Another natural evolution of environmental measurement is changes in observation practices. These changes include variations in spatial and temporal sampling, processing algorithms and the tools used to take the measurements. Sometimes these changes can seriously affect the interpretation of data and products derived from the measurements, particularly climate variability and trend analysis.

These changes are frequently desired in the interest of improved data accuracy. However, it is imperative that the changes be documented in the interest of homogeneity of past, present, and future measurements. This practice cannot be accomplished without a sound data management system that includes documentation of all changes that can introduce data discontinuities.

Let us examine an example of how different observing practices can impact the continuity and accuracy of the climate record. Let's first look at snowfall observations since this topic has been of great concern to scientists and customers alike in recent years.

NWS snowfall measurement guidelines have, for decades, promoted the use of snow measurement boards as the standard measurement surface for the measurement of fresh snowfall amounts. The spirit of the standard is to ensure compatible and consistent snowfall measurements (and data) nationwide.

Unfortunately, until the 2002 winter season, the agency did not provide users with the specified snow measurement boards. The result was data that was collected inaccurately and inconsistently from place to place. Amounts measured vary in the same location with which surface was being used, what the temperature was at observation time, etc.. Was the snowfall amount measured on the lawn (grass), the metal roof of the car, the driveway or sidewalk, the roof of a heated airport terminal building, or something else? In many cases, the metadata does not document this important aspect of the measurement procedure.

Another excellent example of the impact that different observing practices has on the climate record is the frequency with which snowfall measurements are taken. It has long been known that the amount of snowfall one measures is strongly correlated with the frequency of measurement. That is the case where the measurement surface is cleared with each observation.

In recent years, it became evident that the standard frequency of once each six hours (or once daily at COOP stations) was not being followed by all observers and that some were taking observations hourly and then summing the 24 values to produce a 24-hour total. The problem with this non-compliant methodology is that it inflates totals as compared to the standard frequency.

Event and seasonal snowfall decrease with higher frequency of measurements
Figure 20. Snowfall Measurement Frequency Differences.

Snowfall, unlike rain, cannot simply be summed over varying time intervals to obtain a total. Snow settles, melts, and blows around and thus the total accumulation varies significantly with the frequency of the measurement. The more frequently that snow is measured, the snowier a location appears to become. The impacts of non-compliance with measurement standards on the data are 1) incompatibility between stations and 2) inhomogeneity with past data, depending on the methodology used historically.

A similar case can be made for changing the time of observation for temperature. NOAA researchers evaluating temperature trends in the U.S. during the 1950's through 1970's were aware that the NWS was requesting volunteer COOP observers to shift their 24-hour time of observation from afternoon. to morning readings to better support NWS hydrologic services needs. This change introduced a cold monthly bias of up to 2 degrees F into the climate record some parts of the country (Figure 21). It is now well-known that whatever time instruments are reset, those are preferred times for daily maximum and minimum temperatures to also occur, and especially so in frontal or advective (that is, non-solar-dominated) climatic regimes, such as winter and in northern latitudes). Conversely, it has been shown that if a single observation time is adhered to, no matter what the time the records can yield excellent tracking of the temporal variations in climate.

The apparent climate cooling introduced by the change in observing practices was many times greater than the magnitude of the natural climate change signal. In this case, good metadata records allowed the National Climatic Data Center to develop and apply monthly time-of-observation bias corrections to create more homogeneous data sets for research.

Change in January Average Temperature
Figure 21: Change in January Average Temperature Resulting from Changing the Time of Observation from 5 p.m.

Recommended Observing Practices Related Actions

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

1) Ensure that all changes in observing practices are entered into the station metadata record including the data of the change.

2) Ensure that observers understand the importance of compliance with standards for observing methodology with respect to the integrity of the climate record and that any changes they make in their practices need to be reported to the NWS representative and recorded in the station metadata immediately.

3) Coordinate with your RCSPM and other climate services partners (NCDC, RCCs, SCs) on any proposed observation change related issues before implementing the change so that impacts on the climate program and options can be considered.