Thank you for visiting nature.com. You are using a browser version with limited CSS support. Obtain
For the best experience, we recommend using a more recent browser (or the compatibility mode in
Internet Explorer). In the meantime, we are displaying the site without styles for continued support
and JavaScript.

Nature sustainability

(2021) Quote this article

The possibility of a massive oil spill in the Red Sea is becoming ever more likely. The Safer, a decaying oil tanker carrying 1.1 million barrels of oil, has been abandoned near the coast of Yemen since 2015 and threatens an environmental disaster for a country currently in a humanitarian crisis. Here we model the immediate public health effects of a simulated release. We estimate that all of Yemen’s imported fuel would be disrupted via its main Red Sea ports and that the expected leak could affect the supply of clean water, which equates to a daily consumption of € 9.0 € 9.9. ? Millions of people, food supplies for 5.7 million people and 93? 100% of Yemeni fisheries in the Red Sea. We also estimate an increased risk of cardiovascular hospitalization due to environmental pollution between 5.8 and 42.0% over the duration of the spillage. The oil spill and its potentially catastrophic effects remain completely preventable by draining the oil. Our results underscore the need for urgent action to avert this impending disaster.

Since 2015, Yemen has become the site of “the world’s worst humanitarian disaster” due to war and blockade. Tanker moored 4.8 nautical miles off the coast of Red Sea in Yemen. As the safer, which has been classified as alien since 2016 and has not been maintained since the beginning of the conflict, continues to deteriorate, concerns have arisen about a massive spillage. The safer contains 1.1 million barrels of oil, more than four times the amount spilled by the Exxon Valdez2 in 1989. The potential leak threatens to harm the environment, the economy and the public health of the countries bordering the Red Sea.

The possibility of a spill is becoming ever more likely. The visibly dilapidated Safer is envelope-shaped, which means that in the event of a break, the oil on board spills directly into the sea. In May 2020, water leaked into the engine room through a leak in a seawater pipe and the ship’s fire extinguishing system is out of order3. A leak or burn could result in a spillage. A leak could result from continued damage to the ship’s hull or from breakage of the ship’s hull due to bad weather; Combustion could result from the accumulation of volatile gases on board the ship or from direct attack on the ship. Ansar-Allah (colloquially known as the Houthis), a political and armed movement that controls northern Yemen, currently has access to Safer. At this point in time, negotiations between the United Nations and the Houthis over the inspection and repair of the Safer have stalled indefinitely, and no long-term solutions such as offloading the oil have been publicly proposed.

Yemen is due to the Dependence on the major ports near Safer, Hudaydah and Salif, through which 68% of humanitarian aid enters the country, particularly vulnerable to the anticipated oil spill. In the event of a port disruption, the rerouting of humanitarian aid would be logistically difficult due to regional instability, a lack of capacity in other ports and the ongoing blockade, which severely restricts the access of supplies4. Overall, Yemen imports 90-97% of its fuel and 90% of its food supplies5. More than half of the population of Yemen is dependent on humanitarian aid in the ports, with 18 million people in need of clean water and 16 million in need of food aid1.

The expected leakage also threatens the clean water supply of the arid region of the Red Sea. Oil could contaminate the desalination plants along the coast north of the Safer, disrupting the region’s clean water supply. For Yemen in particular, clean water is mostly supplied via groundwater pumps or water trucks, both of which require fuel. Fuel shortages caused by the blockade had far-reaching public health implications: for example, clean water and sanitation systems were shut down, waste collection was blocked, and power cuts resulted in blackouts that affected hospital operations and contributed to a massive cholera outbreak in 20176.

The Yemeni fisheries, which are responsible for supplying 1.7 million people in the country, would also be threatened. Fisheries were Yemen’s second largest export before the conflict began and continue to provide a source of income and food security in a country on the verge of famine. The sector has declined sharply in recent years due to conflict and fuel shortages; a massive oil spill would devastate an industry already struggling to survive7,8,9.

Pollution from spillage, be it from evaporation or smoke from combustion, can lead to cardiovascular and respiratory diseases. Particulate matter (PM2.5) generally increases the risk of hospitalization for cardiovascular and respiratory diseases, and oil spills in particular are known to cause a variety of health problems ranging from psychiatric to respiratory symptoms10,11, 12,13,14 . The resulting pollution from a leak could increase the burden on underserved health systems in Yemen and hinder clean-up operations.

Large oil spills are known to have far-reaching environmental and economic consequences. The danger that safer poses to the unique ecosystem of the Red Sea has been documented2,15. However, the immediate public health impact of safer spilling remains unresolved, and the extent to which the anticipated leak could disrupt humanitarian aid to Yemen is critical to whether and with what urgency interventions are carried out should. Here we are modeling the immediate public health effects that would follow major safer releases. We stochastically simulate oil spills using historical data to assess likely escape and pollution paths, and use these results to estimate impaired access to fuel, food and clean water, as well as initial health effects from air pollution.

