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

The study addresses how hydrological droughts and cooling water shortages affect thermo-electric power plant availability and wholesale electricity prices in Great Britain. In liberalized electricity markets, disruptions that remove low-cost generation shift dispatch up the merit order, raising system marginal prices and imposing welfare losses on consumers. Large-area meteorological hazards like drought can simultaneously affect multiple units, amplifying system-level price impacts. Prior work has highlighted energy sector dependence on reliable water resources for hydropower and cooling, and documented large economic losses from droughts elsewhere. However, few assessments have used probabilistic, spatially explicit approaches linking low flows to plant-level outages and market price impacts. This paper’s purpose is to quantify these risks under historical and future climates, exploring uncertainty and extremes, and to inform resilient energy and water management.

Evidence from Brazil and California shows droughts displacing low-cost hydropower with more expensive thermal generation, incurring substantial price impacts. Numerous studies document electricity’s dependence on water for hydropower and cooling, and evaluate climate risks to power plant reliability, including European-scale analyses of drought-driven price impacts. Risk-based water resources planning frameworks have been advocated, but few studies probabilistically propagate climate and hydrological uncertainties to plant-level low-flow constraints and then to market prices. Existing economic methods (e.g., input-output, CGE) often lack spatial detail and non-linear supply-curve representation necessary for heterogeneous meteorological hazards. This study addresses these gaps by coupling probabilistic hydroclimate simulations, plant-level environmental flow constraints, and an electricity market representation.

The authors develop a coupled hydroclimate–electricity market framework to estimate welfare impacts on wholesale prices from cooling water shortages. Climate and hydrology: They use Weather@Home (W@H) regional climate ensembles (HadAM3P/HadRM3P) to produce 100 unique 30-year daily simulations for three time slices: Baseline (1975–2004), Near Future (NF, 2020–2049), and Far Future (FF, 2070–2099) under RCP8.5. The DECIPHER hydrological model simulates river flows at 24 gauges nearest to 32 freshwater-cooled thermo-electric plants. DECIPHER is calibrated using 1973–2003 observed precipitation (CEH-GEAR), PET (CHESS-PE), and discharge, with 10,000 Monte Carlo parameter sets per gauge evaluated using log Nash–Sutcliffe efficiency and RMSE; the top 100 parameterizations per gauge represent hydrological uncertainty. This yields 720,000 30-year daily flow simulations (100 climate samples × 100 parameter sets × 3 time slices × 24 gauges). Power plant availability: For 32 freshwater-dependent thermal plants (coal, CCGT, biomass, waste-to-energy; 35–2400 MW), daily availability is constrained by environmental flow requirements (EFRs) and a current system plus (CSP) hands-off flow regime. Curtailment begins near Q91, allocations are reduced to protect EFRs at Q90 (10–20% of Q90 reserved depending on ecological sensitivity), with a no-go-below threshold at 75% of Q90, and at Q70 only 10% of allocation remains. Given typical minimum stable generation, availability is effectively zero at or below Q70. Coastal plants and once-through temperature constraints are excluded, as are freshwater temperature effects for evaporatively cooled plants. Electricity supply curve: A bespoke short-run marginal cost (SRMC) supply curve represents 893 generating units totaling 86.88 GW (from DUKES). SRMCs by technology (35 types) are adopted from National Grid’s ELSI model, with unit-level cost variation by age. Wind and solar variability is represented by monthly P10/P50/P90 production adjustments using 2013–2016 daily data, yielding 36 adjusted supply curves (3 per month). Other generation availability is assumed 100% except when thermal capacity is constrained by low flows. Sensitivity to fuel price is examined by ±25% adjustments to coal, gas, biomass, and oil SRMCs. Demand model: Daily GB electricity demand is estimated using a gradient boosting regression trees model (Huber loss), trained on 2012–2017 observed demand (Elexon/Sheffield, via Gridwatch) and meteorological covariates (population-weighted MIDAS observations from 13 major urban stations): min/mean/max temperature, wind speed, windchill, month, week, and weekday/weekend indicator. Public holidays are excluded. Cross-validation yields R² ≈ 0.81, CV ≈ 0.87, and bias −0.08%; load duration curves and seasonal/weekly profiles are well reproduced. W@H daily climate is population-weighted to drive simulated demand for each ensemble member. Price impact simulation: For each day, demand intersects the adjusted supply curve to determine a strike price. When low flows reduce thermal availability, affected capacity is removed from the supply curve, shifting it left and increasing the strike price for that day, holding short-run wholesale price-demand elasticity at zero (retail buffering). Welfare impact is computed as the additional cost from increased strike prices paid over contracted generation, aggregated monthly, annually, and across ensembles. Analyses report cumulative annualized costs over 30-year samples, distributions of extreme (p99) days, impact duration curves, return periods, and sensitivities to renewables output and fuel prices.

