Jump to comment:

• Comments on “Children living in areas with more street trees have lower prevalence of asthma”
  Paul A Zandbergen
  Published on: 27 April 2016

• Published on: 27 April 2016

  Comments on “Children living in areas with more street trees have lower prevalence of asthma”

  • Paul A Zandbergen, Associate Professor

  Lovasi et al. [1]) document the relationship between the density of street trees and the prevalence of childhood asthma in New York City. Their findings suggest street trees are associated with a lower prevalence, although no causality was inferred. I would like to point out a number of methodological issues which should benefit future studies on this subject.

  Prevalence of asthma was determined for 4-year-old a...

  Show More

  Lovasi et al. [1]) document the relationship between the density of street trees and the prevalence of childhood asthma in New York City. Their findings suggest street trees are associated with a lower prevalence, although no causality was inferred. I would like to point out a number of methodological issues which should benefit future studies on this subject.

  Prevalence of asthma was determined for 4-year-old and 5-year-old children using data from school screenings in 1999. Street tree density was derived from the 1995 street tree census completed by the Parks and Recreation Department of the City of New York and expressed as the total number of trees on streets segments divided by the land area. Data was aggregated at the level of United Hospital Fund (UHF) areas. Additional variables used in the analysis were population density, racial/ethnic composition and a measure of proximity to pollution sources. The initial correlation analysis suggested a negative association between street tree density and prevalence of asthma. However, one of the strongest positive associations was between street tree density and population density. This initially appears somewhat counterintuitive, until it is recognized exactly which types of trees are included in the analysis. The street tree census conducted by the Parks and Recreation Department of the City of New York only considered trees along city streets, and trees in parks and open space are not included. As a result, the street tree density derived by Lovasi et al. [1] is a substantial underestimate of the actual number of trees within a UHF area, in particular in areas with large areas of parks and open space. In fact, Nowak et al. [2] estimate that New York City has about 5.2 million trees, while the latest street tree census in 2005-2006 counted 592,130 trees [3]. Street trees, therefore, account only for around 11% of trees within the study area. The strong correlation between street tree density and population density is strongly driven by the fact that the total length of streets segments per unit area increases with population density. Logic suggests that overall tree density and population density are likely negatively associated since many parks and open space occur in areas with lower (average) population density. The pattern in street tree density by UHF areas is therefore a poor representation of the overall tree density.

  Figure 1 below provides an example of a park in Brooklyn adjacent to a residential area. While numerous trees are visible on the residential streets, the number of trees in the park far exceeds the number of street trees. While Figure 1 is not representative for the entire study area, it illustrates how not including trees in parks and open space presents a misleading picture of the potential effects of trees on local air quality in urban areas

  Figure 1 Digital orthophoto of a portion of Prospect Park in Booklyn, New York City adjacent to a residential area.

  Source: United States Geological Survey 2006.

  

  The argument could be made that street trees are more relevant than those in parks and open space since street trees are much closer to the residential homes. However, Lovasi et al [1] do not make this argument and instead aggregate all variables at the level of UHFs. This aggregation does not allow for a determination of street tree density in close proximity to the residential addresses of asthma cases. If street trees in close proximity are deemed of greater relevance than trees in parks and open spaces at greater distances, an individual or street segment level analysis is required.

  A second methodological issue relates to the determination of the measure of proximity to pollution sources. Relying on the methodology presented by Maantay [4], Lovasi et al. [1] create uniform distance buffers around toxic release inventory sites, stationary point sources and major truck routes and then determine the percentage of each UHF falling within one or more of these buffers. While the specific distances and types of sources were derived from the Maantay [4] study, the authors fail to highlight the many limitations of this approach as detailed at length in the original study, including the use of single buffer distances, treating all pollution sources as being similar, and ignoring cumulative effects from multiple sources. Perhaps more importantly, Maantay [4] used the individual geocoded residential locations of asthma hospitalization cases and determined if they fell within a particular buffer or not. Lovasi et al. [1] instead determine the prevalence of asthma as a rate based on the number of children within each UHF and compared this to the percentage of the area of the UHF falling within one of more of the buffers around the pollution sources, without considering the proximity of individual cases to pollution sources. The data aggregation to the level of UHFs represents a very substantial loss of information. No evidence is presented that aggregation at the level of UHF areas is justified given the nature of the research question since it remains unclear at what (spatial) scale the potential effects of street trees on air quality are expected to occur.

  Future research efforts in this area should consider the following three refinements:
  1) Developing a more robust measure of tree density which includes trees in parks and open spaces. This could be addressed by using land use or land cover maps supplemented with field sampling as employed by Nowak et al. [2].
  2) Using individual level analysis instead of aggregation to coarse units. This would involve geocoding individual address locations of asthma cases and creating individual-level measures of tree density and proximity to pollution sources, as well as creating a meaningful sample of non-asthma cases for comparison.
  3) Employing more robust measures of proximity to pollution sources. One approach to accomplishing this is to use cumulative distribution functions as employed by Waller et al. [5] and Zandbergen and Chakraborty [6]. While each of these three elements requires considerable effort, they should contribute to a much improved understanding of the complex relationships between tree density and asthma prevalence.

  Paul Zandbergen
  University of New Mexico
  Department of Geography

  References

  1. Lovasi GS, Quin JW, Neckerman KM, Perzanowksi MS Rundle A. 2008. Children living in areas with more street trees have lower prevalence of asthma. J Epidemiol Community Health 2008;62:647-649.

  2. Nowak DJ, Hoehn III RE, Crane D E, Stevens JC, Walton JT. Assessing urban forest effects and values, New York City's urban forest. Resource Bulletin NRS-9. United States Department of Agriculture, Forest Service, Northern Research Station, 2007.

  3. Peper PJ, Mcpherson EG, Simpson JR, Gardner SL, Vargas KE, Xiao Q. New York City, New York municipal forest resource analysis. Center for Urban Forest Research, United States Department of Agriculture, Forest Service, Pacific Southwest Research Station, 2007.

  4. Maantay J. Asthma and air pollution in the Bronx: methodological and data considerations in using GIS for environmental justice and health research. Health Place 2007;13:32-56

  5. Waller LA, Louis TA, Carlin BP. 2007. Environmental justice and statistical summaries of differences in exposure distributions. J Exposure Analysis and Env Epi 1999;9:56–65.

  6. Zandbergen PA, Chakraborty J. Improving environmental exposure analysis using cumulative distribution functions and individual geocoding. Int J Health Geographics 2006;5:23.

  Show Less

  Conflict of Interest:
  None declared.

  • Back to top