1. Successful execution of the GCOS and GTOS programmes requires that information about numerous variables be obtained for terrestrial environment around the world and converted into homogenous data sets. These data sets are necessary to achieve the GCOS objectives of climate change detection, assessment of seasonal and interannual variability, model validation, etc.; and the GTOS objective of detecting and assessing the impact of climate change on terrestrial ecosystems, among others. To obtain such data sets implies that each variable is measured at many locations around the world, with the necessary temporal frequency, accuracy and consistency so that the needed global data sets can be generated from these measurements. Since most variables of interest vary both in time and in space, often quite rapidly, this poses a formidable challenge.
2. The fundamental constraint of global observing systems is that it is not practically feasible to measure all of the variables all of the time everywhere. This dictates that a sampling design must be used to most effectively utilize available resources in obtaining the information at the global scale. For example, one of the efficient sampling designs is stratified random sampling in which the distribution of samples in the population is based on the distribution of the variability within the population of interest. The highly variable portions (or strata) of the population are then sampled more intensively than those with less variability.
3. Since each variable may have its own distribution of variability globally, a different stratification should in principle be envisioned for each GCOS/GTOS variable of interest. This would imply that many different sites would be required around the world, each providing measurements for one or few variables. Again, this would be practically very difficult to implement, besides being too costly. Fortunately, many variables of interest vary in similar manner, i.e. their stratifications more or less coincide. For example, variables related to biogeochemical cycling are closely related to the vegetation characteristics. In defining a sampling program, one can thus envision stratification based on biome, land cover, or vegetation type. Similarly, some variables describing hydrological processes co-vary at the scale of a watershed or a river basin. Such co-variations have been employed in the past in the establishment of research centres or sites at which numerous processes and variables are investigated. An example are Long-Term Ecological Reserves, agricultural research stations, experimental watersheds, etc.
4. A variable of interest may often be measured in many different ways. For example, soil organic matter content may be measured directly through chemical analysis, indirectly using spectrophotometry, or simply related to soil colour through the use of colour charts. Such procedures usually differ in accuracy, complexity, and costs. It is thus possible to make many (but less accurate) measurements with a simple approach but only few using a complex, expensive procedure.
5. Various measurement techniques also provide results over different spatial domains. For example, direct, detailed measurements can be applied to few locations because of logistics and costs. On the other hand, indirect methods such as space observations provide the means of covering the whole globe, with a spatial resolution that can be selected through sensor design and mission management.
6. The three above aspects of determining the global distribution of a variable: co-variance with other variables of interest, different techniques for measuring the variable, and the different capabilities for spatial coverage can be used to define a comprehensive global monitoring system. In such a system, various measurement approaches for a variable should be combined so as to maximize the quality of the resulting global data products. Furthermore, the measurements of different variables should be coordinated so that the cost effectiveness of the entire system is maximized.
7. With the above considerations in mind, concept of tiers is defined for GCOS/GTOS terrestrial observation purposes. In principle, ‘tier’ characterizes the measurement approach to a variable and is defined in three ways:
the number of GCOS/GTOS variables measured at the same location; this is a reflection of the degree of co-variance among the different variables of interest;
the accuracy/precision of the method used to measure the variable; and
the spatial frequency with which the variable is sampled.
8. For GCOS/GTOS purposes, a five-tier sampling concept is defined using these three criteria. For the lower tiers, the number of co-located measurements is higher, measurement techniques employed for the variable of interest are more complex, and the measurements are made at fewer locations than for higher tiers. Each measurement approach to a variable can be classified as belonging to one of the tiers. For example, precipitation measured continuously or at frequent intervals by (few) automatic recording stations and along with other meteorological variables is tier 1 measurement, while measurement once a day at (many) climatological stations (where no other variables or air temperature only are measured) may be tier 4.
9. Although the tier definition is developed in terms of measurements of individual variables, it can also be employed to ‘sites’, i.e. locations where sets of GCOS/GTOS variables are measured. At sites belonging to the lower tiers, more variables are measured with more complex methods but at fewer sites than at higher tiers. In the five-tier scheme, tier 1 includes sites not initially established for monitoring purposes but which otherwise meet the tier criteria. Tier 1 sites are established by comprehensive research programs aimed at understanding the environmental processes involved in climate and climate-ecosystems interactions and at the development and validation of models needed by GCOS and GTOS. In these studies, unique data sets are obtained which are of value to the global monitoring programs because of the models they underpin. It is to be expected that measurements at many of these sites will continue, thus becoming part of the global monitoring systems. Tiers 2 and 3 exist in many instances, as indicated below. Tier 5 (individual variable typically measured with high spatial frequency and lower accuracy/precision) contains principally satellite-based measurement techniques.
10. Tier 4 exists in some cases but not in most. It may serve two purposes: to provide spatially dense measurements for variables not measured (or not measurable) from satellites, or to provide a basis for statistical extrapolation of more intensive measurements from lower tiers. The design of tier 4 measurements can therefore be complicated.
11. When applied to existing sites on which global GCOS/GTOS observations may be based, the tiers have the following characteristics:
|
Tier characteristics |
|
|
1 |
Large-scale experiments. Very intensive measurement of a large number of variables over a ‘site’ of several hundred square kilometres or a transect of several thousand kilometre length, over a limited period of time. They obtain data to study processes, interactions among processes and ecosystem components, and to develop models describing these. GCOS/GTOS does not create or fund these, it simply ensures that the data and understanding from them are captured for future use. Examples are the IGBP Megatransects, GEWEX large catchment studies, ISLSCP projects, etc. May later continue observations of selected variables (similar to tier 2 or 3 sites). |
|
2 |
Research centres with a large staff (>10 participating scientists), sophisticated and expensive infrastructure (the actual infrastructure varies with the research focus but may include laboratories, data loggers for continuous measurement, flux towers and gauging weirs, etc.). They tend to focus on one biome type for unmanaged ecosystems (e.g. the US-LTERs) or one crop type in the case of agricultural centres (e.g. CIAT, IRRI). GTOS/GTOS relies on these sites to develop process-level understanding and to make continuous measurements of fluxes and numerous other variables. There are thought to be about 100 such sites globally. In principle, there should be one per major crop type, biome type, aquatic system type and cryospheric type. |
|
3 |
Research stations with permanent presence on site. Many hydrological observation sites fall into this category, as do ecological field stations and many agricultural research stations. Their principal function is to characterize the seasonal and interannual variations across the range of variability within each major ecosystem, crop system, aquatic system or cryospheric system. Globally, this is expected to require about 500-1000 sites, the overwhelming majority of which already exist. The observations do not require very expensive, sophisticated or continuous measurements. |
|
4 |
Sites which are revisited regularly but infrequently (once every 5-10 years). Their principal purpose is to provide information on variables that cannot be accurately obtained by remote sensing, or to provide statistically valid spatial information for extrapolating findings from the lower tiers. |
|
5 |
Remote sensing, which is usually quasi-continuous in both time and space, but can provide data for only some of the variables of interest. |
