Ranking critical factors in estimating root longevity and turnover: Comparisons between diameter and order approaches

Background/Question/Methods This study is to determine relative importance of diameter and order approaches in estimating root longevity and turnover. The fine root pool is critical to ecosystem carbon dynamics because of its rapid turnover rates. Understanding differences between the two approaches has implications to sampling strategies and analysis methods. However, previous studies may have been based on limited data and inferior methodology; simulated data should help resolve the issues by allowing for detailed analyses that cannot be done with empirical data alone. The root turnover simulator (RTS) was used to generate both censored and full data of fine root longevity and biomass. Empirical censored data of two temperate trees from Northeastern China were used as reference in analysis and in parameterization of RTS. Two specific analyses were performed. First, to assess its usefulness and limitations, Cox regression for censored data was compared to correlation for full data from the same root populations. Second, to quantify effects of the two factors on estimation accuracies, the same root population data were analyzed by four schemes of different combinations of factors: (1) both diameter and order considered, (2) order only, (3) diameter only, and (4) neither diameter nor order considered. 

Results/Conclusions

The results from empirical data of the two Chinese trees showed that Cox regression may be weak to determine importance of critical factors because differences between factors are often too small to make a valid judgment. In addition, effects of the two factors on estimation accuracies displayed different patterns, which may vary with tree species. The results from simulated data also suggested that Cox regression may have limited use in ranking the importance of critical factors. Among the four schemes of factor combinations, the one considering both factors (the maximum) and the one not considering any (the minimum) define two extremes of a gradient of heterogeneity captured, with the one-factor schemes coming in-between. The results of the simulated data indicated that the order approach tended to be more likely to produce accurate estimates because its estimates were often closer to those of the maximum captured heterogeneity scheme than those of the diameter approach were. The results of this study highlighted the limitations to empirical data and the associated methods and, thus, the needs for simulated data in both method comparisons and significance ranking.