Characteristics of Power Outages from Compound Weather Extremes in Florida

Characteristics of Power Outages from Compound Weather Extremes in Florida | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Characteristics of Power Outages from Compound Weather Extremes in Florida Mohammad Siddiqur Rahman, Thomas Wahl, Meghana Nagaraj, Alejandra R. Enríquez This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6404193/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract The growing number of extreme weather events has contributed to the increasing number and severity of power outages. However, the complex interplay of extreme weather events and their compounding effects on power outage characteristics (e.g., event duration is yet to be explored). Power outage data is often not publicly available, especially at high spatial resolution. Identifying outages related to weather events can also be challenging as various weather variables can trigger or moderate power outages when they occur, in isolation or combined. Here, we use county-level power outage data from EAGLE-I for the state of Florida from 2015 to 2022 to identify moderate and major weather-related outages and analyze their characteristics. We show that total outage counts were higher in metro areas than in non-metro areas. However, the percentage of weather-related power outages was higher in non-metro areas than in metro areas. Spatial variation of grid reliability indicators derived from all weather-related events follows similar patterns as derived when just focusing on tropical cyclone events, highlighting the importance of these types of extremes in creating prolonged outages. Considering six relevant weather variables, we identify univariate and compound events (i.e., when more than one weather variable was extreme at the time of the outage). Univariate events have a homogenous pattern across the state of Florida, while compound events have more localized hotspots. The average duration of the outages also increases when moving from univariate to multivariate events. Our results shed light on the relative importance of different weather variables (in isolation or combination) in creating power outages with different characteristics across Florida. Identifying such causal relationships is an important step in understanding how power outage risk profiles may change when certain extreme weather events become more frequent. Climatology Climate Analysis and Modeling Power outage extreme weather compound events EAGLE-I grid reliability Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Introduction The power grid is essential for the functioning of communities, providing critical services as identified by the Federal Emergency Management Agency (FEMA 2020). Despite its importance, the power grid across the United States faces increasing disruptions. Over the last decade (2014–2023), significant power outages – those affecting ≥ 50,000 customers – doubled compared to the early 2000s (2000–2009) (Climate Central 2024). Extreme weather events caused 50% or more of these outages (Hines et al 2009a , Do et al 2023 ), underscoring the grid's vulnerability to environmental conditions beyond typical weather patterns (Liu et al 2017 ). On the Gulf and Atlantic coast of the United States (U.S.), the most severe power outages are often due to hurricane-induced factors such as extreme wind, rainfall, or storm surge (Ali et al 2023 ). One example was Hurricane Irma in 2017, which left nearly two-thirds of Florida's customers without power due to high winds and flooding (Florida Division of Emergency Management 2017; Chakalian et al 2019 ). Superstorm Sandy in 2012, impacted more than 8.5 million people across several states in the northeastern U.S.(Mukherjee et al 2018). Hurricane Ian in 2022, resulted in more than 2.7 million people without power in Florida (Entress and Stevens 2023 ). Winter storms also have the potential to lead to widespread outages, as exemplified by winter storm Uri in 2021, which impacted more than 4 million customers in Texas (Xu et al 2023). In many of those cases, power restoration took days to weeks, or even months, for many customers. Large-scale power outages are defined by the Department of Energy (DOE) as events that affect ≥ 50,000 customers or cause load loss of ≥ 300 MW. Information on these low-probability but high-impact events is provided through the Electric Emergency Incident and Disturbance Report (DOE OE-417). Small or moderate outage events occur more frequently and affect fewer customers in specific locations. Over time, these high-probability/low-impact events can also lead to substantial disruptions. They are included in high-resolution datasets such as the Environment for the Analysis of Geo-Located Energy Information (EAGLE-I) and PowerOutage.us. DOE OE-417, which only includes large-scale events, reported 605 events between November 2014 and March 2021 (Abdelmalak et al 2023) in the U.S. compared to 17,484 outages (which lasted more than 8 hours) between 2018 and 2020 alone based on high-resolution PowerOutage.us data (Do et al 2023 ). This indicates a disparity in the number of power outages based on the selection criteria, as localized outages are unlikely to affect 50,000 customers or create 300 MW of load loss, while still significantly impacting certain communities (often repeatedly). Previous studies suggest that precipitation, wind, lightning strikes, winter weather, and tropical storms are the most dominant weather and climate events responsible for power outages in the U.S (Hines et al 2009a , Do et al 2023 ) with vegetation cover, and grid infrastructure (Belligoni et al 2025). Hines et al (2009) analyzed DOE OE-417 data from 1984–2006, reporting that wind or precipitation caused 31% of outages, while hurricanes or tropical storms accounted for 10%, and lightning strikes caused about 9%. Similar to that study, most other studies in literature analyzed weather influence on power outages from a univariate perspective. Do et al ( 2023 ), for example, showed that when outages longer than 8 hours co-occurred with at least one extreme weather variable, 75% co-occurred with extreme precipitation, and over a third co-occurred with multiple weather extremes. However, this analysis ignored shorter duration outage events (which happen more frequently and still disrupt people’s lives) and did not explicitly account for the compounding effects of multiple hazards, or hazard drivers, where not all need to be extreme at the same time to create an extreme impact (Zscheischler et al 2018 , 2020, Ali et al 2023 ). Understanding the combined influence of multiple weather variables on outage occurrences is crucial, particularly in regions like Florida, where various storm types create complex weather patterns with the potential for outages of different severity. Here, we overcome some of those knowledge gaps by addressing the following three objectives using Florida as a case study region: (i) use outage data with high temporal resolution paired with data of relevant weather variables to identify counties across the Florida peninsular that are relatively more vulnerable to power outages, in particular outages related to weather events; (ii) assess the relative importance of different weather variables in generating outage events; and (iii) quantify the role of compound events in causing outages and explore the outage event characteristics when they are caused by different types of weather events. Overall, our results lead to a better understanding of the spatial variability and drivers of weather-related outage events, thereby paving the way to develop intervention strategies that best serve the most vulnerable communities. Study area and data 2.1. Study area We chose Florida as a case study region, because power outages occur frequently across the state due to various types of extreme weather events leading to strong wind, precipitation, lightning strikes, storm surge, and flooding, thereby causing widespread disruptions. Previous work has shown that coastal states are often at higher risk of experiencing outages (Mukherjee et al 2018, Casey et al 2020, Do et al 2023 ), and 35 of Florida's 67 counties have coastlines. At the same time, rural areas (non-metro) often face longer power restoration times compared to urban (metro) areas (Spurlock et al 2023); 45 counties in Florida are classified as metro and 22 as non-metro (Fig. 1 ). 2.2. Power outage data This study uses power outage data retrieved from EAGLE-I maintained by Oak Ridge National Laboratory for the Department of Energy. EAGLE-I records encompass over 8 years of outage data from 2015 to 2022 at the county level, captured at 15-minute intervals. It is collected by different web parsers and individual scripts that track utilities' publicly available outage information to estimate the number of customers without power (Abdelmalak et al 2023)through the establishment of a collection pathway for each utility company (Brelsford et al 2024). However, data quality is contingent upon consistent reporting practices by the utility companies. Variations in reporting methods and aggregating differing geographical boundaries of counties and utility companies can introduce spatial inconsistencies into the dataset. Over the years, coverage of this data has improved, increasing from 86 percent in 2018 to 92 percent in 2022, which might also introduce some temporal inconsistencies in trend analyses. Power outages are often considered to fall into one of three categories: (i) Minor power outages, classified as minor system interruptions related to technical causes like short circuits, equipment failure, or any malfunction in the grid. Very few customers are impacted; the power outage duration is relatively short, and less likely to be caused by extreme weather (Abdelmalak et al 2023). (ii) Moderate power outages are more likely to be triggered by extreme weather events but can also be due to accidents and other causes. (iii) Major power outages have the longest durations (Bhusal et al 2020) and impact many customers. Those events are often caused by extreme weather events, including hurricanes, wildfires, or winter storms (Hines et al 2009b , Abdelmalak et al 2023, Dugan et al 2023 ).Different approaches have been proposed to do the outage event classification, usually based on the number of customers impacted or outage duration (Sullivan et al 2012 , 2018 , Abdelmalak et al 2023). For instance, DOE defines major power outages as events that exceed 300 MW of loss and/or affect ≥ 50,000 customers. Here, minor outages are outside the scope of the analysis as they are often not weather-related, and even if they are, the societal impacts are minimal. Our analysis is based on EAGLE-I data, which reports the number of customers without power in each county at 15-minute intervals. Starting with the raw data, we perform five critical pre-processing steps. The first step is setting a threshold to remove reported power outages where the number of customers did not exceed the 85th percentile of all the values for a specific county across the 67 counties in FL, which leads to values between 12 and 423 customers being impacted. Events below that threshold are considered minor and removed from further analysis. The second step is to calculate the power outage duration, defined here as the time over which the 85th percentile threshold from step 1 is being exceeded. The third step is to merge events that are separated by less than one hour (i.e., independent events have to be at least one hour apart). The fourth step is to remove events that are shorter than one hour. The fifth and final step is to assign the maximum number of customers without power during an event and the total event duration to the first day of the outage. This is because extreme weather creates the outage, which occurs at the beginning when the outage is triggered. Still, the outage itself may last much longer than the extreme weather (e.g., power outages of several days or weeks after the passage of a hurricane). 2.3. Weather data Since we are interested in weather-related outages, we include a range of different weather variables in our analysis. Weather information has been retrieved from different data sources to obtain the highest temporal and spatial resolution possible. Table 1 shows the weather variables included in the present study, their spatial and temporal resolution, and the source. CONUS404 data has a four-kilometer spatial resolution, while ERA5 data has ~ 31-kilometer spatial resolution, and both have hourly temporal resolution. For precipitation, wind speed, and CAPE, we use the daily maximum values across a county (because the EAGLE-I data is only available at the county-level), whereas for soil moisture, we use the daily mean, and for lightning strikes, we use the daily accumulated number of strikes across a county. The National Hurricane Center (NHC) reports hurricane paths through HURDAT2 at 6-hour intervals, representing the hurricane's center as a point. NOAA's Coastal Ocean Reanalysis (CORA) reports coastal water levels in a mesh grid. The analysis involves creating an equal-spaced 200-point along Florida’s coastline and extracting the daily maximum water level value for each coastal county from multiple points. Table 1 Weather variables considered in the analysis with spatial and temporal resolution and source. Variable​ Spatial res.​ Temporal res.