Spatial and Temporal Patterns of PM2.5 in Santiago, Chile€¦ · 3D kriging using time stamp as Z...

Post on 18-Jul-2020

0 views 0 download

Transcript of Spatial and Temporal Patterns of PM2.5 in Santiago, Chile€¦ · 3D kriging using time stamp as Z...

Spatial and temporal patterns of PM2.5in Santiago, Chile

Carolina Magri, Penn State MGIS

Presentation Overview• Introduction

• Objectives

• Exploratory Analysis

• Kriging

• Emerging Hot Spot Analysis

• Conclusions

What is PM2.5?

• Particulate matter or PM is a common air pollutant that consists of a mix of solid and liquid particles that varies by location.

• It is described by its mass concentration (micrograms per cubic meter, μg/m3) and is classified according particle diameter

• PM2.5 are particles with diameters less than 2.5 μm.• World Health Organization guidelines:

• Annual average not above 10 μg/m3.• 24-hour mean less than 25 μg/m3, not to be exceeded for more

than 3 days/year.

Health Impacts• Short term and long term exposure of PM2.5 leads to

cardiovascular and respiratory diseases as well as lung cancer.

• It also increases morbidity in patients with preexisting lung or heart diseases, the elderly and children.

• In 2015, papers published in the British Medical Journal showed that high concentrations of PM2.5 are positively correlated with strokes, high anxiety symptoms and premature aging of the brain.

• The black carbon portion of PM2.5 is considered a known carcinogen and a major contributor to global climate change.

The Study Area

• Since 1997 Chile has implemented decontamination plans that have reduced the levels of air pollution in Santiago, but the daily and annual averages of 50 and 20 µg/m3 for PM2.5 are still exceeded.

• In 2005, the Chilean Ministry of the Environment and the World Health Organization estimated that 4,000 premature deaths are the consequence of atmospheric pollution, costing the country 670 million dollars in medical expenses and loss of productivity.

• The major sources of PM2.5 in Santiago are:• Industry and agriculture (33%)• Residential burning of wood (22%, but reaching 30% in winter time)• Other vehicles (36%, including 3% attributable to private vehicles)• Public transportation (8%)• Between 10% and 20% of PM2.5 in Santiago is Black Carbon.

• Examine the spatial and temporal patterns of hourly PM2.5concentration using basic statistics and semivariograms.

• Explore relationships between PM2.5 and environmental factors such as wind speed and direction, relative humidity, air temperature and elevation.

• Interpolate PM2.5 concentrations using 3D kriging.• Identify areas of concentration and levels of exposure that

exceed recommended levels. • Explore spatial and temporal variations of the concentration

throughout the city using Emerging Hot Spot Analysis .

Objetives

Data Source

• 11 measurement stations.

• 8,784 hourly measurements of PM2.5 and weather conditions taken between January 1 and December 31, 2016. (http://sinca.mma.gob.cl/)

Data explorationSUBHEAD INFORMATION

Time Series plot for each station

Summary statistics for all stations

Average 26 to 33

Max 142 to 580

Standard Deviation 17 to 36

Average January -April 14 to 24

Average May-August 38 to 56

Average September -December 14 to 23

General Relative Semivariogram

Main cycle of 24 hours and a second cycle of 10 to 12 hours, probably related to traffic.

Gamma gives an indication of the variability associated with each lag and represents the variance in PM2.5concentrations through time.

Cycles

Regression analysis• Linear regressions were run for each station and then for all

stations together, based on hourly measurements.

• PM2.5 was the dependent variable.

• Relative humidity, temperature, wind speed, wind direction and elevation were used as independent variables.

• The number of nearby polluting sources, pollution per day and distance to main roads were added when the model for all stations was run.

Linear regression model for all stations

* Significant p-value < 0.01

Variable Coefficient Range of values

Adjusted R2 0.113

Intercept 39.47*

Relative Humidity -0.004 2.7 to 105 %

Temperature -1.001* -5 to 37 ºC

Wind Speed -1.62* 0 to 20 m/sec

Wind Direction -0.003* 0 to 359 º

Elevation 0.008* 309 to 789 masl

Number of pollution sources 0.014* 4 to 398

Pollution per day -0.009* 5.5 to 509 kg/day

Distance to main roads 0.005* 173 to 2295 mts.

KrigingSUBHEAD INFORMATION

Experimental Semivariograms3D kriging using time stamp as Z coordinate

Omni directional horizontal Vertical

Semivariograms models

Omni directional horizontal

Vertical

Header for Demo SlideSupporting Text

Kriged surfaces

Emerging Hot Spot AnalysisSUBHEAD INFORMATION

Space Time Cube• 500 x 500 meters and 1 hour

intervals• 7,979 locations• 8,784 time steps intervals (hours)• Area of 50,500 mts. West to East

and 39,500 mts. North to South• 70,087,536 space time bins• Decreasing overall trend of the

mean of PM2.5

Emerging Hot Spot Results

Explanation of the results• Count is constant

• PM2.5 concentrations are cyclical

• Significant trends are difficult to identify because the data is highly variable through time (daily and by season).

• Some categories of the hot spot analysis are based on behavior in the last time step of the cube relative to previous ones. This makes it hard for the tool to identify hot spots in cyclical data such as PM2.5 concentrations

Population density and income level maps

Conclusions PM2.5 concentrations vary through the day and seasonally.

Concentrations are specially high in winter months and most variable in spring.

Variographic analysis shows that PM2.5 concentrations are highly correlated in time and space. General relative semivariograms clearly show that PM2.5 concentration had a main 24 hour cycle and a secondary cycle of 10 to 12 hours for some stations, probably related to traffic patterns.

Kriging surfaces show areas where PM2.5 concentrations are highest, which tend to be in the western part of the city. These areas have higher population density and lower income levels.

Emerging Hot Spot analysis found zones in the PM2.5 concentrations and identified oscillating hot and cold spots.

Environmental variables have negative coefficients but very weak predictive or explanatory power. Highest coefficients were for temperature and wind speed.