Forecasts for anticipatory actions play a pivotal role in mitigating the impact of natural hazards, particularly in regions vulnerable to frequent disasters like Mozambique. These forecasts not only offer invaluable insights for disaster preparedness but also facilitate timely interventions, ultimately safeguarding lives and livelihoods. I agree with the authors of this manuscript about the academic and social significance of AA-tailored forecasts, particularly in a context as critical as Mozambique. While acknowledging the importance of this study, my primary concern is its reproducibility (the method is described in a long and complex way) and ensuring the correct application of methodology, as clarity remains paramount for effective scientific communication and practical implementation.

The abstract provides significant contextual information, yet falls short in delivering a clear explanation of the "ready set go" process. Consider expanding upon this process, especially in the testing triggers section, perhaps through the inclusion of a flow chart to elucidate the operational workflow. Additionally, the practical implications of the results remain ambiguous, particularly regarding the targeting of the first part of the rainy season. Thirdly, there is a confusing focus on two specific provinces in some manuscript sections while in other sections, results over all of Mozambique are shown. Clarifying these points would enhance the abstract's effectiveness.

The introduction effectively outlines the various natural hazards affecting Mozambique but could benefit from clearer terminology regarding the description of multi hazards (meaning consecutive/compounding ones?). Additionally, it is suggested to specify that “impacts” such as flooding and cyclones are caused by these hazards rather than merely being impacts themselves. Moreover, while the introduction mentions the benefits of early action, it is imperative to address in the discussion how anticipatory actions can mitigate impacts cost-effectively, especially considering the limitations of ‘preparedness possibilities’ with respect to water management.

The methodology contains a robust framework, but improvements in clarity, adding relevant citations, and referencing to the framework are needed. Numbering the structure to correspond directly with figures and ensuring consistency between bold text and descriptions in flow charts would enhance readability. Furthermore, some terms require clarification or references, such as "blended precipitation records" and "bilinear interpolations". The core of the analysis revolves around assessing the precision in forecasting precipitation levels one standard deviation below the norm. However, amidst extensive textual explanations, this crucial aspect becomes obscured, overshadowed by peripheral details. Streamlining the method section to prioritize essential components could enhance clarity and comprehension. Also, the nature of the forecasted variables remains ambiguous. Is the SPI3 predicted several months in advance? Including a supplementary list enumerating predictors and predictands would provide invaluable clarification, ensuring transparency and facilitating a deeper understanding of the forecasting methodology.

Additionally, it is recommended to utilize the Stagge et al. (2015) approach for SPI calculation to ensure accuracy, particularly in arid regions. Most importantly, the application of a severe drought threshold for extracting drought probabilities based on SPI lacks clarity. Merely selecting a SPI value inherently provides the probability, unless there are issues with distribution fitting, which would signify a methodological problem rather than a data issue. The rationale behind defining and applying danger thresholds for drought events remains ambiguous; why not simply select values like -1 or -2? Without clear explanation, it raises concerns about methodological integrity. Additionally, the utilization of return period as a quality criterion requires further elucidation, as its significance is not apparent (Table 1).

Another significant issue arises from bias correction based on SPI between CHIRPS and forecast reanalysis data. The process of SPI fitting involves converting values to standardized ones, reflecting standard deviations from a normal distribution. However, using two time series of SPI values for bias correction raises questions about information loss and overall methodological coherence. Furthermore, the inclusion of ENSO in bias correction methodology lacks justification; while rainfall patterns may vary under different ENSO states, assumptions about forecast biases in these states are not clearly articulated. The paragraph on bias correction fails to provide convincing rationale, particularly concerning the absence of transfer functions calculated over raw precipitation data. This oversight could potentially undermine the validity of the approach. Clarification on these aspects is imperative for ensuring methodological soundness and reproducibility.

In the results in line 359, there appears to be an error in the sentence structure. The reference to "frequency" lacks clarity; it's unclear whether it pertains to the occurrences of values below -1. Given that <-1 represents the definition of severe drought, consistency in its frequency is expected across regions. Regarding Figure 4, clarification is needed on whether the counts on the y-axis represent aggregated data across ensemble members. Additionally, the observation that only 24% demonstrate skill improvement with bias correction raises questions about the efficacy of this effort. It may be beneficial to revisit this finding for accuracy. Furthermore, the persistent low hit rate in multiple regions for the "ready set go" approach, as depicted in Figure 7, suggests room for improvement. Consider exploring potential adjustments to enhance the effectiveness of this method.

Please reflect on the identified limitations in the discussion, as I feel the most important bottlenecks, such as on SPI usage, bias correction methodologies, distribution fitting techniques, and threshold selection, are barely discussed. Consider placing greater emphasis on these potential methodological challenges rather than ENSO-related variability.

Overall, enhancing clarity and addressing the identified concerns would significantly strengthen the manuscript for publication.

The Manuscript is excellent, and explores a valuable strategy that is relevant to many Anticipatory Action projects, showing that a "double checking" trigger is valuable at improving triggering skill.

I recommend that the authors only perform minor revisions to make the paper more valuable and impactful for those who may cite it.

1) ECMWF 5.1--The authors utilize ECMWF 5.0, evaluate its skill, and highlight the importance of validation.  If I understand right, ECMWF 5.1 has just been released with no hindcasts between the end of the training period of 2016 and something like 2022.  For validation for humanitarian use, it is important for ECMWF to generate the hindcasts to address that gap.  It would be valuable for the authors to discuss this issue, as it may provide motivation for ECMWF and other models to generate hindcasts for validation.  Of course if they are aware of, or uncover an existing plan for the generation of the data, that would be valuable for the reader to know.

2) Rainfall validation and crop yields, humanitarian drought impacts.  The authors touch on the difference between equating drought to rainfall and further relating it to forage, cropping losses, or humanitarian impacts.  Because there is such a large gap in the relationship of rainfall and drought impacts (perhaps as large as the error between forecasts and rainfall), it would be important for the authors to discuss this in more depth.  For example, the drought of 2015 in nearby Malawi (a year mentioned in the introduction of the paper), where rainfall in a water stress model was used for triggering ARC national insurance, there was a globally publicized gap between the rainfall formulas and agricultural losses, leading to articles in venues such as the Economist.  There is a large literature on this gap for index insurance, it directly using rainfall in more detailed formulae than forecast anticipatory action projects.  It would be valuable for the authors to cite some of the literature on rainfall basis risk/loss mismatches to highlight the importance of investigations that compliment rainfall forecast skill with rainfall relationships will losses.

3) AA projects relevance.  This paper provides a valuable set of insights for anticipatory action projects, well beyond the case study.  The authors should mention that there is a family of project (eg WFP, OCHA) that their findings could be relevant to.