Back to Search
Start Over
Advanced tool for analysing 15N tracing data.
- Source :
-
Soil Biology & Biochemistry . Feb2022, Vol. 165, pN.PAG-N.PAG. 1p. - Publication Year :
- 2022
-
Abstract
- Ntrace is a family of tools for the analysis of 15N tracing data sets, to quantify the simultaneous gross nitrogen (N) transformation rates in terrestrial ecosystems. Starting with the original publication of Ntrace in 2007, a number of specialised tools have been developed that cover a range of applications such as nitrite and gaseous N dynamics. While the capability of the tool has been extended to cover a range of applications, the underlying parameter optimisation algorithm, based on the Metropolis MCMC algorithm, has remained unchanged. However, in more computational demanding applications, this algorithm turned out to be very time consuming, and in some cases it was not even able to identify unambiguous parameter sets. The aim of this study was, to test a new global optimisation algorithm (GlobalSearch), by re-analysing some of the published datasets. In the evaluation, the speed of finding a suitable solution but in particular the accuracy of the new algorithm to obtain an adequate fit to the observations was taken into account. A new uncertainty calculation for the optimised parameters, based on the Levenberg-Marquardt method, has also been implemented. The new Ntrace tool was able to find an adequate fit for all tested datasets, in often a fraction of the time compared to the original MCMC algorithm. For all test cases the accuracy of the fit was at least as good as with the original model. This is partly due to a new algorithm's option to vary the initial pool size. In summary, the new Ntrace tool is quicker in finding suitable parameters for 15N tracing data sets. The fit to measured data is often improved over the original tool. Thus, we recommend the use of the new Ntrace tool for analysing 15N tracing data. The new tool is programmed in MATLAB and can easily be adapted to new applications. • A new parameter optimisation method was tested to analyse complex 15N tracing data. • GlobalSearch (MATLAB) was compared to the original MCMC method in Ntrace. • The new algorithm is quicker and more accurate in finding suitable parameters. • New functions for initial pool sizes and parameter uncertainties were implemented. • The advanced Ntrace tool is recommended for analysing complex 15N tracing data. [ABSTRACT FROM AUTHOR]
- Subjects :
- *MATHEMATICAL optimization
*TRACE analysis
*MARKOV chain Monte Carlo
Subjects
Details
- Language :
- English
- ISSN :
- 00380717
- Volume :
- 165
- Database :
- Academic Search Index
- Journal :
- Soil Biology & Biochemistry
- Publication Type :
- Academic Journal
- Accession number :
- 154617432
- Full Text :
- https://doi.org/10.1016/j.soilbio.2021.108532