​ Source​ Precipitation (Prcp)​ 4*4 km Hourly​ CONUS404​ (Rasmussen et al 2023) Wind Speed (WS)​ 4*4 km Hourly​ Soil Moisture (SMOIS)​ 4*4 km Hourly​ Lightning Strike (LIS)​ Point Count​ U.S. National Lightning Detection Network (NLDN) Convective Available Potential Energy​ (CAPE) 28*28 km ​ Hourly​ ERA5 ECMWF (Hersbach et al 2020) Hurricane Path​ Point 6-hourly​ HURDAT2​ (Landsea and Franklin 2013 ) Water Level (WL)​ 500 meter​ Hourly​ ​NOAA's Coastal Ocean Reanalysis (CORA 2024) Methods As a first step, we analyze the grid reliability across all counties in Florida, overall and specifically for weather-related events. Grid reliability can be assessed in different ways. One popular metric is the System Average Interruption Index (SAIDI), which many utility companies use to report grid reliability. SAIDI represents the total interruption duration for average customers in a given geographical location for predefined time periods (IEEE 2012) and is calculated as follows: $$\:SAIDI=\frac{\sum\:{D}_{i}{N}_{m}}{{N}_{T}}=\frac{\sum\:Customer\:Minutes\:of\:Interruption}{\text{T}\text{o}\text{t}\text{a}\text{l}\:\text{N}\text{u}\text{m}\text{b}\text{e}\text{r}\:\text{o}\text{f}\:\text{C}\text{u}\text{s}\text{t}\text{o}\text{m}\text{e}\text{r}\text{s}\:\text{S}\text{e}\text{r}\text{v}\text{e}\text{d}}$$ 1 D i = Duration for each 1-hour or longer interruption event. N m = Maximum number of interrupted customers for each 1-hour or longer duration event. N T = Total number of customers served in the county. We calculate SAIDI on an annual basis for every county in Florida, separately for weather events and non-weather events, and for tropical events and non-tropical events. We define weather-related outages as events when at least one weather variable exceeded its respective 90th percentile value on the day when the outage started; hence, we assume that this weather event played a role in generating the outage. Tropical events are defined as events when a storm center, as reported in the HURDAT2 database, was within a 500-km buffer of the county's boundary where an outage occurred simultaneously. All other events are considered non-tropical events. Next, we analyze how often individual weather variables triggered outages, either in isolation or in combination. The latter includes compound events, which we define as events where multiple weather variables exceeded their 90th percentile thresholds simultaneously while an outage occurred. First, we identify events where only one weather variable was extreme at the time of an outage started. For example, we find the percentage of all weather-related outage events where only precipitation was extreme or only wind speed was extreme (i.e., above their 90th percentiles) and hence likely the main driver that triggered the outage. Then, we repeat the analysis but focus on bivariate compound events, where two variables exceeded the 90th percentile threshold simultaneously when an outage started. For example, we derive the percentage of weather-related outages where precipitation and wind speed were extreme while other variables did not exceed the threshold. This analysis provides insights into the relative importance of different individual weather variables or bivariate combinations of them in triggering outage events. We exclude CAPE from this part of the analysis to limit the number of possible variable combinations, and because it is an indicator of atmospheric instability, and hence a proxy for the likelihood of a storm to occur not a measure of a specific hydrometeorological variable. Finally, we assess the characteristics of outage events in terms of their typical durations when they are caused by certain weather variables or combinations thereof. For example, are events that are caused by extreme precipitation longer in certain counties than others (e.g., based on grid characteristics or response capabilities), and/or do they have different durations than events that are caused by extreme wind, and/or do compound events display different characteristics than non-compound events? Results 3.1 Grid reliability The total number of power outage events between 2015 and 2022 in Florida exhibits spatial variability across the state, with the lowest number of 112 outages in Calhoun County and the largest number of 1,544 outages in Broward County (Fig. 2 a). More outages occurred in Central Florida (both east and west coast), with a noticeable hotspot in the southeast, while relatively fewer outages occurred in the Panhandle area and counties surrounding Lake Okeechobee. Overall, the spatial pattern of the number of outage events resembles well the metro and non-metro classification (Fig. 1 ), with metro counties often experiencing more power outages compared to non-metro counties. The percentage of weather-related events varies from 39% in Polk County to 68% in Madison County (Fig. 2 b). The Big Bend area in the northwest of Florida is a hotspot of weather-related power outages, where in many counties, up to 60% of all outages were weather-related. Central and northwest Florida have lower percentages of weather-related outages, typically between 40–50%. A contrasting pattern is observed between total outage counts and weather outage percentages. Notably, the Big Bend area experienced fewer outages overall but the highest percentage of weather-related outages. Comparing non-weather (Fig. 2 c) and weather (Fig. 2 d) SAIDI shows distinctly different spatial patterns and overall values. Non-weather SAIDI is highest in the Panhandle area in the northwest, a cluster of counties in the north and southwest, and in Okeechobee County, with values ranging from 150 to 200 minutes of interruption for the average customer per year (Fig. 2 c); note that SAIDI is first calculated for every year from 2015 to 2022 and the average is shown. While weather-related SAIDI is also high in the northwest, it is similarly high across South Florida (except Monroe, Miami-Dade, and Broward counties in the south) and most coastal counties along the Atlantic coast (Fig. 2 d). The SAIDI values in those regions are more than an order of magnitude higher than the non-weather values, in many cases reaching more than 2,000 minutes of interruption per customer and year. Relatively lower weather-related SAIDI values are found in West Central Florida and most northwestern counties, with values ranging from 500-2,000 minutes per customer and year. Overall, non-weather SAIDI shows less spatial variability than weather SAIDI. Non-tropical and tropical SAIDI have similar spatial patterns as the non-weather and weather SAIDI, respectively. The non-tropical SAIDI values are high in the same regions where non-weather SAIDI is high, with values from 100–500 minutes per year (Fig. 2 e). Tropical SAIDI is high where weather-related SAIDI is high, again with similar overall values (Fig. 2 f). One notable exception is a cluster of Monroe, Miami-Dade, and Broward counties in the south. Monroe County, in particular, often exhibits different behavior than most other counties and appears as an outlier; hence, it is not highlighted specifically in the remainder of the results description even though it sometimes shows up as a hotspot (see Discussion for more information). 3.2. Role of different weather variables in outage generation In the following, we show results from analyzing how different weather variables (or combinations thereof) contributed to outages across Florida. First, we focus on univariate events where only one specific weather variable exceeded the 90th percentile threshold (Fig. 3 ). Extreme precipitation shows a relatively homogenous pattern throughout Florida; ~5% of all weather-related outages occurred when only precipitation was extreme and no other weather variable (Fig. 3 a). High winds alone relatively often caused power outages in some Panhandle counties and parts of Central and Southeast Florida (~ 10% of all weather-related outages were only related to high winds) (Fig. 3 b). Lightning strikes were often the main driver of outages, with values reaching more than 15% in hotspot counties that are scattered across Florida but mostly in the interior of the state (Fig. 3 c). Interestingly, soil moisture (SMOIS) was also sometimes the only weather-related variable that was extreme at the time of an outage, especially in the interior counties (Fig. 3 d). We note that this simply shows that SMOIS was the only variable that exceeded its 90th percentile threshold at the time of an outage, while the other variables did not exceed this threshold. But there may still have been, for example, relatively strong wind, which is more likely to lead to tree falls or failure of electrical polls when soil moisture is high. The contribution of CAPE as the only extreme weather variable is small in some counties in the northwest and between 5–10% in most other regions (Fig. 3 e). Extreme coastal water levels alone also often triggered outages in coastal counties (10–15% of all weather-related outages), with lower contributions in the Big Bend region in the northwest (Fig. 3 f). We repeat the same analysis, but instead of focusing on univariate events, we consider bivariate events, showing the fraction of weather-related outages where certain combinations of weather variables were extreme at the same time. In most cases, the percentage values are relatively low, and there is little spatial variability. Precipitation and soil moisture combined were most often related to outages (Fig. 4 c), followed by strong wind and lightning strikes (especially in interior counties and on the east coast) (Fig. 4 e), and strong wind and extreme coastal water levels (especially on the west coast) (Fig. 4 h). However, there are localized hotspots where other bivariate weather events caused a relatively large number of outages. For example, precipitation and wind speed (Fig. 4 b) or lightning strikes and soil moisture (Fig. 4 f) in the Lake Okeechobee area. 3.3 Characteristics of weather-related power outage events We first analyze the durations of power outages that were generated by different univariate events and their associated spatial variation (Fig. 5 ). Power outage duration varies depending on the type of weather variable that triggered it. For example, precipitation-driven univariate events typically result in shorter outages in the north and northwest regions and longer outages in central and southeast Florida (Fig. 5 a). Wind speed and lightning strikes lead to the longest outages (median duration > 5 hours), particularly in south Florida (for wind speed; Fig. 5 b) and central Florida (for lighting strikes; Fig. 5 c), with shorter durations in the northwest Florida (with the exception of Franklin County in the case of lightning). Soil moisture, CAPE, and extreme coastal water levels exhibit comparable spatial patterns with generally less spatial variability and power outage durations increasing toward the south (Figs. 7d–f) and central Florida’s east and west coasts in case of water levels (Fig. 7f). When analyzing the durations of bivariate compound events there is generally more spatial variability (Fig. 6 ). For instance, in almost every bivariate combination, counties in the southeast and southwest experience relatively longer outages. Precipitation combined with lightning strikes, wind speed, or soil moisture led to prolonged power outages in central and south Florida except in counties surrounding Lake Okeechobee (Figs. 6 a-d). Northwest Florida (with the exceptions of the most western counties in some cases) and the Lake Okeechobee area experience relatively shorter outages when they are triggered by the bivariate events analyzed here. Bivariate events of wind speed and soil moisture led to shorter or no power outages in the north and some counties in central Florida, while longer outages occurred in the rest of central Florida and in the south (Fig. 6 d). Wind speed and lightning strikes, and lightning strikes and soil moisture, both have hotspots in south Florida, Tampa Bay, and the central east coast (Fig. 6 e-f). Bivariate events that included coastal water levels generally resulted in longer outages in the southern part of the state, together with some additional counties in the northwest (Fig. 6 g-j). Overall, more counties experienced longer duration outages when they were caused by two weather variables that were extreme at the same time compared to univariate events. So far, we have only focused on events that were triggered by certain individual weather variables or certain combinations. Next, we compare the overall median durations of single driver events, bivariate events, trivariate events, or events where more than three variables were extreme, regardless of which variables were involved. This clearly shows that power outage duration increases with the number of extreme weather variables that were extreme simultaneously when the outage started (Fig. 7). Moving from univariate events to bivariate events the duration mainly increases in the southern part of the state, especially southwest Florida (Figs. 7a-b). Those changes become more pronounced when including trivariate events (Fig. 7c) and expand into the northern and most western parts of Florida when focusing on multivariate events where more than three weather variables were extreme (as is often the case during hurricanes) (Fig. 7d) Finally, we compare the median durations of all weather-related outages with the median durations of non-weather outages (Fig. 7e-f). Weather-related events generally have longer durations than non-weather events while both types of events lead to relatively longer outages in similar parts of the state, mainly in central and south Florida. Discussion Power system operation is complex and the systems are designed to operate under normal weather conditions (Liu et al 2017 ). However, the frequency of weather-related power outages rises as more extreme weather events occur, and it is becoming a concern in many countries and within the U.S., including the state of Florida. The level of impact of weather-related power outages can be measured in different ways, including the interruption