Plant-level availability: Under the baseline, median unavailability across plants is 3.4–4.2% (spread across climate samples ~±3%). Under climate change, medians increase to 5.5–6.9% (NF) and 5.8–11.2% (FF), with greatest impacts at upstream/smaller rivers (e.g., Willington C and Ironbridge). Cumulative capacity impacts: In the baseline, negligible impacts (~1% capacity) occur ~24% of days; 10% of days see ~10% (5–17%) of freshwater thermal capacity unavailable. On extreme (99th percentile) days, median unavailability reaches ~40% (32–47%). Under climate change, negligible-impact days rise to 33% (NF) and 43% (FF), but 10% of days have 20% (NF) and 29% (FF) of capacity unavailable; extreme-day unavailability rises to 46% (NF) and 52% (FF). Extreme-day and annual summaries (Table 1): Median cumulative annual curtailments increase from 4.7 TWh per year (Baseline) to 7.7 (NF) and 11.4 (FF). Median 99th-percentile-day unavailable capacity is 6999 MW (Baseline), 7998 MW (NF), and 9070 MW (FF) (P5–P95: Baseline 5599–8160 MW; NF 6752–9871 MW; FF 7875–10,396 MW). Electricity price and welfare impacts: Baseline annualized cumulative costs are typically £29–66 million per year for the majority of climate samples (0–80th percentile), with ~20% of cases yielding £66–95 million per year. Climate change increases costs substantially: the NF median is ~£93 million per year and the FF median ~£129 million per year. The worst baseline cases (~£95 million per year) are comparable to the best FF cases (~£88 million per year). Seasonal/monthly: Impacts concentrate in late summer–autumn (Aug–Nov). In the baseline, roughly every other year sees autumn impacts with median monthly costs near zero; under climate change, medians reach tens of millions per month and ~3 in 4 years experience impacts. Interannual extremes and return periods: Single-year impacts can exceed £200 million for 1-in-25-year events; the most extreme events can exceed £300 million in the baseline and ~£400 million under climate change. Distributional changes show worsening across the entire year-rank spectrum, including wetter years in future climates. Sensitivities: High monthly renewables output lowers strike prices but increases the incremental price sensitivity to thermal unavailability; net system costs remain lower with high renewables, potentially offsetting drought impacts. Fuel price ±25% produces about +30% and −36% changes, respectively, in the median annualized impact. Correlation analysis identifies 13 plants (6509 MW combined; e.g., Rugeley, Didcot, Ironbridge) with weak inverse correlations (−0.21 to −0.27) between availability and system price impacts. Overall, cooling water shortages impose persistent, accumulating costs even without blackout risk, and climate change shifts both frequency and severity upward.

The findings demonstrate that Great Britain’s electricity system, though resilient to blackouts during droughts, is moderately vulnerable to cooling water shortages that elevate wholesale prices. Spatially heterogeneous low-flow risks translate into plant-specific unavailability that, when aggregated, produces sizeable system-level price effects—particularly during late summer and autumn. Climate change amplifies both cumulative and extreme impacts, pushing median annualized costs above £100 million per year and elevating single-year risk to well over £200 million at moderate return periods. The analysis clarifies how drought-induced supply reductions interact with the merit order and variable renewables, with implications for system planning: high renewables output can buffer total system costs but may heighten the marginal price response to thermal outages. Ongoing coal retirements may reduce exposure, yet freshwater-dependent thermal capacity remains material, and prospective CCS deployment could increase water intensity and concentrate risks in clusters. The probabilistic approach reveals substantial uncertainty from hydroclimate variability, highlighting that decision-relevant, low-probability events can drive high costs. These results inform proportionate adaptation, emphasizing both plant-level and system-level measures to manage evolving drought risks.

This study provides a probabilistic, spatially explicit quantification of how cooling water shortages during droughts affect thermal plant availability and wholesale electricity prices in Great Britain under historical and future climates. By coupling large-ensemble regional climate simulations, a calibrated hydrological model, environmental flow constraints, and an electricity market model, the authors show notable current exposure and substantial increases in cumulative and extreme cost impacts under climate change. The framework advances risk-based assessment for water–energy interdependencies and offers evidence to justify investments and policies that mitigate first-order economic risks from drought-driven cooling water shortages. Future research should incorporate system evolution (generation mix, demand growth), detailed representation of other water users, improved characterization of variable renewables and fuel price dynamics, potential CCS deployment and siting, and development of historical curtailment datasets to validate and refine risk estimates. Evaluating adaptation options (technical retrofits, alternative cooling, regulatory instruments for water allocation, and cross-sector coordination) within this probabilistic framework would further support robust decision-making.

Key limitations include: exclusion of coastal plants and freshwater temperature constraints on once-through cooling; assuming 100% availability for non-thermal and non-impacted generation; no short-run demand response (price elasticity set to zero in wholesale market context); reliance on an approximated SRMC-based supply curve due to commercial sensitivity; static representation of capacity and demand (no long-term system evolution such as retirements, new builds, or socioeconomic change); limited characterization of other water users and cross-sector allocation dynamics; absence of publicly available historical records of drought-related curtailments for validation; and omission of certain indirect or opportunity costs. Uncertainty remains from climate and hydrological modeling despite the large ensembles, and results are sensitive to renewables output and fuel prices.