Analyzing Ambient Air Quality Trends in Ghana

Using data from a fleet of air pollution monitors to visualize patterns in community air quality levels and explore reasons for the pollution trends.
Columbia University
Stats
R
Health
Author
Affiliation

Lewis White

Columbia University

Published

February 25, 2025

Summary

In 2023, our team set up a network of 40 low-cost air pollution monitors across Ghana’s Bono East region—spanning farming communities and the city of Kintampo (population ~111,000), where our Ghanaian research partners are based. While we ran into the typical hurdles that come with rural data collection, such as patchy network connectivity and occasional data gaps, our initial results give us some valuable insights into the area’s air quality.

The fleet of monitors recorded an average daily fine particulate matter (PM 2.5) concentration of about 49 μg/m³—a level higher than the World Health Organization’s primary PM 2.5 target of 5 μg/m3 and interim target of 35 μg/m3. We also saw clear daily spikes in pollution during cooking hours (morning and evening) and a large jump in overall pollution during the Harmattan season, a period roughly from late-November to mid-March when a cool dry wind blows dust from the Sahara over much of West Africa.

Elevated levels of particulate matter can significantly influence both maternal health and early childhood development [1][2]. The measurements shared in this report highlight the scale of this issue in Ghana’s rural farming regions and smaller urban centers, informing future epidemiological studies of air pollution risk.

Monitors

For data collection, we used QuantAQ’s ModulairPM and Modulair devices. Before installing them in each community, we colocated the monitors at the Kintampo Health Research Center (KHRC) to cross-check and adjust for any systematic biases between the units. These initial calibrations ensure consistent, high-quality data across all 40 locations. After the colocation period, we deployed the monitors to their respective communities. The maps below illustrates their distribution.

Code 
# Define the bounding box coordinates
bbox_coords <- matrix(c(-2.2, 7.6, -2.2, 8.8, -1.3, 8.8, -1.3, 7.6, -2.2, 7.6), ncol = 2, byrow = TRUE)
bbox_polygon <- st_polygon(list(bbox_coords))
bbox_sf <- st_sfc(bbox_polygon, crs = st_crs(4326))

# Plot the map with the bounding box
ghana_full <- ggplot() +
  # Add Ghana regions
  geom_sf(data = regions, fill = ifelse(regions$region == "Bono East", "#f5e493", "#dadbe0"), color = "black", alpha = 0.5) +
  # Add monitor points with transparency
  geom_sf(data = monitor_points, aes(geometry = geometry), color = "red", alpha = 0.5, size = 2) +
  # Add bounding box
  geom_sf(data = bbox_sf, fill = NA, color = "blue", lwd = 0.5, linetype = "solid") +
  # Add Ghana country boundary
  geom_sf(data = country, fill = NA, color = "black", size = 1) +
  #labs(title = "Monitor Locations in Ghana") +
  theme_bw() +
  theme(
    axis.text = element_text(size = 11), # Increase axis text size
    axis.ticks = element_line(size = 1), # Increase tick mark size
    plot.title = element_text(size = 16) 
  )


zoomed_map <- ggplot() +
  # Add Ghana regions within the bounding box
  geom_sf(data = regions, fill = ifelse(regions$region == "Bono East", "#f5e493", "#dadbe0"), color = "black", alpha = 0.5) +
  # Add monitor points with transparency within the bounding box
  geom_sf(data = monitor_points, aes(geometry = geometry, color = monitor_type), alpha = 0.5, size = 5) +
  
  # Add Kintampo marker
  geom_sf(data = kintampo_sf, aes(geometry = geometry), shape = 20, size = 3, color = "black") +
  # Add Dwenewoho marker
  geom_sf(data = dwenewoho_sf, aes(geometry = geometry), shape = 20, size = 3, color = "black") +
  # Add new longoro marker
  geom_sf(data = new_longoro_sf, aes(geometry = geometry), shape = 20, size = 3, color = "black") +
  # Add apesika marker
  geom_sf(data = apesika_sf, aes(geometry = geometry), shape = 20, size = 3, color = "black") +
  # Add kurawura akura marker
  geom_sf(data = kurawura_akura_sf, aes(geometry = geometry), shape = 20, size = 3, color = "black") +
  
