We want to thank the referee for the time and effort invested in the constructive review of our work. We appreciate all the suggestions and the points where clarification was needed. Indeed, in our study, we found many non-significant results. Despite that, we did find some treatment effects, even after accounting for the factor contributing the most to plot-to-plot heterogeneity: plantation stand. In the corrected version of the manuscript, we are incorporating the suggestions and making clear the scope of inference of our results (significant and non-significant) which has been highlighted by both referees as the main area of improvement. We are also updating the conclusions to include recommendations based of what we have learn from this experiment, and how to improve large-scale drought experiments. Below we list the response to each of the line-by-line comments. For clarity the comments are in bold font.

Line 21

In the abstract, I had the impression that some results that are later reported as trends but are not significant are emphasised here. Perhaps in this way, the abstract is overselling the story a bit.

Reply // We appreciate this suggestion by the referee with which we agree that the text in the abstract should be clear about the not-significant results. We now proceed to fix the wording, so the abstract is in line with the manuscript results.

Line 112-113

This reference could come in the introduction but also it could be relevant for some parts of the discussion of the results.

https://www.nature.com/articles/s41586-022-05085-2

Reply // We think the referee is on point by suggesting the article by Cuhna et al 2022 as some of their findings are in line with our results. We now are incorporating their findings throughout the text, especially:

• Increase in total productivity.
• Lack of woody growth response

Table S1

It’s not clear what’s the unit of the fertiliser values.

Reply // We apologize for the lack of clarity in the information presented. The values are in kg, which we have now incorporated into the table.

Figure S4

It seems that the significant difference in LAI in 2017 was found in all treatments, but only DR and FR are discussed in figure caption. Can you clarify here, please?

Reply // This figure shows the effect of Hurricane Otto during the month of November in the LAI data. We apologize for the lack of clarity in the text and now made clear that the effects were present across treatments. The figure caption now reads:

“Fig. S4. Leaf area index mean values with 95% confidence intervals for each experimental treatment (CN: Control, DF: Drought+Fertilizer, DR: Drought, FR: Fertilizer) during the month of November. We found a significant year effect (F=21.29; d.f.= 3; p < 0.001) because of tropical storm Nate, in LAI decreased during November 2017 across all treatments. Letters represent multiple comparisons among years within experimental treatments from a Post-Hoc Tukey’s honest significance difference test.”

Line 155/ Figure S2

Indicate somewhere how distant the plots are from each other.

Reply // We now indicate in the methods section the minimum and maximum distances between plots. This varied depending on the treatment, as a through-fall exclusion treatment would not cause a major effect on a plot nearby (minimum distance: 15 m apart). However, we were aware that fertilizer can leach away to other plots. For this reason, we took to measures to avoid this: 1) we placed fertilized plots more than 50 m away from other plots, 2) we always placed fertilizer plots down the slope of the nearest plot. Despite the terrain being relatively flat, there is a small slope that facilitates such an approach. These measures were corroborated when observing water flow during big rainfall events. We apologize for the lack of clarity in this regard. The text in lines 177-178 now reads:

“Nutrient addition rates were targeted to 150 kg N ha-1 yr-1 (Table S1), similar to other large-scale tropical forest fertilization experiments (Wright et al., 2011; Alvarez-Clare et al., 2013; Waring et al., 2019). We placed fertilized plots more than 50 m away from other plots and/or down the slope from control or drought plots whenever we could not find enough trees 50 m away. These measures were considering the possibility of nutrient leaching from one plot to another one…”

Line 159

Any specific design for soil sampling (e.g. corners/centre of the plot)?

Reply // We now incorporated the following text in line 159:

“… by taking 7 to 10 cores (2.5 cm diameter, one on each corner and three to six in the center line of the plot) …”

Line 175

Is there any indication that the study sites were more N or P limited? I see the focus of having comparable N concentrations (like the experiment in Panama), but I just wonder about the other elements, why the choice of a broad-spectrum fertiliser?