duration and total number of customers affected. Certain types of extreme weather events, for example hurricanes, can create extreme precipitation, winds, and storm surges(Ali et al 2023 ) and hence generally lead to the longest and most widespread power outages, as our results confirm (Fig. 2 ). However, moderate outage events that are shorter and affect fewer customers can occur more frequently as the result of different weather variables being extreme, in isolation, or combined. Past studies mainly focused on the most extreme events and neglected moderate events. Using county-level power outage data, this study explicitly considers moderate power outage events on top of major events to study grid reliability and to assess how different combinations of weather extremes contributed to outages and how the characteristics of these outages vary depending on the driving weather variables. We analyze power grid resilience (in terms of SAIDI) considering all power outage events but also separately for weather and non-weather-related events as well as tropical and non-tropical storm events. The derived SAIDI values vary significantly from considering all power outage events to only considering weather-related events (Figs. 2 a-b). Overall, more power outages occurred in central Florida compared to the rest of the state. In contrast, the rate of weather-related power outages is higher in Florida's Big Bend region compared to central Florida (Fig. 2 c). The small number of overall outages but a high proportion of weather-related outages indicates that outages due to equipment failure are rare and that extreme weather events are by far the most important cause of outages in the Big Bend region. Varying patterns of grid reliability also exist when comparing non-weather to weather SAIDI and non-tropical to tropical SAIDI. Weather-related SAIDI varies from 225 minutes to close to 8,000 minutes of interruption on average per year across counties (Fig. 2 d); the average across the state is 2,376 minutes for the 2015–2022 period. In contrast, the Energy Information Administration (EIA) estimated an average SAIDI value for the period 2015–2022 for the state of Florida of 607 minutes (U.S. EIA, 2025). These large spatial variations we identify from our analysis and the differences compared to aggregated state-level data highlight the importance of the spatial and temporal resolution of the outage data from which SAIDI is calculated. In past studies, power outages were primarily evaluated at the state level, with the criteria that either ≥ 50,000 customers were affected or ≥ 300 MW power supply was lost. Multiple studies suggest that many of those reported major outages were in coastal states (Mukherjee et al., 2018; Do et al., 2023 ), which are often prone to hurricanes and tropical cyclones, like Florida. Calculating SAIDI at the county level for non-tropical and tropical events allows us to identify hotspots of outages from different storm types at the sub-state level. Results show that counties in south Florida (except for Monroe, Miami-Dade, and Broward counties), along the Atlantic coast (except Brevard County), and in the Florida Panhandle are hotspots for power interruptions from tropical cyclones (Fig. 2 f). Counties in the Panhandle are also impacted by non-tropical events along with hotspots in Monroe and Miami-Dade counties in the south (Figs. 2 e). The contrasting patterns of a cluster of clusters of three counties in the south (Monroe, Miami-Dade, and Broward) have been observed between the tropical and non-tropical results are striking. As mentioned in the Results section, Monroe County often behaves differently than most other counties (not just in Fig. 2 ). This is likely because it is served by a small independent utility company, which may report outages differently than the large utilities that serve most of the state. Miami-Dade and Broward counties are highly developed areas that were not severely affected by recent hurricanes, leading to small SAIDI values when analyzing outages from tropical cyclones. Aside from those three outliers, the hotspot in south Florida for tropical cyclone SAIDI can be explained partly by Hurricanes Irma in 2017 and Ian in 2022, which made landfall in southwest Florida. EIA also reported elevated SAIDI values for 2017 and 2022 at the state level. Different weather variables contribute differently to power outages. The average contribution of univariate extreme weather events varies from 1–18% across counties, with coastal water levels and lightning strikes contributing the most and precipitation (as the only driver) contributing the least (Figs. 3 a-f). In contrast, a previous study by Do et al ( 2023 ) showed that a significant portion (~ 75%) of power outages nationwide co-occurred with extreme precipitation events. An extreme precipitation event might be a univariate event or be part of a compound event, e.g., with lightning strikes, high wind speed, or other extreme weather. Do et al ( 2023 ) did not explicitly account for that and our bivariate analysis (Fig. 4 ) indicates that different compound events that include precipitation also contributed to the observed outages, especially when paired with high wind speed or soil moisture. Overall, the contributions from bivariate events are smaller than those from univariate events (Figs. 3 and 4 ), as bivariate events are less frequent. The typical durations of power outages varied when different weather variables were involved in univariate or bivariate events (Figs. 5 and 6 ), and also when comparing univariate to multivariate events (Fig. 7). Hines et al (2009) showed that for major events (≥ 50,000 customers or ≥ 300 MW power supply lost), wind speed or precipitation-driven power outages increased across the United States between 1984 and 2006. Here, we did not account for temporal changes in power outages because the number of customers tracked by EAGLE-I has increased from 2015 to present; this would likely introduce biases in trend estimates. However, our results suggest that weather variables other than precipitation and wind speed, like lightning strikes or high coastal water levels, can also generate longer-duration outages (Fig. 5 ), in isolation or when paired with other weather variables (Fig. 6 ). Overall, our results highlight that compound weather extremes lead to longer outrages (Fig. 7). While this general result is intuitive and expected, especially for the most extreme events related to hurricanes, our analysis identifies spatial variability across the state of Florida and identifies hotspots that are relatively more affected by certain types of compound events (Figs. 7a-d). Finally, we show that non-weather-related outages, i.e., those related to equipment failure or other system disruptions, are much shorter than weather-related outages, in many cases less than half (Figs. 7e-f). Limitations of the study include that the power outage data covers varying percentages of customers tracked in different counties, varying methods being used for reporting power outages across different utilities, and details about the grid characteristics not being publicly available. This is likely one of the reasons for differences in our results between urban (metro) and rural (non-metro) areas, with generally fewer outages in rural areas despite there being similar occurrences of extreme weather events as in urban areas. It could be due to underreporting but also due to the fact that fewer people live in rural areas, which means fewer overhead power lines exist. This, in turn, means, for example, that the likelihood of falling trees affecting the grid is smaller as compared to densely populated areas with many overhead lines and generally denser grids; the same is true for lightning strikes. There is also more permeable surface in rural areas, which means less flooding under the same rainfall conditions compared to highly urbanized areas. That explains why there is often extreme weather but, in many cases, without causing outages. Overall, the grid in those areas is still poorly maintained, leading to outages that are not due to weather, and when those happen, the repair times are longer since they have less priority and/or take more time to get crews on site. Our study does not capture many of these, given that those events are mostly isolated and very few customers are impacted. At the same time, utility companies might not track customers as high as metro areas in non-metro areas. Conclusions This study introduced multiple ways to assess power grid risk and reliability in Florida by processing open-access EAGLE-I power outage data in conjunction with extreme weather data. Power outage data processing involved isolating big enough power outage events based on selected thresholds to eliminate small events that are likely generated by equipment failure or other system interruptions rather than triggered by extreme weather events. This allowed us to calculate the duration of individual events based on consecutive power outages in the 15-minute time series. We then characterized weather-related power outages based on those durations and extreme weather information. We showed that total outage counts were higher in metro areas than in non-metro areas. However, interestingly, the percentage of weather-related power outages was higher in non-metro areas than in metro areas. Weather-related power outages dominated the spatial variability in outage frequencies across Florida. We also identified localized hotspots in terms of the drivers and characteristics of power outages from different weather variables and their combinations. The framework adopted in this study identifies the most relevant drivers of weather-related power outages and how each driver or combination of more than one driver either triggered or modified observed power outages. Understanding the complex interplay of different combinations of weather variables and their associated impact on the grid is the first step to develop predictive models that account for those complex relationships and can help inform future grid development to increase resilience against extreme weather. Declarations Acknowledgments Mohammad Siddiqur Rahman and Thomas Wahl were supported by the U.S. Department of Energy's Office of Energy Efficiency and Renewable Energy (EERE) under the Solar Energy Technologies Office Award Numbers DE-EE0010418. We acknowledge DOE-CESER, the agency responsible for funding the creation of EAGLE-I data. Support for DOI 10.13139/ORNLNCCS/1975202 dataset is provided by the U.S. Department of Energy, project EAGLE-I under Contract DE-AC05-00OR22725. Project EAGLE-I used resources of the Oak Ridge Leadership Computing Facility at Oak Ridge National Laboratory, which is supported by the Office of Cybersecurity, Energy Security, and Emergency Response of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725. Financial support This research has been supported by the U.S. Department of Energy's Office of Energy Efficiency and Renewable Energy (EERE) under the Solar Energy Technologies Office Award Numbers DE-EE0010418. Code availability Final processed data and codes are available in GitHub at https://github.com/CoRE-Lab-UCF/PowerOutageFL (last access: 2 April 2025). Data availability EAGLE-I power outage data is available and downloaded from https://doi.ccs.ornl.gov/dataset/d26451af-7ea4-577c-8285-78b72cd8a2dc. Precipitation, soil moisture, and wind speed data used in this paper can be downloaded through USGS at https://www.sciencebase.gov/catalog/item/6372cd09d34ed907bf6c6ab1. The lightning strikes dataset of U.S. National Lightning Detection Network is obtained and available at https://ghrc.nsstc.nasa.gov/home/lightning/index/data_nldn. CAPE is obtained and available through ERA5 at https://cds.climate.copernicus.eu/datasets/derived-era5-single-levels-daily-statistics?tab=download. HURDAT 2 reanalysis data is downloaded available through National Oceanic and Atmospheric Administration at https://www.aoml.noaa.gov/hrd/hurdat/hurdat2-nepac.html. Water levels data is accessed and available at https://registry.opendata.aws/noaa-nos-cora. Metro and non-metro classification used in this paper is downloaded and available at https://www.ers.usda.gov/data-products/rural-urban-continuum-codes. Author contributions This study was conceived by MSR and TW. MSR developed the methodology, undertook the analysis, and wrote the first draft of the paper under the guidance of TW. MN and AE contributed by generating ideas, providing valuable insights during technical discussions, and editing the manuscript. References Abdelmalak M, Cox J, Ericson S, Hotchkiss E and Benidris M 2023 Quantitative Resilience-Based Assessment Framework Using EAGLE-I Power Outage Data IEEE Access 11 7682–97 Ali J, Wahl T, Enriquez A R, Rashid M M, Morim J, Gall M and Emrich C T 2023 The role of compound climate and weather extreme events in creating socio-economic impacts in South Florida Weather Clim Extrem 42 Belligoni S, Trader E, Li M, Rahman M S, Ali J, Enriquez A R, Nagaraj M, Aksha S K, Stevens K A, Wahl T, Emrich C T, Qu Z and Davis K O 2025 Transdisciplinary research promoting clean and resilient energy systems for socially vulnerable communities: A review Renewable and Sustainable Energy Reviews 213 Bhusal N, Abdelmalak M, Kamruzzaman M and Benidris M 2020 Power system resilience: Current practices, challenges, and future directions IEEE Access 8 18064–86 Brelsford C, Tennille S, Myers A, Chinthavali S, Tansakul V, Denman M, Coletti M, Grant J, Lee S, Allen K, Johnson E, Huihui J, Hamaker A, Newby S, Medlen K, Maguire D, Dunivan Stahl C, Moehl J, Redmon D, Sanyal J and Bhaduri B 2024 A dataset of recorded electricity outages by United States county 2014–2022 Sci Data 11 Casey J A, Fukurai