The manuscript describes the development details of a future fractional urban impervious surface area (ISA) dataset for 2015-2100 at a 5-year interval. I think the newly developed future urban ISA dataset will be very useful for understanding the impact of future urbanization on the ecosystem. I have reviewed the revised manuscript and the point-by-point responses to the comments. The authors have revised the manuscript following the suggestions and comments closely. They did a lot of work in quantifying the uncertainties of data harmonization, which increased the reliability of the model and dataset. Overall, the authors have done a good job in addressing these comments, and the manuscript has been improved a lot. But I still have several small suggestions and provide them in the specific comments.
1. P4, Line 16-19. “Given that there are currently no urban fractional ISA dataset in high spatial resolution (e.g., 1km) directly obtained from satellite observations, here we adopted the commonly used strategy through spatial aggregation from high-resolution (e.g., 30m) urban extent data to derive the ISA time series data for modeling.”
As I know, there are at least two fractional impervious surface area datasets have been developed. For example, the Global Man-made Impervious Surface (GMIS) developed by NASA Socioeconomic Data and Applications Center could be available since 2017, which can be accessed at https://sedac.ciesin.columbia.edu/data/set/ulandsat-gmis-v1.
2. P7, Line 15, the stochastic disturbance item is missed in equation (4), and no description of the ‘SP’ item.
3. P9, Line 4-6. “Here we assumed the trend of urban sprawl at the state level is consistent with that at the country level, as population and GDP change are commonly estimated at the country and regional scale”.
This is a simple downscaling method to get the future urban land area demand of each state, and may result in some uncertainties as the urbanization stage varies. It is also contradictory to the description in the first paragraph of section 3.1, indicating the urbanization stages information was not used in the future urban land area prediction. The better way to downscale the future urban land area from country to state is to set the urbanization stage as a weight.
4. One of the corresponding authors published related work in 2019 and 2021 (Li et al., 2019; Li et al., 2021). The two papers also simulated the future urban land expansion based on the nightlight data derived urban land. So, what are the improvements of the newly developed dataset compared with previous work? It can be included in the discussion.
5. Same as Reviewer 2, Fig. S6. I note that there will be no low-density ISA area in the city you show after 2060, and it seems that most of the urban area have the same ISA fraction. It also existed in other metropolitan areas (e.g., Fig. 10 and 11, New York city).
This may be resulted from the spatial allocation algorithm. Specifically, the grid cell with high suitability always has more ISA increment. On the other hand, no enough newly developed urban land grid cells to allocate the increased ISA. Thus, there should be a balance between urban land expansion and ISA increase in the existing urban land pixels. It will be good to improve the spatial allocation model by constraining the filling of urban inner space and expansion of urban bound. Thus, there should be some discussions about the uncertainties of the spatial allocation model.