  # Add Kintamp annotation text
  annotate("text", x = -1.73, y = 8.10, label = "Kintampo", color = "black", size = 5) +
  # Add Dwenewoho annotation text
  annotate("text", x = -1.69, y = 7.74, label = "Dwenewoho", color = "black", size = 5) +
  # Add new longoro annotation text
  annotate("text", x = -2.047, y = 8.214, label = "New Longoro", color = "black", size = 5) +
  # Add apesika annotation text
  annotate("text", x = -1.459, y = 7.96, label = "Apesika", color = "black", size = 5) +
  # Add atta akura annotation text
  annotate("text", x = -1.65, y = 8.75, label = "Kurawura Akura", color = "black", size = 5) +
  
  # Zoom in to the bounding box
  coord_sf(xlim = c(-2.2, -1.3), ylim = c(7.6, 8.8), expand = FALSE) +
  labs(x = "", 
       y = "",
       color = "Monitor Type") +
  theme_bw() +
  theme(
    axis.text = element_text(size = 11), # Increase axis text size
    axis.ticks = element_line(size = 0.5), # Increase tick mark size
    plot.title = element_text(size = 16), # Increase title size
    legend.text = element_text(size = 10), # Increase legend text size
    legend.title = element_text(size = 12) # Increase legend title size
  )

# Create a joint title
title <- textGrob("Monitor Locations in Ghana", gp = gpar(fontsize = 20))

# Arrange plots with title
grid.arrange(
  arrangeGrob(ghana_full, zoomed_map, ncol = 2),  # Arrange plots
  top = title  # Add title at the top
)

Air Quality Index (AQI) Context

Before diving into the results, I wanted to provide some context on air quality measurements. The United States Environmental Protection Agency (U.S. EPA) Air Quality Index (AQI) is a scale that measures pollution levels in the air. The scale is divided into six categories, with each associated with a different color and level of concern. While the final score is calculated using five major air pollutants—particulate matter, ozone, carbon monoxide, sulfur dioxide, and nitrogen dioxide—this analysis will focus on PM 2.5, fine particulate matter with a diameter of 2.5 micrometers or less.

The table below shows the EPA’s cutoffs [3] for each health concern category, including both the standard AQI values (the numbers commonly seen in air quality reports and weather apps) and the specific PM 2.5 concentration ranges that contribute to the index.

AQI Category and Index Value Corresponding fine particle (PM 2.5) Measurement (µg/m³)
Good (0 – 50) 0.0 to 9.0
Moderate (51 – 100) 9.1 to 35.4
Unhealthy for Sensitive Groups (101 – 150) 35.5 to 55.4
Unhealthy (151 – 200) 55.5 to 125.4
Very Unhealthy (201 – 300) 125.5 to 225.4
Hazardous (301+) 225.5+

Results

PM 2.5 Concentrations Over Time

Code 
# Define AQI levels with ordered factors
aqi_levels <- data.frame(
  ymin = c(0, 9.1, 35.5, 55.5, 125.5, 225.5),
  ymax = c(9, 35.4, 55.4, 125.4, 225.4, 450),
  fill = factor(c("Good", "Moderate", "Unhealthy for Sensitive Groups", 
                  "Unhealthy", "Very Unhealthy", "Hazardous"), 
                levels = c("Good", "Moderate", "Unhealthy for Sensitive Groups", 
                           "Unhealthy", "Very Unhealthy", "Hazardous")),
  color = c("green", "yellow", "orange", "red", "purple", "maroon")
)