M, Diana Hernández &, Satchit Balsari &, Kiang M V, Hernández D and Balsari S 2020 Power Outages and Community Health: a Narrative Review Curr Environ Health Rep 7 371–83 Online: https://doi.org/10.1007/s40572-020-00295-0 Chakalian P M, Asce S M, Kurtz L C and Hondula D M 2019 After the Lights Go Out: Household Resilience to Electrical Grid Failure Following Hurricane Irma Online: https://orcid.org Climate Central 2024 Weather-related Power Outages Rising. Accessed on September 24, 2024, from https://www.climatecentral.org/climate-matters/weather-related-power-outages-rising Do V, McBrien H, Flores N M, Northrop A J, Schlegelmilch J, Kiang M V. and Casey J A 2023 Spatiotemporal distribution of power outages with climate events and social vulnerability in the USA Nat Commun 14 DOE OE-417 Office of Electricity Delivery and Energy Reliability, Electric Disturbance Events (OE-417), https://openenergyhub.ornl.gov/explore/dataset/oe-417-annual-summaries/information/ Dugan J, Byles D and Mohagheghi S 2023 Social vulnerability to long-duration power outages International Journal of Disaster Risk Reduction 85 Entress R M and Stevens K A 2023 Public values failure associated with Hurricane Ian power outages Frontiers in Sustainable Energy Policy 2 FEMA 2020 Community Lifelines. Published July 27, 2020. Accessed January 11, 2025. https://www.fema.gov/emergency-managers/ practitioners/lifelines. Florida Division of Emergency Management 2017 “Outage reports” https://www.floridadisaster.org/info/outage_reports/ Hersbach H, Bell B, Berrisford P, Hirahara S, Horányi A, Muñoz-Sabater J, Nicolas J, Peubey C, Radu R, Schepers D, Simmons A, Soci C, Abdalla S, Abellan X, Balsamo G, Bechtold P, Biavati G, Bidlot J, Bonavita M, De Chiara G, Dahlgren P, Dee D, Diamantakis M, Dragani R, Flemming J, Forbes R, Fuentes M, Geer A, Haimberger L, Healy S, Hogan R J, Hólm E, Janisková M, Keeley S, Laloyaux P, Lopez P, Lupu C, Radnoti G, de Rosnay P, Rozum I, Vamborg F, Villaume S and Thépaut J N 2020 The ERA5 global reanalysis Quarterly Journal of the Royal Meteorological Society 146 1999–2049 Hines P, Apt J and Talukdar S 2009a Large blackouts in North America: Historical trends and policy implications Energy Policy 37 5249–59 Hines P, Apt J and Talukdar S 2009b Large blackouts in North America: Historical trends and policy implications Energy Policy 37 5249–59 IEEE Guide for Electric Power Distribution Reliability Indices" in IEEE Std 1366-2012 (Revision of IEEE Std 1366-2003) (2012) vol., no., pp.1-43, doi: 10.1109/IEEESTD.2012.6209381. Landsea C W and Franklin J L 2013 Atlantic hurricane database uncertainty and presentation of a new database format Mon Weather Rev 141 3576–92 Liu Y, Zhong J and Kong H 2017 Risk Assessment of Power Systems under Extreme Weather Conditions-A Review M. Sullivan, M. Perry, J. Schellenberg, J. Burwen, S. Holmberg and S. Woehleke (2012) Pacific gas & electric company’s 2012 value of service study. Mukherjee S, Nateghi R and Hastak M 2018 A multi-hazard approach to assess severe weather-induced major power outage risks in the U.S. Reliab Eng Syst Saf 175 283–305 NOAA's Coastal Ocean Reanalysis (CORA) 2024 Dataset was accessed on September 16, 2024, from https://registry.opendata.aws/noaa-nos-cora. Rasmussen R M, Chen F, Liu C H, Ikeda K, Prein A, Kim J, Schneider T, Dai A, Gochis D, Dugger A, Zhang Y, Jaye A, Dudhia J, He C, Harrold M, Xue L, Chen S, Newman A, Dougherty E, Abolafia-Rosenzweig R, Lybarger N D, Viger R, Lesmes D, Skalak K, Brakebill J, Cline D, Dunne K, Rasmussen K and Miguez-Macho G 2023 CONUS404 The NCAR–USGS 4-km Long-Term Regional Hydroclimate Reanalysis over the CONUS Bull Am Meteorol Soc 104 E1382–408 Spurlock T, Sewell K, Sugg M M, Runkle J D, Mercado R, Tyson J S and Russell J 2023 A spatial analysis of power-dependent medical equipment and extreme weather risk in the southeastern United States International Journal of Disaster Risk Reduction 95 Sullivan, Michael, Collins, Myles T., Schellenberg, Josh, & Larsen, Peter H. (2018) Estimating Power System Interruption Costs: A Guidebook for Electric Utilities. https://doi.org/10.2172/1462980 U.S. Department of Agriculture, Economic Research Service. (January 2024). Rural-Urban Continuum Codes U.S. Energy Information Administration 2015 Form EIA-861, Annual Electric Power Industry Report. Available at https://www.eia.gov/electricity/annual/html/epa_11_04.html; Accessed on January 13th, 2025 Xu J, Qiang Y, Cai H and Zou L 2023 Power outage and environmental justice in Winter Storm Uri: an analytical workflow based on nighttime light remote sensing Int J Digit Earth 16 2259–78 Zscheischler J, Martius O, Westra S, Bevacqua E, Raymond C, Horton R M, van den Hurk B, AghaKouchak A, Jézéquel A, Mahecha M D, Maraun D, Ramos A M, Ridder N N, Thiery W and Vignotto E 2020 A typology of compound weather and climate events Nat Rev Earth Environ 1 333–47 Zscheischler J, Westra S, Van Den Hurk B J J M, Seneviratne S I, Ward P J, Pitman A, Aghakouchak A, Bresch D N, Leonard M, Wahl T and Zhang X 2018 Future climate risk from compound events Nat Clim Chang 8 469–77 Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6404193","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":440182329,"identity":"c2817202-443d-49f7-aadf-7b3b8cee78b4","order_by":0,"name":"Mohammad Siddiqur Rahman","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA4ElEQVRIiWNgGAWjYDACZgYDiYQDDAlAJuMDIMHDR4oWZgOQFjYi7DGQYIBoYZMAcQlqkW9n3njjwRmGPPMZuc8qv+bYybAxMD98dAOfFYfZii0SbjAUy9xIN7stuy0Z6DA2Y+McfFqYecwkEj4wJM6QSGO7LbmNGaiFh00anxb5ZiQtxZLb6glrYTgM0nIDooXx47bDhLVA/HJGoliC5xmzNOO24zxszAT8It9/eOPNH8ds8iTY0xg//txWbc/P3vzwMV6HQQA4RhiYecAkYeUIwPiDFNWjYBSMglEwYgAAE5tAcMzRpFcAAAAASUVORK5CYII=","orcid":"https://orcid.org/0000-0002-0638-9083","institution":"University of Central Florida","correspondingAuthor":true,"prefix":"","firstName":"Mohammad","middleName":"Siddiqur","lastName":"Rahman","suffix":""},{"id":440182330,"identity":"6958d06e-4562-4a8e-ad08-50d6f72b0cd5","order_by":1,"name":"Thomas Wahl","email":"","orcid":"","institution":"University of Central Florida","correspondingAuthor":false,"prefix":"","firstName":"Thomas","middleName":"","lastName":"Wahl","suffix":""},{"id":440182331,"identity":"f3df784b-2501-4119-86d8-89ed6e3e7efe","order_by":2,"name":"Meghana Nagaraj","email":"","orcid":"","institution":"University of Central Florida","correspondingAuthor":false,"prefix":"","firstName":"Meghana","middleName":"","lastName":"Nagaraj","suffix":""},{"id":440182332,"identity":"598a160a-dfc1-44c8-a48c-c5c57735f98c","order_by":3,"name":"Alejandra R. Enríquez","email":"","orcid":"","institution":"University of South Florida","correspondingAuthor":false,"prefix":"","firstName":"Alejandra","middleName":"R.","lastName":"Enríquez","suffix":""}],"badges":[],"createdAt":"2025-04-08 14:14:56","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-6404193/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6404193/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":80197552,"identity":"0be7f209-8e26-4da4-b8a9-285d596969e4","added_by":"auto","created_at":"2025-04-09 06:07:19","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":639929,"visible":true,"origin":"","legend":"\u003cp\u003e67 counties across Florida are classified into metro and non-metro counties based on the U.S. Department of Agriculture Economic Research Services.\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6404193/v1/906455a02ed7a2c7ee0c82fe.jpeg"},{"id":80197551,"identity":"d14de660-caa8-4957-9b9d-4a24bf91197c","added_by":"auto","created_at":"2025-04-09 06:07:19","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":214792,"visible":true,"origin":"","legend":"\u003cp\u003e(a) Total number of power outages (\u003cstrong\u003e\u0026gt;\u003c/strong\u003e85\u003csup\u003eth\u003c/sup\u003e percentile, \u0026gt;1 hour duration, independent events \u0026gt;1 hour apart from each other) between 2015 and 2022 across 67 Florida counties; (b) percentage of weather-related outages; (c) SAIDI from non-weather events; (d) SAIDI from weather events; (e) SAIDI from non-tropical weather events; (f) SAIDI from tropical weather events.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6404193/v1/b8e83a23c27f4e4f8d2249fa.png"},{"id":80198884,"identity":"c9af1f54-2595-4d25-99e3-bb0e87353173","added_by":"auto","created_at":"2025-04-09 06:23:19","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":93068,"visible":true,"origin":"","legend":"\u003cp\u003eThe fraction of weather-related outages that were related to only one weather variable that exceeded the 90\u003csup\u003eth\u003c/sup\u003e percentile threshold at the time when the outage started. Contribution from (a) Prcp, (b) WS, (c) LIS, (d) SMOIS, (e) CAPE, and (f) WL univariate events to weather-related events. The light gray color in (f) represents inland counties unaffected by extreme coastal water levels.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6404193/v1/734f24b120983e2cbc9167a1.png"},{"id":80197554,"identity":"9ff2df6e-46f2-4705-8560-1ff2dbf48880","added_by":"auto","created_at":"2025-04-09 06:07:19","extension":"jpeg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":572802,"visible":true,"origin":"","legend":"\u003cp\u003eThe fraction of weather-related outages that were related to certain bivariate weather events, where two specific variables exceeded their 90\u003csup\u003eth\u003c/sup\u003e percentile thresholds at the time of the outage start (while all other variables did not).\u0026nbsp; (a) Prcp and LIS, (b) Prcp and WS, (c) Prcp and SMOIS, (d) WS and SMOIS, (e) WS and LIS, (f) LIS and SMOIS, (g) Prcp and WL, (h) WS and WL, (i) SMOIS and WL, (j) LIS and WL. The light gray colors in (g), (h), (i), and (j) represent inland counties unaffected by extreme coastal water levels.\u003c/p\u003e","description":"","filename":"floatimage4.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-6404193/v1/2d04234f726a06615b4a4317.jpeg"},{"id":80198350,"identity":"fd42b393-b203-4b59-b42f-3510beca690d","added_by":"auto","created_at":"2025-04-09 06:15:19","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":91371,"visible":true,"origin":"","legend":"\u003cp\u003eMedian durations of power outages caused by different individual weather variables: (a) Prcp, (b) WS, (c) LIS, (d) SMOIS, (e) CAPE, (f) and WL. The light gray color in (f) represents inland counties not impacted by extreme coastal water levels.\u003c/p\u003e","description":"","filename":"floatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-6404193/v1/cb2432dbe8e6dbea747d6ca9.png"},{"id":80197576,"identity":"1af87d96-fc80-4efe-ba76-afb828b765ec","added_by":"auto","created_at":"2025-04-09 06:07:20","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":101162,"visible":true,"origin":"","legend":"\u003cp\u003eMedian durations (in hours) of different bivariate events: (a) Prcp and LIS, (b) Prcp and WS, (c) Prcp and SMOIS, (d) WS and SMOIS, (e) WS and LIS, (f) LIS and SMOIS, (g) Prcp and WL, (h) WS and WL, (i) SMOIS and WL, (j) and LIS and WL. The dark gray represents counties without any events triggered by the specific bivariate combinations, while light gray represents non-coastal counties not impacted by extreme coastal water levels.\u003c/p\u003e","description":"","filename":"floatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-6404193/v1/149d5717c34a72dc766843df.png"},{"id":80197559,"identity":"006e0993-a2bb-4e9c-abeb-51613652ac9c","added_by":"auto","created_at":"2025-04-09 06:07:19","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":96971,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 9. Median durations of power outage events when (a) only one weather variable was extreme (univariate), (b) when any two variables were extreme (bivariate), (c) when any three variables were extreme (trivariate), (d) any four or more variables were extreme (multivariate). (e) Median duration of weather-related power outages and (f) median duration of non-weather outages.\u003c/p\u003e","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-6404193/v1/f5d1a0241711ed5fac40e558.png"},{"id":80199659,"identity":"7ec9bab6-2ee1-4796-a56b-7a5ce594baf9","added_by":"auto","created_at":"2025-04-09 06:31:24","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2316484,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6404193/v1/9be6888f-2d69-433e-b425-1f6d194e4096.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003eCharacteristics of Power Outages from Compound Weather Extremes in Florida\u003c/p\u003e","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe power grid is essential for the functioning of communities, providing critical services as identified by the Federal Emergency Management Agency (FEMA 2020). Despite its importance, the power grid across the United States faces increasing disruptions. Over the last decade (2014\u0026ndash;2023), significant power outages \u0026ndash; those affecting\u0026thinsp;\u0026ge;\u0026thinsp;50,000 customers \u0026ndash; doubled compared to the early 2000s (2000\u0026ndash;2009) (Climate Central 2024). Extreme weather events caused 50% or more of these outages (Hines et al \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2009a\u003c/span\u003e, Do et al \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), underscoring the grid's vulnerability to environmental conditions beyond typical weather patterns (Liu et al \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2017\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eOn the Gulf and Atlantic coast of the United States (U.S.), the most severe power outages are often due to hurricane-induced factors such as extreme wind, rainfall, or storm surge (Ali et al \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). One example was Hurricane Irma in 2017, which left nearly two-thirds of Florida's customers without power due to high winds and flooding (Florida Division of Emergency Management 2017; Chakalian et al \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Superstorm Sandy in 2012, impacted more than 8.5\u0026nbsp;million people across several states in the northeastern U.S.