We simulate the Safer Spilling over a variety of historical weather conditions and finding that most of the simulated spills tend to move towards the northwest coast of Yemen (Fig. 1). We observe seasonal fluctuations in our models: in summer the leaks tend to move southeast and further along the coast of Yemen, but in winter they tend to move north along the coast of the Red Sea. The uncertainty in the estimates suggests a wide range of possible trajectories over the Red Sea, showing the possibility of movement in both directions for both seasons. We estimate it will take 6-10 days for the oil to reach the west coast of Yemen. We estimate that 51% (95% uncertainty interval: 46-54%) of the oil will evaporate within 24 hours of leaking from the container, with the heavier components remaining on the water (Fig. 2). Modeled cleaning attempts – skimming, in situ burns, and dispersants – remove a negligible amount of oil within the first six days. Under optimistic conditions, remedial measures initially reduce evaporation rates slightly (probably due to the delayed action of dispersants), and after six days we estimate that 39.7% of the oil is floating, compared with 38.2% under evaporation alone (Fig. 2.) . ). If the oil spill spreads unhindered for three weeks, the oil will likely obstruct the passage through the Gulf of Aden (Fig. 1).

aâ ??? f, Average surface oil concentration of 1,000 simulated spills in winter (a, b , c) and in summer (d, e, f). The columns show the progress of the 1,000 spills after one week (a, d), two weeks (b, e), and three weeks (c, f). Colored contours represent percentiles of the average surface concentration over 1,000 simulated spills and can be interpreted as the expected surface concentration in relation to other grid cells in the exposed area. The shaded area represents the area in which about 90% of the trajectories are expected to fall. Blue dots represent desalination plants.

a, evaporation of oil spills over time. The black line shows the mean, the gray band shows the 95% uncertainty interval, which is called the 2.5. and 97.5. Percentile of the values ​​is defined over 1,000 Monte Carlo simulations. b, Oil fate under optimistic clean-up conditions. Colored areas indicate different oil removal processes.

Ports and desalination plants, which are crucial for the supply of fuel, food and water, could be destroyed by the leak. We estimate that Yemen’s main ports, Hudaydah and Salif, are likely to be directly affected two weeks after a leak, with average surface oil concentrations in the 90th percentile compared to other exposed areas (Supplementary Table 1). We estimate that a leak could reach the port of Aden (outside the Red Sea) and the desalination plants and ports in Eritrea and Saudi Arabia in three weeks (Fig. 1, additional table 1 and additional figures 1a? ??? 6).

The leak and subsequent port closings will disrupt maritime traffic across the Red Sea and divert many shipments across Africa. We estimate that for every month the Red Sea port closes, the delivery of 200,000 (180,000 – 250,000) tons of fuel to Yemen will be interrupted, which is 38% of the national fuel requirement16. As a result, we expect fuel prices to rise in Yemen; When the blockade to completely close the ports was tightened in November 2017, fuel prices rose sharply across the country, with prices in Hudaydah rising by 72% in the following month (Fig. 3).

The oil spill also threatens supplies with clean water, equivalent to the daily consumption of an estimated 1.0 to 1.9 million people from the potential contamination of desalination plants. In the Red Sea region, we estimate the potential disruption of desalination plants, which in summer provide a total of 77,000 m3 of clean water and 362,000 m3 of clean water. ??? dâ ???? 1 in winter (supplementary table 2). We also believe that water access in Yemen will be severely affected by fuel shortages if the leak closes ports; During the full port closings in November 2017, 8 million people in Yemen lost access to running water as access to water usually relies on fuel pumps or watercraft5,17.

Likewise, food security is being negatively impacted by potential food aid disruptions and closings Fisheries endangered. Should Yemen’s ports on the Red Sea close within two weeks of the leak, food aid will be cut off to an estimated 5.7 (3.7 to 8.1) million people currently in need of food aid. We estimate that if the port of Aden also closes, a total of 8.4 (5.4 11.9) million people will not receive food aid. We estimate that the spill threatens 66.5 85.2% of Yemeni fisheries in the Red Sea within one week and 93 100% of Yemeni fisheries in the Red Sea within three weeks on the season. In summer even 2.6% of Yemeni fisheries in the Gulf of Aden are threatened within three weeks (Supplementary Table 3).

We forecast moderate short-term health effects from air pollution, with the average increased risk of cardiovascular and respiratory hospital admissions ranges from 5.8% (0-7.5%) in 11.3 (0-27) million people. Days for a slow-releasing winter spill to 31.2% (6.5–50.5%) over 19.5 (0.4–24.2) million person-days for a fast-releasing summer spill (Table 1). Incineration would increase pollution, with estimates of the average increased risk of cardiovascular and respiratory hospital admissions of 6.7% (5.2–7.9%) over 22.3 (1.2–41.8) million person-days for a slow release in winter to 42.0% (21.9-61.4%) over 22.7 (17.0 26.0) million person-days for a fast-releasing summer disaster. Seasonality effects exist, with simulated air pollution moving east into Yemen in summer and west into the sea in winter (Fig. 4). In some winter simulations, the pollution did not reach the Yemeni population at all. Air pollution is highest in the immediate vicinity of the oil and reaches PM2.5 values ​​of up to 1,600 µg / µm³, which corresponds to a value of 530% (460 µg). ? 590%) increased risk of cardiovascular and respiratory hospital stays for people who have been directly exposed to the oil, such as cleaners.

aâ ???? d, Projected average 24-hour air pollution concentration at the end of the leak in Winter (a, b) and summer (c, d) for fast release spills (a, c) and slow release spills (b, d). e, population density graph.