# Plot with legend inside the graph
daily_pollutants_long %>%
  filter(pollutant == "Mean PM 2.5") %>% 
  ggplot(aes(x = date, y = measurement)) +
  # Add background rectangles for AQI levels
  geom_rect(data = aqi_levels, 
            aes(xmin = min(daily_pollutants_long$date), 
                xmax = max(daily_pollutants_long$date), 
                ymin = ymin, 
                ymax = ymax, 
                fill = fill), 
            color = NA, alpha = 0.2, inherit.aes = FALSE) +
  # Add the PM 2.5 line
  geom_line(color = "black") +
  theme_bw() +
  labs(title = "Daily Mean PM 2.5 Suggests Levels of Concern for Sensitive Groups Year Round \nwith Unhealthy Air Quality for All During the Harmattan.",
       x = "Date",
       y = bquote("Fleet Mean " ~ PM[2.5] ~ (µg/m^3)),
       fill = "US-EPA AQI Risk Categories") +  # Add legend title for the AQI levels
  theme(
    plot.title = element_text(size=20), # set title size 
    axis.title = element_text(size = 16),   # Increase axis titles size
    axis.text = element_text(size = 14),    # Increase axis labels size
    strip.text = element_text(size = 14),    # Increase facet group text size
    legend.position = c(0.77, 0.73),  # Position legend inside the graph (top right)
    legend.background = element_rect(fill = rgb(1, 1, 1, alpha = 0.7), color = "black"),
    legend.title = element_text(size = 14),  # Adjust legend title size
    legend.text = element_text(size = 12)    # Adjust legend text size
  ) +
  scale_fill_manual(values = aqi_levels$color) +
  guides(fill = guide_legend(reverse = TRUE))  # Reverse the order of the legend

The figure above shows that PM 2.5 levels often exceed the ‘good’ to ‘moderate’ AQI range, particularly from December through February—roughly the Harmattan season.

We see similar trends for other sizes of particulate matter, as illustrated in the three-panel plot below. The first panel highlights PM1 concentrations (mean 37 µg/m³), the second panel shows PM 2.5 (mean 43.2 µg/m³), and the third panel displays PM10 (mean 101 µg/m³). All three exhibit noticeable peaks during the dry, dusty Harmattan months, suggesting that the same sources (such as windblown dust) drive increases across all particle sizes.

Code 
daily_pollutants_long %>%
  ggplot(aes(x = date, y = measurement)) +
  geom_line() +
  geom_hline(aes(yintercept = mean_measurement), color = "#cc1414", linetype = "dashed") +
  geom_text(data = mean_values, aes(x = max(daily_pollutants_long$date), y = mean_measurement, label = paste0(pollutant, ": ", round(mean_measurement, 2), " µg/m³")),
            color = "#cc1414", hjust = 1, vjust = -0.9, size = 4.5) +
  theme_bw() +
  labs(title = "Grand Averages Patterns Consistent Across Particle Sizes",
       x = "Date",
       y = "Pollutant Measurement (µg/m³)") +
  facet_grid(factor(pollutant, levels = c("Mean PM 1", "Mean PM 2.5", "Mean PM 10" )) ~ ., scales = "free_y") +
  theme(
    plot.title = element_text(size = 20),
    axis.title = element_text(size = 16),   # Increase axis titles size
    axis.text = element_text(size = 14),    # Increase axis labels size
    strip.text = element_text(size = 14)  # Increase facet group text size
  )

Hourly PM 2.5 Results

Code 
pm25_community_hourly <- pm25_community_hourly %>%
    mutate(date = ymd(date))  # Use lubridate to ensure date conversion
  
heat_map_data <- pm25_community_hourly %>%
    group_by(date, hour) %>%
    summarize(mean_pm25 = mean(mean_pm25, na.rm = TRUE), .groups = 'drop') %>%
    ungroup() %>%
    mutate(day = day(date),
           month = factor(format(date, "%b"), levels = c("Sep", "Oct", "Nov", "Dec", "Jan", "Feb", "Mar", "Apr", "May", "Jun", "Jul", "Aug")),
           year = year(date)) %>%
  mutate(log_mean_pm25 = log(mean_pm25))


