Enhanced solar and wind potential during widespread temperature extremes across the U.S. interconnected energy grids

We use hourly atmospheric data from the European Centre for Medium-Range Weather Forecasts ERA5 reanalysis (Hersbach et al 2020). We use 2 m dry-bulb temperatures between 0–23 UTC to calculate daily maximum temperature, surface solar radiation downwards (SSRD) to calculate daily cumulative solar potential, and 100 m U and V wind components to calculate daily maximum wind speeds (WS) and wind potential (100 m represents the approximate hub height of modern wind turbines; Hartman 2022).

We quantify widespread hot and cold extremes across the six NERC regional entities (figure 1(a))—Western Electricity Coordinating Council (WECC), Midwest Reliability Organization (MRO), Texas Reliability Entity (TRE), Southeastern and Central Regional Reliability Corporation (SERC), Reliability First Corporation (RFC), and Northeast Power Coordinating Council (NPCC). NERC regions are used because local electricity grids within these regions are interconnected to enable resource sharing, energy reliability, and security, as well as to coordinate operations, monitoring, and infrastructure resilience (NERC 2023).

Zoom In Zoom Out Reset image size

We define hot extremes in summer (June–August) and cold extremes in winter (December–February) as standardized daily temperature anomalies exceeding ±1.5 σ . These standardized anomalies are calculated by removing the 15 day local climatological (1991–2020) mean centered on that date and dividing by the climatological standard deviation of daily temperatures for that 15-day window. This is done to remove the seasonal cycle and account for the local conditions residents are acclimated to. Cold anomalies are multiplied by −1 to facilitate easier comparison with hot anomalies. For each day, we calculate the fractional area in each region experiencing hot or cold extremes. Widespread extremes are days in the top quintile of area affected by extremes, calculated from all days in a region when extremes occur in at least one grid cell. The top quintile threshold was selected to have sufficient sampling of events for calculating statistics. We define frequency as the number of widespread extremes in a season, extent as the fraction of the region under temperature extremes, duration as the number of consecutive widespread extremes, and cumulative intensity as the sum of the standardized temperature anomalies multiplied by area (km²) across all grid cells with extremes within each region.

Degree days is a measure of heat or cold in a location (Quayle and Diaz 1980, Heim et al 2003, Shaffer et al 2022) that is commonly used to represent the weather-driven component of energy demand (Waite and Modi 2020, Doss-Gollin et al 2021, Lee and Dessler 2022), as the two are correlated (Quayle and Diaz 1980, Heim et al 2003, Lee and Dessler 2022). We use cooling degree days (CDD) in summer and heating degree days (HDD) in winter as proxies for energy demand following the widely-used American Society of Heating, Refrigerating, and Air-Conditioning method (Thevenard 2011). These are calculated as follows:

$$\begin{align}{\text{HDD}}& = \left\{ \begin{array}{*{20}{c}} {{18}\,\,^ \circ }{\text{C}} - T,& {\text{ for }}T \unicode{x2A7D} {{18}\,\,^ \circ }{\text{C}} \\ 0,& {\text{ for }}T > {{18}\,\,^ \circ }{\text{C}} \end{array}\right.\nonumber\\{\text{CDD}}& = \left\{ \begin{array}{*{20}{c}} T - {{18}\,\,^ \circ }{\text{C}},& {\text{ for }}T \unicode{x2A7E} {{18}\,\,^ \circ }{\text{C}} \\ 0,& {\text{ for }}T < {{18\,\,}^ \circ }{\text{C}} \end{array}\right.\end{align} \tag{ 1 }$$

where $T$ is dry-bulb temperature. This widely used metric assumes energy use for heating (cooling) when temperatures are below (above) 18 °C (64.4° F), and thus provides an approximation of energy demand that does not require representing the technological, social, and other factors that shape local electricity demand (Doss-Gollin et al 2023). Degree days are multiplied by the 2020 population (CIESIN 2016) for each grid cell to calculate population-weighted energy demand as in Doss-Gollin et al (2021). Total degree days are summed over the entire region for each widespread extreme and non-extreme day.

