Scientists call for careful use of time scales, reference dates and statistical approaches in analyzing climate change trends to avoid distortion and hampering of response
Demonstrating that the use of different time scales, reference dates, and statistical approaches can generate highly disparate results in climate reports, scientists at the University of Alaska Anchorage argue that careful use of these tools is critical for correctly interpreting and reporting climatic trends in Alaska and other polar regions.
In an open-access paper published in the ACS journal Environmental Science & Technology, Dr. Kalb Stevenson and colleagues acknowledge that climate change is of great interest to climatologists, agencies, media outlets, policy makers, and the general public, but note that there are different ways to assess change, including the study of temperature trends. What may seem to be simple or arbitrary choices in these matters could potentially infuse significant bias into the interpretation of data, they suggest, thereby distorting the representation of climate variability in Alaska and handicapping potential strategies for response.
...media reports or summaries are often based on the latest perceived or reported trends (e.g., “Is it warming?”, “Is it cooling?”), and these outlets are often the major information source for the general public and policy makers on the issue of climate. It is important that agencies, scientists, and media outlets (television, radio, online, and print media) are cautious of the potential to introduce erroneous information and bias into reports on climate due to a lack of consideration of appropriate statistical methods or approaches.
...The application of different reference start dates and statistical approaches to multidecade climate data can result in drastically different temperature trend estimations.—Stevenson et al.
In their paper, the authors note that several cautions must be taken in analyzing long-term climatic data in order to avoid misleading or erroneous conclusions.
Time series. The use of different time series can have a strong impact on climate change reports, thereby influencing the understanding of the magnitude and duration of change. Short time series can isolate specific events, but may overlook long-term climatic patterns; long time series show trends over a greater duration of time but may overgeneralize patterns or obscure important events.
To demonstrate the effect of different time series on climate change analysis, the authors showed the change in mean annual temperature throughout Alaska for three different spans: 1949−2009, 1949− 1976, and 1977−2009 (e.g., the graphics above).
The effects of the PDO become apparent when observing temperature changes at varying time scales across a large region such as Alaska. These observations are important to consider in the context of climate change. Mean annual temperature is changing throughout Alaska. It has both increased and decreased over the past 60 years. However, the trend over the shorter scales is generally cooling, while the trend over the longer scale has generally been one of increasing temperature.
Large-scale climate shifts, such as the PDO, should be addressed when climate change is reported by media outlets or through agency or international reports; start and end dates should also be referenced when discussing temperature change trends. Time periods used for analyzing changes should account for these oscillations, with differences in scale acknowledged.
Reference dates. A poor choice of reference dates could introduce bias into temperature trend estimations; specific climatic events or temperature anomalies can skew statistical means or cause a disproportionate influence when estimating trend lines.
As an example, observed temperature increases over a 50-year period in Alaska from 1951 to 2001 can be viewed as temperature declines when estimating trends within the period of 1951−1975 and within the period of 1977−2001 to account for the 1976 PDO event.
Statistical approach. The choice of statistical approach to analyze data sets and generate temperature trend estimations can also allow for biases in trend estimations and misleading information to be reported due to the window of values used in estimation and how these values are treated, the authors point out.
Recommendations. Stevenson et al. make a number of recommendations for the use of analytical tools:
Special consideration should be given to major climatic events, such as the 1976 PDO shift, which has the potential to interact with nearby reference start dates and introduce erroneous information into reports. Using the 1976 Having a time range that covered well before and after the 1976 PDO would help to minimize the effects of this event; a method that de-emphasized extremes, such as a running mean, could be used to estimate long-term temperature trends.
Caution should be taken when comparing temperature trends from multiple studies that do not use identical methods or when considering the selection of reference start date from a year exhibiting a temperature extreme.
Different types of climatic patterns and anomalies are captured when using various local and global statistical methods, suggesting that caution is needed when comparing estimates from these different classes of methods.
The chosen statistical method should be able to capture trends in a data set without being overly sensitive to variation.
The implementation of a comprehensive and comparative analysis of time series and sensitivities to start dates and statistical methods for future estimations of multidecadal temperature trends. Applying multiple methods with different start and end dates, along with varying window size and filters used, should be considered and outputs tested using sensitivity analysis in order to determine how sensitive results are to extreme variations. Methods that over- emphasize extreme variations are likely to be less useful to climate scientists and policy makers interested in long-term temperature trends.
Comparisons of studies carried out by different groups would be best served by using standardized time periods. This is especially important when different methods are used to analyze data, the authors note.
There are a number of complex drivers underpinning arctic and subarctic climate, and therefore caution should be used against making large, broad-scale, or sweeping statements about climate and climate change. In the recent past, some regions of Alaska have been warming while others appear to have been cooling. However, the summarized or reported direction and degree of change depends heavily upon the choice of time scale, reference date, and statistical approach. For Alaska, greater variation in microclimates could lead to temperature trend estimates being more sensitive to reference start dates, and thus greater discrepancy between temperature changes reported by different statistical methods. This has implications for management practices that rely either on historical trend estimates or on anticipated temperature trajectories.
...As policy makers contend with developing responses to climate change and its impacts in Alaska and beyond, it is imperative that the use and interpretation of scientific studies to support policy development minimizes any potential for bias by giving due consideration to the methods used to estimate temperature change.—Stevenson et al.
Kalb Stevenson, Lilian Alessa, Mark Altaweel, Andrew D. Kliskey, and Kacy E. Krieger (2012) Minding Our Methods: How Choice of Time Series, Reference Dates, and Statistical Approach Can Influence the Representation of Temperature Change. Environmental Science & Technology doi: 10.1021/es2044008