ggplot(heat_map_data, aes(day, hour, fill = mean_pm25)) +
    geom_tile(color = NA, size = 0) + 
    scale_fill_gradientn(
        colors = c("#221f26", "#3f2363", "#872CA2", "#D44292", "#F98477", "#EDD9A1", "#fffceb"), # Specify your custom colors
        values = scales::rescale(c(0, 5, 12, 20, 40, 60, 100)), # Set breaks; adjust these values based on your data range
        name = "PM 2.5 \n(µg/m³)"
    ) + 
    facet_grid(~month) +
    scale_y_continuous(trans = "reverse", breaks = unique(heat_map_data$hour)) +
    scale_x_continuous(breaks = c(1, 15, 30)) +
    theme_minimal(base_size = 8) +
    labs(title = "Heatmap Shows Strong Seasonal and Daily Patterns in PM 2.5",
         x = "Day",
         y = "Hour") + 
    theme(legend.position = "right",
          plot.title = element_text(size = 20, hjust = 0),
          axis.text.y = element_text(size = 12),
          strip.background = element_rect(colour = "white"),
          strip.text = element_text(size = 15),   # Increase facet group text size
          axis.ticks = element_blank(),
          axis.text = element_text(size = 9),
          legend.title = element_text(size = 14),
          legend.text = element_text(size = 13),
          axis.title.y = element_text(size = 16),
          axis.title.x = element_text(size = 16),
          panel.background = element_blank())

The heatmap illustrates how both season and time of day affect PM 2.5 concentrations across all 40 monitors. During the Harmattan season (December–February), PM 2.5 levels stay elevated, peaking in the early morning (6–8 a.m.) and early evening (5–9 p.m.). These spikes likely coincide with cooking and other daily activities that generate or stir up particulate matter. As the Harmattan subsides around March, average PM levels drop, and remain lower into the late spring and early summer months.

Code 
base_plot <- pm25_community_hourly %>% 
  mutate(harmattan = ifelse(date >= harmattan_start & date <= harmattan_end, "Harmattan", "Not Harmattan")) %>%
  mutate(
    cooking_period = factor(case_when(
      hour >= 16 & hour <= 19 ~ "Evening Cooking Hours", #4-9pm
      hour >= 5 & hour <= 8 ~ "Morning Cooking Hours", #5-9am
      (hour < 16 | hour > 19) &  (hour < 5 | hour > 8) & hour != 3 ~ "Other Hours",
      hour == 3 ~ "Control Hour"
    ), levels = c("Evening Cooking Hours", "Morning Cooking Hours", "Control Hour", "Other Hours"))
  ) %>%
  filter(!is.na(hour)) %>%
  ggplot(aes(x = harmattan, y = mean_pm25, fill = cooking_period)) +
  geom_boxplot() +
  theme_classic() +
  labs(y = "PM 2.5 Concentration", 
       x = "Season",
       fill = "Cooking\nPeriod") +  # Change legend title and add line breaks
  theme(legend.position = "right") +
  scale_fill_manual(values = c(
    "Evening Cooking Hours" = "goldenrod", 
    "Morning Cooking Hours" = "#7db569", 
    "Control Hour" = "#b672e0",
    "Other Hours" = "#52b3eb"
  ),
  labels = c("Evening Cooking Hours (4pm-8pm)", "Morning Cooking Hours (5am-9am)", "Control Hour (3am)", "Other Hours (Excluding Cooking \nand Control Hours)"))  # Add line breaks in legend labels


base_plot +
  geom_segment(aes(x = 0.72, xend = 0.90, y = 515, yend = 515), color = "black") +  # Harmattan 
  annotate("text", x = 0.81, y = 518, label = "***") +
  
  geom_segment(aes(x = 0.91, xend = 1.09, y = 515, yend = 515), color = "black") +  # Harmattan 
  annotate("text", x = 1, y = 518, label = "***") +
  
  geom_segment(aes(x = 1.10, xend = 1.28, y = 515, yend = 515), color = "black") +  # Harmattan 
  annotate("text", x = 1.19, y = 518, label = "**") +
  
  geom_segment(aes(x = 0.72, xend = 1.09, y = 525, yend = 525), color = "black") +  # Harmattan 
  annotate("text", x = 0.9, y = 529, label = "***") +
  
  geom_segment(aes(x = 0.91, xend = 1.27, y = 536, yend = 536), color = "black") +  # Harmattan
  annotate("text", x = 1.09, y = 540, label = "***") +
  
  geom_segment(aes(x = 0.72, xend = 1.27, y = 548, yend = 548), color = "black") +  # Harmattan
  annotate("text", x = 1, y = 552, label = "***") +
  