The public health impact of a leak in the Safer oil tanker is expected to be catastrophic, particularly for Yemen. A fuel import disruption is expected to shut down hospitals and key services at a time when Yemen is already facing fuel shortages and only 50% of its health services are operational18,19. Both fuel shortages and contamination of desalination plants are likely to exacerbate an existing water crisis and potentially lead to a resurgence of waterborne infectious diseases20,21,22,23. An interruption in food aid would likely raise food prices and exacerbate a persistent famine1. The leak threatens to disrupt almost all Yemeni fishing in the Red Sea, which would worsen food security and exacerbate Yemen’s displacement crisis as workers look for new jobs7,8,9,19. The import of medical supplies from aid agencies would also likely be disrupted, further destabilizing health services22,24. Our modeling showed that even under optimistic conditions, the cleanup would be slow; The actual clean-up work would possibly take longer and be logistically difficult given the conflict in the region and the sea mines in the water. Ports might remain closed until adequate cleanup has taken place; Should ports reopen prematurely, ships risk mobilizing the oil again and damaging the environment. Air pollution from the leak can be moderate compared to the supply disruptions from the leak, but the cleaners essential to contain the effects of the leak can be at high risk of hospitalization13,25,26 and the the resulting air pollution and an increase in hospital admissions for respiratory diseases could put an already underserved healthcare system under further strain27. Personal protective equipment could significantly mitigate damage from pollution, but the ongoing shortage of medical equipment would likely be exacerbated by port closings28,29.

The long-term and global effects of the oil spill, while outside the scope of our model analysis, are also potential serious. Ecological and environmental impacts from wildlife endangerment and coastal contamination from major oil spills can last for years or decades30,31. The spillage particularly threatens the coral reefs of the Red Sea, which have been studied for their unique resistance to seawater warming32. In addition, the leak could impede world trade through the vital Bab-el-Mandeb Strait, which is 29 kilometers wide at its narrowest point and through which 10% of world shipping goes. Exclusion zones created for the cleanup could divert traffic and deliveries will be delayed as ships potentially exposed to oil need to be cleaned.

Kleinhaus et al. warned of the threat safer poses to the Red Sea ecosystem and conducted a single simulation analysis which found particles moving north in February and particles moving south in August2. Our analysis of the spill trajectories strengthens their results by explicitly modeling the oil properties of Safer and thoroughly evaluating the range of possible spill trajectories through thousands of Monte Carlo simulations based on historical data. We believe that their simulated trajectories match ours, and we confirm their claim that winter accidents tend to migrate north and summer accidents tend to migrate south. However, we emphasize that our primary finding that a Safer leak could lead to catastrophic public health effects is true regardless of the time of year due to the wide range of pathways we observe in our uncertainty analyzes. Accordingly, misinterpretation of seasonality can lead malicious actors to cause an oil spill under supposedly favorable conditions; We strongly claim that our analysis shows that despite the seasonality, there is considerable uncertainty as to the course of events and therefore all parties involved in the conflict must bear the burden of the spill.

Although our model has a number of uncertainties related to a future Taking into account the oil leak, some uncertainties remain. Our oil spill model is averaged through simulations based on historical data, but oil spills often occur due to extreme conditions, so the actual oil spill may manifest itself differently than the scenarios we presented. The data that we use to model the spill and downstream effects are of different quality, so that our results can be influenced by measurement errors. Stored supply volumes could alleviate bottlenecks in the event of a port disruption, but these volumes are likely to be minimal as supply bottlenecks already exist in Yemen and price spikes would make supplies inaccessible to many5,24. Our model only takes into account the full exit scenario; For example, we are not modeling the effects of a small leak in the hull. We do not give any expected durations for the closure of ports and desalination plants, as we cannot predict when the clean-up work will take place. In particular, it is difficult to predict the international response to the expected oil spill: some parties may prefer to keep ports closed to minimize environmental damage, and others may prefer to open earlier to minimize supply disruptions. Our cleanup analysis does not take into account how containment booms can alter the trajectory, but due to the size of the leak, the wave heights in the Red Sea, and the lack of pre-existing capacity, cantilevers would have limited ability to respond immediately33,34. Our spill models do not explicitly model tides and instead assume that tidal currents are implicitly measured in the current data; the Red Sea has a small tidal range, so we do not expect any significant effects on our model estimates35. Our plotted regions of uncertainty for potential exit paths identify the outer particles of uncertainty about where the leaks might go, but can overestimate the areas within the region as there may be pockets of sea where oil is less likely to be transported due to wind or current patterns. However, our models also do not fully take into account the wave turbulence, so that the affected overflow area can be larger than we show. We are also conservative on the assumption that ports will only be disrupted if oil reaches them directly. In reality, oil only needs to hinder nearby maritime traffic to threaten a port closure to prevent the environmental hazard of ships remobilizing the oil. Although we cannot predict the exact likelihood of a port closure, our leak analysis confirms the claim that port closings are a definite possibility, as authorities familiar with the situation have warned36. We therefore do not believe that the uncertainty in our spill models should seriously alter our assessment of the potential downstream public health impact.