(Mukherjee \u003cem\u003eet al\u003c/em\u003e 2018). Hurricane Ian in 2022, resulted in more than 2.7\u0026nbsp;million people without power in Florida (Entress and Stevens \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Winter storms also have the potential to lead to widespread outages, as exemplified by winter storm Uri in 2021, which impacted more than 4\u0026nbsp;million customers in Texas (Xu \u003cem\u003eet al\u003c/em\u003e 2023). In many of those cases, power restoration took days to weeks, or even months, for many customers.\u003c/p\u003e \u003cp\u003eLarge-scale power outages are defined by the Department of Energy (DOE) as events that affect\u0026thinsp;\u0026ge;\u0026thinsp;50,000 customers or cause load loss of \u0026ge;\u0026thinsp;300 MW. Information on these low-probability but high-impact events is provided through the Electric Emergency Incident and Disturbance Report (DOE OE-417). Small or moderate outage events occur more frequently and affect fewer customers in specific locations. Over time, these high-probability/low-impact events can also lead to substantial disruptions. They are included in high-resolution datasets such as the Environment for the Analysis of Geo-Located Energy Information (EAGLE-I) and PowerOutage.us. DOE OE-417, which only includes large-scale events, reported 605 events between November 2014 and March 2021 (Abdelmalak \u003cem\u003eet al\u003c/em\u003e 2023) in the U.S. compared to 17,484 outages (which lasted more than 8 hours) between 2018 and 2020 alone based on high-resolution PowerOutage.us data (Do et al \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). This indicates a disparity in the number of power outages based on the selection criteria, as localized outages are unlikely to affect 50,000 customers or create 300 MW of load loss, while still significantly impacting certain communities (often repeatedly).\u003c/p\u003e \u003cp\u003ePrevious studies suggest that precipitation, wind, lightning strikes, winter weather, and tropical storms are the most dominant weather and climate events responsible for power outages in the U.S (Hines et al \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2009a\u003c/span\u003e, Do et al \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) with vegetation cover, and grid infrastructure (Belligoni \u003cem\u003eet al\u003c/em\u003e 2025). Hines et al (2009) analyzed DOE OE-417 data from 1984\u0026ndash;2006, reporting that wind or precipitation caused 31% of outages, while hurricanes or tropical storms accounted for 10%, and lightning strikes caused about 9%. Similar to that study, most other studies in literature analyzed weather influence on power outages from a univariate perspective. Do et al (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), for example, showed that when outages longer than 8 hours co-occurred with at least one extreme weather variable, 75% co-occurred with extreme precipitation, and over a third co-occurred with multiple weather extremes. However, this analysis ignored shorter duration outage events (which happen more frequently and still disrupt people\u0026rsquo;s lives) and did not explicitly account for the compounding effects of multiple hazards, or hazard drivers, where not all need to be extreme at the same time to create an extreme impact (Zscheischler et al \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2018\u003c/span\u003e, 2020, Ali et al \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Understanding the combined influence of multiple weather variables on outage occurrences is crucial, particularly in regions like Florida, where various storm types create complex weather patterns with the potential for outages of different severity.\u003c/p\u003e \u003cp\u003eHere, we overcome some of those knowledge gaps by addressing the following three objectives using Florida as a case study region: (i) use outage data with high temporal resolution paired with data of relevant weather variables to identify counties across the Florida peninsular that are relatively more vulnerable to power outages, in particular outages related to weather events; (ii) assess the relative importance of different weather variables in generating outage events; and (iii) quantify the role of compound events in causing outages and explore the outage event characteristics when they are caused by different types of weather events. Overall, our results lead to a better understanding of the spatial variability and drivers of weather-related outage events, thereby paving the way to develop intervention strategies that best serve the most vulnerable communities.\u003c/p\u003e"},{"header":"Study area and data","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Study area\u003c/h2\u003e \u003cp\u003eWe chose Florida as a case study region, because power outages occur frequently across the state due to various types of extreme weather events leading to strong wind, precipitation, lightning strikes, storm surge, and flooding, thereby causing widespread disruptions. Previous work has shown that coastal states are often at higher risk of experiencing outages (Mukherjee \u003cem\u003eet al\u003c/em\u003e 2018, Casey \u003cem\u003eet al\u003c/em\u003e 2020, Do et al \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), and 35 of Florida's 67 counties have coastlines. At the same time, rural areas (non-metro) often face longer power restoration times compared to urban (metro) areas (Spurlock \u003cem\u003eet al\u003c/em\u003e 2023); 45 counties in Florida are classified as metro and 22 as non-metro (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Power outage data\u003c/h2\u003e \u003cp\u003eThis study uses power outage data retrieved from EAGLE-I maintained by Oak Ridge National Laboratory for the Department of Energy. EAGLE-I records encompass over 8 years of outage data from 2015 to 2022 at the county level, captured at 15-minute intervals. It is collected by different web parsers and individual scripts that track utilities' publicly available outage information to estimate the number of customers without power (Abdelmalak \u003cem\u003eet al\u003c/em\u003e 2023)through the establishment of a collection pathway for each utility company (Brelsford \u003cem\u003eet al\u003c/em\u003e 2024). However, data quality is contingent upon consistent reporting practices by the utility companies. Variations in reporting methods and aggregating differing geographical boundaries of counties and utility companies can introduce spatial inconsistencies into the dataset. Over the years, coverage of this data has improved, increasing from 86 percent in 2018 to 92 percent in 2022, which might also introduce some temporal inconsistencies in trend analyses.\u003c/p\u003e \u003cp\u003ePower outages are often considered to fall into one of three categories: (i) Minor power outages, classified as minor system interruptions related to technical causes like short circuits, equipment failure, or any malfunction in the grid. Very few customers are impacted; the power outage duration is relatively short, and less likely to be caused by extreme weather (Abdelmalak \u003cem\u003eet al\u003c/em\u003e 2023). (ii) Moderate power outages are more likely to be triggered by extreme weather events but can also be due to accidents and other causes. (iii) Major power outages have the longest durations (Bhusal \u003cem\u003eet al\u003c/em\u003e 2020) and impact many customers. Those events are often caused by extreme weather events, including hurricanes, wildfires, or winter storms (Hines et al \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2009b\u003c/span\u003e, Abdelmalak \u003cem\u003eet al\u003c/em\u003e 2023, Dugan et al \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).Different approaches have been proposed to do the outage event classification, usually based on the number of customers impacted or outage duration (Sullivan et al \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2012\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2018\u003c/span\u003e, Abdelmalak et al 2023). For instance, DOE defines major power outages as events that exceed 300 MW of loss and/or affect\u0026thinsp;\u0026ge;\u0026thinsp;50,000 customers.\u003c/p\u003e \u003cp\u003eHere, minor outages are outside the scope of the analysis as they are often not weather-related, and even if they are, the societal impacts are minimal. Our analysis is based on EAGLE-I data, which reports the number of customers without power in each county at 15-minute intervals. Starting with the raw data, we perform five critical pre-processing steps. The first step is setting a threshold to remove reported power outages where the number of customers did not exceed the 85th percentile of all the values for a specific county across the 67 counties in FL, which leads to values between 12 and 423 customers being impacted. Events below that threshold are considered minor and removed from further analysis. The second step is to calculate the power outage duration, defined here as the time over which the 85th percentile threshold from step 1 is being exceeded. The third step is to merge events that are separated by less than one hour (i.e., independent events have to be at least one hour apart). The fourth step is to remove events that are shorter than one hour. The fifth and final step is to assign the maximum number of customers without power during an event and the total event duration to the first day of the outage. This is because extreme weather creates the outage, which occurs at the beginning when the outage is triggered. Still, the outage itself may last much longer than the extreme weather (e.g., power outages of several days or weeks after the passage of a hurricane).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Weather data\u003c/h2\u003e \u003cp\u003eSince we are interested in weather-related outages, we include a range of different weather variables in our analysis. Weather information has been retrieved from different data sources to obtain the highest temporal and spatial resolution possible. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e shows the weather variables included in the present study, their spatial and temporal resolution, and the source.\u003c/p\u003e \u003cp\u003eCONUS404 data has a four-kilometer spatial resolution, while ERA5 data has ~\u0026thinsp;31-kilometer spatial resolution, and both have hourly temporal resolution. For precipitation, wind speed, and CAPE, we use the daily maximum values across a county (because the EAGLE-I data is only available at the county-level), whereas for soil moisture, we use the daily mean, and for lightning strikes, we use the daily accumulated number of strikes across a county. The National Hurricane Center (NHC) reports hurricane paths through HURDAT2 at 6-hour intervals, representing the hurricane's center as a point. NOAA's Coastal Ocean Reanalysis (CORA) reports coastal water levels in a mesh grid. The analysis involves creating an equal-spaced 200-point along Florida\u0026rsquo;s coastline and extracting the daily maximum water level value for each coastal county from multiple points.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eWeather variables considered in the analysis with spatial and temporal resolution and source.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable​\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSpatial res.​\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eTemporal\u0026nbsp;res.​\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSource​\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePrecipitation\u0026nbsp;(Prcp)​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4*4 km\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHourly​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\" morerows=\"2\" rowspan=\"3\"\u003e \u003cp\u003eCONUS404​ (Rasmussen \u003cem\u003eet al\u003c/em\u003e 2023)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWind Speed\u0026nbsp;(WS)​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4*4 km\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHourly​\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSoil Moisture\u0026nbsp;(SMOIS)​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4*4 km\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHourly​\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLightning Strike (LIS)​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePoint\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eCount​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eU.S. National Lightning Detection Network (NLDN)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eConvective Available Potential Energy​ (CAPE)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e28*28 km ​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHourly​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eERA5 ECMWF (Hersbach \u003cem\u003eet al\u003c/em\u003e 2020)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHurricane Path​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePoint\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e6-hourly​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eHURDAT2​ (Landsea and Franklin \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2013\u003c/span\u003e)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWater Level (WL)​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e500 meter​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003eHourly​\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e​NOAA's Coastal Ocean Reanalysis (CORA 2024)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Methods","content":"\u003cp\u003eAs a first step, we analyze the grid reliability across all counties in Florida, overall and specifically for weather-related events. Grid reliability can be assessed in different ways. One popular metric is the System Average Interruption Index (SAIDI), which many utility companies use to report grid reliability. SAIDI represents the total interruption duration for average customers in a given geographical location for predefined time periods (IEEE 2012) and is calculated as follows:\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:SAIDI=\\frac{\\sum\\:{D}_{i}{N}_{m}}{{N}_{T}}=\\frac{\\sum\\:Customer\\:Minutes\\:of\\:Interruption}{\\text{T}\\text{o}\\text{t}\\text{a}\\text{l}\\:\\text{N}\\text{u}\\text{m}\\text{b}\\text{e}\\text{r}\\:\\text{o}\\text{f}\\:\\text{C}\\text{u}\\text{s}\\text{t}\\text{o}\\text{m}\\text{e}\\text{r}\\text{s}\\:\\text{S}\\text{e}\\text{r}\\text{v}\\text{e}\\text{d}}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eD\u003csub\u003ei\u003c/sub\u003e = Duration for each 1-hour or longer interruption event.