  
  
  geom_segment(aes(x = 1.72, xend = 1.90, y = 360, yend = 360), color = "black") +  # Not Harmattan 
  annotate("text", x = 1.81, y = 364, label = "***") +
  
  geom_segment(aes(x = 1.91, xend = 2.09, y = 360, yend = 360), color = "black") +  # Not Harmattan 
  annotate("text", x = 2.0, y = 364, label = "***") +
  
  geom_segment(aes(x = 2.10, xend = 2.28, y = 360, yend = 360), color = "black") +  # Not Harmattan 
  annotate("text", x = 2.19, y = 364, label = "***") +
  
  geom_segment(aes(x = 1.72, xend = 2.09, y = 372, yend = 372), color = "black") +  # Not Harmattan 
  annotate("text", x = 1.9, y = 376, label = "***") +
  
  geom_segment(aes(x = 1.91, xend = 2.28, y = 384, yend = 384), color = "black") +  # Not Harmattan 
  annotate("text", x = 2.09, y = 388, label = "***") +
  
  geom_segment(aes(x = 1.72, xend = 2.27, y = 396, yend = 396), color = "black") +  # Not Harmattan 
  annotate("text", x = 2, y = 400, label = "***") +
  
  
  geom_segment(aes(x = 1, xend = 2, y = 590, yend = 590), color = "black") +  # Harmattan vs Not Harmattan
  annotate("text", x = 1.5, y = 600, label = "***") +
  
  labs(title = "Higher Ambient PM 2.5 Concentrations During Cooking Periods Persist Across Season",
    caption ="Comparisons and their significance are denoted by the horizontal lines: \n* indicates p <0.05, ** indicates p < 0.01, and *** indicates p < 0.001") +
  
  theme(
    plot.title = element_text(size = 20),
    axis.text = element_text(size = 13), # Increase axis text size
    axis.ticks = element_line(size = 0.7), # Increase tick mark size
    legend.text = element_text(size = 12), # Increase legend text size
    legend.title = element_text(size = 14), # Increase legend title size
    plot.caption = element_text(size = 13),
    axis.title.x = element_text(size = 15),
    axis.title.y = element_text(size = 15),
  )

The boxplots above further examine the relationship between seasonal Harmattan, peak cooking times, and air pollution. For this analysis, Harmattan was defined from December 1st to March 1st, morning cooking hours between 5–9am, and evening cooking hours between 4–8pm, based on our research partners’ local knowledge of the region. As a comparison to the cooking hours, I looked at a control hour (3am) in the middle of the night as it should be completely unaffected by human behavior, as well as a group of ‘other hours’ that excluded the cooking hours and the control hour, leaving me with a total of 4 groups.

I utilized ANOVAs to test for differences in the time groupings for the entire data, and then for each season separately. The results were highly significant for each test, so I moved forward with Tukey’s HSD to further examine the differences in means for each group. Partially due to the large number of observations in the data (and thus small standard deviations), all differences between groups were highly significant.

Regarding seasonality, as expected, PM 2.5 concentration was higher (M = 84, SD = 60) on average during Harmattan months than non-Harmattan months (M = 26, SD = 24). During the Harmattan, evening cooking hours (M = 131, SD = 77) were associated with the highest pollution levels, followed by morning cooking hours (M = 95, SD = 64), with other hours (M = 74, SD = 48) and the control hour (M = 70, SD = 45) experiencing lower levels. The same order held outside of Harmattan, with evening cooking hours (M = 47, SD = 34) associated with the highest PM 2.5 concentration, then morning cooking hours (M = 25, SD = 19), then other hours (M = 21, SD = 19), and finally the control hour (M = 16, SD = 13).

