The cofecha-KINSYS Quality Control Method
We include in our international data sets only the samples that can be verified or judged to be correctly measured and dated. The rest of the data (undated, uncertain dating, strangely formed samples etc.) are not included there, but they are available by separate request.
There is sometimes a need of filtering data, e.g. for improving climatic signal in a chronology. This is the case especially in a very noisy data (e.g. Juniperus communis), where there are just short segments (<50 years) reasonable to date and accept to the final data. We have developed a special iterative process for improving quality of tree-ring data. This approach, utilizing Cofecha’s output files and letting to filter tree-ring data according to a preset user-defined quality criteria, is called the KINSYS-Cofecha Quality Control and Dating. This procedure was developed by the author (Mauri Timonen) mainly in 1996-1999 in the Advance-10K project,
Data quality improvement starts usually from Cofecha’s Descriptive statistics (Part 7). We have found useful to check the total number of flags in each sample and their proportion from the total number of segments in a sample. In an ideal case of high-quality data there would be zero or just a few flags and high correlation (greater than 0.4 in a sample with more than 100 tree-rings, depends of course on tree species, in this case Scots pine in Northern Finland). Using the number of flags as a criteria for high-quality data, a rule of thumb for the Finnish timberline pine would be less than 3% flags in the data.
The quality-criteria is based on threshold values defined by Cofecha’s “Correlation with master” table (statistics in Part 7), number of segments and number of rings in a sample. Only those samples that exceed all the chosen criteria will be accepted for the final data set.
The basic qualifying work is carried out by a program called COFCOR. Its ascii output file allows editing, which makes it possible also to add and remove samples also manually. After the Cofcor analysis the next step is to bring the suggested dating data (Cofcor's output file) to a KINSYS program called KINDATA that creates a new dated Tucson file containing the suggested datings.
The filtering system can be applied both for improving data quality and developing new chronologies. If the idea is just to improve data quality, the next iterative operation would be a new Cofecha Quality Control run, a new sample selection in Cofcor, and a new data set creation in Kindata. This iterative process can be continued until the result is satisfactory.
In the case of building a basic chronology, the method is the same, but new dating material has to be imported from a Cofecha-based dating output file.
The main data we utilize from Cofecha’s output files are listed in Part 5 (Correlation by segment of each series with Master), Part 6 (Potential problems: low correlation, divergent year-to-year changes),and Part 7 (Descriptive statistics). Flags play an important role in choosing the samples. They are classified into two categories: A-type flags indicate low correlation with Master; B-type flags warn about possible shifts in datings. Closely related to these parameters, poor correlations, even minus-signed, appear in the data. A closer look in Part 6 provides a detailed information for the origin of flags. This information is very useful for assessing next steps in chronology analysis. See also: Dating report (in Finnish) and Saima's version of Kinsys-Cofecha dating, where there are also a lot of examples, how the system works in practice. A more detailed version of Kinsys-Cofecha dating here.