Drought and climate change impacts on cooling water shortages and electricity prices in Great Britain

The risks of cooling water shortages to thermo-electric power plants are increasingly studied as an important climate risk to the energy sector. Whilst electricity transmission networks reduce the risks during disruptions, more costly plants must provide alternative supplies. Here, we investigate the electricity price impacts of cooling water shortages on Britain’s power supplies using a probabilistic spatial risk model of regional climate, hydrological droughts and cooling water shortages, coupled with an economic model of electricity supply, demand and prices. We find that on extreme days (p99), almost 50% (7GWe) of freshwater thermal capacity is unavailable. Annualized cumulative costs on electricity prices range from £29–66m.yr-1 GBP2018, whilst in 20% of cases from £66-95m.yr-1. With climate change, the median annualized impact exceeds £100m.yr-1. The single year impacts of a 1-in-25 year event exceed >£200m, indicating the additional investments justifiable to mitigate the 1st-order economic risks of cooling water shortage during droughts.


Review
The manuscript presents a novel integrated approach that includes a physically-based modelstatistical model and a power system model to evaluate the evolving economic cost of droughts on the UK electricity sector. The subject is of high interest to the Nature Communications audience. The approach includes a robust probabilistic analytics, careful attention has been given to evaluating models, and the no-gohands-off approach to derate the thermo-electric power plants generating capacity based on water availability is an improvement with respect to existing literature. Overall this is a very well performed analytics that requires clarification in some of the assumptions and reporting of the results, and a sensitivity analysis of the results with respect to fuel prices which is important since the system scale cost is at the center of the analytics.
Couple comments in no particular order: 1) Based on the supplementary material, it seems that there is minimal reservoir regulation on the observed flow at the monthly time scale. Since the derating is however performed at the daily time scale, is there an assessment on how reservoir operations could alleviate some of the derating in the results? While this is not necessary to add in the manuscript, it seems worth mentioning as a limitation.
2) Are droughts over the UK a year-long process? Some of the results are at the annual time scale while the seasonal and monthly analysis is taken into account to describe the inter-actions with the renewables. While the analysis allows for the author to comment on how renewables can compensate for droughts in terms of cost, it might not be a "drought" process anymore? Could you clarify?
3) Use of $US currency in the introduction and UK pounds in the remaining of the text -while I understand, it makes the paper a bit confusing. 4) Page 2: Voisin et al. 2016 and 2018 used distribution of capacity derating as a way to evaluate potential vulnerabilities of the grid and associated the risk with system-scale economic cost, with an explicit "risk-based approach". While those applications were not done under climate change and neither with as many runs and robust statistics as done in this analytics, those papers should however be mentioned to nuance the claim of the paper that this is the first attempt to use this risk-based approach.
[ presentation of the distribution and risk-based approach] Voisin, N., M.  5) The authors made careful sensitivity analysis. Beside clarifications throughout the paper, I would recommend the authors to perform a sensitivity analysis on the results with volatility in prices. While the focus is on climate change, it would be more transparent for the electricity sector to also address/ comment on how fuel price volatility affects the perceived cost of drought. Cost of fuel is a major source of uncertainty for long term planning. It may also affect the statement on how renewables counter-balance the cost of droughts. Or there might be UK price fluctuation regulation in place that already address this point. Please consider. This comment is in relation to a paper published this year on the sensitivity of the system-scale and regional scale cost to fuel prices. O'Connell et al. (2019) demonstrated a significant sensitivity with the cost of drought being as significant as extreme natural gas price volatility. It might be more significant over the UK , or not, but it brings more perspective on the findings of the analytics for long term planning purposes. 7) In the discussion, please clarify that the demand is changing only in response to temperature and not to changes in population or technology innovation. Another assumption important to clarify with respect to expectation for long term planning in the electricity sector is that the generation portfolio is not changing -no new plants are added and old / aging plants are not retrofitted. While not change in the analytics is needed, this is worth mentioning. 8) Clarification of the numerical experiment in the introduction: the baseline is presented as " no derating" but there is inter-annual derating -please clarify. In the UK are the Hydros considered renewable? I ask this because in Brazil in reason of the impacts of large dams, many Hydros are no more acceptable as renewable (in the sense of being environmentally friendly). For this classification, we consider only the run of a river power projects or the small ones . 4. 3. The scales and explanations related to figure 7 deserve to be improved. 5. 4. I recommend a broader explanation for the 1st. Paragraph of item 5. 3 Finally I recommend the publication of the article considering the small revisions suggested.