Spatial Variability

Code 
# DURING HARMATTAN ----

#PM 2.5
pm25_summary_harmattan <- pm25_corrected %>%
  filter(date >= harmattan_start & date < harmattan_end) %>%
  group_by(monitor) %>%
  summarise(mean_pm25 = mean(pm25, na.rm = TRUE),
            median_pm25 = median(pm25, na.rm = TRUE)) %>%
  right_join(monitor_points)


harmattan_spatial_plot <- ggplot() +
  # Add Ghana regions within the bounding box
  geom_sf(data = regions, fill = ifelse(regions$region == "Bono East", "#f5e493", "#dadbe0"), color = "black", alpha = 0.5)  +
  
  # Add road data
  geom_sf(data = roads_filtered, color = "gray", size = 0.3) +  # Add roads data
  
  # Add monitor points with transparency within the bounding box
  geom_sf(data = pm25_summary_harmattan, aes(geometry = geometry, color = mean_pm25), alpha = 0.7, size = 6) +

  # Zoom in to the bounding box
  coord_sf(xlim = c(-2.2, -1.3), ylim = c(7.6, 8.8), expand = FALSE) +
  
  # Customize
  labs(title = "Mean PM 2.5 for Each Monitor During Harmattan",
       x = "", 
       y = "") +
  theme_bw() +
  scale_color_gradient(low = "#0540a1", high = "#eb4c2d",
                       breaks = c(min(pm25_summary_harmattan$mean_pm25), 85, 110, max(pm25_summary_harmattan$mean_pm25)),
                       labels = c(paste(round(min(pm25_summary_harmattan$mean_pm25),0)), "85", "110", paste(round(max(pm25_summary_harmattan$mean_pm25),0))),
                       name = "PM 2.5 \n(µg/m³)") +
  theme(
    axis.text = element_text(size = 11), # Increase axis text size
    axis.ticks = element_line(size = 0.7), # Increase tick mark size
    plot.title = element_text(size = 14), # Increase plot title size
    legend.text = element_text(size = 12), # Increase legend text size
    legend.title = element_text(size = 12), # Increase legend title size
    legend.key.size = unit(1.3, "lines") # Increase legend tick mark size
  )


# OUTSIDE OF HARMATTAN ----

pm25_summary_not_harmattan <- pm25_corrected %>%
  filter(date < harmattan_start | date >= harmattan_end) %>%
  group_by(monitor) %>%
  summarise(mean_pm25 = mean(pm25, na.rm = TRUE),
            median_pm25 = median(pm25, na.rm = TRUE)) %>%
  right_join(monitor_points)

not_harmattan_spatial_plot <- ggplot() +
  # Add Ghana regions within the bounding box
  geom_sf(data = regions, fill = ifelse(regions$region == "Bono East", "#f5e493", "#dadbe0"), color = "black", alpha = 0.5)  +
  
  # Add road data
  geom_sf(data = roads_filtered, color = "gray", size = 0.3) +  # Add roads data
  
  # Add monitor points with transparency within the bounding box
  geom_sf(data = pm25_summary_not_harmattan, aes(geometry = geometry, color = mean_pm25), alpha = 0.7, size = 6) +

  # Zoom in to the bounding box
  coord_sf(xlim = c(-2.2, -1.3), ylim = c(7.6, 8.8), expand = FALSE) +
  
  # Customize
  labs(title = "Mean PM 2.5 for Each Monitor Outside of Harmattan",
       x = "", 
       y = "") +
  theme_bw() +
  scale_color_gradient2(low = "#0540a1", mid = "#eb4c2d", high = "goldenrod", 
                        midpoint = 55,  # Set the midpoint value
                        breaks = c(min(pm25_summary_not_harmattan$mean_pm25, na.rm = TRUE),35, 55, max(pm25_summary_not_harmattan$mean_pm25)),
                        labels = c(paste(round(min(pm25_summary_not_harmattan$mean_pm25),0)),"35", "55", paste(round(max(pm25_summary_not_harmattan$mean_pm25),0))),
                        name = "PM 2.5 \n(µg/m³)") +
  theme(
    axis.text = element_text(size = 11), # Increase axis text size
    axis.ticks = element_line(size = 0.7), # Increase tick mark size
    plot.title = element_text(size = 14), # Increase plot title size
    legend.text = element_text(size = 12), # Increase legend text size
    legend.title = element_text(size = 12), # Increase legend title size
    legend.key.size = unit(1.3, "lines") # Increase legend tick mark size
  )

