The authors present an interesting article improving water management options in the global hydrological model Xanthos where these novel options for global hydrological models forsee irrigation, hydropower and flood control reservoirs. The spatial resolution is 0.5 °. Different operation rules were applied to the different reservoir type after optimizing long-term hydropower production. Model results were compared to measured streamflow data and the agreement was improved for most basins. The paper can be published after minor revisions:

Minor:

• Setup a list of abbreviations
• Check the figure numbers starting in sec. 3.2
• Several times you mention several basins (e.g. l. 445f) or rivers (Fig 5); it would be interesting to see were they are located in the world; try to include that in the figures
• Further minor comments are in a pdf

General comments:

This is a nice paper which demonstrates the benefits of implementing reservoir operation schemes designed specifically for hydropower reservoirs and an improved version of the Xanthos model. I have a couple of minor comments but otherwise think this will make a nice contribution to the wider literature.

Specific comments:

L35 – the text here switches between using km³ and m³ to describe the reservoir capacity, it might be easier to compare the numbers if you use consistent units.

Have the authors looked into whether the model performance is much worse downstream of lumped reservoirs compared to single reservoirs? I’d be interested to know whether combining multiple reservoirs (and reservoirs of multiple types) has a significant impact on the simulated flow.

L212- the new hydropower scheme is described as being trained with naturalized inflow generated by MRTM without the water management option, and I wondered if you might be able to give some context to the quality of these reservoir inflow simulations? Often bias in model inflows makes running online reservoir simulations tricky, is that something the authors found impacted this work?

L371- Do you know when all the reservoirs represented in this study were built? Presumably pre 1971 but if not the calibration period could be calibrating the model to an unimpacted timeseries.

L332 – this seems like a really useful metric, but could you give some example units for X_m and ? Initially it wasn’t clear to me that  would be cumulative.

It’s quite hard to distinguish between flood control and other reservoirs in figure 3a, perhaps you could consider using a different color.

L384-  Does selecting the most downstream gauge in each basin not mean that you exclude a lot of the closer, more impacted gauges from your evaluation? It would be interesting to see whether your gains in model performance are larger at the closer, more impacted gauges.

L447-448-  is the improvement in high and low flow periods one that has only been observed visually or have you quantified this?

L517-  could the authors clarify how the RII has been used here? My understanding is that here the metric is only accounting for the largest hydropower reservoir in the basin, but if there were several large reservoirs, would it not be important to account for the cumulative capacity? And equally, to contextualize the RII associated with the hydropower reservoirs with the RII calculated from any reservoirs of another type.

There isn’t much discussion of the impacts of the surface water extraction on the model performance. Did this have any noticeable effect or did most of the improvements come from the reservoir operations?

Technical corrections:

L215-  this sentence is accidentally repeated