Our air pollution models are also subject to several levels of uncertainty. Our evaporation and combustion estimates are based in part on data from the Deepwater Horizon, which contained light crude that was similar, but not equivalent, to the oil the Safer was carrying. It is possible that the tanks on board the Safer are already leaking and that some of the light constituents of the oil have already evaporated, which would reduce air pollution from the leak. We based our health impact assessments on health outcomes of individual pollutants and do not explicitly model volatile organic compounds due to lack of data, but PM2.5 from an oil spill is likely to be different from PM2.5 from other sources. Our estimates of the size of the population affected by pollution may be subject to measurement errors in the original population dataset due to displacement or regional instability. We also only consider the risk of cardiovascular and respiratory hospitalization due to oil exposure; Oil spills are also known to have neurological, hematological, dermatological and psychiatric effects13,14,26. Our modeling therefore does not capture the full extent of the health effects of oil exposure. Additionally, our modeling assumed that the pollution is emitted from a single location (the location of the leak), but in reality it would be emitted from anywhere the oil spills. Hence, our estimates of hospital admission rates are skewed downwards as the pollution would likely be a little closer inland.

Despite the uncertainty inherent in our modeling, our evidence shows that an oil spill from the safer is an extreme risk to the public Represents the health of people in the area, with Yemen having the greatest impact. Our results show that the leak will jeopardize access to fuel, food and water in Yemen, a country already facing a shortage of all three. Other countries bordering the Red Sea are also burdened with port closures and disruption of desalination plants. This public health disaster could be averted by finding a long-term solution to managing the oil on board the Safer, underscoring the need for urgent action by the international community.

For gridded wind data, we use the ERA5- Data sets of the surface winds 2019 and 2020 of the European Center for Medium-Range Weather Forecasts (ECMWF) with a temporal resolution of 1 hour and a spatial resolution of 1/4 ° 37. For rasterized current, sea temperature and salinity data, we use data from 2019 and 2020 from the hybrid coordinate ocean model with a temporal resolution of 3 hours and a spatial resolution of 1/12 ° 38. For data on the properties of the oil, we use the Oil Library Project of the National Oceanic and Atmospheric Administration (NOAA) 39. For data on fuel prices, we use a data set from the World Food Program40. We use fuel import data from the United Nations Verification and Inspection Mechanism for Yemen41. We have used various sources for data on the locations and capacities of desalination plants (Supplementary Table 2) 42,43,44. For Yemen there were neither locations for all desalination plants, nor were the water capacities of the known plants available. Therefore, to estimate the water capacity for each of the known plants in Yemen, we used the latest available data on the nationwide desalination capacity45 and divided this evenly among the known plants. The food imports estimates were derived from the port data of Yemen46.

We modeled the spill with NOAA’s pyGNOME library to use their GNOME model47. GNOME is a widely used Euler / Lagrangian spill trajectory model that models spills with Lagrangian elements within flow fields. Like most operational response tools, GNOME is able to model the oil transport and weathering processes of advection, diffusion, dispersion-entrainment, emulsification, evaporation, propagation, oil-coast interaction, and dissolution. We chose NOAA’s modeling tools because they have been operationally implemented in the past and validated against real environmental disasters, as well as because of their widespread use by civil protection agencies48. As model inputs, we used the properties of the crude oil on board the Safer, Marib Light, as well as the historical currents and wind data of the region. We performed 1,000 Monte Carlo simulations each for summer (June – August) and winter (December – February) with different times of the day and date. The seasons were selected based on known current patterns in the Red Sea49. We limited the simulations to a three week timeframe due to the predictability limits inherent in modeling oil spills and the uncertainty in the cleanup efforts50.

Each simulation had 1,000 particles representing the oil track based on historical weather conditions. The particles were sizeless and were represented by surface oil concentration values ​​and coordinates on a width-length grid rounded to three decimal places, which is approximately 100 µm. For each season we simulated 1,000 leaks with three-week schedules. At the end of weeks 1, 2, and 3, we have calculated the average surface concentration value at each point on the grid. We then used bilinear interpolation to calculate values ​​between the points. Given the uncertainty about the amount of oil that will be spilled, we have converted the average surface concentration values ​​from absolute values ​​(average surface oil thickness, measured in meters) to relative values ​​(average surface oil thickness relative to other exposed areas, measured in percentiles).

Each simulation also had 1,000 uncertainty particles simulating the oil track through parameter settings that assume extreme weather conditions. According to the GNOME documentation, the area enclosed by the uncertainty particles represents where about 90% of the trajectories are expected to fall51. We calculated the area enclosed by the uncertainty particles by calculating the convex hull from the positions of all uncertainty particles from all 1,000 simulations and plotting it as the uncertainty area.

We repeated our spill analyzes, using the spill duration, the Season of the year, grid resolution, and number of particles varied to assess how the model output would change. Our original models assumed a 7-day spilling; We repeated the models for a 24 hour spill. We repeated the spill models for spring (March – May) and fall (September – November). We have also resimulated the leaks with a coarser spatial resolution (latitude and longitude rounded to 2 decimal places, which is approximately 1.1 km corresponds) and higher number of particles (10,000 instead of 1,000) (Supplementary Fig. 3â ???? 5).

