Observation System Intercomparison-Examples from Bomex

The basic and readily observable meteorological variables are related to each other through partial differential equations in a fourdimensional space-time domain. With these one can compute unmeasured quantities, such as the vertical velocity and sea surface evaporation. Several estimates of the same derived quantity can be made by using several observation systems, such as a rawinsonde network, a fleet of aircraft, and instrumented ships. Intercomparison of the final results provide one measure of the error in the derived quantity. However, more extensive intercomparisons and cross-calibrations are required to understand the error content of each result in order to weight them before reaching a final conclusion, to avoid systematic errors or biases among the many observations taken by the component parts of any single observation system, and to substitute observations from one system in place of those deficient or missing from another system. In a future project, such as the GARP Atlantic Tropical Experiment, the international character of the effort and the desire to work on smaller space scales and shorter time intervals than in the Barbados Oceanographic and Meteorological Experiment (BOMEX) places especially stringent demands on the designed system accuracy and the intercomparisons necessary to ensure that, in fact, these demands have been met and preserved throughout the experiment.