Climatic Research Unit : Papers
Thomas M. Melvin and Keith R. Briffa
Dendrochronologia (2014) 32, 7-20.
doi: http://dx.doi.org/10.1016/j.dendro.2013.06.002
This is the first of a two-part description of a new software tool CRUST (Climatic Research Unit Standardisation of Tree-ring data). This program has been designed primarily to allow the convenient, routine application of "Signal-Free Regional Chronology Standardisation" (SF RCS) to different types of tree-ring data. The program also enables the use of other popular standardisation methods. A series of experiments is described in which the ability of simple RCS and SF RCS to recover known tree-growth forcing signals is tested. In the comparatively rare situation where many sub-fossil data are distributed over a wide time range and there is no slope in the overall common-growth forcing signal, simple RCS is satisfactory. Simple RCS produces distortion in all other examples explored here. SF RCS is superior to simple RCS and in all cases examined. SF RCS works well except when the span of starting dates of sample trees is too narrow, a situation for which a test is available. Based on the results of the tests explored here, we conclude that Signal-Free RCS should be used as the standard method of RCS processing.
Thomas M. Melvin and Keith R. Briffa
Dendrochronologia (2014) 32, 343-356.
doi: http://dx.doi.org/10.1016/j.dendro.2014.07.008
A number of processing options associated with the use of a "regional curve" to standardise tree-ring measurements and generate a chronology representing changing tree growth over time are discussed. It is shown that failing to use pith offset estimates can generate a small but systematic chronology error. Where chronologies contain long-timescale signal variance, tree indices created by division of the raw measurements by RCS curve values produce chronologies with a skewed distribution. A simple empirical method of converting tree-indices to have a normal distribution is proposed. The Expressed Population Signal, which is widely used to estimate the statistical confidence of chronologies created using curve-fitting methods of standardisation, is not suitable for use with RCS generated chronologies. An alternative implementation, which takes account of the uncertainty associated with long-timescale as well as short-timescale chronology variance, is proposed. The need to assess the homogeneity of differently-sourced sets of measurement data and their suitability for amalgamation into a single data set for RCS standardisation is discussed. The possible use of multiple growth-rate based RCS curves is considered where a potential gain in chronology confidence must be balanced against the potential loss of long-timescale variance. An approach to the use of the "signal-free" method for generating artificial measurement series with the 'noise' characteristics of real data series but with a known chronology signal applied for testing standardisation performance is also described.
Citations:
Melvin T. M. and Briffa K.R. (2013) CRUST: Software for the implementation of Regional Chronology Standardisation: Part 1. Signal-Free RCS. Dendrochronologia 32, 7-20, doi: 10.1016/j.dendro.2013.06.002
Melvin T. M. and Briffa K.R. (2013) CRUST: Software for the implementation of Regional Chronology Standardisation: Part 1. Further RCS options and recommendations. Dendrochronologia 32, 343-356, doi: 10.1016/j.dendro.2014.07.008
The papers, supplementary material, data, and software (source code as well as executables) are provided (in this zip file) to document in detail the tree-ring data and the multiple stages of data processing and analysis.
We have created a reduced-functionality version of the CRUST software that is specific to the RCS Parts 1 and 2 papers (i.e. it does the specific processing necessary to construct the tree-ring chronologies, undertake the various analyses reported in the papers, and create the figures shown in the papers). This version of CRUST is available in the zip file linked to above. It contains all the raw data, together with executable versions for Linux and Windows. It also includes the Fortran source code, from which it can be compiled. We are not able to offer any technical support, but it does include a "makefile" to assist with compiling the software (though we recommend using the pre-compiled, executable versions if they work on your computer system). The zip file contains the TRW measurement data and pith offsets files used in the paper.
The full-functionality version of CRUST is available from the CRUST software page.