An evaluation of Terra-MODIS sampling for monthly and annual particulate matter air quality assessment over the Southeastern United States

Abstract

Although satellites provide reliable and repeated measurements on a global basis, particulate matter air quality information can be derived from satellites only when clouds are absent and when surface conditions are favorable. However, ground measurements provide particulate matter information irrespective of cloud cover and surface conditions. Therefore there could be a sampling bias when using satellite data for air quality research. To examine this issue, we calculate particulate matter (PM2.5) mass concentration from daily ground-based measurements (ALLPM) on monthly to yearly time scales and compare these against the same ground measurements for only those days when satellite data is available (SATPM). To accomplish this, we use six years of PM2.5 mass concentration data from 38 stations along with Terra-MODIS satellite data over the Southeastern United States. Our results indicate that satellite data are generally available less than 50% of the time over these locations, although the interregional variability of data availability is between 32% and 57%. However, the mean differences between the ALLPM and SATPM, over monthly to yearly time scales over the Southeastern United States, is less than 2 μgm⁻³ indicating that low sampling from satellites due to cloud cover and other reasons is not a major problem for studies that require long term PM2.5 data sets. These results have important implications for satellite studies especially over areas where ground-based measurements are not available.

Introduction

Particulate matter (PM) is a mixture of both solid and liquid particles suspended in air and is usually classified as fine (PM2.5, d < 2.5 μm) and coarse (PM10, 2.5 < d < 10 μm), where d is the aerodynamic diameter. In this paper, we are primarily concerned with PM2.5 that could be from various sources including dust, vehicle and industrial emissions, forest and agricultural fires. PM2.5 air quality continues to degrade throughout the world due to increasing pressures of urbanization that has serious implications for health, climate, visibility, and hydrology (Kaufman et al., 2002). Although some countries have a dense network of PM2.5 monitoring stations (Al-Saadi et al., 2005), worldwide, there are limited ground measurements thereby creating a challenge for monitoring and studying air pollution. With the launch of Terra and Aqua polar orbiting satellites, there has been an increased emphasis for using satellite data to study PM2.5 to alleviate some of the problems due to the unavailability of ground measurements (Al-Saadi et al., 2005). While satellites can provide reliable, repeated measurements from space, monitoring surface level air pollution continues to be a challenge since most satellite measurements are column-integrated quantities. However, several studies have shown that satellite data can be a good surrogate for ground measurements provided appropriate adjustments are made for converting columnar quantities to surface values (van Donkelaar et al., 2006, Liu et al., 2004). With new satellites that can now provide vertical distribution of aerosols and clouds, we are poised to make significant advances in using satellite data for particulate matter air quality research (Engel-Cox et al., 2006).

The link between PM exposure and adverse health recently prompted the United States Environmental Protection Agency (EPA) to tighten its 24-h fine particle standard from 65 μgm⁻³ to 35 μgm⁻³ (Federal Register, 2006). Studies show that long term particulate matter exposures are associated with death due to heart failure, and cardiac arrest (Pope et al., 2002). However, it is difficult to obtain long term estimates in large spatial scales from the limited number of ground measurements and therefore the use of satellite data could be beneficial.

Several research papers have outlined the methods by which satellite data can be used to obtain surface PM2.5 (e.g. Wang and Christopher, 2003, Engel-Cox et al., 2006, Hutchison et al., 2005, Gupta et al., 2006, Liu et al., 2004, van Donkelaar et al., 2006). In summary, first the columnar satellite-derived aerosol optical depth (AOD) values are related to surface PM2.5 mass measurements. Then this AOD-PM2.5 relationship can be used to convert the satellite measurements to air quality indices based on EPA guidelines. These values are then color coded for dissemination to the public where Green is for Good air quality and Orange and Red are poor quality. A good example of this can be seen at http://alg.umbc.edu/usaq/.

Given the links between PM2.5 and health, and the scarcity of monitoring stations throughout the world, satellite remote sensing appears to be the only viable method to monitor PM2.5 air pollution over large spatial scales. However, satellite retrievals of AOD rely on cloud-free conditions and favorable surface conditions to obtain PM2.5 air quality thereby limiting the number of days where satellite data can be used over a certain location. Also satellite retrievals are sometimes not available due to various retrieval issues such as bright surface backgrounds and data dropouts. What the satellite retrievals lose in terms of cloud cover limitations, it makes up in terms of the wide spatial coverage that is often useful for assessing how the pollution plumes move from one area to another (Hoff et al., 2005). Even with these limitations, satellite data sets due to their global coverage are a valuable asset for monitoring PM2.5 air quality (Gupta et al., 2007).

In contrast, ground measurements of PM2.5 are available regardless of cloud cover and depending upon the location, measurements are made available every hour or as 24-h averages. While this is extremely useful, ground measurements are limited due to lack of spatial coverage or unavailability. Since continuous monitoring of PM2.5 is essential and monthly and annual averages of PM2.5 air quality is vital for assessing global air quality, it is important to assess whether satellite data can provide the sampling necessary to monitor air quality over these time scales. We assume that satellite data sets are good surrogate for monitoring surface PM2.5 air quality while recognizing that there are indeed some research limitations that are currently being addressed using new satellite data, meteorology, and other tools (e.g. Engel-Cox et al., 2006).

Since PM2.5 mass is measured from the ground irrespective of cloud cover while satellite data only provide AOD information during cloud-free and favorable retrieval conditions, we ask the following questions, ‘What is the difference between ground-based PM2.5 (ALLPM) and the PM2.5 for only those days where satellite data are available (SATPM) on monthly and yearly time scales?’ How many days of satellite data are available due to cloud cover contamination and other limitations for PM2.5 air quality research? Understanding these differences are important to address the utility of satellite data in mapping PM2.5 air quality over monthly and yearly time scales especially since long term exposure studies require global data sets on yearly time scales (Pope and Dockery, 2006). Note that we are not using the satellite-derived AOT in this paper, rather we simply examine the PM2.5 during the time of the satellite overpass. To examine this issue, we selected the EPA region 4 in Southeast United States (Fig. 1) where previous research has shown that satellite data is indeed a robust surrogate for PM2.5 estimation (Wang and Christopher, 2003). This region was also selected due to the numerous ground-based PM2.5 measurements that are available to address the aforementioned questions.