In order to calculate the fate of the oil, we used the NOAA tool for automatic data retrieval of oil contamination. We used data on the type of oil, grid-shaped winds, grid-shaped currents, water salinity and water temperature as inputs. We modeled the oil fate for two scenarios: one without cleaning (only evaporation) and one with extremely optimistic cleaning conditions. We have limited the running time of all weathering models to six days, as they do not take into account longer-term factors such as biodegradation or photo-oxidation, which can influence the weathering rate47. To get a series of evaporation estimates, we ran 1,000 Monte Carlo simulations by randomly varying the time and date. We calculated the mean and the 95% uncertainty intervals, defined as the 2.5. and 97.5. Percentile across all simulations. For the cleanup analyzes, we made the optimistic assumptions that (1) the cleanup would begin immediately after the oil spill and (2) oil production over six days would be comparable to the 2010 Deepwater Horizon oil spill cleanup – high. We modeled a skimmer with a recovery rate of 14 barrels per hour and an efficiency of 100%, an in-situ incineration with an area of ​​70,000 m² and an efficiency of 50% and a 15% oil spray with dispersant at an efficiency of 20 %. at 29 ° C and 42% salinity. The cleaning parameters were chosen so that they correspond to the cleaning rate in the Deepwater Horizon oil spill. We used the amount of oil recovered by various cleaning methods in the Deepwater Horizon oil spill, scaled it from its 85 day cleanup timeframe to our 6 day timeframe, and selected the cleanup parameters that would allow this estimated recovery of oil volume over 6 days52. The weather values ​​were chosen to match the values ​​from our previous Monte Carlo simulation analysis, which maximized oil evaporation.

For air pollution modeling, we used the Hybrid Single-Particle Lagrangeian Integrated Trajectory (HYSPLIT) model from NOAA53. HYSPLIT is a widely used framework for modeling atmospheric transport and dispersion that uses a hybrid Eulerian / Lagrange approach to calculate the trajectory of airborne particles as well as pollutant concentrations. We performed simulations in four scenarios with different seasons (summer and winter) and spill duration (24 hours for a fast release spill and 72 hours for a slow release spill). The duration of the burial corresponds to the duration of the pollutant emission. We used HYSPLIT’s daily run feature to run simulations every three hours from January to March and June to August. Each simulation ran for 144 hours with pollutants emitted at sea level. Each simulation had 2,500 particles with a concentration value for each particle. For each combination of season and duration of the leak, we have calculated the average pollutant concentration value at each point on a latitude-longitude grid rounded to three decimal places. We used bilinear interpolation to calculate values ​​between points. We assumed a complete leak of 150,000 tons in all scenarios.

To obtain air pollution values ​​in the form of fine dust, we converted the initial oil release from barrels to micrograms, multiplied by that of Middlebrook et al multiplied by that of the model estimated concentration values. We calculated the population-weighted average increased risk for cardiovascular and respiratory hospital stays from air pollution by multiplying the air pollution values ​​by the increased risk and population percentage at each interpolated grid point. Our risk function for PM2.5 exposure during cardiovascular and respiratory hospital admissions was derived from Burnett et al.12. The risk was averaged over person-days in order to compare it over different spill durations. We calculated the person-days by calculating the number of people exposed to at least 10 µg µm3 PM2.5 and the duration of exposure.

Due to the uncertainty in the rate of particulate conversion and the increased risk of cardiovascular and respiratory mortality through fine dust, we repeated this process through Monte Carlo simulation and varied these two parameters over 1,000 simulations each in order to spread the uncertainty. To spread the uncertainty, we varied two parameters during the simulations: the rate of conversion of surface oil to particulate matter and the increased risk (IR) of particulate matter with regard to cardiovascular and respiratory hospital stays. Based on the point estimates and confidence intervals (CIs) from the available literature, we adopted the methodology of Khomenko et al.55 and calculated the IR standard deviations as follows (where qnorm is the inverse of the cumulative standard normal distribution function):

We took normal distributions and took samples from the reported mean and the estimated standard deviation. The standard deviation for the particulate matter conversion rate is given, so we used the reported estimate. We calculated the mean IR from our simulations and constructed 95% uncertainty intervals. If the lower uncertainty interval for the simulated IR was less than 0, we did not assume an increased risk of hospitalization. We also calculated the mean value and the 95% uncertainty intervals for the affected population over time, measured in person-days.

