General Comments

This paper introduces a method for calculating luminescence ages using a dose rate distribution in combination with an equivalent dose distribution. It is a fantastic new tool to have for luminescence dating and I am pleased to see that it is readily available for community use in the R Luminescence package. As advances in dose rate characterisation continue, dose rate distributions could become more common, and this method could indeed be part of a paradigm shift as suggested in the paper’s discussion section.

Most of my comments are suggestions for very minor word changes either for grammatical reasons (I often tried to list these under the technical corrections but I’m sure I sometimes listed them under specific comments instead) or to improve clarity. One comment that is repeated at various places throughout the paper is that using the word ‘sample’ to refer to both a sediment sample for dating and a sample of numbers from a distribution can at times be confusing. Also that using ‘De_Dr Age’ instead of ‘apparent age’ would be more specific and help readers to know that the number being referred to was calculated using the De_Dr age calculation method introduced in this paper.

One of the tenets of the De_Dr method presented is the identification and removal of outliers based on the shape of the Dr distribution (once the internal overdispersion of the equivalent dose distribution has been incorporated), and a significant proportion of this paper focuses on testing the ability of the method to identify outliers introduced to simulated datasets. Whilst I think there is exciting potential for this De_Dr method to identify outliers in real datasets, in some ways it feels like introducing this concept already is skipping a step in the development of the De_Dr method, which first needs to test on real samples the assumptions often stated in this paper regarding whether the shape of equivalent dose distributions mirrors the shape of dose rate distributions. It is not yet known how well a dose recovery test overdispersion value fully characterises the natural internal overdispersion or whether that internal overdispersion is symmetrical. However, these are questions to be addressed by experiments on real samples and what this paper introduces is a tool that can be used to address them, which is very exciting.

I find it curious that the De_Dr method does not expect uncertainties on individual dose rate values input for the dose rate distribution. I haven’t personally used numerical sediment models for predicting dose-rates but based on analysing this paper I presume that they don’t calculate such uncertainties. Even still I wonder if further advances in such models or in experimental dose rate distribution measurement techniques might lead to uncertainties being generated, but then given such circumstances, this De_Dr method could also be further developed to account for them.

Overall, I very much enjoyed reviewing this paper and I think it’s ideas and (readily available!) method will stimulate further research and advance luminescence dating techniques. I’m happy to recommend it for publication following minor revisions.

This is quite an interesting manuscript on the implementation of new age model for use with SG-OSL data. The approach that is advocated makes sense and seems to be implemented well. The ability to fit any pre-defined distribution of ‘expected’ burial-doses is needed (or should be one day, once we know how to estimate them). The model achieves this using an outlier detection routine that is quite innovative. However, the details of the model are not the focus of this manuscript (it seems to be published elsewhere). Here there is some introduction to the model concept and software, but the main purposes of study seems to be to test the model using simulated datasets. This makes a good, justifiable basis for a paper, but at the moment the central part of the paper is underwhelming: the number of tests performed is limited with some obvious questions not addressed; and there is no discussion to speak of (and there really should be, for a paper like this). I have given some points below that might provide some bulk to the manuscript; some of these might be worth considering in detail, particularly on the uncertainty of the true burial-dose distribution.

For the model testing (Section 3):

• All the simulations so far assume that the chosen distribution is correct (e.g. the Dr distribution model is tested on simulations that use that exact distribution +outliers). In reality, we are unlikely to ever have perfect knowledge of the burial-dose distribution. The obvious question is how does the model perform when the true burial-dose distribution is not known? E.g. if the burial dose is assumed to be lognormal in the simulation, but normal in the De_Dr analysis?
• How does the model compare to existing models/approaches? This is of practical importance for dating, because users want to know whether the new model makes a significant difference in accuracy or precision.
• What about averaging of De estimates? The assumption is that each De relates to a single grain, but in practice there might be more than one grain in some grain holes. What is the consequence for the model- does it create a significant error or not?
• What is the sensitivity to the number of grains? The simulations use 100 grains form the burial-dose population, which is more grains than are normally measured with good precision. How does the model perform when the number of ‘good’ grains is limited? Is it better or worse than alternatives.

Some things worth considering for the discussion:

In what situation does/would the model fail, or be worse than simpler alternatives?

Can stratigraphic information be used to help constrain the age estimates (i.e. can it be incorporated into an age-depth model)

Given the current state of knowledge, what distributions would be most suitable for the Dr? what factors are significant in determining the distribution? Is there any benefit of using the model if the distribution is not well known? How would users go about selecting the correct distribution? What research is now needed to resolve the question of the distribution?