Assessing Urban Growth Using NDVI in the Greater Seattle Metropolitan Area
Remote sensing has been used to determine vegetative biomass and plant communities in a variety of regions. One method of determining the vegetative biomass in a community is through the normalized difference vegetation index.
INTRODUCTION
Remote sensing has been used to determine vegetative biomass and plant communities in a variety of regions. One method of determining the vegetative biomass in a community is through the normalized difference vegetation index (NDVI). NDVI compares different multispectral bands to reveal where vegetation is. By understanding where vegetative and non-vegetative materials are one can show where urban growth is occurring. The change in NDVI in a certain temporal scale indicates urban growth as indicated by Howarth and Boassan (1983).
Seattle has been experiencing increasing population growth since 1974 (Robinson et al. 2005). This growth continued to increase in the 90’s and early 2000 as high-tech jobs brought more people to the region (Glaese and Kahn. 2001). This resulted in the demand for more housing and infrastructure resulting in suburban sprawl. Urban growth and suburban sprawl results in less green spaces as vegetation is removed in place of housing and other developments. Fewer green spaces can cause reduced biodiversity, loss of habitat, and habitat fragmentation (Liu et al. 2015) Knowing where urban growth is occurring can help land managers manage existing green spaces and predict where future environmental stresses could begin.
Objectives:
The objective of this study is to assess urban growth over an extended period using Landsat 5 data. This urban growth was assessed by looking at the difference in NDVI for the Greater Seattle Metropolitan Area located in Washington State. The study focused on a time period of 27 years with images captured in June and July.
During a 27-year time frame, the NDVI should show a decrease in the Greater Seattle Metropolitan area compared to natural regions outside of the cities. This hypothesis is based on the NDVI models from Ramadan et al. (2004) in which change in NDVI is a representation of urban growth.
BACKGROUND AND DESCRIPTION OF STUDY AREA
The area of focus for this study is around Seattle which is located along the Puget Sound in Washington State. For this study the Seattle Metropolitan Area is considered the region from the city of Tacoma north to the city of Everett and east to the foothills of the Cascade Mountains. This region of interest can be seen in figure (1).
DESCRIPTION OF DATA AND METHODS
The data used for this study was obtained from Landsat 5. Landsat 5 was deployed in 1984 and retired in 2013. The satellite carried a multispectral scanner (MSS) and Thematic Mapper (TM) which span 7 spectral bands for analysis (USGS a). The bands available from Landsat 5 along with their description and wavelengths can be seen in table 1. It orbited the earth with a 16 day repeat cycle at 705 km altitude.
| Band Number | Description | Wavelength μm |
| 1 | Blue | 0.45-0.52 |
| 2 | Green | 0.52-0.6 |
| 3 | Red | 0.63-0.69 |
| 4 | Near infrared | 0.76-0.90 |
| 5 | Short-wave infrared | 1.55-1.75 |
| 6 | Thermal | 10.4-12.3 |
| 7 | Short-wave infrared | 2.08-2.35 |
Table 1: Landsat 5 spectral Bands (USGS b)
For the study two images for the years 1984 and 2011 are used to show the urban growth for the study area during a time of great growth. The images were captured in the months of June and July to allow for peak plant growth in the region. NDVI is calculated for each of the images revealing the vegetative biomass for the region in the given years (Figure 2).
The NDVI formula was applied to the Landsat 5 data bands using band math in ENVI. This calculation results in an image showing NDVI. The NDVI image from 2011 and 1984 can be seen in figure 3. The two images are then subtracted from another to create a change in NDVI image. This change of NDVI will reveals where the most change of vegetative biomass has occurred. The image from 2011 from 1984 using (band math) in ENVI. This resulted in the image seen in Figure 4. Using the subtracted image of the area a new color classification was added to emphasize the changes.
RESULTS
In order to analysis urban growth for the Seattle region NDVI was calculated for the year 1984 and 2011. Data was captured in the months of June and July to allow for peak plant growth in the region. This also helped to reduce seasonal variation in plant growth for the region. The resulting output images for NDVI can be seen in figure 3. These images show regions of high NDVI as white and low NDVI areas as black. High NDVI values are associated with spaces with high vegetative biomass. The change in NDVI between 1984 and 2011 was needed to evaluate if urban growth was accruing in the region. The two NDVI images subtracted show the relative difference between NDVI. The raw resulting change in NDVI image can be seen in figure 4. To better visually represent this information a modified color classification was added to the raw subtracted image. The modified color classification image can be seen in figure 5. This color image emphasizes the change in NDVI for the region.
