Burrowing Parrots Cyanoliseus patagonus as Long-Distance Seed Dispersers of Keystone Algarrobos, Genus Prosopis, in the Monte Desert

Round 1

Reviewer 1 Report

This is a fascinating account of seed dispersal that adds significantly to the literature on seed dispersal by parrots. It gave me way of thinking about how a seed predator familiar to me could also be dispersing seeds through wasteful feeding and stomatochory, although the system is quite different from yours. The narrative is well-written and clear, except perhaps on line 585, where I think a comma would help between “…quick germination[,] rarely developing…” Likewise, the tables and figures are well-executed and informative. The article was a pleasure to read.

Author Response

Thank you very much for your positive appreciation of our study. We have included the comma suggested by the reviewer.

Reviewer 2 Report

This is an interesting study. Readability is generally good, but terminology is opaque in places. My suggestions for alternatives to your terminology include: consumption rather than exploitation; branch bark (?) rather than branch flesh; absent or missing rather than lacking. I think Readability would also be improved with less unnecessary terminology (e.g. fruiting rather than fructification). At the very least they should be defined for those not experts in this field (epizoochory, endozoochory, stomatochory?).

Your methods involve measurements of various parameters of food seed pod selection and processing either before or after transport by parrots. Listing these by term, definition and units in a table would improve clarity as I rather got lost reading the text.

Your reasoning behind your use of MCMCglmm (presumably the R-package) over alternatives such as REML GLMMs is not given and would help justify your choice to the reader. I have not used this package, but you mention a fixed factor but no random factors (so a GLM) – is that correct?

You carried out a bird survey using a fixed route (roads – what distance covered and how many times) and used this to report a relative abundance value. However, by not taking into account distance of records from the transect route this is not a particularly useful figure, other than to show there were lots of them (a density estimate would have been more ecologically informative). Could you compare relative abundance between different habitat categories (or even paved vs. track if those are associated with degree of development/agriculture) – is there evidence that parrots are more abundant where there were higher densities of food trees?

Did distance vary with the nature of the destination of the parrots on each movement? Did they primarily move between seed food trees (potentially resulting in local seed tree gene flow but not necessarily impact on forest community composition, or did they frequently move to non-food perches where wasted seeds may alter local tree communities in time? I assume that seeds consumed by parrots are destroyed. Thus, parrots can only be short distance dispersers (<300m, Fig.5) unless there is observational bias in these data. Even though these parrots may show much larger-scale diurnal or seasonal movements, you present no evidence that seeds are carried over such distances. Finally, you demonstrate that the parrots disperse seeds (and potentially in ecologically meaningful quantities), but what might the impact of this be on their habitat if, for example these parrots were to change in abundance, or shift their range southwards in response to climate change?

Author Response

This is an interesting study. Readability is generally good, but terminology is opaque in places. My suggestions for alternatives to your terminology include: consumption rather than exploitation; branch bark (?) rather than branch flesh; absent or missing rather than lacking. I think Readability would also be improved with less unnecessary terminology (e.g. fruiting rather than fructification). At the very least they should be defined for those not experts in this field (epizoochory, endozoochory, stomatochory?).

Response.Thank you very much for your positive appreciation of our study. We have included the corrections proposed by the reviewer, including the reduction of terms that could be redundant, and the definition of the terms that have to do with the mechanisms of seed dispersal.

Your methods involve measurements of various parameters of food seed pod selection and processing either before or after transport by parrots. Listing these by term, definition and units in a table would improve clarity as I rather got lost reading the text.

Response. We have included some indications to more clearly interpret the terms to which the reviewer refers, both in the text and in the legends of the tables where these terms are mentioned. We have considered better to include these specifications in the text than in a new table, in order to try to avoid repeating these specifications.

Your reasoning behind your use of MCMCglmm (presumably the R-package) over alternatives such as REML GLMMs is not given and would help justify your choice to the reader. I have not used this package, but you mention a fixed factor but no random factors (so a GLM) – is that correct?

Response. MCMCglmm is the name of the package (R-package) used for models. Although at first glance it could seem that we are dealing with a mixed model, this package can include random terms but it can be also used without it as a simple glm (our procedure). This is similar to other packages (glmmTMB, for example) which run glm or glmm depending on the inclusion (or not) of the random term.

