Shanghai Urban Expansion, 1996-2008-2017
This post discovers the land cover change in Shanghai, China from 1996 to 2008 and to 2017. I chose this project for two reasons. One, I wanted to explore urban expansion in the last 20 years of one of the world’s largest cities because this information can be used to improve the city’s sustainable development. Second, I chose to study Shanghai because this city has played a large role in my life. In 1996, my parents moved from their post in Uzbekistan to the US. On their journey, they boarded the Trans-Continental Railroad and made their way to Beijing, and later Shanghai. In 2008, I went to Shanghai to visit my close family friends. Finally, in 2017 I worked in Shanghai. Thus, I wanted to track how Shanghai has changed between my visits. I downloaded images from Shanghai, China in row 118, column 38. I had one image from 1996, one from 2008, and the last from 2017. I used supervised classification to study the percentages of land cover types in each image. I looked at water, urban cover, lush farmland (appears bright green in satellite images), farmland (appears as a forest green crop in satellite images), bare soil, and urban greenery (parks and plants in Shanghai). The satellite images themselves provide strong evidence that the urban area of Shanghai has rapidly grown, while the classified images show concretely where man-made structures and buildings have increased around the city. II. Methods I obtained the satellite images from the United States Geological Survey. I searched Shanghai, China and narrowed the search to row 118, column 38. I then filtered the search by 1996, then 2008, and finally 2017. For each year, I looked for an image unobstructed by cloud cover. Furthermore, I ensured that I found images taken in April of each year. This way, seasonal blooms would show the location of vegetation around the city. Had I chosen images taken during the winter, the leaf-less vegetation would create a challenge when classifying the land cover. To avoid this confusion, I chose to only look at spring-time images. Next, I downloaded each image with the full range of bands. One at a time, I open each image in Envi. I stacked the bands and then showed each image in True Color. Because the images from 1996 and 2008 used Landsat 5, I I designated Red as band 3, G as band 2, and B as band 1. For the image taken in 2017, Red took band 4, G took band 3, and B took band 1. After stacking the images, I created Regions of Interest (ROIs). I created at least 8 ROIs per category on each image to maximize accuracy. In total, I had 6 categories for each image: water, urban cover, lush farmland, farmland, bare soil, and urban greenery. After saving the XML and shapefile of each ROI, I created a maximum classification. Because bands 2, 6, and 7 were the least correlated, I selected those bands in the classification tool. Finally, I loaded each map with the land classification types.
III. Results Images from 1996
This image shows Shanghai in 1996. Notice the lush farmland that dominated the majority of the land.
These images show the land cover types from the image in 1996. Notice the dominance of lush farmland and farmland from the image. Notably, 2,370,339 pixels make up urban areas.
This satellite image shows clearly the expansion of Shanghai’s buildings since 1996. The white-grey areas represent urban areas.
Urban expansion clearly occurred, with 4,332,870 pixels representing urban areas. This is nearly double the number of urban area pixels from 1996.
The vast amount of buildings in this image demonstrates a rapid urban expansion since 1996 and 2008.
In the 2017 image, 32,438,979 pixels represent urban areas. This is nearly eight times what it was in 2008. However, though the urban expansion has been quite high, I believe that the cloud cover at the bottom right of the image increased the urban pixel count. Thus, I believe the actual percentage of urban areas is lower than in this image, but it is still quite high.
These images clearly demonstrate a massive increase in urban areas between 1996 and 2017. In addition, the percentage of all farmland has decreased in the areas around Shanghai, China. In 1996, farmland plus lush farmland took up 3,426,438 pixels. In 2017 the same land cover types took up 1,972,763 pixels. This displacement of farmland around Shanghai presents the question of where Shanghai sources its food now, and what has happened to the farmers since 1996. Further studies can use these images to investigate food production and urban cover in Shanghai. In addition, as Shanghai sits on the coast, these maps can aid in urban development with regards to climate change and rising sea levels. This information can aid the city’s adaptation to changing times!
This investigation encountered several difficulties. For one, the 1996 and 2008 images presented discernment challenges. The pixels and images were less sharp than those of 2017, so classifying pixels into ROIs posed a challenge or the earlier images. This could have led to false- classifications of different pixels. Furthermore, I had trouble finding an image from April 2017 without cloud cover. This presented a severe challenge in that Envi classified the cloud cover as urban area each time I attempted a Maximum Likelihood Classification. While urban expansion did indeed occur in Shanghai between 2996 and 2017, the classification percentages likely inflated in this investigation due to classification error.
V. References All data sourced from USGS in Shanghai, China. Images: 1996 ● LT05_L1TP_118038_20080425_20170613_01_T1_ 2008 ● LT05_L1TP_118038_19960424_20170105_01_T1_ 2017 ● LC08_L1TP_118038_20170418_20170501_01_T1_