Today, a primary concern of many climate scientists is the homogeneity of the data sets they use, especially relative to some long-term baseline such as 30-year decadal climate normals. Inhomogeneous data sets are the result of changes in the biases associated with data measurement (discontinuities).

Homogenized data sets have been adjusted for discontinuities and provide the means for data users to be confident that any changes and variations identified in the data are natural and not artificial. The adjustments are normally made so that the entire data set is based on the most recent observing system. This concept is an extremely important one to the climate community and requires a great deal of effort.

Inhomogeneous data sets can arise from a variety of factors, but in general, can be classified into three basic groups related to:

1) Changing the precision and/or accuracy of measurements,
2) Changes of temporal and/or spatial sampling and/or processing, and,
3) Micro-climatic changes of the local environment.

Changes in the accuracy and precision of the operational observing systems can lead to significant inhomogeneities. Discontinuities in precipitation measurements provide a good example.

Figure 22 illustrates numerous equipment and exposure-related precipitation discontinuities (Karl, et. Al. 1993). Note the apparent climate shifts (inhomogenities or discontinuities) have been introduced as a result of changing precipitation gauges, adding or removing wind shields, changing the height of the gauge orifices, and changing observing practices (10 to 1 snow-to-liquid equivalent ratio used in Canada).

International Precipitation records
Figure 22: Discontinuities in international precipitation records (from Karl, et al.1993b).

Many of the noted countries sport expensive, high quality precipitation gauges. These discontinuities also show up with climatological isoline precipitation analysis at international boundaries.

The ratio of 10:1 snowfall-to-liquid precipitation has dropped from about 30-40 percent of all U.S. snowfall observations at the turn of the century, to about 10 percent today (Figure 23). The frequency was derived for each individual station and then averaged for several hundred long-term U.S. stations.

Frequency of a Daily Ratio of 10:1 of Snowfall to Liquid Equivalent Precipitation.
Figure 23. Frequency of a Daily Ratio of 10:1 of Snowfall to Liquid Equivalent Precipitation. (prepared by Ken Kunkel at the Midwestern Regional Climate Center/Illinois State Water Survey).

An example of two types of other data continuity problems are shown in Figures 24a and 24b.

Locations of Reno Runways, Urban Heat Bubble, Previous and Installed/ Relocated and Re-relocated ASOS Instruments.
Figure 24a. Locations of Reno Runways, Urban Heat Bubble, Previous and Installed/ Relocated and Re-relocated ASOS Instruments.
Changes in Mean Annual Minimum Temperatures with ASOS Instrument Relocations at Reno, Nevada.
Figure 24b. Changes in Mean Annual Minimum Temperatures with ASOS Instrument Relocations at Reno, Nevada.

Reno's busy urban airport has seen the growth of an urban heat bubble on its north end. The corresponding graph of mean annual minimum temperature (average of 365 night-time minimums each year) has as a consequence been steadily rising. When the new ASOS sensor was installed, the site was moved to the much cooler south end of the runway. Nearby records indicate that the two cool post-ASOS years should have been warmer rather than cooler. When air traffic controllers asked for a location not so close to nearby trees (for better wind readings), the station was moved back. The first move was documented, the second was not. The climate record shows both the steady warming of the site, as well as the big difference in overnight temperature between one end of this flat and seemingly homogeneous setting, an observation borne out by automobile traverses around the airport at night.

Another bias can creep into the climate record simply by changing the sampling frequency or spatial averaging algorithms of the measuring and processing systems. This bias can be removed readily as long as it is quantified (by taking overlapping parallel observations between the old and new systems) and documented in time in the station's history file (metadata).

Perhaps the most difficult biases to compensate for occur when local or micro-climatic changes around the instrumentation introduce non-representative (but physically accurate spot scale) changes in the data record. These inhomogenities can settle into the data quickly (a few days), as with parking lot expansions, or gradually (months or longer), as in the case of slowely maturing trees. The best known examples include the growth of big-city urban heat islands, changes of land use, and stations relocations (Figure 25). Any of these can blur or completely mask regional larger scale climate variations and/or change.

