Weather and climate observing networks tend to change over time with the development of new instrumentation, changes in stations, and even changes in observing practices. All of these factors can impact analyses of the data. Some of these changes can improve the analyses. But all too often, changes in observing networks will adversely impact climate analyses if the analyses don't account for the changes. For example, changes from one type of thermometer to another can add an artificial discontinuity to a regional time series. But if the changes are documented and the information on these changes are readily available, analyses can be improved by using these metadata, for example, to remove data from problem stations prior to producing regional time series.

Partly for this reason, the U.S. National Climatic Data Center is developing a system to monitor the health of several observing networks. This system uses a variety of network performance indicators to track network changes. Some of these indicators are very simple, such as the receipt of data from a station or the changes in station history forms that describe the station's instrumentation. However, other indicators, such as assessing time-dependent biases, are very complex statistical tests.

The system is designed to function as near to real time as possible. For certain performance measures at some networks, this will mean monitoring incoming data on an hourly basis. However, the sensitivity of the time dependent bias tests increases with the length of data available on both sides of potential discontinuities. Therefore, some problems will only be able to be identified months or even years after they occur. However, even detecting a time dependent bias in a station two years after it occurs is still a very valuable contribution to improving the fidelity of the results of climate change analyses.

These performance indicators also provide valuable information to network managers. Lists of stations experiencing various problems are being sent routinely to network managers. Coupling these metadata with map-based visualizations can assist managers in focusing their network improvement efforts where they are most important. Also, special alerts can be automatically sent out to network managers and data users alike when problems are observed.