8 using traditional methods over the same period. The calibration gains of different bands derived from the cluster-based approach are similar to the traditional PICSbased method but with lower uncertainty.

Furthermore, the Extended PICS Absolute Calibration Model (ExPAC) model was developed using Landsat 8 OLI Cluster 13 data and the knowledge of geometry illumination angles; solar zenith, sun azimuth, satellite zenith, and satellite azimuth angles to create BRDF. Cluster 13 hyperspectral profile is used to create a scale factor to match any satellites to Landsat 8 OLI. The validation of ExPAC model with Landsat 8, Landsat 7, Sentinel 2A and 2B shows that the model can predict satellite measurements with accuracy to be within 2% for all spectral bands with random uncertainties better than 3% for all bands except Coastal Aerosol and Blue band with 4%. The ExPAC model can increase temporal resolution for calibration points to be on a daily basis. It can be an alternative independent technique, reliable and less expensive to monitor the performance of satellites; especially the short-lived satellites, thanks to the coverage of Cluster 13 in Northern Africa.

This new expansion of PICS to EPICS, is an exciting advancement for calibration, addressing many of the challenges radiometric calibration on the quantity, and shorter lived small satellites on orbit.