Higher landuse in 1992 than in 1973
Figure 25: Changing Land-use Patterns in Atlanta, Georgia (Urbanization).

Station relocations are unavoidable on occasion due to loss of an observer, property ownership transfers, poor data quality, etc. When this situation occurs, for whatever reason, the NWS field person is responsible for determining whether the new site is "climatologically compatible" with the old site and whether the data sets are thereafter treated as different time series.

A frequently asked question is how is climate compatibility determined? NWS policy (NWS Observing Handbook No. 6: Cooperative Program Operations) states that:

"compatibility is always determined by comparing the new to the original equipment location for the station as described on Rendition [original site, not last site] of the station's WS Form B-44. With some exceptions, a move is considered compatible if the new equipment location is within 5 miles of the original equipment location and the difference in elevation is 100 feet or less. However, take great care to assure that moves made within these limits are not, for example, from a hilltop to a valley bottom or subject to other large magnitude influences such as large water bodies, pavements, etc."

This somewhat arbitrary policy can result in the inappropriate continuation of long-term data sets at climate stations which in fact may be incompatible. Discontinuities such as these can then lead to serious misinterpretation of climate variability and trends at the locale by data users.

Figures 26 a and 26 b illustrate an example of a relocation that resulted in a large climate discontinuity although the station name remained the same. This COOP station was moved from about 400 yards from this 100 foot-high south-facing rock face to within about 100 feet of it. The original site was an open sagebrush exposure. That summer, the station set or tied four consecutive all time monthly extreme maxima.

This illustrates the importance of station exposure in determining climate continuity. In this case, although the station's relocation fell well under guidelines that require renaming (5 miles, 100 feet), the new setting is completely incompatiable with the original site and needs to be started over for climate purposes.

Station Relocations are a Common Sources of Data Discontinuities and Non-homogeneous Climate Records.
Figure 26a: Station Relocations are a Common Sources of Data Discontinuities and Non-homogeneous Climate Records.
Relocated COOP Station With Same Name but Large Climate Discontinuity
Figure 26 b: Relocated COOP Station With Same Name but Large Climate Discontinuity.

NWSH Climate Services Division plans to revisit this policy in the near future (FY05) to consider options for a more appropriate approach. However, at this time, we discuss the situation within the limitations of the existing policy.

Climate compatibility is best determined by running overlapping observations at both the existing and new site for a minimum of a year or so and then comparing differences. However, even when this approach is satisfied, a quantitative definition of climate compatibility does not exist. Thus, for the time being, it must be stated that climate continuity is subjectively determined. Traditionally, this determination has been made on the spot, on the day the station is moved, without benefit of analysis of actual data behavior. Evidence is now suggesting that a number of those determinations are in error and that many relocated stations should in fact be considered as separate stations.

Recommended Station-Relocation Related Actions

The following actions are recommended to minimize unwanted discontinuities in the climate record:

1) When considering station relocation climatological compatibility, strongly weight topographic characteristics of the old and new sites in addition to the 5 mile/100 feet policy. In some cases the relevant tolerance may be as low as 5-15 feet vertically and a few tens of feet horizontally. Understand that topographic setting differences (i.e., slope orientation and setting; crest, valley bottom, slope, plateau) can have a much greater impact on data continuity than, say, the relocation coordinates in the context of the 5-horizotal, 100-vertical feet rule. The chances are high that the two stations are climatologically incompatible if there are any significant topographic exposure differences.

2) Coordinate with your RCSPM and other climate services partners (NCDC, RCCs, SCs) before making the final determination on station relocation climate compatibility. Although current NWS policy gives you, the NWS representative, responsibility for determining climatological compatibility for station relocations, consultation and input from our climate partners will assist you in making the best call possible.