Reply // In general the soils in the area tend to be P limited as highlighted by Waring et al. 2019. Regarding the fertilizer choice, there are a couple of factors that influenced our decision. First, as with other fertilization experiments (Wright et al., 2011; Alvarez-Clare et al., 2013; Waring et al., 2019) responses can be limited by multiple factors. Second, in nearby secondary forests where we studied how increases in rainfall lead to increases in net primary productivity (ANPP), we have found that soils with lower total elements (P, N, and cations except Mn), lower cation exchange capacity, and lower pH tend to limit how ANPP increases with higher precipitation (Becknell et al. 2021). This suggests that multiple nutrients might be limiting such responses. Since our interest was to understand how nutrient availability limits responses to soil moisture, we decided to use a broad-spectrum fertilizer rather than looking at the interaction of different nutrients with soil moisture changes. Our goal was to boost nutrient availability overall, and hopefully overcome limitation by any nutrient, irrespective of the identity of that nutrient. However, we acknowledge that this limits the potential of pointing out which specific nutrient limit plants’ responses to soil moisture.

Line 185

I thought that due to its tree-centric approach in building the plots, only the planted trees were going to be taken into account when measuring productivity. Couldn’t the panels at the throughfall exclusion limit new plant appearance/recruitment?

Reply // Panels can limit new plant appearance recruitment indeed. However, we placed the panels in a way that we avoided the standing understory plants. Regarding recruiting understory plants, we did move some of the panels in two plots to allow plants to grow. In this case, it only required moving one panel 15-30 cm to the right or the left. Whenever we did that, we placed a portion of a panel on the other side of the recruiting tree. However, it is possible that in the future this can be evaluated, as it is a commonality and a limitation among throughfall exclusion experiments. We consider that at the time of examining these results is too early (~5 years) to look for an effect on seedlings/saplings' recruitment responses.

Line 200-201

There is, perhaps, a mistake in this sentence, as “leaves, and, reproductive litterfall” appear twice in the description.

Reply // We corrected the text, and it now reads:

“… total litterfall (leaves, small branches, flowers, and fruits), only leaves, and reproductive litterfall (flowers and fruits) …”

Line 238

I understand how difficult it must be to sample these cores in the dry season. One thing that needs clarification here then, is if root productivity in kg m-2 yr-1 was calculated extrapolating those 2-month interval sampling periods to one year, or if root productivity between, for example, November to June next year was taken into account. If the second is true, I would imagine that during these 6 months root productivity might not be accurate, as root mortality and recruitment could have happened. On the other hand, by using root productivity only during the wet/growth season, perhaps root productivity on a yearly basis could be overestimated (assuming there is higher root productivity in the wet and lower in the dry season). In year 1, you sampled along the whole year, right? How does the data between different seasons compare for this year?

Another point, so in every sampling, you installed new ingrowth cores in different locations. Did I get that right? If so, would you also have root stocks in those same sampling dates, or when freeing the soil from roots to install the ingrowth cores, these roots were discarded? Just curious to know a bit more about the root dynamics here.

Reply // We appreciate the interest of the referee in our root sampling protocol. In these highly seasonal forests, both belowground and aboveground biomass is limited by rainfall, and during the dry season the production of roots is minimal or non-existent (Waring et al. 2019). However, to avoid overestimating or providing inaccurate data on root productivity as highlighted by the referee, we installed the root ingrowth core that was going to be collected in June during the November collection. In this way, we did consider root growth, if any, during the dry season. The one limitation is that we cannot make inferences on root seasonal dynamics. However, that was not the objective of the experiment as we focused only on the total annual flux. To make this clear we modified the text in lines 238-239 as follows:

“…The cores were collected two months after deployment and a subsequent new set of cores was installed after collection. While deploying the cores, we filled them with sieved, root-free soil collected on-site. During the first year of the experiment, cores were sampled the dry season. However, the clay-rich soils harden greatly during the dry season, which increased the difficulty of deploying new bags during these times. Therefore, for three years ingrowth bags were harvested in June, August, and November with the modification that the bags harvested in June were deployed in November. We acknowledge that roots may have grown, died and/or decomposed during the dry season (Kummerow et al. 1990), however we think that this effect would lead to minimal bias in annual productivity totals, as root growth and decomposition are expected to be very small during the dry season (Kavanagh and Kellman 1992).”