We repeated our air pollutant analyzes under different scenarios. In addition to modeling 24-hour emissions and 72-hour emissions during different seasons, we have modeled air pollution with and without combustion. For combustion air pollution estimates, we added the rate of conversion of burned oil to particulate from Middlebrook et al.54 to the existing rate of evaporation. We also varied the function of increased risk, as the estimate used by Burnett et al.12 may not reflect the population in Yemen demographically. Therefore, we performed the above calculations using the respiratory hospitalization rates from Wei et al.11 and short-term PM2.5-related mortality rates from Kloog et al.10. Wei et al.11 used a Medicare population (mostly 65 years old) who are more vulnerable to air pollution, which may reflect some subgroups of the Yemeni population, many of whom are malnourished and lack adequate health care. Kloog et al.10 used complete mortality records from the state of Massachusetts, used more modern methods than Burnett et al.12, and can more accurately reflect the younger age structure of Yemen than the Medicare population studied by Wei et al.11. We present the results in terms of increased hospitalization risk instead of increased hospitalization risk, since the increased risk we calculated is relative to a baseline level of hospitalization risk that we are not aware of for the population of Yemen (Supplementary Tables 4 and 5).

We estimated the fuel interruption by calculating the average and 95% uncertainty intervals of monthly fuel imports through Red Sea ports from January to May 2020 (nâ ??? = â ???? 5) before fuel imports were restricted became. We calculated the fuel price increase due to the port closings in Al Hudaydah in November 2017 by taking the median price between diesel, gasoline and gas and comparing it between the price data from October 15, 2017 and November 15, 2017.

We estimated the disruption of desalination capacity by compiling a data set with locations and water capacities of all known plants in the region and identifying locations that were reached by the simulated oil spills (Supplementary Table 1). The total water consumption equivalents were calculated by multiplying each affected country’s share of the water volumes by their respective daily per capita consumption.

We estimated average and 95% uncertainty intervals of food disturbances based on historical data on imports into Yemeni Ports. To calculate the extent of the food aid disruption we used 2019 data showing percentages of total food aid in Yemen from Hudaydah and Aden46. We then multiplied these percentages by the average of the total population destined for food aid based on the available World Food Program situation reports from March 2020 to February 2021 (nâ ??? = â ???? 10). We used linear interpolation through the quantile algorithm in R to construct the uncertainty intervals. See supplementary Table 6 for the reported and originally calculated estimates.

We estimated the fish yield loss based on the gridded annual fish yield in the Red Sea and Gulf of Aden in 201656. We initially only filtered the fish yield for Yemen as Fishing unit and then totaled across all types of fishing sectors (artisanal, industrial, and so on). We then summed up Yemen’s fish yield across each gridded cell achieved by the simulated spillage for each week and season. By default, we have only included cells if the oil concentration in them was in the tenth percentile of the surface concentration or higher in order to exclude cells with traces of oil. To assess how thresholds would affect estimates of fish yield loss, we repeated the analysis at no threshold and a threshold of the 20th percentile (Supplementary Table 3).

For more information on research design, see the link linked to this article Nature Research Reporting Summary.

The raw data used in this study are publicly available and are described in the main text. The simulated data are available at https://doi.org/10.7910/DVN/XPESLB.

The code used to support this study is available from the author on request. Due to concerns about possible abuse in the event of ongoing conflicts, the code is not publicly available.

Kleinhaus, K. et al. A closing window of time to save a unique marine ecosystem. Front. Mar. Sci. 7, 615733 (2020).

ArticleÂ

Google Scholar®

A rusting oil tanker off the coast of Yemen is an environmental disaster that is still a long way off. Can anyone prevent that? Time (May 14, 2021); https://time.com/6048436/fso-safer-yemen-oil-tanker-disaster/

Fink, M. D. Sea blockade and the humanitarian crisis in Yemen. Neth. Int. Law Rev. 64, 291 ???? 307 (2017).

ArticleÂ

Google Scholar®

Sowers, J. & Weinthal, E. Humanitarian Challenges and the Targeting of Civil Infrastructure in the Yemen War. Int. Aff. 97, 157 ???? 177 (2021).

ArticleÂ

Google Scholar®

Al-Fareh, A. M. The Impact of the War in Yemen on Artisanal Fishing of the Red Sea (LSE, 2018); http://www.lse.ac.uk/Middle-East-Centre

Sanders, MJ & Morgan, GR Review of the Fisheries Resources of the Red Sea and Gulf of Aden FAO Fisheries Technical Paper 304 (FAO, 1989 ).

Tesfamichael, D. et al. (Ed.) The Red Sea Ecosystem and Coral Reef Fisheries of the World Vol. 2, No. 7 (Springer, 2016).

Kloog, I., Ridgway, B., Koutrakis, P., Coull, B.A. & Schwartz, J. D. Long and short term exposure to PM2.5 and mortality. Epidemiology 24, 555 ???? 561 (2013).

ArticleÂ

Google Scholar®

Wei, Y. et al. Short-term exposure to particulate matter and hospitalization risks and costs in the Medicare population: time-stratified case crossover study. Gebr. Med. J. 367, 16258 (2019).

ArticleÂ

Google Scholar®

Burnett, R.T., Smith-doiron, M., Stieb, D., Cakmak, S. & Brook, J.R. Effects of particulate and gas air pollution on cardiovascular hospital stays. Arc. Surroundings. Health 54, 130 ???? 139 (1999).

CAS
Article

Google Scholar®

Goldstein, B.D., Osofsky, H.J. & Lichtveld, M.Y. The Gulf Oil Spill. N. Engl. J. Med. 364, 1334 ??? 1348 (2011).

