The manuscript outlines a methodology for constructing a time-dependent fixed asset model applicable to China. It establishes a model using fixed asset data at the provincial level and further refines it to a more detailed grid level with remote sensing ancillary data. The findings are validated with results from other studies. The methodology is purported to assist in quick post-earthquake loss estimations.

The topic is interesting, and the manuscript is well-written. However, it can benefit from some comments I suggest below to improve its clarity and contribution to the field.

General comment:

1. The methodology and results seem useful for assessing economic losses due to any hazard, not only due to earthquakes. Even though the introduction includes references about the importance of earthquakes and previous earthquake loss estimations, the methodology and results do not confront loss estimations from past earthquake events. Furthermore, Fig. 2 has no input that one could identify with an earthquake hazard or risk (e.g., ground motion fields, earthquake occurrence rate, or fragility model). The methodology and results focus on modelling the observed economic and demographic growth in the study areas consistently, considering the data quality limitations in terms of spatial resolution and time-frequency. I think the paper can be stronger (and closer to the journal's scope) by adding some statements in the discussion and conclusion about how the methodology and results can help in the loss estimation due to an earthquake event or other natural hazards. It would also be interesting to add to the introduction a review of studies of other natural hazards, such as floods or extreme wind events.

Specific comment:

1. The conclusion says the methodology can be extended to more recent years once the data is available. However, considering that the methodology aims to help in quick loss assessment for future earthquake events, can the methodology with the available data today (in 2024) provide a prediction, for example, of the losses after an earthquake event in 2030?

2. I suggest mentioning the future availability of grid-level fixed asset data only in the section “code and data availability”, as it is done and justified in this version of the manuscript, and only mentioning it in the abstract and conclusion in a later version, when the data is effectively available.

Technical corrections

1. Tables 2 and 3, as well as Figs. 5, 7-9: Consider changing the monetary units to billion Chinese yuans, as done in Figs. 11-12

2. Fig. 2: There is a typo in one of the charts: “Harmonized” instead of “Harmanized”

3. Although described in the text, the delta in Eq. 3 has a different meaning than the delta in Eq. 4. I suggest using a different symbol for one of them.

4. Table 1: There is a typo in the 2nd column, 8th row: “Population density data” instead of “Population dentsity data”.

1. The overall quality of the preprint (general comments)

The overall quality of the paper is high. The topic of developing a novel fixed asset model to improve seismic loss estimation is significant for the science community. By mapping fixed assets at a 1 km × 1 km grid level, the model better serves rapid seismic loss assessments and informs emergency response plans. The research is well structured and explained. The authors made an effort to combine various data sources and techniques.

While the model represents a significant advancement, several limitations could impact its effectiveness, particularly in high-stakes applications like earthquake response. The paper’s scientific contributions justify publication with minor revisions to handle specific data assumptions better and further validate the model’s application. These adjustments would help ensure the model’s broader applicability and robustness.

1. Individual scientific questions/issues (specific comments)

Here are some topics which the authors could discuss in more detail:

Reliance on Historical Investment Data and Simplified Depreciation Rates. The model bases its estimates on historical investment data and applies a uniform 5% depreciation rate across all provinces, regardless of variations in asset type, economic condition, or regional maintenance practices. The uniform deprecation rate can introduce inaccuracies, especially for assets with different service lives or conditions. The simplified approach to depreciation may lead to skewed asset values, particularly in provinces with unique economic trajectories or asset compositions. For instance, in industrialized regions, assets may have a shorter useful life than in less industrialized provinces, affecting the accuracy of economic loss projections. Can the model be refined by including a variable depreciation rate based on more detailed asset-specific and regional data, if available?

Inconsistencies in the Data Sources for Ancillary Datasets. The model relies on ancillary datasets (e.g., nighttime lights, population, built-up areas) which are not consistently available across all years. This results in the use of alternative data types to approximate missing data. For instance, population data alone is used in the early years when nighttime light data is unavailable. These proxies may not accurately represent economic activity, especially in rural areas or less-developed regions, leading to potential over- or under-estimations in asset distribution. Could the model be strengthened by incorporating more recent, high-resolution satellite data or by exploring alternative disaggregation methods that do not depend solely on proxies like nighttime lights?

Lack of Structural Detail in Asset Composition. The model's focus on fixed capital should differentiate between asset types (e.g., residential vs. industrial buildings) in disaggregation. This lack of structural specificity reduces the model's utility for applications that require asset type differentiation, such as insurance underwriting or infrastructure resilience planning. Different asset types respond differently to seismic events; for instance, infrastructure like bridges and roads may sustain different levels of damage compared to residential buildings. This generalization could lead to misaligned resource allocations during emergency responses. Introducing asset type categorization, possibly by incorporating land use or building inventory data, would enhance the model's accuracy for specific asset loss estimations.

1. Technical corrections

The article's language quality is overall sound, with a few areas where readability and formality can be improved. Here are specific suggestions regarding grammar, spelling, and phrasing.

• Maintain past tense in descriptions of the completed study. For example, in the sentence, "The fixed asset model to be developed in this paper is also based on the Level 1 data."
• Remove redundant phrases. For example, "To summarize, the nighttime light data and GHS-POP data are used to generate the lit-pop disaggregation indexes from 1991 to 2020, while the GHS-BUILT-S data and GHS-POP data are used to construct area-pop disaggregation indexes from 1971 to 1990..."
• Avoid Informal Language. For example, replacing "Luckily" with "Fortunately" is more formal.

Revise for Consistency in Abbreviations and Acronyms. Introduce abbreviations consistently upon first mention, ensuring they are used uniformly throughout.