Page P# in new manuscript
Reviewer 1

1.1
The manuscript presents a novel integrated approach that includes a physically-based model -statistical model and a power system model to evaluate the evolving economic cost of droughts on the UK electricity sector. The subject is of high interest to the Nature Communications audience. The approach includes a robust probabilistic analytics, careful attention has been given to evaluating models, and the no-go-hands-off approach to derate the thermo-electric power plants generating capacity based on water availability is an improvement with respect to existing literature. Overall this is a very well performed analytics that requires clarification in some of the assumptions and reporting of the results, and a sensitivity analysis of the results with respect to fuel prices which is important since the system scale cost is at the center of the analytics.
Thank you for the rigorous review, complements and suggestions for improvement.
We have endeavored to implement all recommended changes.
1.2 1) Based on the supplementary material, it seems that there is minimal reservoir regulation on the observed flow at the monthly time scale.
It's correct, the potential effects of reservoir regulation to alleviate curtailment is not specifically assessed. All of the reservoirs on the rivers concerned are water supply reservoirs, which we can expect to be at low levels during times of low flow.

Page P# in new manuscript
Since the derating is however performed at the daily time scale, is there an assessment on how reservoir operations could alleviate some of the derating in the results? While this is not necessary to add in the manuscript, it seems worth mentioning as a limitation.
These reservoirs are not operated to alleviate water shortage at power plants. The effect of reservoir regulation at headwaters water supply reservoirs on power plants in the lower reaches of catchments is small . We have added this caveat to the discussion: "These may be both technical adaptations at the unit level, or regulatory instruments to optimise water allocation within catchments, for example allocation trading between users and reservoir operation, though we do not expect these to have a significant effect on flows for the power plants considered" Discussion, page 11

1.3
2) Are droughts over the UK a year-long process? Some of the results are at the annual time scale while the seasonal and monthly analysis is taken into account to describe the inter-actions with the renewables. While the analysis allows for the author to comment on how renewables can compensate for droughts in terms of cost, it might not be a "drought" process anymore? Could you clarify?
Droughts in the UK have been experienced over multiple years (Barker et al 2019).
In the paper we have considered all occasions of low flows, whether they occur during a "drought" (however it is defined), or not. Impacts were calculated on an annualized basis to present an amortized risk -although as we see (Fig 5 in paper) some years would be more expensive than the others. One option was to isolate the drought events and present the cost of each event, although this was not done as some difficulties arise in doing this, e.g.: 1) de-rating also occurs outside of droughts, so how to account for this effect?
2) definition of drought over the spatial area, and would we consider plants not in the drought zone but that are de-rated due to low flows?
3) in the Far Future scenario, depending on the definition, the UK would be experiencing drought, somewhere, a lot of the time.
As you note, the renewables variability is substantial, which is why we have presented the effects of this, and centered the analysis using the median renewables case. Another option would have been to fix the renewables production The impacts of drought on power plant cooling water shortages on electricity prices Byers, E.A., et al.

REVIEWER COMMENT RESPONSE (new and modified text sections shown in blue)
Page P# in new manuscript to an annual mean. Whilst this would make clearer specifically the impact of drought, it reduces the realism of the study and risks over-estimating the economic impacts.
In line with your comment, we have made more careful use of the terms "low flows" to describe any single day with flows below the Q 90 and "drought" relating to more prolonged events.
Throughout text

1.4
3) Use of $US currency in the introduction and UK pounds in the remaining of the text -while I understand, it makes the paper a bit confusing.
We agree. We reported the values as reported in the original sources. These are converted to current 2020 values and British Pounds for better context. used distribution of capacity derating as a way to evaluate potential vulnerabilities of the grid and associated the risk with system-scale economic cost, with an explicit "risk-based approach". While those applications were not done under climate change and neither with as many runs and robust statistics as done in this analytics, those papers should however be mentioned to nuance the claim of the paper that this is the first attempt to use this risk-based approach.

