This paper deals with the classical theme of vegetation dynamics: Will communities converge towards a steady state, or diverge towards multiple states, over time? In contrast to numerous papers with space-for-time (chronosequence) comparisons, the authors used data from a long-term (22-year) experiment that manipulated both disturbance and nutrient addition. On the basis of this unique long-term data set, the authors demonstrate that regardless of disturbance, communities within a given nutrient treatment converged in composition, and communities receiving different nutrient treatments diverged through time, indicating communities tended towards resource-mediated equilibrium states.

In general, this is an important data set and the analyses are generally straightforward and appropriate. I do, however, have two concerns about the data presentation.

(1) Nitrogen gradient. This experiment included nine levels of nutrient addition treatments, from 0 to 27.2 g N year−1. However, only the results of the first four levels of treatments (0 to 9.5 g N year−1) were reported here. The authors suggested that higher N addition would introduce toxicity. I agree with this explanation but it would not hinder the authors to analyze the full N gradient. Many agricultural fields have experienced a high level of N enrichment and have been toxic, and understanding the dynamics of these communities would be quite interesting. Therefore, rather than omitting data from high-level nutrient addition treatments, I strongly recommend the authors present analyses of the full N gradient, to see whether a higher or lower N addition rate would show similar or different temporal patterns.

(2) Among and within-field variation. Since the three fields have different initial conditions, it would be critical to distinguish the contribution of among and within-field variation to the overall trends the authors found. For example, the convergence of communities within a given nutrient treatment could either be driven by the increased similarity among the fields, or the increased similarity of the 6 replicate plots within the same field. It would be not surprising that different fields converged in community composition, as they were different at the initial ages (especially for the undisturbed grids). However, if the 6 replicate plots within the same field also converged over time, it would be much stronger evidence for the emergence of equilibria. Therefore, it is important to show the community trajectories of different treatments within the same field.

Minor comments:

Line 144: If disturbance occurred annually while the nutrient additions were applied once at the beginning, will communities still shift towards resource-mediated equilibria? In other words, whether the different effects of disturbance and nitrogen fertilization were the consequence of how the treatments were applied (annually vs. once)? I think this point should be mentioned in the discussion.
Line 164: How about the community dynamic of E001 after 2004? Were these plots continuously surveyed until now? If so, I strongly recommend the authors report the whole time-series data (~40 years). That is, whether resource-mediated equilibria could persist after 40 years of succession?
Line 260-272: Rather than reporting the names of species in the plots, I think it is more important to show whether community compositions shifted from short-lived weedy herbs to longer-lived perennials? Similarly, whether the decreased speed of succession was driven by these life history changes? I feel this information is needed to provide a context for interpreting the results.
Figure 2: Is it possible to merge panels a, c, e, g, i into the same one to show the trajectories of different treatments? For example, you can use different colors to represent different N treatments, and use the arrow to show the trajectory of each treatment (see Figures 1 & 2 of Fukami et al. 2005, Ecology Letter, 8: 1283–1290 as an example). If so, I think figure 4 c & d would be not necessary.

Overall, I think the framework and analysis of this paper are excellent, but I strongly encourage the authors to report results from the whole data set across the full N gradient and survey years, rather than restricting analyses to low N treatments.

I would like to express my appreciation for the thorough response from the authors. I am now satisfied with the validity of the data selection and robustness of the results. Thus, I suggest that this manuscript is accepted in its current form or following just some minor revisions.

It would be valuable if the authors could further consider the initial ages of the fields in their analysis and discussion. The recent findings suggest that the convergence of communities under a given nutrient treatment was primarily driven by increased similarity among the fields rather than increased similarity within the 6 replicate plots in the same field. This is a noteworthy outcome, does it suggest that initial ages may have played a greater role in driving convergence rather than nutrients?