Scientific Quality Assurance
The scientific quality of Terrascope data products is regularly assessed. Such quality assessment can generally be performed through comparison with:
For the validation of satellite retrievals of land surface properties, such as the Sentinel-2 vegetation indicators provided by Terrascope, the Committee on Earth Observation System Land Product Validation (CEOS-LPV) has set-up guidelines for standardized intercomparison and validation across products from different satellite, algorithms, and agency sources.
Comparison with other satellite retrievals
Ideally, one wants to compare a retrieved satellite-measured quantity with ground-based observations. However, such ground-based observations are not always available, due to for example insufficient spatial and temporal data coverage. Therefore inter-comparisons with the same or similar retrievals from other satellite instruments can be performed.
In such comparisons, the following differences between the two (or more) retrieval datasets need to be accounted for:
- Spatial resolution
- Spectral resolution
- Temporal sampling/aggregation (e.g. differences in overpass time of the satellites)
- Algorithmic differences (e.g. different corrections/outlier filtering, etc.)
One should keep in mind that these inter-comparisons, although generally providing quality indications over longer time periods and large geographical areas, only provide an indication on the differences between the inter-compared satellite datasets. As such, they do not provide indications on the absolute quality. The absolute quality can only be expressed through validation with ground-based observations.
Comparison with ground-based observations
A real validation of a retrieved gephysical quantity is done by comparing with ground-based observations. It should be kept in mind that also in satellite-surface comparisons differences in representativeness play a role:
- Spatial sampling
The satellite retrievals’ footprint depends on the nominal resolution of the underlying observations and ranges from ~100 m2 to several km2, whereas for most ground-based observations this range is mostly several times smaller. - Spectral sampling
… - Temporal sampling
…