CAS
Article

Google Scholar®

Alexander, M. et al. The Deepwater Horizon Coast Guard Cohort Study of the Oil Spill: A Cross-sectional Study of Acute Respiratory Disease. Surroundings. Res. 162, 196 ???? 202 (2018).

CAS
Article

Google Scholar®

Liu, X. & Wirtz, K. W. The economics of oil spills: direct and indirect costs depending on the size of the oil spill. J. Danger. Mater. 171, 471 ???? 477 (2009).

CAS
Article

Google Scholar®

Huynh, B. Q. & Basu, S. Predicting the Migration Patterns of Internally Displaced Persons in Syria and Yemen. Disaster med. Preparation for public health. 14, 302â ???? 307 (2019).

ArticleÂ

Google Scholar®

Camacho, A. et al. Cholera epidemic in Yemen, 2016–18: an analysis of surveillance data. Lancet Glob. Health 6, e680â ?????? e690 (2018).

ArticleÂ

Google Scholar®

Minissale, A., Chandrasekharam, D. & Al-Dubai, MFM in Oceanographic and Biological Aspects of the Red Sea (Ed. Rasul, NMA & Stewart, ICF) 195â ???? 213 (Springer, 2019) .

El Bcheraoui, C., Jumaan, AO, Collison, ML, Daoud, F. & Mokdad, AH Health in Yemen: Losing Ground in Time of War. Globe. Health 14, 42 (2018).

ArticleÂ

Google Scholar®

Burki, T. Health situation in Yemen – from crisis to disaster “. Lancet 385, 1609 (2015).

ArticleÂ

Google Scholar®

Kimball, A. M. & Jumaan, A. Yemen: The Challenge of Providing Assistance in an Active Conflict Area. Globe. Secure. Health Sci. Directive 5, 65â ???? 70 (2020).

ArticleÂ

Google Scholar®

Moore, R. & Burns, C. M. The Effect of Oil Spills on Workers Involved in Containment and Control: The Role of the Occupational Health Nurse. AAOHN J. 59, 477 ???? 482 (2011).

Google Scholar®

Dâ ????? Andrea, M.A. & Reddy, G.K. The Development of Long-Term Negative Health Effects in Oil Pollution Workers of the Deepwater Horizon Offshore Oil Rig Disaster. Front. Public Health 6, 117 (2018).

ArticleÂ

Google Scholar®

Looi, M.-K. COVID-19: Deaths in Yemen are five times the global average as healthcare collapses. Gebr. Med. J.370, m2997 (2020).

ArticleÂ

Google Scholar®

Elnakib, S. et al. Care under Extreme Adversity: The Impact of the Yemen Conflict on the Personal and Professional Life of Health Workers. Social Science Med. 272, 113751 (2021).

Kingston, P. F. Long-term environmental effects of oil spills. Spill Sci. techn. Bull. 7, 53 ???? 61 (2002).

CAS
Article

Google Scholar®

Peterson, C.H. et al. Long term ecosystem response to the Exxon Valdez oil spill. Science 302, 2082 ???? 2086 (2003).

CAS
Article

Google Scholar®

Savary, R. et al. Fast and penetrating transcriptomic resilience and acclimatization of extremely heat tolerant coral holobionts from the northern Red Sea. Proc. Natl Acad. Science USA 118, e2023298118 (2021).

CAS
Article

Google Scholar®

Al-Majed, A.A., Adebayo, A.R. & Hossain, M. E. A Sustainable Approach to Oil Spill Control. J. Environment. Administer. 113, 213 ???? 227 (2012).

ArticleÂ

Google Scholar®

Langodan, S. et al. The climatology of the Red Sea – Part 2: The waves. Int. J. Klimatol. 37, 4518 ???? 4528 (2017).

ArticleÂ

Google Scholar®

Madah, F., Mayerle, R., Bruss, G. & Bento, J. Characteristics of Tiden in the Red Sea Region, a numerical model study. Open J. Mar. Sci. 5, 193 (2015).

ArticleÂ

Google Scholar®

Yemen: Health, environment and economy remain threatened by damaged oil tankers. UN News (June 3, 2021); https://news.un.org/en/story/2021/06/1093352

Hersbach, H. et al. The global ERA5 reanalysis. Q.J.R. Meteorol. Social 146, 1999-2049 (2020).

ArticleÂ

Google Scholar®

Chassignet, E.P. et al. The data assimilation system HYCOM (HYbrid Coordinate Ocean Model). J. Mar. Syst. 65, 60 ???? 83 (2007).

ArticleÂ

Google Scholar®

United Nations Verification & Inspection Mechanism for Yemen (UNVIM) (UNVIM Dashboard, accessed April 1, 2021); https://www.vimye.org/statistics

Bin Marshad, S. M. H. Economic Assessment of Seawater Desalination: A Case Study Analysis of the Cost of Water Production from Seawater Desalination in Saudi Arabia. PhD thesis, Heriot Watt Univ. (2014).

CAS
Article

Google Scholar®

Ahmed, A.-E.-W. Cooperation Council for the Arab Gulf States (GCC). In the Max Planck Encyclopedia of International Law (Oxford Univ. Press, 2014); https://doi.org/10.1093/law:epil/9780199231690/e604

Frenken, K. (Ed.) Irrigation in the Middle East Region in Figures: AQUASTAT Survey – 2008 (FAO, 2009).

