| 24 Jul 2026
Assessing and reducing dry-season water-scarcity risk on a tourism-exposed small island: a multi-source water-balance and robust-optimisation framework for Si Chang Island, Thailand
Abstract. Dry-season freshwater scarcity is a recurrent hazard on small tropical islands, where seasonal rainfall, limited storage and tourism-driven demand coincide. We assess and reduce this risk for Si Chang Island, Thailand, with an integrated framework that resolves inter-month storage carry-over across coupled sources. Connected catchment is estimated from a land-use layer and satellite imagery, separating rooftop harvesting to household cisterns from watershed runoff to the reservoir–pond–bank system, while tourism demand enters through an equivalent-resident formulation. Supply reliability is quantified with reliability–resilience–vulnerability metrics under a downscaled CMIP5 ensemble, from which RCP4.5 is selected by bias screening. A mixed-integer optimisation, extended to a two-stage robust formulation, sizes capacity against stochastic dry years. Rooftop harvesting is the dominant source, and the present system is reliable but limited by small cistern capacity; under projected demand a structural annual deficit renders the island supply- rather than storage-limited, so additional storage alone cannot secure supply and harvestable catchment must expand. Critically, a design optimised on mean rainfall delivers markedly lower reliability under realistic variability, and drought-robust sizing requires substantially more catchment. The framework yields transferable, drought-robust design targets for small islands.
The paper presents a strong methodology that could be widely applicable or adaptable for similar islands or water management systems. I see a need for improvement in the paper's presentation, especially in the methods section. I also think the paper would benefit from adding an SI section, where some of the tables and figures should be moved. Code and data availability should also be shared during the review process so the reviewers can check it as well. I recommend publishing this paper but it could benefit from some substantial editing.
Figure 1 is called location of the Si Chang Island but I think you cannot read the coordinates well not the legends with the scale, there is no north arrow and no way to understand where in regards to the rest of Thailands land mass it is located.
Table 1 and line 97: City/Town/Commercial class for land and named Land Development Department classification from which country? Is this information necessary maybe remove the Code. Table 1 should be moved to SI and also the area in m2 and km2 conversion is unnecesary. Table 1 you calculate a sum but where do you use this value and what for?
Line105, 165 : where did you take the runoff coeffcient from?
Figure 2 includes the demand and demand future which is only calcuted in the 2.4 and there Figure 2 is not cited anymore.
Figure3 and 4: move to SI or combine Figure 1 and Figure 4 for example?
Line 131: how is the future population growth calculation justified?
Line 142-145: you give two methods here but the units between the two methods to compare the values are not the same on is m2 and the other ha. In the end which value did you choose and why?
Figure 5 and chapter 3 methods. I would rather you describe the method in written form and have less information in the figure. It is hard for me to follow the methodology in this way. The subheadings don’t match with the wording in the Figure 5 so it is hard for me to follow which chapter explains which part of the methodology
Line 175: whats the self suficiency criterion is it included in the equations (9) and three equations are given but only one index.
Chapter 5.3 is maybe better to redistrubute along the other chatpers and add to the introduciton.
Line 182: do not understand why this comment on physical units in necessary. Why did you add it?
Line 190: Please add a graphical representation of the probability density function used for the Monte Carlo sampling to the Supplementary Information.
Line 201: comment of data availability should be mentioned in the end of the manuscript. Consider removing.
Figure 6 and text 3.6, similar ot text Figure 5 I think the figure describes the method better to me than the written text. The section tries to do too many jobs at once, method, selection criterion, results, a tie-break, a robustness check, and a CMIP5-vs-CMIP6 defense, all in one dense block. There's also a criterion-switching problem. The text says selection is "made on annual-rainfall bias... not day-to-day correlation," and screens out MPI-ESM-LR on that basis. But to choose between the two survivors, it switches to monthly dry-season fidelity without acknowledging the switch or explaining why that criterion wasn't part of the original screen.
Section 3.6: is it climate forcing? Just say forcing makes me wonder what you’re doing.
Table 4, 5 and 8 should be moved to the SI. As well as Table 9 and 10 because the data can be already seen from Figure 11. Same for Table 11.
Explain Figure 8 only in chapter 4.4. why need to add chapter 4.3 then?
Section 4.3: why did you only try to add 20,000m3 of added storage how do you justify this number (only explained in chapter 5.3) and can you add two or three other scenarios to the figure 8 to compare different outcomes and influences. You already did all the work to make these calculations so adding more scenarios would be a great way to extend on your current work.
Code and data availability should be shared also during the review process so the reviewers can check it as well.
Author contributions: Tanachaichoksirikun, P. developed the optimization model, performed simulations, and wrote the original manuscript why is ne not the first author?