We use daily cumulative SSRD and daily maximum 100 m WS data to calculate solar potential and wind potential, respectively. These metrics are proxies for how much renewable energy can be theoretically produced each day. Wind potential is measured by the wind capacity factor, which is the wind power per turbine (in kW) divided by the maximum power generated, and solar potential for a photovoltaic panel (PV) is similarly measured by solar capacity factor. These metrics are calculated following (Bett and Thornton 2016, Amonkar et al 2022):

$$\begin{align}& {\text{Wind power per turbine }}\left( {{\text{kW}}} \right)\nonumber\\&\quad = \left\{ \begin{array}{*{20}{c}} {{ }0;{\text{ for WS}} < 3\,{\text{m}}\,{{\text{s}}^{ - 1}}{ }or{\text{ WS}} \unicode{x2A7E} 25\,{\text{m}}\,{{\text{s}}^{ - 1}}} \\ {a - b*{\text{WS}} + c*{\text{W}}{{\text{S}}^2} - d*{\text{W}}{{\text{S}}^3} + e*{\text{W}}{{\text{S}}^4} - f*{\text{W}}{{\text{S}}^5} + g*{\text{W}}{{\text{S}}^6} - h*{\text{W}}{{\text{S}}^7} + i*{\text{W}}{{\text{S}}^8};{\text{ for }}3\,{\text{m}}\,{{\text{s}}^{ - 1}} \unicode{x2A7D} {\text{WS}} < 13\,{\text{m}}\,{{\text{s}}^{ - 1}}} \\ {2000;{\text{ for }}13\,{\text{m}}\,{{\text{s}}^{ - 1}} \unicode{x2A7D} {\text{WS}} < 25\,{\text{m}}\,{{\text{s}}^{ - 1}}} \end{array}\right.\end{align} \tag{ 2 }$$

where WS = daily maximum windspeed, a = 634.228, b = 1248.5, c = 999.57, d = 426.224, e = 105.617, f = 15.5487, g = 1.3223, h = 0.0609186, and i = 0.001162565. a–i are coefficients from the wind power curve for a V90-2.0 MW Vestas turbine from Amonkar et al (2022). These coefficients and the cut-in (3 m s⁻¹; speed at which the turbine starts to generate power), rated (13 m s⁻¹), and cut-out (25 m s⁻¹; speed at which the turbine is mechanically limited and needs to shut down to avoid damage) speed thresholds are turbine-specific and could vary for different turbines. We convert the wind power into a unitless wind capacity factor by dividing the calculated wind power by the maximum operating power output of 2000 kW. Since we are using a wind power curve to calculate the wind capacity factor, the mechanical efficiency is incorporated. However, wind power efficiency losses due to temperature or icing are not accounted for.

Solar capacity factor is calculated using the following equation:

$$\begin{equation}{\text{Solar Capacity Factor}} = {\eta_{{\text{rel }}}}*{ }\frac{G}{{{G_{{\text{STC}}}}}}\end{equation} \tag{ 3 }$$

where G is the irradiance or the SSRD; G_STC = irradiance at standard testing conditions (STC) of 1000 W m⁻² and ${\eta _{{\text{rel}}}}\left( {G,T} \right)$ is the relative efficiency that accounts for variations in performance of the PV with temperature. ${\eta _{{\text{rel}}}}\left( {G,T} \right)$ is calculated from the PV module temperature (T_mod), standard testing temperature (T_STC = 25°C), ambient temperature (T₀ = 20°C), and nominal operating cell temperature (T_NOCT = 48°C). The module temperature is calculated as follows:

$$\begin{equation}{T_{{\text{mod}}}} = \left( {T + {T_{{\text{NOCT}}}} - { }{T_0}} \right)*{ }\frac{G}{{{G_0}}}\end{equation} \tag{ 4 }$$

