|Review of “Deduction of reservoir operating rules for application in global hydrological models” by H. Coerver, M. Rutten and N. van de Giesen for potential publication in HESS|
Clarifications have been added and the comparison with respect to the generic rules is adding tremendously to the paper. I strongly support the fuzzy approach to improve the representation of reservoir regulation of river flows. Some conclusions with respect to the overall contribution of the approach (improvement upon GRAND database, application to GHMs) remain however still unsupported.
1) Applications to GHMs:
- Significant errors in inflow;
The approach still needs some support for claiming that it can be applied to GHMs. GHMs can only be calibrated to a certain extent and in specific locations, not at 40,000 nor 6,000 dam locations. GHMs have significant errors in the seasonality and the mean annual balance of the river flows. Those errors get even larger when irrigation and other sector demand need to be represented. Reservoir characteristics are set, i.e. if there is a 10% overestimation of flow, some small reservoirs can only regulate as much as the set reservoir capacity allows for it. The current approach seems to assume that the errors in flow are not taken into consideration, or are very small. Page 25 second paragraph does not address the point of the errors in flow and how it could affect the decision in particular.
- Mass balance : P25 first paragraph: the system says that there is no mass balance check, which is a no-go for GHMs. In some areas with groundwater-surface water interactions that are still a challenge to simulate, reservoirs dry up in the simulation but not in reality. For node-based water resources management models, the inflow input into the system is typically bias corrected in order to apply the observed regulation rules. Based on the previous point, if there is a consistent under/overestimation, how can you ensure that you are not creating a source/sink of water in the system.
Either the approach needs more clarification, or/and a mass balance check is necessary to claim application to GHMs in particular, and for integration in any hydrology model in general.
2) Physical/operational insight
An improvement is the analysis by type of impoundment and some conclusions now relate to this classification. It really improves the paper and the analysis! There are remaining concerns:
- The current conclusion based on below and above median level of impoundment presently does not support the conclusion because the median is based on the 11 selected dams. Based on GRAND or ICOLD database, if you were to derive a level of impoundment for the 6000 dams, where are the dams selected for this paper?
- The authors have complemented the discussion based on Hejazi et al. (2008) which discusses the human decision making process based on reservoir characteristics. There is still no discussion on the type of hydrological regime for which the fuzzy approach improves upon the generic rules. Even for specific level of impoundment, the seasonality in flow can affect drastically the performance of the generic rules, and the fuzzy rules. For example, in the context of a reservoir containing 30% of the annual flow and the spring snowmelt is 60% of the annual flow, the variations in reservoir storage from the generic operating rules are certainly further from operations rules as they will drive to drying and uncontrolled spilling. When the reservoir can store only 10% of the mean annual flow, in the same situation, the generic operating rules have much lower impact on the overall regulation. This is the sum of all the small reservoirs that make an impact on the regulation. With this perspective, the generic operating rules offer an advantage in that they allow representing all the reservoirs, even the small ones ( which can be created without the GrAND database), and with less data constraints than the fuzzy approach. In the discussion of the improvement upon the generic operating rules, the system effect of many small reservoirs should be mentioned even if it does not get evaluated. This discussion would provide more insights on the actual improvement of the approach - only the reservoir scale is evaluated but not the system scale. This is okay, but it needs to be mentioned.
3) Improvement upon previous approaches/Grand database
I am not convinced that the approach here allows to improve upon the GRAND or ICOLD databases. The fuzzy approach can be applied to the GRAND database, and future satellite-based emerging databases. And so does the generic operating rules approach as well. Both approaches need storage characteristics, and information on inflow, (and demand, reservoir purpose if possible, etc, as applicable). I suggest removing this section (in order to avoid confusion and divergence from the main objective of the paper around the fuzzy roles and improvement wrt the generic rules. Else you could clarify that both generic and fuzzy approach can be used with both GRAND, ICOLD and other databases. Generic rules need only long term mean monthly flow while the fuzzy approach need a longer training periods ( time series instead of average) with possibly a longer time series when the inter-annual variability is large (and this point could benefit from being discussed).
The improvement with respect to previous approaches should likely focus on the new analysis quantifying the improvement with respect to the generic rules only, which is already provided.
Minor: the text refers to figure 11 then jumps to figure 15.