Beegle-Krause, J. General NOAA Oil Modeling Environment (GNOME): a new spill trajectory model. In Proz. International Oil Spill Conference 865 ???? 871 (American Petroleum Institute, 2001).

Keramea, P., Spanoudaki, K., Zodiatis, G., Gikas, G. & Sylaios, G. Oil Spill Modeling: a critical overview of current trends, perspectives and challenges. J. Mar. Sci. Closely. 9, 181 (2021).

ArticleÂ

Google Scholar®

Taqi, A. M., Al-Subhi, A. M., Alsaafani, M. A. & Abdulla, C. P. Estimation of geostrophic currents in the Red Sea based on sea level anomalies derived from expanded satellite altimetry data. Ocean Sci. 15, 477 ???? 488 (2019).

ArticleÂ

Google Scholar®

Barker, C.H. et al. Advances in operational modeling to aid oil spill response. J. Mar. Sci. Closely. 8, 668 (2020).

ArticleÂ

Google Scholar®

Kerr, R. A. Lots of oil on the loose, not that much to be found. Science 329, 734 ?? 735 (2010).

CAS
Article

Google Scholar®

Draxler, R. R. & Rolph, G. D. Modeling results from the HYSPLIT Atmospheric Transport and Dispersion Model (NOAA, 2010); https://www.ready.noaa.gov/hysplit_metadata.html

Middlebrook, A.M. et al. Air quality impact of the Deepwater Horizon oil spill. Proc. Natl Acad. Science USA https://doi.org/10.1073/pnas.1110052108 (2011).

Khomenko, S. et al. Premature mortality from air pollution in European cities: a health impact assessment. Lancet Planet. Health 5, E121 ???? E134 (2021).

Palomares, M.L.D. et al. Trends in the fishing biomass of exploited fish populations in marine ecoregions, climates and sea basins. Mouth. Coast. Regal Sci. 243, 106896 (2020).

ArticleÂ

Google Scholar®

B.Q.H. acknowledges support from the National Science Foundation’s graduate research grant under grant no. DGE 1656518 and the National Library of Medicine under Training Grant T15 LM 007033. E.T.C. acknowledges support from the National Science Foundation’s graduate research grant under grant no. DGE 1656518. A.M.M. received grant from the National Institutes of Health (NIH) NIAID T32AI007433. P.G. was awarded by the National Center for Advancing Translational Sciences of the NIH under Price No. KL2TR003143. The authors are solely responsible for the content of this article and do not necessarily represent the official views of the NIH. Funding sources were not a factor in the preparation of this manuscript or the decision to submit it for publication. We thank the Refugee and Asylum-seeker Health Initiative at the University of California, San Francisco, for facilitating co-author collaboration for this work.

Department of Primary Care and Population Health, Department of Medicine, Stanford University School of Medicine, Stanford, CA, USA

Refugee and Asylum Seekers Health Initiative, Department of Medicine, University of California San Francisco, San Francisco, CA, USA

You can search for this author in as well
PubMed Google Scholar

You can also search for this author in
PubMed Google Scholar

You can also search for this author in
PubMed Google Scholar

You can also search for this author in
PubMed Google Scholar

You can also search for this author in
PubMed Google Scholar

You can also search for this author in
PubMed Google Scholar

You can also search for this author in
PubMed Google Scholar

You can also search for this author in
PubMed Google Scholar

You can also search for this author in
PubMed Google Scholar

You can also search for this author in
PubMed Google Scholar

B.Q.H. conceived the first study design, performed the analysis, and wrote the first draft. L.H.K., M.V.K., E.T.C., S.B., P.G. and D.H.R. helped revise the study design. A.M.M., A.O.J. and F.M.K. provided important contextual information. All authors contributed significantly to the interpretation of the results and to the preparation of the final version of the manuscript.

Correspondence
Benjamin Q. Huynh.

Peer-reviewed information Nature Sustainability thanks Alesia Ferguson, Tor Nordam, Raúl Periñez, and the other anonymous reviewers for contributing to the peer review of this work.

Editor’s note Springer Nature remains neutral with regard to legal claims in published maps and institutional connections.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits its use, distribution, adaptation, distribution and reproduction in any medium or format as long as you adequately identify the original author (s) and source. Include a link to the Creative Commons license and indicate if any changes have been made. The images or other third-party material in this item are included under the item’s Creative Commons license, unless otherwise stated in a line of credit for the material. If any material is not under the item’s Creative Commons license and your intended use is not permitted by law or exceeds permitted use, you must obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Huynh, B. Q., Kwong, L. H., Kiang, M. V. et al. Public health impact of an impending Red Sea oil spill.
Nat Sustain (2021). https://doi.org/10.1038/s41893-021-00774-8

Provided by the Springer Nature SharedIt content sharing initiative

Advanced Search

Nature Sustainability (Nat Sustain)

ISSN 2398-9629 (online)

Sign up for the Nature Briefing newsletter â ???? what’s important in science in your inbox every day, free of charge.

Ref: https://www.nature.com