where the irradiance G₀ = 800 W m⁻². Using T_mod, the relative efficiency ${\eta _{{\text{rel}}}}$ is calculated using the following equation:

$$\begin{align}& \eta_{{\text{rel }}}\left( {G,T} \right) = \nonumber\\ &\quad \left[ {1 +{\alpha} {{ \Delta }}{T_{{\text{mod}}}}} \right] * [ {1 + {c_1}{\text{ln }}G{^{^{\prime}}} + {c_2}{\text{l}}{{\text{n}}^2}{ }G{^{^{\prime}}} + {\beta}{{ \Delta }}{T_{{\text{mod}}}}} ]\end{align} \tag{ 5 }$$

${\text{where }}\,\,\alpha\,\, =\,\, { }4.2\,\, \times\,\, {10^{ - 3}}\,\,{ }{{\text{K}}^{ - 1}}$; $\beta \,\,=\,\, - 4.6 \times {10^{ - 3}}{ }{{\text{K}}^{ - 1}}$; ${c_1}$ = $0.033$; ${c_2}$ = $- 0.0092$; $\Delta {T_{{\text{mod}}}}$ = $({T_{{\text{mod}}}} -$ ${T_{{\text{STC}}}}$), and G' = G/G_STC.

Annual summaries of electrical disturbances compiled by the U.S. Department of Energy (DOE) Office of Cybersecurity, Energy Security, and Emergency Response (ISER—electric disturbance events (DOE-417), 2023) are used to examine power outages associated with 'severe weather' from 2012–2021. Within each region, we identify the days on which weather-related outages coincide with widespread extremes and refer to them as extreme-related power outages. Non-weather related outages and unmatched outages are excluded from our analysis. Power outages during widespread temperature extremes could occur due to supply shortages from reductions in power generation capacity or failures in the transmission or distribution infrastructure such as sagging power lines due to heat or ice accumulation or strong winds during winter storms. The outage data does not include detailed information on the cause of the outage beyond categorizing them as 'severe weather' disturbances nor the precise location of outages within the NERC region in which they occurred. Linking the outage to the specific weather conditions during widespread extremes would require the outage location to co-locate with weather conditions. In the absence of such data, we focus on the coincidence of outages with widespread temperature extremes in each region.

To examine the influence of modes of variability on widespread extremes characteristics, we use standardized indices of the Arctic Oscillation (AO), North Atlantic Oscillation (NAO), El Niño-Southern Oscillation (ENSO; Niño3.4), and Pacific/North American pattern (PNA) (https://psl.noaa.gov/data/climateindices/list/). The positive and negative phases for each mode are defined as months with standardized anomalies exceeding ±0.5.

First, we quantify the frequency, extent, intensity, and duration of widespread hot and cold extremes (1980–2021). Median characteristics vary substantially across seasons and NERC regions, as does their variability (figure 1). WECC, MRO, and SERC have the highest median frequencies of widespread cold extremes (11.5, 9, and 7.5 d-per-year) and hot extremes (13, 9.5, and 11 d-per-year, figure 1(b)). However, individual years can have substantially more events. For instance, during the 1983–1984 winter, TRE and MRO recorded their highest number of widespread cold extremes, ∼5 times and 2 times higher than their median, respectively. Similarly, the record high frequency of widespread hot extremes in TRE and SERC in 2011, in MRO, NPCC, and RFC in 1988 and in WECC in 2021, exceeded their corresponding medians by a factor of 4.

Every region has experienced at least one widespread cold or hot extreme that affected nearly the entire region. Widespread cold extremes typically affect larger fractions of each region than widespread hot extremes, and have higher cumulative intensities (figures 1(b) and (c)). NPCC and TRE have the largest extents of widespread cold extremes, with their medians exceeding 85%. Some events that affected the entire extent of these region lasted multiple days. For example, the February 2021 cold wave included several days with 100% of TRE affected and the highest cumulative intensity day. In winter 1983–84, MRO, SERC, and RFC experienced their most intense cold extremes. This winter also produced the largest cold events for SERC and RFC. While extent is standardized by region size making their maximum extents similar across regions, cumulative intensity depends on region size. Therefore, the regions with the highest extents do not have the highest maximum or median intensities—these are typically experienced in WECC and MRO.

