GEOSS

The philosophy of combining many geospatial data sources in order to retrieve more and better information from Earth observation data is expressed in the GEOSS 10 year implementation plan [37], which states:

“Under GEOSS, national, regional, and international policy makers are collectively harmonizing observations, real- or near real-time monitoring, integration of information from in situ, airborne, and space-based observations through data assimilation and models.”

How the concept of GEOSS could be applied in practice is illustrated in [this figure] which shows a modelling system that simulates images recorded by various sensors on board Earth observation satellites. The heart of the system is a generic RS (observation) model that takes data from a GIS as input and produces as output simulated imagery at the correct spatial resolution and for the spectral bands of the simulated sensor. The RS model includes atmospheric effects and produces top-of-atmosphere (TOA) radiance images for all required spectral bands. The satellite data distributor also provides calibrated TOA radiance data, so this product can be compared to the simulated data.

This comparison is illustrated by the scale symbol, to illustrate the balance between the noise characteristics of the sensor, on one hand, and uncertainty in the surface properties on the other. If simulated and actual satellite images do not correspond sufficiently well, the GIS information is adjusted until the error becomes acceptable. In the GIS, multiple layers of vector and raster data are stored and, in combination with attribute information and values of physical quantities expressing the surface properties, this information is used as input for the radiative transfer sub-models (e.g. for soils, leaves, vegetation canopies and the atmosphere) of the RS model. In other words, the GIS system provides the surface properties as well as their geographic location.

Both spatial data and attribute data (properties) can be in error, and different actions should be taken according to the kind of error. Geometric errors require a correction of the geographic position of one or more objects, whereas errors in surface properties only require the adjustment of these properties. Although the system as sketched is very complex, it has a high degree of flexibility with regard to sensors and geometries, so it would be possible to bridge gaps among the variety of sensor systems that are orbiting the Earth, thereby facilitating the assimilation of data from different sources(as promoted by GEOSS).