Hurricane Idalia made landfall at Keaton Beach in Florida’s Big Bend as a Category 3 storm at 8 a.m. ET on August 30th, 2023. The storm had been downgraded from a Category 4 but still maintained winds of 125 mph.

The impact of the storm has resulted in flooded streets, closed airports, canceled flights, and widespread power outages.

The National Weather Service in Tallahassee warned that certain locations might be uninhabitable for weeks or even months due to wind damage and storm surge barriers.

CrisisReady, a joint effort of Harvard University and Direct Relief, is creating situation reports to inform relief efforts. These reports analyze anonymized and aggregated data on the movement of Facebook users to track changes in population densities, highlighting displacement in affected areas of Florida. They also contain information about baseline population vulnerabilities and local healthcare infrastructure, such as facilities, outpatient centers, and hospitals.

The report below provides data on population density changes in northern and central Florida as of 7:00 p.m. local time on Tuesday, August 29, 2023, the evening prior to Hurricane Idalia’s landfall. It highlights population displacement during the period evacuation orders were issued by state and county officials.

Additionally, the report provides information on baseline population vulnerabilities and the status of local healthcare infrastructure.

Key Observations

Counties with Population Decline Prior to Landfall

Predominantly in the Big Bend and inland areas:

• Madison
• Suwanee
• Levy
• Lafayette
• Wakulla
• Echols
• Dixie
• Taylor
• Franklin

Counties with Significant Declines

Coastal regions with declines of 10% or more where evacuation orders were issued:

• Dixie
• Taylor
• Franklin

Areas registering population drops of 45% or more before the storm:

• Panacea
• Steinhatchee
• St. Mark’s
• Crystal River
• Homosassa

Note: Steinhatchee reported storm surge levels over 8 feet as of the morning of August 30th, indicating extreme risk.

Social Media Usage & Mobility Data:

Florida communities in the Big Bend area typically exhibit:

• High Facebook usage rates with location services (15%-25%).
• Consequently, there is a high degree of representativeness in mobility data.

The post Hurricane Idalia: Tracking Evacuations and Population Movement appeared first on Direct Relief.

Heavy rainfall in Southern Pakistan and melting glaciers in the country’s northern mountains have caused massive floods and flash floods that continue to devastate districts across the region. The floods have killed at least 1,191 people, 399 of whom were children, since the flooding began in mid-June at the beginning of the monsoon season. In addition to this, a reported 3,554 individuals have been injured. The National Disaster Management Authority stated that as of August 29, 2022, more than 33 million people have been impacted and more than 1 million houses have been destroyed.

The floods have spawned a critical humanitarian crisis as damage and displacement increase across the country. So far, 66 districts have been officially declared “calamity hit.” An estimated 50,000 people have been evacuated since rescue efforts began. Pakistan’s meteorological office has predicted that more flash floods are expected throughout September.

CrisisReady has published an interactive map that shows population movement patterns driven by the floods between August 13, 2022, and September 5, 2022. Data reflecting population movement originated from selected level 2 administrative units of Pakistan, including Karachi, Larkana, Malakand, Quetta, and Sukkur.

The red arrows (shown below) on the map show the directional patterns of population movement. The size (width) of the arrows correlates with the volume of individuals displaced from the selected origins. The larger the arrow, the greater number of movement vectors. Transparency of arrows indicates the baseline population traveling between the origin and the destination under the pre-crisis situation.

Explore the dashboard here.

The maps were generated using data provided by Data for Good at Meta. For more information about the disaster population maps provided by Data for Good at Meta. Data on flood extent is gathered using the Visible Infrared Imaging Radiometer Suite (VIIRS), an instrument that collects visible and infrared images and global observations of the land, atmosphere, cryosphere, and oceans.

The post Catastrophic Floods Devastate Southern Pakistan: CrisisReady Responds With New Data Reports appeared first on Direct Relief.

This post was authored by Caleb Dresser, Neil Singh Bedi, and Akash Yadav of CrisisReady.

As California prepares for another wildfire season, safety net health facilities are preparing to meet the needs of populations that will be affected in coming months. Given the continued effects of the COVID-19 pandemic and ongoing resource limitations, those efforts need to be focused where impacts are likely to be greatest.

CrisisReady is informing Direct Relief’s effort to assess how current wildfire risk, historical wildfire patterns, and historical use of shelters by affected populations overlap with the locations of safety net health centers in California. While predicting exact health care facility impacts and population movements for future disasters is not feasible, this analysis is able to provide information on patterns from past wildfire seasons that can inform current planning efforts.

