The Global Hierarchical Observing Strategy (GHOST) is an information gathering system shared by both the terrestrial and climate observing systems.

GHOST responds to the need for reliable, representative, long-term data about the earth's land and freshwater ecosystems. The system works on a global scope because many of the processes affecting the quality of the environment occur beyond national boundary lines - climate change, desertification and transboundary pollution are examples. The system uses a long-term strategy because the processes involved may be slow, but cumulatively dangerous. The need for a long observation period and global coverage means that the measurements must be made very consistently if they are to be comparable. The scope of the effort requires that the observing system be designed for maximum efficiency.

GHOST is a hierarchical strategy. On the macro level a few variables are measured regularly in a large number of places, and on the micro level a large number of variables are measured in a few locations for a limited period. The hierarchy divides into five tiers, each with unique characteristics and roles, although existing facilities often straddle more than one tier (see Table 1). The concept is applicable to the three main areas which GHOST is concerned with - the land surface, freshwater ecosystems and ice surfaces - each with its own hierarchy, but all share tier five (Table 1). The observing system is built largely out of existing national and international observation systems and research centres, with modest additions of stations and sites where representation was previously inadequate. GHOST was designed with a heavy emphasis on methodological consistency and efficient data management.

Table 1 Examples of terrestrial carbon information users

Tier	Role	Characteristics	Indicative Numbers
1. Large area experiments e.g., IGBP transcets, large catchment studies.	Understanding of spatial structure and processes.	Cover a linear dimension of >100 km, very intensive sampling, highly integrated data sets.	10
2. Research centres e.g., Large LTERS, large agricultural research stations.	Understanding processes, experimentation, method development, data synthesis.	Fundamental research on a crop, ecosystem or cryosphere type, one per major type. Generally expensive complex instruments.	100 a
3. Stations e.g., Biosphere reserves, smaller national agricultural and ecosystem research sites, research catchments, small polar stations.	Long-term measurement of variables which vary over periods from weeks to years. Calibration and validation of remotely-sensed variables. Trends of variables.	Secure existence, representative of the range within a type, but not statistically unbiased. Frequent measurement of variables.	1000 a
4. Sample sites e.g., US EMAP programme, UK country survey.	Direct measurement of variables not observable by remote sensing, calibration and validation of remotely-sensed variables, status and trends of biome health.	Infrequently visited (once per year to once per decade), large sample, statistically unbiased.	1000 a
5. Remote sensing e.g., AVHRR, SPOT, Landsat.	Spatial and temporal interpolation at scales down to 1 day and 30 metres. Extent of biome, ice sheets, etc., status and trends of a biome health.	Frequent, complete coverage, variables mostly indirectly observed.	Not applicable
Notes: a. The numbers indicated are estimates only and may vary significantly especially at tier 3. Approximately in the ratios: 35 natural ecosystems; 35 agroecosystems; 10 rivers; 10 lakes; 10 cryosphere.

For more information on GHOST, please refer to two published reports from the planning phase: GTOS-12 and GTOS-13.