We installed the new ingrowth cores in the same location within the plot (corners and center line), but in a different soil core. We indeed discarded all plant/fungal material from the soil core when preparing the new ingrowth core to ensure we measured new root growth. However, we did not quantify the weight of roots present in those soil cores when preparing the new ingrowth core.

Line 315-316

Could you please add to this sentence if such reported differences are statistically significant?

Reply // We now modified the text to read as follows:

“At 40 cm depth, there was a significant reduction in soil moisture as a function of the pre-treatment period in the plots with a throughfall exclusion structure (~ -13%), contrary to a non-significant reduction in the plots without throughfall exclusion (~ -4%) (Figure 2)…”

Lines 315-318

My main concern is related to the extent to which the throughfall exclusion really worked, and how this may have affected the general lack or weak trends you find in your study. This should be acknowledged a bit more in the discussion, perhaps in a way to provide advice for future research. Many of the results you show in the supplementary material, for example, regarding the production of flowers and seeds, are marginally non-significant, despite the big difference in productivity. I imagine that in addition to specific species differences, the relative short nature of the experiment was not enough to capture significant changes.

Reply // Our response to this comment is twofold. First, we agree here with the perspective of the referee and we tried to touch on this point from the perspective that perhaps the observed changes in soil moisture are not strong enough to induce significant changes in productivity patterns in response to soil moisture (lines 397-398, and 409-423 in the discussion). Second, we did not manipulate atmospheric water demand (e.g., VPD) and this may be an equally important component of drought than soil moisture. We are now including in the conclusion a line on how to improve throughfall exclusion structures in this type of ecosystem. Lines 502-503 now reads:

“…Studying the role of soil moisture on plant nutrient acquisition dynamics remains a largely unexplored venue in TDF ecology. Considering the observed patterns, a total throughfall exclusion will be necessary to cause a soil moisture decrease greater than 15 % and manipulations of the atmospheric water demand (e.g., vapor pressure deficit) could help to improve of understanding of drought in tropical forests.…”

Line 321

Since there were not such big differences between mean soil moisture comparing control plots and throughfall exclusion, I wonder if this signal is small because of a potential effect of the years, meaning that perhaps stronger patterns could be seen towards the end versus the start of the experiment. Have you tested for the changes in soil moisture between years and treatments?

Reply // Besides the control plots (Fig. S6), we did not evaluate the year-to-year soil moisture variation in all the treatments. We now are including this analysis for all the treatments. This must be done with the wet season data only as the dry season variation is larger than any experimental signal

Line 330-331

Can you please add to this sentence if such interaction resulted in a positive or negative trend?

Reply // The interaction occurred in that when applying fertilizer trees in drought plots showed an increase in RGR while trees in non-drought plots a decrease, particularly for non-N-fixing plants. The text now reads:

“…We found moderate evidence of an interaction between drought and fertilizer for plantation trees (F-v = 5.16, d.f. = 1, p-v = 0.0499) but not for understory trees (F-v = 5.04, d.f. = 1, p-v = 0.0659). This interaction shows that plantation trees in drought plots showed an increase in RGR while trees in non-drought plots a decrease after the application of fertilizer (new supplementary figure)…”

We are also adding a depurated version of the following supplementary figure

Line 870

I think it should read “median”, not “media” (or mean?). Same for some figures in the supplementary material.