\u003c/p\u003e \u003cp\u003eN\u003csub\u003em\u003c/sub\u003e = Maximum number of interrupted customers for each 1-hour or longer duration event.\u003c/p\u003e \u003cp\u003eN\u003csub\u003eT\u003c/sub\u003e = Total number of customers served in the county.\u003c/p\u003e \u003cp\u003eWe calculate SAIDI on an annual basis for every county in Florida, separately for weather events and non-weather events, and for tropical events and non-tropical events. We define weather-related outages as events when at least one weather variable exceeded its respective 90th percentile value on the day when the outage started; hence, we assume that this weather event played a role in generating the outage. Tropical events are defined as events when a storm center, as reported in the HURDAT2 database, was within a 500-km buffer of the county's boundary where an outage occurred simultaneously. All other events are considered non-tropical events.\u003c/p\u003e \u003cp\u003eNext, we analyze how often individual weather variables triggered outages, either in isolation or in combination. The latter includes compound events, which we define as events where multiple weather variables exceeded their 90th percentile thresholds simultaneously while an outage occurred. First, we identify events where only one weather variable was extreme at the time of an outage started. For example, we find the percentage of all weather-related outage events where only precipitation was extreme or only wind speed was extreme (i.e., above their 90th percentiles) and hence likely the main driver that triggered the outage. Then, we repeat the analysis but focus on bivariate compound events, where two variables exceeded the 90th percentile threshold simultaneously when an outage started. For example, we derive the percentage of weather-related outages where precipitation and wind speed were extreme while other variables did not exceed the threshold. This analysis provides insights into the relative importance of different individual weather variables or bivariate combinations of them in triggering outage events. We exclude CAPE from this part of the analysis to limit the number of possible variable combinations, and because it is an indicator of atmospheric instability, and hence a proxy for the likelihood of a storm to occur not a measure of a specific hydrometeorological variable.\u003c/p\u003e \u003cp\u003eFinally, we assess the characteristics of outage events in terms of their typical durations when they are caused by certain weather variables or combinations thereof. For example, are events that are caused by extreme precipitation longer in certain counties than others (e.g., based on grid characteristics or response capabilities), and/or do they have different durations than events that are caused by extreme wind, and/or do compound events display different characteristics than non-compound events?\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Grid reliability\u003c/h2\u003e \u003cp\u003eThe total number of power outage events between 2015 and 2022 in Florida exhibits spatial variability across the state, with the lowest number of 112 outages in Calhoun County and the largest number of 1,544 outages in Broward County (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea). More outages occurred in Central Florida (both east and west coast), with a noticeable hotspot in the southeast, while relatively fewer outages occurred in the Panhandle area and counties surrounding Lake Okeechobee. Overall, the spatial pattern of the number of outage events resembles well the metro and non-metro classification (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), with metro counties often experiencing more power outages compared to non-metro counties.\u003c/p\u003e \u003cp\u003eThe percentage of weather-related events varies from 39% in Polk County to 68% in Madison County (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003eb). The Big Bend area in the northwest of Florida is a hotspot of weather-related power outages, where in many counties, up to 60% of all outages were weather-related. Central and northwest Florida have lower percentages of weather-related outages, typically between 40\u0026ndash;50%. A contrasting pattern is observed between total outage counts and weather outage percentages. Notably, the Big Bend area experienced fewer outages overall but the highest percentage of weather-related outages.\u003c/p\u003e \u003cp\u003eComparing non-weather (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec) and weather (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed) SAIDI shows distinctly different spatial patterns and overall values. Non-weather SAIDI is highest in the Panhandle area in the northwest, a cluster of counties in the north and southwest, and in Okeechobee County, with values ranging from 150 to 200 minutes of interruption for the average customer per year (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec); note that SAIDI is first calculated for every year from 2015 to 2022 and the average is shown. While weather-related SAIDI is also high in the northwest, it is similarly high across South Florida (except Monroe, Miami-Dade, and Broward counties in the south) and most coastal counties along the Atlantic coast (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed). The SAIDI values in those regions are more than an order of magnitude higher than the non-weather values, in many cases reaching more than 2,000 minutes of interruption per customer and year. Relatively lower weather-related SAIDI values are found in West Central Florida and most northwestern counties, with values ranging from 500-2,000 minutes per customer and year. Overall, non-weather SAIDI shows less spatial variability than weather SAIDI.\u003c/p\u003e \u003cp\u003eNon-tropical and tropical SAIDI have similar spatial patterns as the non-weather and weather SAIDI, respectively. The non-tropical SAIDI values are high in the same regions where non-weather SAIDI is high, with values from 100\u0026ndash;500 minutes per year (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ee). Tropical SAIDI is high where weather-related SAIDI is high, again with similar overall values (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ef). One notable exception is a cluster of Monroe, Miami-Dade, and Broward counties in the south. Monroe County, in particular, often exhibits different behavior than most other counties and appears as an outlier; hence, it is not highlighted specifically in the remainder of the results description even though it sometimes shows up as a hotspot (see Discussion for more information).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Role of different weather variables in outage generation\u003c/h2\u003e \u003cp\u003eIn the following, we show results from analyzing how different weather variables (or combinations thereof) contributed to outages across Florida. First, we focus on univariate events where only one specific weather variable exceeded the 90th percentile threshold (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Extreme precipitation shows a relatively homogenous pattern throughout Florida; ~5% of all weather-related outages occurred when only precipitation was extreme and no other weather variable (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea). High winds alone relatively often caused power outages in some Panhandle counties and parts of Central and Southeast Florida (~\u0026thinsp;10% of all weather-related outages were only related to high winds) (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003eb). Lightning strikes were often the main driver of outages, with values reaching more than 15% in hotspot counties that are scattered across Florida but mostly in the interior of the state (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ec). Interestingly, soil moisture (SMOIS) was also sometimes the only weather-related variable that was extreme at the time of an outage, especially in the interior counties (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ed). We note that this simply shows that SMOIS was the only variable that exceeded its 90th percentile threshold at the time of an outage, while the other variables did not exceed this threshold. But there may still have been, for example, relatively strong wind, which is more likely to lead to tree falls or failure of electrical polls when soil moisture is high. The contribution of CAPE as the only extreme weather variable is small in some counties in the northwest and between 5\u0026ndash;10% in most other regions (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ee). Extreme coastal water levels alone also often triggered outages in coastal counties (10\u0026ndash;15% of all weather-related outages), with lower contributions in the Big Bend region in the northwest (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ef).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWe repeat the same analysis, but instead of focusing on univariate events, we consider bivariate events, showing the fraction of weather-related outages where certain combinations of weather variables were extreme at the same time. In most cases, the percentage values are relatively low, and there is little spatial variability. Precipitation and soil moisture combined were most often related to outages (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ec), followed by strong wind and lightning strikes (especially in interior counties and on the east coast) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ee), and strong wind and extreme coastal water levels (especially on the west coast) (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eh). However, there are localized hotspots where other bivariate weather events caused a relatively large number of outages. For example, precipitation and wind speed (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003eb) or lightning strikes and soil moisture (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003ef) in the Lake Okeechobee area.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Characteristics of weather-related power outage events\u003c/h2\u003e \u003cp\u003eWe first analyze the durations of power outages that were generated by different univariate events and their associated spatial variation (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e). Power outage duration varies depending on the type of weather variable that triggered it. For example, precipitation-driven univariate events typically result in shorter outages in the north and northwest regions and longer outages in central and southeast Florida (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ea). Wind speed and lightning strikes lead to the longest outages (median duration\u0026thinsp;\u0026gt;\u0026thinsp;5 hours), particularly in south Florida (for wind speed; Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003eb) and central Florida (for lighting strikes; Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003ec), with shorter durations in the northwest Florida (with the exception of Franklin County in the case of lightning). Soil moisture, CAPE, and extreme coastal water levels exhibit comparable spatial patterns with generally less spatial variability and power outage durations increasing toward the south (Figs.\u0026nbsp;7d\u0026ndash;f) and central Florida\u0026rsquo;s east and west coasts in case of water levels (Fig.\u0026nbsp;7f).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eWhen analyzing the durations of bivariate compound events there is generally more spatial variability (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). For instance, in almost every bivariate combination, counties in the southeast and southwest experience relatively longer outages. Precipitation combined with lightning strikes, wind speed, or soil moisture led to prolonged power outages in central and south Florida except in counties surrounding Lake Okeechobee (Figs.