For widespread hot extremes, the largest median extents (∼60%) also occur in NPCC, followed by RFC (∼44%), whereas the median extent in all other regions is <32%. The relatively large extent of both widespread hot and cold extremes in these regions is likely due to their small sizes and regionally homogenous response to synoptic-scale temperature extremes. The largest region—WECC, has the smallest median fractional extent of widespread hot extremes (typically <25%). However, in 2021 WECC experienced multiple heat days that affected up to ∼70% of the region simultaneously and contributed to the highest cumulative intensity during 1980–2021.

The median duration of widespread extremes is ∼1–2 d, but every region has experienced protracted events that have resulted in substantial damages (figure 1(e)). For example, during the 1983–84 winter, MRO and TRE experienced 14- and 9 day stretches of widespread cold extremes, respectively, during a cold wave that broke duration records in many cities, resulting in economic damages of ∼6.2 billion USD (in 2023 dollars; NOAA Billion Dollar Weather disaster database; Ludlum 1984)). While event extents and intensities are typically greater in winter, frequencies, and durations are generally greatest in summer. Multiple regions experienced widespread hot extremes lasting >10 d, with the longest event lasting 22 d in SERC in 2011. MRO and TRE experienced their longest lasting widespread hot extremes (14 and 11 d, respectively) in 1980, during a prolonged hot, dry summer which resulted in economic losses of at least 39.7 billion USD (in 2023 dollars; NOAA Billion Dollar Weather disaster database; Karl and Quayle 1981). In WECC, the longest hot event (18 days in 2020) contributed to severe wildfire activity and a multi-week widespread air pollution episode across the western U.S. (Kalashnikov et al 2022).

Hot and cold extremes often drive peak demands for energy, the prediction of which is an important component of resource planning on seasonal time scales (Zamuda et al 2018). To compare the energy demand during days with widespread extremes to normal days, we calculate the population-weighted HDD and CDD (equation (1)). Widespread extremes are associated with significantly higher median energy demand (HDD and CDD) in all regions, relative to demand on non-extreme days (figures 2(a) and (b)). Differences between degree-day distributions on extreme and non-extreme days are larger for widespread cold extremes, compared to widespread hot extremes, likely due to their greater extents and cumulative intensities (figures 1(c) and (d)). During winter, the median HDD is greatest in RFC, nearly 2 times that on non-extreme days. Similar ratios are also seen in TRE and SERC (figure 2(a)). During summer, the median CDD is greatest in SERC, WECC, and RFC, and widespread hot extremes also significantly amplify CDD compared to non-extreme days (figure 2(b)).

Zoom In Zoom Out Reset image size

The increase in energy demand along with direct impacts to electricity infrastructure (Dawson et al 2018, Climate Central 2022) have resulted in power outages in all NERC regions in at least one season in the past decade (figures 2(c) and (f)). Using data between 2012–2021, we quantify widespread extreme days with power outages, fraction of days with weather-related outages that coincide with widespread extremes, and the annual fraction of widespread extremes with an outage. Across all years, SERC and TRE have experienced the highest number of widespread cold extremes with outages, accounting for ∼19% and 53% of all winter days experiencing weather-related outages, respectively (figure 2(c)). In 2021, ⩾42% of widespread cold extremes were associated with power outages in every region except NPCC, including 100% of events in TRE and RFC (figure 2(e)). In summer seasons, RFC has experienced the highest number of widespread hot extremes with outages, accounting for ∼26% of all summer days with weather-related outages (figure 2(d)). In four of the past ten summers, >50% of widespread hot extremes in RFC have been associated with outages, the highest among all 6 regions (figure 2(f)). These results suggest that the electricity infrastructure in RFC is considerably more vulnerable to widespread hot extremes than other regions whereas SERC and TRE are relatively more vulnerable to widespread cold extremes. Although WECC has a relatively lower absolute number of outages than these regions, widespread heat extreme-related outages account for 41% of all summer days with weather-related outages in this region, a majority of which occurred in 2020 (figures 2(d) and (f)). Notably, NPCC has experienced no outages associated with widespread cold extremes during this period.