# Create a joint title
title <- textGrob("Maps of Pollution Levels for PM 2.5 Show Spatial Variability in the Monitors.", gp = gpar(fontsize = 20))

# Arrange plots with title
grid.arrange(
  arrangeGrob(harmattan_spatial_plot, not_harmattan_spatial_plot, ncol = 2),  # Arrange plots
  top = title  # Add title at the top
)

The maps above display the mean PM 2.5 levels for each monitor in the study area. The plots show that PM concentrations vary across the region, with certain monitors recording higher mean pollutant levels. These spatial patterns suggest that local environmental factors and/or human activities may influence pollution levels, leading to the observed variations in PM concentrations.

While there is variation in pollution levels across the communities, I used Moran’s I test to look for spatial correlation and found no evidence that the variation can be explained by the location alone. I also tried running regressions to see if proximity to road, type of road, and population of the community could explain the variation, but found no evidence of an association with any of these predictors.

Conclusions

These results underscore the significant day-to-day and seasonal fluctuations in air quality across Ghana’s Bono East region, particularly during the Harmattan months and around cooking hours. They point to an urgent need for both seasonal and local interventions—whether through cleaner fuels, improved ventilation, or other strategies—to reduce PM 2.5 exposure.

Building on these insights, my ongoing research at Columbia focuses on supporting communities as they transition from charcoal to cleaner-burning LPG, which holds promise for substantially lowering pollution and improving maternal and child health outcomes. This evidence strengthens the case for adopting cleaner cooking practices and highlights the value of continued monitoring to guide effective, data-driven solutions.

References

[1]: Quinn, A. K., Adjei, I. A., Ae-Ngibise, K. A., Agyei, O., Boamah-Kaali, E. A., Burkart, K., Carrión, D., Chillrud, S. N., Gould, C. F., Gyaase, S., Jack, D. W., Kaali, S., Kinney, P. L., Lee, A. G., Mujtaba, M. N., Oppong, F. B., Owusu-Agyei, S., Yawson, A., Wylie, B. J., & Asante, K. P. (2021). Prenatal household air pollutant exposure is associated with reduced size and gestational age at birth among a cohort of Ghanaian infants. Environment International, 155, 106659.https://doi.org/10.1016/j.envint.2021.106659

[2]: Lacasaña, M., Esplugues, A., & Ballester, F. (2005). Exposure to ambient air pollution and prenatal and early childhood health effects. European Journal of Epidemiology, 20(2), 183–199.https://doi.org/10.1007/s10654-004-3005-9

[3] EPA. Final Updates to the Air Quality Index (AQI) for Particulate Matter Fact Sheet and Common Questions - Policy Brief. Available from: https://www.epa.gov/system/files/documents/2024-02/pm-naaqs-air-quality-index-fact-sheet.pdfttps://www.epa.gov/system/files/documents/2024-02/pm-naaqs-air-quality-index-fact-sheet.pdf

Citation

BibTeX citation:
@online{white2025,
  author = {White, Lewis},
  title = {Analyzing {Ambient} {Air} {Quality} {Trends} in {Ghana}},
  date = {2025-02-25},
  url = {https://lewis-r-white.github.io/posts/2025-01-26-heat-stress/},
  langid = {en}
}
For attribution, please cite this work as:
White, Lewis. 2025. “Analyzing Ambient Air Quality Trends in Ghana.” February 25, 2025. https://lewis-r-white.github.io/posts/2025-01-26-heat-stress/.