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ea-d). Northwest Florida (with the exceptions of the most western counties in some cases) and the Lake Okeechobee area experience relatively shorter outages when they are triggered by the bivariate events analyzed here. Bivariate events of wind speed and soil moisture led to shorter or no power outages in the north and some counties in central Florida, while longer outages occurred in the rest of central Florida and in the south (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ed). Wind speed and lightning strikes, and lightning strikes and soil moisture, both have hotspots in south Florida, Tampa Bay, and the central east coast (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003ee-f). Bivariate events that included coastal water levels generally resulted in longer outages in the southern part of the state, together with some additional counties in the northwest (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003eg-j). Overall, more counties experienced longer duration outages when they were caused by two weather variables that were extreme at the same time compared to univariate events.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eSo far, we have only focused on events that were triggered by certain individual weather variables or certain combinations. Next, we compare the overall median durations of single driver events, bivariate events, trivariate events, or events where more than three variables were extreme, regardless of which variables were involved. This clearly shows that power outage duration increases with the number of extreme weather variables that were extreme simultaneously when the outage started (Fig.\u0026nbsp;7). Moving from univariate events to bivariate events the duration mainly increases in the southern part of the state, especially southwest Florida (Figs.\u0026nbsp;7a-b). Those changes become more pronounced when including trivariate events (Fig.\u0026nbsp;7c) and expand into the northern and most western parts of Florida when focusing on multivariate events where more than three weather variables were extreme (as is often the case during hurricanes) (Fig.\u0026nbsp;7d)\u003c/p\u003e \u003cp\u003eFinally, we compare the median durations of all weather-related outages with the median durations of non-weather outages (Fig.\u0026nbsp;7e-f). Weather-related events generally have longer durations than non-weather events while both types of events lead to relatively longer outages in similar parts of the state, mainly in central and south Florida.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"Discussion","content":"\u003cp\u003ePower system operation is complex and the systems are designed to operate under normal weather conditions (Liu et al \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). However, the frequency of weather-related power outages rises as more extreme weather events occur, and it is becoming a concern in many countries and within the U.S., including the state of Florida. The level of impact of weather-related power outages can be measured in different ways, including the interruption duration and total number of customers affected. Certain types of extreme weather events, for example hurricanes, can create extreme precipitation, winds, and storm surges(Ali et al \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) and hence generally lead to the longest and most widespread power outages, as our results confirm (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). However, moderate outage events that are shorter and affect fewer customers can occur more frequently as the result of different weather variables being extreme, in isolation, or combined. Past studies mainly focused on the most extreme events and neglected moderate events. Using county-level power outage data, this study explicitly considers moderate power outage events on top of major events to study grid reliability and to assess how different combinations of weather extremes contributed to outages and how the characteristics of these outages vary depending on the driving weather variables.\u003c/p\u003e \u003cp\u003eWe analyze power grid resilience (in terms of SAIDI) considering all power outage events but also separately for weather and non-weather-related events as well as tropical and non-tropical storm events. The derived SAIDI values vary significantly from considering all power outage events to only considering weather-related events (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ea-b). Overall, more power outages occurred in central Florida compared to the rest of the state. In contrast, the rate of weather-related power outages is higher in Florida's Big Bend region compared to central Florida (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ec). The small number of overall outages but a high proportion of weather-related outages indicates that outages due to equipment failure are rare and that extreme weather events are by far the most important cause of outages in the Big Bend region. Varying patterns of grid reliability also exist when comparing non-weather to weather SAIDI and non-tropical to tropical SAIDI. Weather-related SAIDI varies from 225 minutes to close to 8,000 minutes of interruption on average per year across counties (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ed); the average across the state is 2,376 minutes for the 2015\u0026ndash;2022 period. In contrast, the Energy Information Administration (EIA) estimated an average SAIDI value for the period 2015\u0026ndash;2022 for the state of Florida of 607 minutes (U.S. EIA, 2025). These large spatial variations we identify from our analysis and the differences compared to aggregated state-level data highlight the importance of the spatial and temporal resolution of the outage data from which SAIDI is calculated.\u003c/p\u003e \u003cp\u003eIn past studies, power outages were primarily evaluated at the state level, with the criteria that either \u0026ge;\u0026thinsp;50,000 customers were affected or \u0026ge;\u0026thinsp;300 MW power supply was lost. Multiple studies suggest that many of those reported major outages were in coastal states (Mukherjee et al., 2018; Do et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e), which are often prone to hurricanes and tropical cyclones, like Florida. Calculating SAIDI at the county level for non-tropical and tropical events allows us to identify hotspots of outages from different storm types at the sub-state level. Results show that counties in south Florida (except for Monroe, Miami-Dade, and Broward counties), along the Atlantic coast (except Brevard County), and in the Florida Panhandle are hotspots for power interruptions from tropical cyclones (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ef). Counties in the Panhandle are also impacted by non-tropical events along with hotspots in Monroe and Miami-Dade counties in the south (Figs.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003ee). The contrasting patterns of a cluster of clusters of three counties in the south (Monroe, Miami-Dade, and Broward) have been observed between the tropical and non-tropical results are striking. As mentioned in the Results section, Monroe County often behaves differently than most other counties (not just in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). This is likely because it is served by a small independent utility company, which may report outages differently than the large utilities that serve most of the state. Miami-Dade and Broward counties are highly developed areas that were not severely affected by recent hurricanes, leading to small SAIDI values when analyzing outages from tropical cyclones. Aside from those three outliers, the hotspot in south Florida for tropical cyclone SAIDI can be explained partly by Hurricanes Irma in 2017 and Ian in 2022, which made landfall in southwest Florida. EIA also reported elevated SAIDI values for 2017 and 2022 at the state level.\u003c/p\u003e \u003cp\u003eDifferent weather variables contribute differently to power outages. The average contribution of univariate extreme weather events varies from 1\u0026ndash;18% across counties, with coastal water levels and lightning strikes contributing the most and precipitation (as the only driver) contributing the least (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003ea-f). In contrast, a previous study by Do et al (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) showed that a significant portion (~\u0026thinsp;75%) of power outages nationwide co-occurred with extreme precipitation events. An extreme precipitation event might be a univariate event or be part of a compound event, e.g., with lightning strikes, high wind speed, or other extreme weather. Do et al (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e) did not explicitly account for that and our bivariate analysis (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e) indicates that different compound events that include precipitation also contributed to the observed outages, especially when paired with high wind speed or soil moisture. Overall, the contributions from bivariate events are smaller than those from univariate events (Figs.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e and \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e), as bivariate events are less frequent.\u003c/p\u003e \u003cp\u003eThe typical durations of power outages varied when different weather variables were involved in univariate or bivariate events (Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e and \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e), and also when comparing univariate to multivariate events (Fig.\u0026nbsp;7). Hines et al (2009) showed that for major events (\u0026ge;\u0026thinsp;50,000 customers or \u0026ge;\u0026thinsp;300 MW power supply lost), wind speed or precipitation-driven power outages increased across the United States between 1984 and 2006. Here, we did not account for temporal changes in power outages because the number of customers tracked by EAGLE-I has increased from 2015 to present; this would likely introduce biases in trend estimates. However, our results suggest that weather variables other than precipitation and wind speed, like lightning strikes or high coastal water levels, can also generate longer-duration outages (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e), in isolation or when paired with other weather variables (Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). Overall, our results highlight that compound weather extremes lead to longer outrages (Fig.\u0026nbsp;7). While this general result is intuitive and expected, especially for the most extreme events related to hurricanes, our analysis identifies spatial variability across the state of Florida and identifies hotspots that are relatively more affected by certain types of compound events (Figs.\u0026nbsp;7a-d). Finally, we show that non-weather-related outages, i.e., those related to equipment failure or other system disruptions, are much shorter than weather-related outages, in many cases less than half (Figs.\u0026nbsp;7e-f).\u003c/p\u003e \u003cp\u003eLimitations of the study include that the power outage data covers varying percentages of customers tracked in different counties, varying methods being used for reporting power outages across different utilities, and details about the grid characteristics not being publicly available. This is likely one of the reasons for differences in our results between urban (metro) and rural (non-metro) areas, with generally fewer outages in rural areas despite there being similar occurrences of extreme weather events as in urban areas. It could be due to underreporting but also due to the fact that fewer people live in rural areas, which means fewer overhead power lines exist. This, in turn, means, for example, that the likelihood of falling trees affecting the grid is smaller as compared to densely populated areas with many overhead lines and generally denser grids; the same is true for lightning strikes. There is also more permeable surface in rural areas, which means less flooding under the same rainfall conditions compared to highly urbanized areas. That explains why there is often extreme weather but, in many cases, without causing outages. Overall, the grid in those areas is still poorly maintained, leading to outages that are not due to weather, and when those happen, the repair times are longer since they have less priority and/or take more time to get crews on site. Our study does not capture many of these, given that those events are mostly isolated and very few customers are impacted. At the same time, utility companies might not track customers as high as metro areas in non-metro areas.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study introduced multiple ways to assess power grid risk and reliability in Florida by processing open-access EAGLE-I power outage data in conjunction with extreme weather data. Power outage data processing involved isolating big enough power outage events based on selected thresholds to eliminate small events that are likely generated by equipment failure or other system interruptions rather than triggered by extreme weather events. This allowed us to calculate the duration of individual events based on consecutive power outages in the 15-minute time series. We then characterized weather-related power outages based on those durations and extreme weather information.\u003c/p\u003e \u003cp\u003eWe showed that total outage counts were higher in metro areas than in non-metro areas. However, interestingly, the percentage of weather-related power outages was higher in non-metro areas than in metro areas. Weather-related power outages dominated the spatial variability in outage frequencies across Florida. We also identified localized hotspots in terms of the drivers and characteristics of power outages from different weather variables and their combinations. The framework adopted in this study identifies the most relevant drivers of weather-related power outages and how each driver or combination of more than one driver either triggered or modified observed power outages. Understanding the complex interplay of different combinations of weather variables and their associated impact on the grid is the first step to develop predictive models that account for those complex relationships and can help inform future grid development to increase resilience against extreme weather.