Reply // We fixed this mistake in all the figures as it should read the mean (in the case of Figure 3).

Line 356

There is a discrepancy between the unit described for root productivity in the methods section (annual basis) and here in the results.

Reply // We added corrected the units as mistakenly we omitted the year. It now reads (kg m^-2 yr^-1)

Line 381

You state there are no significant differences but the p value is 0.0431, can you clarify this please?

Reply // Indeed the ANOVA test used reported a p-value < 0.05. However, when studying the pairwise comparisons with Tukey’s HSD we found no differences between the drought and the control treatments. Now the text reads:

“…Although we observed a 14% NPP increase in the drought plots (F-v = 5.29, d.f. = 1, p-v = 0.0431), when applying Tukey’s HSD for multiple comparisons we found no evidence this was different from the control plots (Figure 5)…”

Lines 421-423

Great argumentation here!

Reply // Thank you!

Line 426

Again, I think that this recent paper could make it to the discussion (strong and direct evidence of P limitation in Amazon forests). This paper contradicts the statement you also make on line 500 in the conclusion, since the authors found strong NPP responses after 2 years of fertilisation in a mature forest. https://www.nature.com/articles/s41586-022-05085-2

Reply // We have now cited the suggested manuscript multiple times, with the consideration that the nutrient limitation present in the Amazon forests from that study is way larger than the one present in our study site. Therefore, I do consider it possible that the observed responses might differ in magnitude. Regarding the contrasting conclusions, we were trying to be cautious as the magnitude of the change relative to the amount of fertilizer added is small. However, that does not mean there are no effects of nutrient additions. We change the text to make this clear. Lines 499-501 now read:

“However, despite adding both macro- and micro-nutrients, our results confirm that tropical dry forest trees show modest short-term responses to fertilization…”

Lines 434-436

I would suggest toning down this sentence a bit, as this is a trend, and no real strong evidence of colimitation by water and nutrients were found in your study.

Reply // We understand the point made here by the referee, given the fact that we did not find a strong interaction signal. We modified the text, it now reads:

“Moreover, the increase in productivity as a function of fertilization showed a smaller, yet not significant, increase with the presence of throughfall structures when compared to non-drought plots (Figure 5, panel b). This result is in line with observed patterns in nearby stands of TDF, where forests in fertile soils show greater increases in productivity concomitantly with rainfall than forests in infertile soils (Becknell et al., 2021).”

Line 441

Insert . after “treatments”.

Reply // Thank you for pointing out this mistake. Now the text reads:

“…response to the experimental treatments. The timing…”

Line 456

Better to use the past tense here: “experienced”.

Reply // We fixed the grammatical mistake. Now the text reads:

“…which these three species were present experienced the highest biomass losses due to mortality…”

Line 463

Could you specify which resources you refer to here, perhaps light?

Reply // Although we cannot point out a single resource, light availability is one of the possible resources limiting plant growth. We modified the text to state that in those plots there is in general less competition. The text now reads:

“…the availability of resources such as light and/or less competition could be the cause…”

Line 475

It could be useful to acknowledge the fact that in these dry forests, and especially by increasing drought experimentally, roots can go really deep/deeper in search for water and nutrients. The lack of responses found for root productivity in your study is limited to the 0-15cm, and if we think that you only captured changes in soil moisture at the 40cm depth, it’s plausible that roots could be changing down the soil profile as well.

Reply // This is a really important point highlighted by the referee, to which we agree. Quantification of the vertical root profile in relation to experimental treatments is certainly one of the next steps in our future work in this experiment. We modified the text and now reads:

“…Moreover, with reductions in soil moisture trees tend to rely more on deeper water sources with less access to nutrients (Querejeta et al., 2021). This allocation of root biomass might also enhance nodulation in legumes as there might be changes in the vertical profile of nutrients in the soil. Collectively, our data and these studies suggest that the effects of soil moisture reduction go beyond ecosystem water/carbon balance and could cause a domino effect that might alter forest biogeochemistry.”