The research team combined data from CalFire on wildfire risk zones (Fire Hazard Severity Zones) with information on historical emergency shelter utilization from the National Shelter System to provide assessments for 2059 safety net health centers in California. Of these, 23 are located in areas of Moderate fire risk, 9 are located in areas of High fire risk, 10 are located in areas of Very High fire risk by CalFire (see above image). While this approach provides information about direct risk to the facilities themselves, it likely underestimates risk to the populations they serve when facilities in low-risk zones are located near higher-risk areas.

Historical emergency shelter utilization data can provide proxy information about population displacement during wildfires and the scale at which vulnerable persons need aid, including medical care, which they may access at nearby health centers. Spatial data on shelter use related to wildfires in California was obtained from the National Shelter System and was manually verified against media reports; sheltered population information for a small subset of fires was not available, and as a result, numbers presented here are very likely to be underestimated in some locations (Figure 2).

Figure 2 – Average Annual Wildfire Event-Related Shelter Stays, per Capita (2007 – 2019)

For the years 2007 to 2019, the total number of shelter stays, mean annual stays, and median annual stays were computed for a 25-mile radius around each of the 2059 safety net health centers. Use of shelters varied considerably by year, with a small number of major disasters such as the Camp Fire accounting for a substantial fraction of total shelter use. When data was examined on an annual basis, the team found that safety net health centers with high mean numbers of persons sheltering nearby often had been adjacent to a single devastating event that skewed the average upward. Information on median annual usage likely provides a more stable metric of expected local population displacement and relative population vulnerability in any given year, but both metrics have been provided given that exceptionally devastating events do represent the greatest demand on facilities, and should be considered in planning for future events. (Figure 3)

Figure 3 – Shelter Usage within a 25mi Radius of Safety Net Health Centers

Within the limitations of these data sets, many safety net health care facilities in Los Angeles, San Diego, Butte, Glenn, Sonoma, Marin, and Napa counties have historically seen exceptionally large numbers of persons sheltering nearby, and should be particularly alert to needs for preparation to serve displaced populations during the upcoming fire season. In addition, facilities in rural parts of Northern California and inland Southern California may see high numbers of displaced persons relative to the baseline population, which also creates the potential to strain limited local resources.

Key steps should be taken now to prepare health facilities for what promises to be a difficult 2021 fire season. According to the National Interagency Fire Center’s National Significant Wildland Fire Potential Outlook, issued in May 2021, California is seeing warm, dry conditions and can expect above-normal fire risk in many areas. Similar dynamics contributed to a record-setting scale of wildfire activity in California in 2020. Long-term, the problem is getting worse, in part because of warmer, drier conditions related to climate change, as well as development and land use decisions.

As conditions become more challenging for safety net health centers and the populations they serve, resources need to go where they are most needed and facilities in high risk locations or serving high risk populations need to improve planning for future needs. Assessments of hazards and human impacts such as those presented here can help decision makers understand how to optimize resources and advocate for additional investment or support. Additional information on Direct Relief’s safety net preparedness efforts will be available at https://www.directrelief.org as fire season progresses.

The post Where are Health Resources Most Needed to Prepare for the 2021 California Wildfire Season? appeared first on Direct Relief.

This post was authored by Dr. Jennifer Chan – Northwestern University, with support from Lauren Bateman and Kate Bruni. This work is part of the User Feedback Project for the Covid-19 Mobility Data Network, which has been generously supported by the Institute for Global Health at Northwestern University.

The interactive dashboard can be found here, although it is not currently being updated.

Syracuse is a modest sized city of just over 140,000 people in upstate New York. It sits within the heart of the Finger Lakes region, not far from Lake Ontario. Although far from the worst hit areas of New York state by COVID-19, Syracuse saw alarming increases in cases and deaths during the initial wave of the pandemic. Daily cases in Onondaga County, where Syracuse is located, hit their daily peak of 145 on the 19th of May, with nearly 5,000 total cases and over 200 deaths recorded as of October 13, 2020.

The city of Syracuse, like many others around the United States, sought to use digital mobility data as part of their means to implement, monitor and adjust physical distancing policies in order to mitigate the spread of COVID-19. A small data team within the city government took it upon themselves, with support from the COVID-19 Mobility Data Network, to learn how to analyze digital mobility data in order to inform public health decision-making, and in the process made significant advances in data applications for the control of Covid-19 in their area.