The risk of power outages during extremes could be exacerbated by coincident reductions in energy supply. The meteorological processes that drive widespread temperature extremes also affect surface solar radiation and wind speeds, and thus have the potential to affect the capacity of solar and wind energy systems. Therefore, we examine how energy potential from these sources is affected during widespread extremes. The median regional-average solar capacity factor is significantly higher in most regions during widespread cold extremes, with the largest increases over RFC, SERC, and NPCC and smallest increases over WECC and MRO (figure 3(a)). TRE is the exception where the solar capacity factor is significantly lower during widespread cold extremes. Similarly, the median regionally averaged solar capacity factor is significantly higher during widespread hot extremes than during non-extreme days in all regions, with the largest increase in TRE (figure 3(b)). These results suggest that all regions could likely benefit from expanding solar capacity to enhance electricity generation during widespread temperature extremes.

Zoom In Zoom Out Reset image size

The response of wind energy potential on widespread temperature extreme days is more varied across regions than solar potential. NPCC is the only region with significantly higher median wind potential during widespread cold extremes relative to non-extreme days (figure 3(c)). Conversely, wind potential across SERC and MRO is significantly lower during widespread cold extremes, whereas across WECC and TRE the differences are statistically indistinguishable (figure 3(c)). In contrast, wind potential during widespread hot extremes is significantly higher over all regions except WECC relative to non-extreme days. The largest increases in median wind potential during widespread hot extremes occurs over TRE and MRO, suggesting that these regions could likely benefit most from increased wind generation capacity during hot events.

Enhanced solar potential during widespread hot extremes is because such events are primarily associated with atmospheric ridges (Horton et al 2015, Grotjahn et al 2016, Loikith et al 2017, Xie et al 2017, Agel et al 2021). Ridges are typically associated with clear skies that allow more solar radiation to reach the surface, however these conditions can also decrease wind potential in some areas. For example, during the hottest day of the 2021 Pacific Northwest heat wave (29th June), solar radiation was higher than average across areas of WECC experiencing the largest positive temperature anomalies (figures 3(e)–(g)). Near-surface winds were below-normal across much of the northern and eastern part of WECC (figure 3(f)), the leading edge of the associated ridge. Due to the large size of WECC compared to other regions, temperature, wind, and solar conditions vary spatially during such extremes. For example, solar anomalies were below normal over southeastern WECC where cool anomalies persisted while wind anomalies were above normal over Arizona. Such heterogeneity likely explains why WECC shows some of the smallest differences between median solar and wind potential anomalies on widespread extreme versus non-extreme days (figures 3(a)–(d)).

Widespread cold extremes across the U.S. are associated with cold air outbreaks (Smith and Sheridan 2020). These systems are associated with strong winds at their leading edge during cold air advection, but calmer conditions at their center under the surface high pressure system, co-located in space and time with the coldest conditions (Robert De 1923). These atmospheric conditions explain the increased solar potential and suppressed wind potential across most regions during cold extremes (figures 3(a) and (c)). For instance, during the coldest day of the exceptional 2021 North American cold wave across the central U.S. (15th February), MRO and TRE experienced above-normal solar radiation, and much of MRO experienced below-normal winds (figures 3(h)–(j)). Conversely, TRE saw relatively weak but positive windspeed anomalies over most of the region. Concurrently, stronger than normal winds were observed including over parts of SERC and WECC and weaker winds were observed over NPCC (figure 3(j)). However, these negative anomalies over NPCC are not typical during cold extremes and are likely due to the winter storm's extreme southward extent (Bolinger et al 2022). Typically, positive median wind anomalies in this region during widespread cold extremes (figure 3(c)) are likely because the region generally lies at the leading edge of cold air outbreaks that are most often centered around central US. For similar reasons, solar radiation anomalies were above normal over TRE even though typically, solar radiation is suppressed during widespread cold extremes. Such event-to-event variability in energy potential is observed for solar and wind across all regions (figures 3(a)–(d)).