Line 490

Maybe replace ; by , after “disturbed soil”.

Reply // We fixed the grammatical mistake. The text now reads:

“…This trend suggests that when large rainfall events occur in disturbed soil; a decrease in microbial CUE could potentially lead to a stronger Birch Effect and enhance the soil C loss (Schimel, 2018)….”

Line 498

It seems there is some word missing between “responses to” and “is sensitive to”.

Reply // We modified the text to add clarity and a better flow of ideas:

“Our results highlight that forest productivity is sensitive to soil fertility and that this might interact with changes in soil moisture...”

References

We really appreciate the time and effort by the referee in performing this review, which is allowing us to improve our manuscript. One aspect that both referees highlighted was that the scope of inference in the conclusions we made had to be adjusted to our results. We are now addressing this in the text, as we agree that the message must be clear on what the data is telling us.

We also want to apologize for the lack of clarity in the description of the statistical analysis. In our analyses, we accounted for all the known sources of variation either through fixed or random effects. We analyzed the fixed effects by evaluating the main effect of drought, the main effect of fertilization, and the interaction in a two-way factorial experimental design. We then included random effects in this model to account for other sources of variation such as plot, stand, and species identity (when analyzing tree relative growth rates). This approach ensures that we do not ignore the influence of these factors in the observed patterns. When studying the changes in RGR, we accounted for the unbalanced design and the interactions by calculating the type III sum of squares when obtaining the F values in the analysis of variance. Having said this, we agree that the approach of before-after-control impact could be helpful. However, besides soil moisture and initial soil characteristics collected one month after the panels we installed, we did not have the ability to collect pre-treatment data on productivity, although we agree that this would have been desirable. Our expectations were that that treatment effects would increase in magnitude through time. We now clarify that in the revised manuscript.

We thank the reviewer for the comment on the focus of the manuscript and we have tried to present a coherent, streamlined central theme in the revised manuscript. Our focus is on the responses of TDF to reduction in throughfall, and whether this depends upon soil nutrients availability. During the writing of this manuscript, we considered the option of two separate manuscripts or one manuscript telling a holistic “microbial to ecosystem” level story. We decided to put together an ecosystem-level study that quantifies responses at various biological levels, with the challenge of increased complexity on how to tell a cohesive story. We appreciate the suggested work as it has been informative to the observed patterns in our experiment.

We appreciate the work of the referee to make the point of how soil nutrients covary with certain tree species. While there is evidence of nutrient deposition in tropical forests (Wang et al. 2017, Lu et al. 2018), forest composition might also influence soil properties and vice-versa as highlighted by the referee. We addressed this in the statistical analysis by adding the plantation species combination stand as a random effect, and now we also touch base on this in the discussion of the results. Below we list the responses to the line-by-line comments with the comments in bold for clarity.

Line comments:

67) because? rather than ‘as’?

Reply // We modified the text as follows:

“…high nutrient availability could alleviate drought stress because plants with higher leaf nitrogen…”

70) first use of this undefined abbreviation

Reply // Perhaps it went unnoticed in the reviewing process as we defined the abbreviation on line 43.

Section 1.1 This is a little chicken before the eggish. Inverting the paragraph will allow you to first establish that innate variation in soil fertility drives variation in growth-response to rainfall. Then second, the opposite is also true (nutrient limitation negatively affects water use efficiency). This then logically leads to the third point that high nutrient availability could alleviate drought stress BUT at a potential cost to NPP (which is where I assume you are headed).