The use of mobility data by the city of Syracuse from March through June, 2020, is an example of a multi-member collaboration between Harvard researchers in the COVID-19 Mobility Data Network and a city-level data team, which was tasked with responsibilities in public health analysis despite not itself sitting within the department of health.

Facebook mobility data was integrated into broader situational awareness efforts including an interactive dashboard that was made available to the mayor and others. Different users and decision-makers viewed mobility data trends after social distancing measures during phases of reopening, holidays, and spring temperature changes.

“I think it’s (mobility data) pushing the decision makers to think about things in ways that we haven’t thought about before.” – Neha Majety

The all-woman data team worked closely with Nishant Kishore, a PhD student affiliated with the Center for Communicable Disease Dynamics at Harvard University.

The Syracuse team learned about the COVID-19 Mobility Data Network and Facebook mobility data through their chief data officer as well as a group call with other cities in the United States. Virtual conversations followed between Nishant and the team, where they discussed both the potential added value of mobility data and its limitations for COVID-19 response activities. The data team was working to support the mayor’s interests in seeing how people were responding to the executive orders to stay at home.

“Our mayor is very adamant about us making data driven decisions, as a government, as a whole. And being able to remind him that there are different, obscure ways of addressing that data and supporting our populations (is important).” – Amanda Darcangelo

After two calls, the team decided to use Facebook population data given the delayed baseline period and subsequent lack of movement data. Nishant created a template showing population changes as a proxy for movement at the census tract level, as well as population changes during different times of day. The situation reports were shared in PDF format and frequently reviewed by the data team and chief data officer. Joanna Bailey, the data team’s GIS expert, soon had access to geographic shapefiles, enabling them to adapt and determine how best to use the data to fit their needs.

Mobility on Easter Sunday and Changes in Temperature

The Syracuse team integrated the Facebook data into a broader situational awareness dashboard for the county. They were interested in visualizing the trends in case data in relation to spikes in mobility. The purpose of the dashboard was to correlate spikes in mobility to weather pattern trends.

At the team’s request, Joanna Bailey received comma-separated-values (CSV) files of anonymized aggregated data so they could continue to monitor changes and analyze the data in combination with other data including temperature, COVID-19 symptom emergency call locations, and the number of COVID-19 cases and tests. During Easter weekend they anticipated that there would be increased movement during that time due to family gatherings and warmer weather.

Their analysis confirmed that there was a spike in movement during the holiday weekend. The structured data files enabled Joanna and Jennifer Glass, the data intern, to create maps that the team believed provided an easier way for them to monitor changes and work with other datasets to see potential relationships between movement near business areas during reopening phases as a potential proxy for rates of economic recovery.

Envisioning Use Cases and Collaborations

Like many other collaborative teams in the mobility network, groups working within government agencies envisioned that the mobility data would help them better understand the impact of social distancing policies during the early phases of the pandemic. What many groups learned over time was that iterations and adaptations with the mobility data were crucial in aligning it with their specific needs.

“I also feel like this is kind of giving us the opportunity to look deeper into the different parameters that govern the society… You know, it kind of pushes us into more diverse tracks of thought. So I think it’s helping us in a lot of ways.” – Neha Majety

More often than not, defining those needs with an often unfamiliar data source takes time – a journey of learning between all groups involved. The Syracuse collaboration “kicked off” in the first week of April with PDF situation reports, and evolved over time to maps created internally by the Syracuse data team and eventually an online dashboard. Using the Facebook population data as a proxy for movement data helped the data team in Syracuse support the broader needs of the government during the response to COVID-19.

“It’s definitely made our leadership reassess what using data means… [for example] here are some people who are mobile, and here are tracks in which we may have more mobile people but we don’t know because they don’t have internet.” – Amanda Darcangelo

While some of the outcomes included an evolution of maps, data products, and insights over time, the use of Facebook data had meaningful influence on how the data team envisioned the role of data generally in the crisis, including perceptions of how it can play a meaningful role in the ongoing digital transformation within their organization.

The post Digital Mobility Data, Physical Distancing, and COVID-19 in Syracuse appeared first on Direct Relief.

This post was authored by Andrew Schroeder, Nishant Kishore, Navin Vembar, and Caleb Dresser.

Key points:

• The area of displacement from Calcasieu post-Laura stretched in all directions from the epicenter, principally to Orleans Parish and Baton Rouge in the east, and to Houston in the West.
• Rates of return to Calcasieu have been relatively slow, with only ⅓ of displaced persons having returned 25 days post-storm.
• Displacement and return have been disproportionately gendered, with greater numbers of women being displaced and also returning at higher rates than men.
• Shelter usage patterns show significantly lower utilization after Laura in the past.

