Introduction

        A LUR model of ambient nitrogen dioxides (NO₂) was developed for Metro Vancouver in 2003 and it has been recently updated to 2009/10. Both models followed the same methodology, except input variables, i.e. 2003 model census 2001 data while 2009/10 model used data from census 2006. It was thought that the differences between the modeled surfaces should reveal change in the pollution level over time, as well as shift in the pollution patterns. However, the resulting maps showed unexpected artifacts which were caused by different predictor variables included in the models, rather than real change in the physical world. This section includes maps showing change in pollution surface predicted by LUR models from 2003 to 2009/10, as well as interpretation of the maps.

Methods

         Two LUR models have been developed for 2003 and 2009/10 respectively (referred as 2003 model and 2009/10 model-1). And maps have been rendered from the two models. Details for the model development can be found at Henderson et al., 2007. Variables and coefficients included in each model are listed in Table 1. To visualize the change over time, the 2003 surface was subtracted from the 2009/10 surface-1, using 'Spatial Analyst Tools - Math - Minus'.

        However, the resulting surface showed artifact caused by different predictor variables included in the two models, thus another 2009/10 model (named as 2009/10 model-2) was built using the same set of variables that was used in 2003.  And a second surface for 2009/10 (2009/10 surface-2) was mapped and compared with the 2003 surface to generate another image of the change in the modeled pollution surface over time.  Variables and coefficients of the second 2009/10 model are also included in Table 1.

Table 1. Variables with associated coefficients in the three models

* Variables are presented in Name.Buffersize. For sample: Highway.100 means the total length of highway within a 100m buffer

Results

Difference between the modeld 2003 surface and 2009/10 surface-1

        The map shows the difference in modeled pollutant concentrations from 2003 to 2009/10. Overall, pollution levels have been significantly reduced. The darker the green color is, the larger the reduction. However, some misleading scenes, however, were noticed. In the map, there are blue lines along road sides (not roads). These blue lines are not real reduction in the pollution surfaces, but as a result of different predictor variables included in the two models. The 2009/10 model-1 includes 'total length of major road within a 100m buffer' as one predictor variable while the 2003 model included the variable with a larger buffer size (200m). The width of those lines on map was measured as 100m, equal to the difference between the two buffer sizes. Thus, it is unquestionable that the difference in the buffer size caused the illusionary "reduction" on the road sides.   Another artifact is the trend along the latitude that the reduction increases as the latitude decreases. This is also caused by the difference in predictor variables included in the two models. Latitude is included in 2003 model but not the 2009/10 model

jiohtops

Difference between 2003 surface and 2009/10 surface-2 (same variables)

         This map shows the difference between the modeled 2009/10 surface-2 and 2003 surface, using the same set of variables. Compared with the previous map, the "road edge" effect was eliminated. However, other "changes" shown in the previous map, disappeared as well.  The influence by latitude/longitude is more distinct in the second map.

Discussion

        This part demonstrated the power of visualization. It revealed some aspects that otherwise might be missed and greatly facilitated understanding. The unrealistic trend shown along the latitude indicates change in regional patter that is not explained by the LUR model. As a predictor variable, longitude/latitude represents other potentially associated variables (but not tested in the LUR model). Nevertheless, an overall reduction in pollution surface, particularly on roads, is evident in the maps.

Content last updated: Dec 9, 2010