DISCUSSION
The individual NDVI images for the year 1984 and 2011 show the regions that have high vegetation and low vegetation for the given year. Regions that appear black have low vegetation and areas of white have high amounts of vegetation. Comparing them side by side as seen in figure 3 shows that between the given years subtle changes have occurred. More black regions are visible as you look at the 2011 image than in the 1984 image. The changes are more apparent when the two images are subtracted. The raw subtracted NDVI image is a black and white scale image that visually shows the change in values image and can be seen in figure 4. This subtracted image reveals where vegetation biomass has increased as well as decreased in the 27-year time span. Areas where the vegetation biomass change has grown appears black and areas with decreased vegetation appears white. Regions of grey represent where a change in NDVI did not occur.
To better visually represent this information a modified color classification was added to the raw subtracted image. The modified color classification image can be seen in figure 5. This color image emphasizes the change in NDVI for the region. Areas with red show the decrease in NDVI and areas with green show an increase in NDVI. Grey still represents regions where no NDVI change occurred. The image shows a decrease in vegetation biomass around the Seattle city metropolitan and spreading and expanding out from the city centers. There are also some small patches of increased vegetation on the outskirts of the metropolitan which corresponds to agricultural or timberlands. Between 1950 and 1970 harvest from logging increased resulting in clear cuts in this study region (Forest Service. 1996). These clear-cut regions correspond to many of the green regions in figure 5 as forests regenerated. Further analysis of each of the negative NDVI value patches would help to determine exactly what is occurring. It would be beneficial to further classify the identified patches of negative NDVI to assess whether patches are due to urbanization or a more natural change (e.g. fire) as done by Kennedy et al. 2105.
Possible short comings of the project:
To increase the accuracy of determining where urban growth has occurred in the region of interest additional use of various sensors would be needed. Investigating regions that displayed low NDVI scores using surface mineralogy applications to determine the soil or substrates in these regions. This would help to identify if the low NDVI regions are indeed human influenced or a more natural phenomenon.
One source of error of this study could be in using only 1 day of satellite data for 1984 and 2011. Using an average NDVI over a couple months could reduce the chance of false readings. By averaging the sensor data, you would minimize errors due to possible phenological cycles (e.g. Drought) for the chosen years in the study. Another source of error could be due to atmospheric conditions. Atmospheric conditions could result in anomalous values that create false results. Atmospheric conditions can be filtered to alleviate these negative influences
CONCLUSION
This project examined urban growth in the Greater Seattle Metropolitan area between a 27-year span of time. This 27-year time span between 1984 and 2011 represented a time period of rapidly expanding population for the Seattle Metropolitan region. NDVI was used in order to capture this rapid population growth and urban development to observe where vegetation and green spaces were being lost. The analysis for the change in NDVI values between 1984 and 2011 revealed urban development expanding outward from the city centers into regions where human development had been nonexistent or minimal previously. Using difference indexes in conjunction with NDVI would improve accuracy in assessing urban growth.
References:
Forest Service. U.S. Department of Agriculture. (1996). Status of the interior Columbia Basin: Summary of scientific findings. General Technical Report (GTR), pnw-gtr-385, 55. doi:10.2737/pnw-gtr-385
Howarth, P. J., & Boasson, E. (1983). Landsat digital enhancements for change detection in urban environments. Remote Sensing of Environment, 13(2), 149-160. doi:10.1016/0034-4257(83)90019-6
Glaeser, E., & Kahn, M. (2001). Decentralized Employment and the Transformation of the American City. Brookings-Wharton Papers on Urban Affairs 2. doi:10.3386/w8117
Kennedy, R. E. Yang, Z. Braaten, J. Copass, C. Antonova, N. Jordan, C. and Nelson P. (2015). Attribution of disturbance change agent from Landsat time-series in support of habitat monitoring in the Puget Sound region, USA. Remote Sensing of Environment, 166(C), 271–285. https://doi.org/10.1016/j.rse.2015.05.005
Ramadan, E., Feng, X., & Cheng, Z. (2004). Satellite remote sensing for urban growth assessment in Shaoxing City, Zhejiang Province. Journal of Zhejiang University-SCIENCE A, 5(9), 1095-1101. doi:10.1631/jzus.2004.1095
Robinson, L., Newell, J. P., & Marzluff, J. M. (2005). Twenty-five years of sprawl in the Seattle region: Growth management responses and implications for conservation. Landscape and Urban Planning, 71(1), 51-72. doi:10.1016/j.landurbplan.2004.02.005
USGS. (n.d.a). Landsat Missions: Landsat 5. Retrieved from https://www.usgs.gov/land-resources/nli/landsat/landsat-5?qt-science_support_page_related_con=0#qt-science_support_page_relate
USGS (n.d. b). What are the band designations for the Landsat satellites?. Retrieved from https://www.usgs.gov/faqs/what-are-band-designations-landsat-satellites-0?qt-news_science_products=7#qt-news_science_products