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgments\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMohammad Siddiqur Rahman and Thomas Wahl were supported by the U.S. Department of Energy\u0026apos;s Office of Energy Efficiency and Renewable Energy (EERE) under the Solar Energy Technologies Office Award Numbers DE-EE0010418. We acknowledge DOE-CESER, the agency responsible for funding the creation of EAGLE-I data. Support for DOI 10.13139/ORNLNCCS/1975202 dataset is provided by the U.S. Department of Energy, project EAGLE-I under Contract DE-AC05-00OR22725. Project EAGLE-I used resources of the Oak Ridge Leadership Computing Facility at Oak Ridge National Laboratory, which is supported by the Office of Cybersecurity, Energy Security, and Emergency Response of the U.S. Department of Energy under Contract No. DE-AC05-00OR22725.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFinancial support\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research has been supported by the U.S. Department of Energy\u0026apos;s Office of Energy Efficiency and Renewable Energy (EERE) under the Solar Energy Technologies Office Award Numbers DE-EE0010418.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCode availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFinal processed data and codes are available in GitHub at https://github.com/CoRE-Lab-UCF/PowerOutageFL (last access: 2 April 2025).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEAGLE-I power outage data is available and downloaded from https://doi.ccs.ornl.gov/dataset/d26451af-7ea4-577c-8285-78b72cd8a2dc. Precipitation, soil moisture, and wind speed data used in this paper can be downloaded through USGS at https://www.sciencebase.gov/catalog/item/6372cd09d34ed907bf6c6ab1. The lightning strikes dataset of U.S. National Lightning Detection Network is obtained and available at https://ghrc.nsstc.nasa.gov/home/lightning/index/data_nldn. CAPE is obtained and available through ERA5 at https://cds.climate.copernicus.eu/datasets/derived-era5-single-levels-daily-statistics?tab=download. HURDAT 2 reanalysis data is downloaded available through National Oceanic and Atmospheric Administration at https://www.aoml.noaa.gov/hrd/hurdat/hurdat2-nepac.html. Water levels data is accessed and available at https://registry.opendata.aws/noaa-nos-cora. Metro and non-metro classification used in this paper is downloaded and available at https://www.ers.usda.gov/data-products/rural-urban-continuum-codes. \u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was conceived by MSR and TW. MSR developed the methodology, undertook the analysis, and wrote the first draft of the paper under the guidance of TW. MN and AE contributed by generating ideas, providing valuable insights during technical discussions, and editing the manuscript.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eAbdelmalak M, Cox J, Ericson S, Hotchkiss E and Benidris M 2023 Quantitative Resilience-Based Assessment Framework Using EAGLE-I Power Outage Data IEEE Access 11 7682\u0026ndash;97\u003c/li\u003e\n \u003cli\u003eAli J, Wahl T, Enriquez A R, Rashid M M, Morim J, Gall M and Emrich C T 2023 The role of compound climate and weather extreme events in creating socio-economic impacts in South Florida Weather Clim Extrem 42\u003c/li\u003e\n \u003cli\u003eBelligoni S, Trader E, Li M, Rahman M S, Ali J, Enriquez A R, Nagaraj M, Aksha S K, Stevens K A, Wahl T, Emrich C T, Qu Z and Davis K O 2025 Transdisciplinary research promoting clean and resilient energy systems for socially vulnerable communities: A review Renewable and Sustainable Energy Reviews 213\u003c/li\u003e\n \u003cli\u003eBhusal N, Abdelmalak M, Kamruzzaman M and Benidris M 2020 Power system resilience: Current practices, challenges, and future directions IEEE Access 8 18064\u0026ndash;86\u003c/li\u003e\n \u003cli\u003eBrelsford C, Tennille S, Myers A, Chinthavali S, Tansakul V, Denman M, Coletti M, Grant J, Lee S, Allen K, Johnson E, Huihui J, Hamaker A, Newby S, Medlen K, Maguire D, Dunivan Stahl C, Moehl J, Redmon D, Sanyal J and Bhaduri B 2024 A dataset of recorded electricity outages by United States county 2014\u0026ndash;2022 Sci Data 11\u003c/li\u003e\n \u003cli\u003eCasey J A, Fukurai M, Diana Hern\u0026aacute;ndez \u0026amp;, Satchit Balsari \u0026amp;, Kiang M V, Hern\u0026aacute;ndez D and Balsari S 2020 Power Outages and Community Health: a Narrative Review Curr Environ Health Rep 7 371\u0026ndash;83 Online: https://doi.org/10.1007/s40572-020-00295-0\u003c/li\u003e\n \u003cli\u003eChakalian P M, Asce S M, Kurtz L C and Hondula D M 2019 After the Lights Go Out: Household Resilience to Electrical Grid Failure Following Hurricane Irma Online: https://orcid.org\u003c/li\u003e\n \u003cli\u003eClimate Central 2024 Weather-related Power Outages Rising. Accessed on September 24, 2024, from https://www.climatecentral.org/climate-matters/weather-related-power-outages-rising\u003c/li\u003e\n \u003cli\u003eDo V, McBrien H, Flores N M, Northrop A J, Schlegelmilch J, Kiang M V. and Casey J A 2023 Spatiotemporal distribution of power outages with climate events and social vulnerability in the USA Nat Commun 14\u003c/li\u003e\n \u003cli\u003eDOE OE-417 Office of Electricity Delivery and Energy Reliability, Electric Disturbance Events (OE-417), https://openenergyhub.ornl.gov/explore/dataset/oe-417-annual-summaries/information/\u003c/li\u003e\n \u003cli\u003eDugan J, Byles D and Mohagheghi S 2023 Social vulnerability to long-duration power outages International Journal of Disaster Risk Reduction 85\u003c/li\u003e\n \u003cli\u003eEntress R M and Stevens K A 2023 Public values failure associated with Hurricane Ian power outages Frontiers in Sustainable Energy Policy 2\u003c/li\u003e\n \u003cli\u003eFEMA 2020 Community Lifelines. Published July 27, 2020. Accessed January 11, 2025. https://www.fema.gov/emergency-managers/ practitioners/lifelines.\u003c/li\u003e\n \u003cli\u003eFlorida Division of Emergency Management 2017 \u0026ldquo;Outage reports\u0026rdquo; https://www.floridadisaster.org/info/outage_reports/\u003c/li\u003e\n \u003cli\u003eHersbach H, Bell B, Berrisford P, Hirahara S, Hor\u0026aacute;nyi A, Mu\u0026ntilde;oz-Sabater J, Nicolas J, Peubey C, Radu R, Schepers D, Simmons A, Soci C, Abdalla S, Abellan X, Balsamo G, Bechtold P, Biavati G, Bidlot J, Bonavita M, De Chiara G, Dahlgren P, Dee D, Diamantakis M, Dragani R, Flemming J, Forbes R, Fuentes M, Geer A, Haimberger L, Healy S, Hogan R J, H\u0026oacute;lm E, Janiskov\u0026aacute; M, Keeley S, Laloyaux P, Lopez P, Lupu C, Radnoti G, de Rosnay P, Rozum I, Vamborg F, Villaume S and Th\u0026eacute;paut J N 2020 The ERA5 global reanalysis Quarterly Journal of the Royal Meteorological Society 146 1999\u0026ndash;2049\u003c/li\u003e\n \u003cli\u003eHines P, Apt J and Talukdar S 2009a Large blackouts in North America: Historical trends and policy implications Energy Policy 37 5249\u0026ndash;59\u003c/li\u003e\n \u003cli\u003eHines P, Apt J and Talukdar S 2009b Large blackouts in North America: Historical trends and policy implications Energy Policy 37 5249\u0026ndash;59\u003c/li\u003e\n \u003cli\u003eIEEE Guide for Electric Power Distribution Reliability Indices\u0026quot; in IEEE Std 1366-2012 (Revision of IEEE Std 1366-2003) (2012) vol., no., pp.1-43, doi: 10.1109/IEEESTD.2012.6209381.\u003c/li\u003e\n \u003cli\u003eLandsea C W and Franklin J L 2013 Atlantic hurricane database uncertainty and presentation of a new database format Mon Weather Rev 141 3576\u0026ndash;92\u003c/li\u003e\n \u003cli\u003eLiu Y, Zhong J and Kong H 2017 Risk Assessment of Power Systems under Extreme Weather Conditions-A Review\u003c/li\u003e\n \u003cli\u003eM. Sullivan, M. Perry, J. Schellenberg, J. Burwen, S. Holmberg and S. Woehleke (2012) Pacific gas \u0026amp; electric company\u0026rsquo;s 2012 value of service study.\u003c/li\u003e\n \u003cli\u003eMukherjee S, Nateghi R and Hastak M 2018 A multi-hazard approach to assess severe weather-induced major power outage risks in the U.S. Reliab Eng Syst Saf 175 283\u0026ndash;305\u003c/li\u003e\n \u003cli\u003eNOAA\u0026apos;s Coastal Ocean Reanalysis (CORA) 2024 Dataset was accessed on September 16, 2024, from https://registry.opendata.aws/noaa-nos-cora.\u003c/li\u003e\n \u003cli\u003eRasmussen R M, Chen F, Liu C H, Ikeda K, Prein A, Kim J, Schneider T, Dai A, Gochis D, Dugger A, Zhang Y, Jaye A, Dudhia J, He C, Harrold M, Xue L, Chen S, Newman A, Dougherty E, Abolafia-Rosenzweig R, Lybarger N D, Viger R, Lesmes D, Skalak K, Brakebill J, Cline D, Dunne K, Rasmussen K and Miguez-Macho G 2023 CONUS404 The NCAR\u0026ndash;USGS 4-km Long-Term Regional Hydroclimate Reanalysis over the CONUS Bull Am Meteorol Soc 104 E1382\u0026ndash;408\u003c/li\u003e\n \u003cli\u003eSpurlock T, Sewell K, Sugg M M, Runkle J D, Mercado R, Tyson J S and Russell J 2023 A spatial analysis of power-dependent medical equipment and extreme weather risk in the southeastern United States International Journal of Disaster Risk Reduction 95\u003c/li\u003e\n \u003cli\u003eSullivan, Michael, Collins, Myles T., Schellenberg, Josh, \u0026amp; Larsen, Peter H. (2018) Estimating Power System Interruption Costs: A Guidebook for Electric Utilities. https://doi.org/10.2172/1462980\u003c/li\u003e\n \u003cli\u003eU.S. Department of Agriculture, Economic Research Service. (January 2024). Rural-Urban Continuum Codes\u003c/li\u003e\n \u003cli\u003eU.S. Energy Information Administration 2015 Form EIA-861, Annual Electric Power Industry Report. Available at https://www.eia.gov/electricity/annual/html/epa_11_04.html; Accessed on January 13th, 2025\u003c/li\u003e\n \u003cli\u003eXu J, Qiang Y, Cai H and Zou L 2023 Power outage and environmental justice in Winter Storm Uri: an analytical workflow based on nighttime light remote sensing Int J Digit Earth 16 2259\u0026ndash;78\u003c/li\u003e\n \u003cli\u003eZscheischler J, Martius O, Westra S, Bevacqua E, Raymond C, Horton R M, van den Hurk B, AghaKouchak A, J\u0026eacute;z\u0026eacute;quel A, Mahecha M D, Maraun D, Ramos A M, Ridder N N, Thiery W and Vignotto E 2020 A typology of compound weather and climate events Nat Rev Earth Environ 1 333\u0026ndash;47\u003c/li\u003e\n \u003cli\u003eZscheischler J, Westra S, Van Den Hurk B J J M, Seneviratne S I, Ward P J, Pitman A, Aghakouchak A, Bresch D N, Leonard M, Wahl T and Zhang X 2018 Future climate risk from compound events Nat Clim Chang 8 469\u0026ndash;77\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"University of Central Florida","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Power outage, extreme weather, compound events, EAGLE-I, grid reliability","lastPublishedDoi":"10.21203/rs.3.rs-6404193/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6404193/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe growing number of extreme weather events has contributed to the increasing number and severity of power outages. However, the complex interplay of extreme weather events and their compounding effects on power outage characteristics (e.g., event duration is yet to be explored). Power outage data is often not publicly available, especially at high spatial resolution. Identifying outages related to weather events can also be challenging as various weather variables can trigger or moderate power outages when they occur, in isolation or combined. Here, we use county-level power outage data from EAGLE-I for the state of Florida from 2015 to 2022 to identify moderate and major weather-related outages and analyze their characteristics. We show that total outage counts were higher in metro areas than in non-metro areas. However, the percentage of weather-related power outages was higher in non-metro areas than in metro areas. Spatial variation of grid reliability indicators derived from all weather-related events follows similar patterns as derived when just focusing on tropical cyclone events, highlighting the importance of these types of extremes in creating prolonged outages. Considering six relevant weather variables, we identify univariate and compound events (i.e., when more than one weather variable was extreme at the time of the outage). Univariate events have a homogenous pattern across the state of Florida, while compound events have more localized hotspots. The average duration of the outages also increases when moving from univariate to multivariate events. Our results shed light on the relative importance of different weather variables (in isolation or combination) in creating power outages with different characteristics across Florida. Identifying such causal relationships is an important step in understanding how power outage risk profiles may change when certain extreme weather events become more frequent.\u003c/p\u003e","manuscriptTitle":"Characteristics of Power Outages from Compound Weather Extremes in Florida","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-04-09 06:07:15","doi":"10.21203/rs.3.rs-6404193/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"cd816018-f814-4b04-83cd-cbf722d2cc85","owner":[],"postedDate":"April 9th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":46859448,"name":"Climatology"},{"id":46859449,"name":"Climate Analysis and Modeling"}],"tags":[],"updatedAt":"2025-04-09T06:07:15+00:00","versionOfRecord":[],"versionCreatedAt":"2025-04-09 06:07:15","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6404193","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6404193","identity":"rs-6404193","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}