The 2020 hurricane season has been unusually active, with 23 named storms as of September 24th, six greater than the previous annual record set in 2005. As Hurricane Sally arrived on the Gulf Coast across the Florida panhandle and into Mississippi and Alabama, people once again had to evacuate at-risk locations and seek safety from the storm in the midst of a pandemic. Hurricane Laura landed in Louisiana and Texas less than three weeks previously, and some evacuees from that storm were still staying in temporary shelters, hotels, and other types of short-term housing. The succession of events and their potential geographic overlap raised serious concerns about safe evacuation and long term displacement. Risks remain high for similar events throughout the rest of this year, making reflection and learning from the recent past a particular imperative now.

Where Did People Go?

Data from Facebook and Camber Systems highlights patterns of movement as people moved to shelter from the hurricanes that landed on the Gulf Coast in the early part of September.

First, we can see the estimated change in the population of devices from Facebook data. The difference in densities between counties can be seen by comparing the percentage change of the population with an estimated absolute change extrapolated using the American Community Survey (ACS) data from the US Census.

Cameron Parish, in the southwest corner of the state, is highlighted red in both pictures, indicating a significant percentage change of people as well as a significant absolute change. But just south, Jefferson Parish, is not red in absolute change, as it has a far lower density of people.

The overall pattern of dispersal from Calcasieu Parish in the face of Laura is shown in the below animation. Calcasieu Parish is useful to focus on given that it hosts the city of Lake Charles, which bore the brunt of the storm.

According to reports from the Red Cross over 8,000 homes were destroyed or badly damaged, the majority located in the Lake Charles area. Throughout the region, the locations with the greatest reduction in population as of September 7th (the last date of Facebook mobility data available)ranged from Harris County, TX on the highest end through the following set of cities across southern Louisiana.

Gender and Displacement

According to long-term displacement data from Facebook, population displacement from Calcasieu Parish has been gender skewed to a degree that demands further investigation. Although the overall Facebook data sample appears to be skewed towards women, with the user base in the region, those with location services enabled, being composed of 56.2% women, the proportion of persons displaced from Calcasieu Parish was tilted to an even greater degree towards women. Whereas 55.1% of all displaced persons in the region were female, according to Facebook, as of September 10th, 2020 that number rose to 63.3% for women from Calcasieu, a difference of 8.2%.

The proportion of returnees was also skewed towards greater proportions of women. As of September 20th 33% of the population displaced from Calcasieu had returned. However the rate for women was much higher at 37% as compared with 28% for men. The rate of return has also increased steadily for women relative to men over the past 10 days. These gender differences in rates of return potentially indicate greater challenges for women with longer term displacement.

Reasons for the gender disparities in displacement are likely due to underlying social vulnerabilities in the area. For example, the rate of households headed by a single parent in Calcasieu is more than three times greater for women as compared to men, with 10.5% of family households being headed by a single female parent as opposed to 2.8% for men. These types of gendered household disparities tend to reflect underlying income differences as well.

Shelter Utilization

Historically, the American Red Cross and community partners have provided shelter for tens of thousands of people during major hurricanes. This year, however, the ongoing COVID-19 pandemic has fundamentally altered the process of evacuation and sheltering. Response agencies are now operating smaller shelters, conducting COVID-19 screening, and working to house evacuees in hotels, where they can self-isolate, rather than in mass shelters. Many evacuees have been also choosing to avoid shelters altogether, in some cases choosing to sleep in vehicles instead.

While past hurricanes in the region have resulted in many thousands of people staying in shelters at the height of a storm, usage numbers peaked at a few hundred per night following Hurricane Laura and quickly downtrended. Most of the people who would have been expected to use shelters during a major storm chose not to do so during and after Hurricane Laura. Based on this, it seemed likely that evacuees from Hurricane Sally would also use shelters far less than they would have been expected to in years past.

Epidemiological modeling shows that hurricane evacuations can impact the spread of COVID-19 as evacuees are displaced from their homes. In seeing the spread of people away from the hurricanes and that people had yet to return home from Laura before Sally hit, the long-term impact of transmission from evacuees needs to be studied. The information here, especially that showing where people went – and then stayed – is important for future planning. There are nearly impossible tradeoffs to be made here, given that evacuation from hurricanes is necessary – but, appropriate shelter planning can potentially reduce the spread of COVID-19 across counties.

The post Displacement, Gender Disparities, and Shelter Utilization after Hurricane Laura appeared first on Direct Relief.