To understand the changing nature of widespread hot and cold extremes, we calculate trends in event characteristics over 1980–2021 (figures 4 , S1 and S2). All regions show a decline in the frequency of widespread cold extremes, with significant trends (p-value < 0.1) over TRE (−79.9%) and NPCC (−97.8%; figure 4(a)). There are no significant changes in the extent, duration, or cumulative intensity of widespread cold extremes over any region, although there is a tendency towards less frequent, less intense, and smaller extent events in SERC, and less intense events in RFC (p-value ⩽0.17, figures 4(b)–(d) and S1). Contrastingly, half of the regions showed trends in one or more characteristics of widespread hot extremes (figures 4(e)–(g) and S2). WECC is the only region with significant trends in all metrics—increases in frequency (123.6%), average extent (32.5%), average duration (55.2%), and average cumulative intensity (29.3%) of widespread hot extremes. Of the six regions, TRE experienced the largest increase in frequency of widespread hot extremes (131.6%), along with small, insignificant trends in other characteristics. In contrast, SERC experienced a significant decline in the extent (−30.4%) and intensity (−37.4%), along with a substantial but insignificant decline in duration (p-value = 0.15; figures 4(f), (h) and S2). MRO and NPCC are the only regions with no significant changes in the characteristics for either widespread hot or cold extremes. These observed trends are consistent with greater warming in the western U.S. relative to other regions (van Oldenborgh et al 2019, Keellings and Moradkhani 2020, Perkins-Kirkpatrick and Lewis 2020, Rogers et al 2021, Wanyama et al 2023), increased summertime ridging over western states (Dong et al 2021), and the 'warming hole' in summer and winter over the southeastern and midwestern U.S., partly driven by anthropogenic aerosols, agricultural intensification, and natural climate variability (Leibensperger et al 2012, Banerjee et al 2017, Mascioli et al 2017, Vose et al 2017, Partridge et al 2018, Coffel et al 2022)

Zoom In Zoom Out Reset image size

Next, we examine the association between widespread extremes and four known modes of natural climate variability that represent potential sources of predictability—AO, NAO, ENSO, and PNA (figures 5 and S3). All modes significantly influence the frequency, extent, or intensity of widespread cold extremes in at least one region (figure 5). La Niña and negative PNA conditions favor more frequent widespread cold extremes across WECC and negative NAO conditions favor more frequent widespread cold extremes across MRO (figure 5(a)). Widespread cold extremes are significantly more intense and larger over WECC during positive AO and negative PNA conditions, over MRO during negative PNA conditions, and over SERC and TRE during La Niña and negative PNA conditions (figures 5(b) and (c)). In addition, negative AO and NAO conditions are associated with significantly larger cold extremes over NPCC.

Zoom In Zoom Out Reset image size

Compared to cold extremes, fewer regions exhibit robust relationships between the characteristics of widespread hot extremes and the concurrent phase of climate modes (figure S4). The frequency of widespread hot extremes was significantly higher than normal during positive NAO conditions over WECC, negative NAO conditions over SERC, and positive PNA conditions over RFC. Positive AO, NAO, and PNA conditions favor significantly larger extremes across WECC. In addition, negative PNA conditions favor significantly larger and more intense heat extremes across RFC, but positive PNA favors an increase in the frequency of events. La Niña and negative AO conditions were also associated with significantly larger and more intense heat extremes over SERC. These relationships between natural climate variability modes and widespread hot and cold extremes are largely consistent with the previously identified influence of modes on the pattern of temperature extremes across parts of the U.S. (e.g. (Westby et al 2013, Loikith and Broccoli 2014, Grotjahn et al 2016, Yu et al 2019, Shi et al 2021).