There are several approaches in R to model censored data. We choose MCMCglmm over other REML options because despite REML is fast and easy to use and MCMC can be slow and technically more challenging (in particular when you use priors, which is not our case), REML based procedures use approximate likelihood methods that may not work well sometimes. MCMC is also an approximation but the accuracy of the approximation increases the longer the analysis is run for, being exact at the limit. In addition, REML uses large sample theory to derive approximate confidence intervals that may have very poor coverage, especially for variance components. Again, MCMC measures of confidence are exact, up to Monte Carlo error, and provide an easy and intuitive way of obtaining measures of confidence on derived statistics such as ratios of variances, correlations, and predictions. Anyway, results should be similar using any of them as there are no strong constraints regarding assumptions.

We have included a clarification on this in the new version of the manuscript.

You carried out a bird survey using a fixed route (roads – what distance covered and how many times) and used this to report a relative abundance value.

Response. As cited in the manuscript, we conducted 529 km of roadside surveys. We have included in the new version of the manuscript that each road transect was surveyed only once (thus avoiding pseudoreplication).

However, by not taking into account distance of records from the transect route this is not a particularly useful figure, other than to show there were lots of them (a density estimate would have been more ecologically informative).

Response. We also recorded distances of detection for estimating parrot density (individuals / km²) through distance sampling modelling, but only reported the obtained density in Discussion. We have now incorporated this information in Methods (lines 180-181ines) and Results (lines 256-257).

Could you compare relative abundance between different habitat categories (or even paved vs. track if those are associated with degree of development/agriculture) – is there evidence that parrots are more abundant where there were higher densities of food trees?

Response. Unfortunately, we cannot assess it since it was not among our many objectives. Our aim conducting roadside surveys was just to obtain overall estimates of parrot abundance to compare them with other abundance estimates of the same and other parrot species across the world, showing an unusual high abundance of burrowing parrots in the Monte Desert.

Did distance vary with the nature of the destination of the parrots on each movement? Did they primarily move between seed food trees (potentially resulting in local seed tree gene flow but not necessarily impact on forest community composition, or did they frequently move to non-food perches where wasted seeds may alter local tree communities in time?

Response.The fruits were dispersed to different places such as other algarrobo trees, other plant species, poles and power lines used as perches, or in open country when parrots released fruits in flight. Seed dispersal under the fruiting trees occurs mainly due to the waste of food, while the dispersions outside the crown of the tree of origin can occur in different places inside and outside the wooded areas. Unfortunately we do not record the place where each fruit was dispersed associated with each dispersal distance.

I assume that seeds consumed by parrots are destroyed.

Response. As expressed in the manuscript, when parrots feed on unripe pods they focussed on the consumption (predation) of seeds (soft) in all cases. When feeding on ripe pods, they only consumed the pulp but never the seeds (hard) or other parts of fruits in both algarrobo species.

Thus, parrots can only be short distance dispersers (<300m, Fig.5) unless there is observational bias in these data. Even though these parrots may show much larger-scale diurnal or seasonal movements, you present no evidence that seeds are carried over such distances.

Response. We are sorry we do not understand this comment. We are providing evidence that burrowing parrots disperse undamaged seeds (within pods) at distances averaging 238 m for Prosopis alba and 418 m for Prosopis nigra. These distances are well above 100 m—a distance threshold often used to define long-distance seed dispersal [see Jansen et al. 2012 and references there]. We have clarified this point in lines 503-505 of the Discussion. Seed dispersal by epizoochory could reach longer distances.

Finally, you demonstrate that the parrots disperse seeds (and potentially in ecologically meaningful quantities), but what might the impact of this be on their habitat if, for example these parrots were to change in abundance, or shift their range southwards in response to climate change?

Response. The questions raised by the reviewer are very interesting, but we currently have no answer for them. Specific studies would be necessary to address the impact of parrot population reduction on the habitat, and vice versa, as well as the possible impact of climate change on parrots and their habitat, but these issues require long-term studies and specific methodologies to address them.