The design and use of climatological data banks, with emphasis on the preparation and homogenization of surface monthly records

Abstract

The procedures involved in constructing data banks for use in climatological research are described, using examples from work done in the Climatic Research Unit. Such data banks will normally have two component parts: the meteorological records themselves, and the accompanying documentary and information systems.

As a first step, meteorological records appropriate for the intended application of the data bank must be collected and stored, commonly in a computer. Individual records must then be merged into a form convenient for the user. For example, all records from one geographical region may be stored in one computer file. Procedures for quality control of the data are discussed. We emphasize the need to ensure that records are homogeneous, i.e., that they do not contain spurious jumps or trends caused by non-climatic factors such as site change or urbanization. Some techniques to correct inhomogeneities in meteorological records are described.

The documentation accompanying the meteorological records enables users of the data bank to assess the suitability and reliability of the data. It has three components: first, information on the individual records such as start year, end year, altitude of the site, and geographical position; second, a list of data sources used in the compilation process; third, station histories which detail any known changes in site, instrumentation, etc. The station histories will be added to as work progresses on the data bank, to describe any attempts to homogenize records, and ultimately to give the compiler's assessment of the reliability of each record.

User needs must be considered at all stages of data bank design and construction. Only in this way will a well-documented and easy-to-use system result.