Reply // We appreciate the suggestion made by the referee as this improves the flow of ideas. The text now reads:

“Soil fertility is an important factor modulating the responses of forest productivity to rainfall variation. For instance, TDF stands growing in more fertile soils tend to show higher increases in productivity with higher rainfall when compared to forest stands in nutrient-poor soils (Medvigy et al., 2019; Becknell et al., 2021). At the same time, nutrient limitation negatively affects water use efficiency in crop species and tropical seedlings (Santiago, 2015). High nutrient availability, therefore, alleviates drought stress as plants with higher leaf nitrogen maximize water use efficiency at the cost of photosynthetic nitrogen use efficiency (Lambers et al., 2008). However, this might have a potential cost to productivity.”

74) remove ‘main’

Reply // We addressed the suggestion now the text reads:

“Leaves, and more precisely canopy cover, are the center for carbon assimilation”

80) replace the slash with ‘or’

Reply // We addressed the suggestion now the text reads:

“…the timing of leaf flushing or shedding may create a…”

82) “Specifically”, rather than “Moreover”

Reply // We appreciate the suggestion as this improves the text:

“…1998). Specifically, changes in the timing of leaf…”

89) Why are transpiration rates increasing? Is this a generally accepted phenomenon in response to nutrient deposition?

Reply // In general with more soil fertility there is an increase in photosynthesis, hence an increase in transpiration rates, which ultimately leads to a decrease in water use efficiency (Lambers 2008). Indeed, there is evidence suggesting that this is occurring in tropical forests as suggested by Lu et al. 2018 in their work on long-term responses to high nitrogen deposition in the Dinghushan Biosphere Reserve in southern China. Among their main results are:

• Aceleration of soil acidification
• Reduction of biologically available cations
• Increases in plant transpiration

94) Perhaps “current thinking” would be more appropriate than “Theory”, as the later implies an established and tested precedent in the field

Reply // We appreciate the suggestion and agree with it. We modified the text accordingly:

“…nutrient availability. Current knowledge suggests that microbes with high…”

259) Priming is of interest here too?! Unless there is a specific hypothesis that you are testing, I don’t see that this adds to the manuscript.

Reply // We appreciate this comment by the referee, because it allowed us to improve the introduction and the presentation of our research objectives in this study. Indeed priming is of interest in our study. In general in our study we were looking at responses to experimental manipulation from microbes to ecosystem productivity. Besides microbial carbon use efficiency there are other aspects of soil microbial communities affected by drought or modulated by soil nutrient availability, priming being one of those. We now added the following text to line 104:

"Other aspects of soil microbial communities maybe affected by drought or modulated by soil nutrient availability (Ahmed et al. 2018). Priming refers to the decomposition older recalcitrant organic matter following stimulation of the soil microbial community by addition of labile organic matter (Liu et al. 2020). If drought alters patterns of fine root growth and of rhizodeposition (Preece and Peñuelas 2016), this may lead to altered priming with altered consequences of soil organic carbon storage."

294) This is a major flag, if soil properties were so variable that something as basic as volumetric water content cannot be compared across plots, how are any of the other comparisons valid? Especially given that the previous sentence claims that soils were saturated when the plots were established – does this mean that the volumetric content at saturation was meaningfully different for each plot?

The current approach requires either a strong validation or for the analysis to be shown using both response variables, allowing the reader to weigh the value of each.

Table S2 as two sets of boxplots, one by treatment and one by stand type

Reply // Our response to this concern is twofold. First, we may have over-emphasized the heterogeneity. In fact, out of 16 plots there are seven clay loams, four silty clay loams, four loams, and one silty loam. We are including now a supplementary figure with the soil texture triangle and the placement of each plot in among soil types. We apologize for the lack of clarity as the message did not go the way it was intended. Second, we did include all plots in the same linear mixed effects model, but instead of comparing the volumetric water content raw values we were comparing the change in water content in reference to the wet-season and dry season values. We decided to implement this approach as changes in the volumetric content of water provide a measure of the effect on soil moisture after the onset of the throughfall exclusion structure, an approach that has been tested in other large scale manipulation experiments (Reid et al. 2015). Once again, we apologize for the lack of clarity in the text.

309) Below 0, I believe is meant.

Reference