Additional costs in electricity prices have been estimated to be in the order of US$41 million(Almeida
Thanks for drawing these important studies to our attention, and it is appropriate to cite them in this section. The text is slightly modified and these references have been added.: However, very few studies have applied probabilistic methods and risk assessment approaches [18,19]  Thanks for this suggestion, which is a very important point. To explore this issue we ran two additional sensitivity simulations, with fuel prices for coal, oil, gas and biomass 25% higher and lower than the base case. Depending on the technology, subsequent short-run marginal costs for the technologies increased in the order of 15-22%, given that non-fuel costs also vary by technology.

REVIEWER COMMENT RESPONSE (new and modified text sections shown in blue)
Page P# in new manuscript would be more transparent for the electricity sector to also address/ comment on how fuel price volatility affects the perceived cost of drought. Cost of fuel is a major source of uncertainty for long term planning. It may also affect the statement on how renewables counter-balance the cost of droughts. Or there might be UK price fluctuation regulation in place that already address this point. Please consider. This comment is in relation to a paper published this year on the sensitivity of the system-scale We have added the result to the text of section 3.4 and a new plot in Figure 6c to highlight the results of this sensitivity analysis: "Fuel prices also have the potential to augment or dampen the economic impacts. For the Baseline scenario, it was found that +/-25% change in all fuel prices, i.e. coal, gas, biomass and oil, resulted in, respectively, +30% and -36% change in the median annualized impact. This is similar to a finding for the US that found that natural gas price volatility to be as significant as the impacts of drought [26]."

REVIEWER COMMENT RESPONSE (new and modified text sections shown in blue)
Page P# in new manuscript 6) Drought definition is missing. Authors seem to use a plant-scale drought definition ( flow percentile), yet the drought definition at the system scale is based on the overall derating or weighted flow at the different stations? Please clarify.
This is a good point. We consider anything below the historical Q 90 to be a low flowand this is the point at which plant curtailment begins. We have specified this now early in section 3.1 at beginning of results: Aggregated over the Baseline period, individual powerplant unavailability due to low river flows (Figure 2, a) varies between 1-8%. Here, we characterise low flows as days where the river discharge is below the historical Q 90 When this is prolonged (with no specific definition made on the duration), then this would be considered a drought.
In the introduction and discussion, we have generally used drought as we are discussing the aggregated risks, that would be prolonged, spatially extensive, affecting multiple plants, and accumulating impact.
In reporting the results, we have replaced about 8 mentions of drought with low flows as this is more precise i.e. there would be some days (e.g. 1 day low flow), that results in curtailment of a plant. This curtailment counts towards the aggregated impact cost estimates, but the 1-day event was not a drought. So, for example, caption of Figure  Section 3.1, page 3 Section 3.3, page 7 7) In the discussion, please clarify that the demand is changing only in response to temperature and not to changes in population or technology innovation. Another assumption important to clarify with respect to expectation for long term planning in the electricity sector is that the generation portfolio is not changing -This is an important point worth reiterating in the discussion. As you are aware, the reasons these weren't included comes down to two main points: 1. Our desire to explicitly quantify the drought risk on the basic operation of the current system, without obfuscating this with additional scenarios of capacity change and changing societal demands. 2. The difficulty in knowing the locations of any future plants, and of what type they will be. Thanks, it's a relevant point that perhaps becomes clearer in the Methods but is not clear at this stage in the manuscript. The first mention of renewables, (section 3.3), we clarify that we mean the non-thermal renewables.:

REVIEWER COMMENT RESPONSE (new and modified text sections shown in blue)
Note that different dynamics of energy demand and the availability of non-thermal renewables (wind, solar, hydro, hereafter, renewables) have a role in the daily and monthly impacts.

Page P# in new manuscript
Finally I recommend the publication of the article considering the small revisions suggested.