Minimize OLI & TIRS - Monthly Quality Reports
Radiometric Accuracy Stability Monitoring

 

Method Description

 

The methodology consists of monitoring the Top of Atmosphere (TOA) reflectance acquired over a bright stable site as a function of time for the same view and sun illumination acquisitions. Essentially, the procedure consists of TOA reflectance inter-comparisons for data acquired under the same conditions. The Libyan site (referred to as Libya4) is the reference test site for this analysis. It is located at latitude 28.55N, longitude 23.39E, at altitude 118m. The following main outputs are given and the performance reporting is organized accordingly:

  • Dispersion of TOA measurements around each center wavelength value
  • TOA reflectance values against date since launch
  • Inter-comparison of OLI data with TM and ETM+ data
  • Inter-comparison of OLI data with MERIS, MODIS Aqua and MODIS Terra data

Input Data

 

The reference input data used for the latest tests is available here.

 

Results

  • Number of Products : 21
  • Period : [2014-01-29 - 2015-03-21]
  • Mean/STD TOA reflectance blue band : 0.2119/0.26%
  • Mean/STD TOA reflectance green band : 0.3285/0.42%
  • Mean/STD TOA reflectance red band : 0.4444/0.64%
  • Mean/STD TOA reflectance NIR band : 0.5682/0.80%
  • Mean/STD TOA reflectance SWIR1 band : 0.6219/1.38%
  • Mean/STD TOA reflectance SWIR2 band : 0.5991/1.43%

 

 

 

 

 

 

 

 

Geolocation Accuracy Stability Monitoring Method Description The multi temporal geo-positional accuracy monitoring is associated to a dataset observed over a region of interest (ROI) with temporal data involved. Two test sites located in southern France are used for this analysis: La Crau and Grenada. The accuracy of each product is assessed by considering, as reference, one product selected from the dataset. Validation with an external reference (Absolute accuracy) has been done on a site-by-site basis but is not reported herein as it is not performed on a monthly basis. The standard statistics, applicable at point/product/dataset level used to report the geolocation accuracy, are defined as follows;

  • The standard deviation
  • The Root Mean Square Error (RMSE) in one direction
  • Two dimensional RMSE
  • The Circular Error probable error at 90 percentile (CE90) or 95 percentile (CE95)
The performance monitoring methodology considers a temporal data stack observed over different regions of interest as the input. Any input images are compared with a pre-defined reference image for which an in depth validation has been performed. Every new cycle, the newly observed data are assimilated and the multi temporal accuracy is updated. From assimilating a new product up to the time series up date, there are basically three main stages:
  1. The dense matching process between an input images and a reference
  2. The filtering of correlation grid and visual inspection of displacement images
  3. The accuracy analysis at a product level.
Before releasing the new accuracy specification, it is the overall consistency of the time series accuracy that is analysed, any deviations are investigated and actions taken if required. From a given input multi temporal data stack; the following main outputs are reported:
  1. Overall statistics: This is a summary of the time series accuracy to be communicated to the general audience. The statistics are based on accuracy obtained for each product. The data stack, as input, has been filtered in a way that selected results are relevant for the GCP covering the full extent of the scene and a multi temporal GCP set is considered.
  2. GCP based Circular Error (CE) at 90 percentile: The input sample for accuracy analysis includes errors on multi temporal GCPs (a GCP set common in all images of the same ROI). It is considered as a more informative approach, and there is no accuracy specification output from this approach.
  3. Product based Circular Error (CE) at 90 percentile: The input sample for accuracy analysis includes errors on each product. It is a standard approach. The accuracy specification is computed based on the filtered sample data.
  4. Multi-temporal analysis: Analysis of geolocation accuracy reports considering the observation date. The graphic plot also includes the accuracy report in the Landsat product metadata file.
  5. In depth analysis of a sample: Dedicated to products that have been, in most cases, discarded from the overall statistics (1) because of quality issues.
Input Data The reference input data used for the latest tests is available here. Results The overall statistics shown below are all data merged together, without taking into account the test site.
  • Number of Products : 10
  • Period : [2013-08-31 - 2014-02-23]
  • Mean Easting Direction : -1.214 m
  • Mean Northing Direction : -1.443 m
  • Standard Deviation Easting Direction : 6.838 m
  • Standard Deviation : 6.552 m
  • Root Mean Square : 9.75 m
  • Root Mean Square Easting Direction : 7.023 m
  • Root Mean Square Northing Direction : 6.712 m
  • Circular Error : 3.4156 m

 

 

Interband Registration Accuracy Monitoring

 

 

Method Description

 

The interband registration is assessed by means of correlation processing between two consecutive bands and is conducted on test products from the four test sites: Saragossa (Spain), Egypt, Libya4 and Belgium. A sub-sampling of 20% is done to compute correlation maps. The panchromatic band is resampled to 30m when comparing with visible and NIR/SWIR bands. It is the same ROI that is kept for temporal analysis and trending. The following categories of information are the main outputs of this analysis:

  • Statistics (mean displacements, rms) for VIS/NIR/PAN bands
  • Statistics (mean displacements, rms) for VIS/TIR bands
  • Error budget results

Input Data

 

The reference input data used for the latest tests is available here.

 

Results