DigitalGlobe MDA Radiant Solutions SSL

Detecting 30-Year Urban Change Using Persistent Change Monitoring

Dr. François G. F. Smith, Senior Principal Scientist, Radiant Solutions
Apr 11 2018

More than half of the world’s population resides in urban areas—a concentration expected to grow to almost 70 percent by 2050. The ability to see, understand, and anticipate urban change across the globe provides important insights, which can be used in urban planning and infrastructure development to ensure people living in these areas have access to vital services and disaster response.

At Radiant Solutions, we’ve developed a change detection algorithm for satellite imagery that can analyze change in urban centers throughout the world. In this blog, I will share an example of a change detection study I did of Houston, Texas, and the insights it revealed about the connections between weather and development in that location over time.

Decades of Landsat Imagery

Houston has undergone enormous change in the past 30 years, and has grown immensely around the metropolitan area. Much of the farmland and forest that had surrounded the city have been replaced by residential areas. Some of these areas are particularly prone to flooding, but have nevertheless been developed into neighborhoods.

To locate instances of urban change in order to understand anticipate vulnerabilities created by that change , I analyzed imagery collected by the Landsat program, which launched its first satellite in 1972 and is the longest-running initiative to collect Earth imagery. Leveraging Radiant’s Persistent Change Monitoring (PCM) algorithm, I used 30 years of Landsat imagery (1987 to present) over two Landsat footprints covering the metropolitan Houston area. To begin, I looked at the spatial and temporal distribution of change in the area, and see where human development is occurring within that distribution.

Persistent Change Monitoring

“Persistent change” refers to any change to the Earth’s landmass that becomes permanent. New building construction is an example of persistent change, while vegetation varying with the seasons is not. Persistent Change Monitoring (PCM) is Radiant’s primary technology for identifying persistent change. Performed over a time-series of imagery, PCM delineates areas where change has persisted in at least three dates of imagery collection, and flags the changes based on the date they occurred. Although all change types are recorded during the process, the final PCM product specifically reports change associated with urban features. PCM’s primary purpose is to monitor human development anywhere in the world spatially and through time using satellite imagery.

Featuring products with 30 x 30 meter resolution and imagery dating back to 1987, Radiant PCM determines persistent change by confirming that change is constant for at least three dates of imagery. To do that, it examines three images prior and three images after a target date to verify the change is real (and not just sensor or algorithmic noise or a passing cloud). PCM would not detect maximum flood waters, but it could identify damage to human settlement caused by flooding or other disasters.

Radiant Solutions developed several derived products from PCM, including the National Urban Change Indicator (NUCI). NUCI is a persistent change index of urban areas across the United States and is used by the Federal Emergency Management Agency (FEMA) to support its Risk Mapping Assessment and Planning, known as Risk MAP.

What PCM Reveals About Houston

We ran a PCM for Houston using data from 1987 to 2016 as part of the NUCI product. The results are shown in Figure 1. The white outline is the perimeter of the city of Houston. The colored pixels represent land-cover changes observed in urban-associated features sometime during the respective year. Magenta and blue covers mainly the 1990s. Yellow and red indicates more recent changes, mostly in the last decade.

At this scale, we can see a predictable pattern of outward growth over time. Notice how more land is changed the further out we go from the center of Houston until it quickly tapers off. This is a classic example of urban expansion. The rate of expansion appears to vary over time. We observe intense yellow to red colors at the outermost rim of the city, more so to the west side.

We can see when major roads leading into the center of Houston were built or expanded. We can also get a sense of how much of the porous vegetated surface surrounding the city has become impervious to water. The increase in impervious surface equates to an increase in the magnitude of precipitation runoff, which increases the risk of flooding.

Figure 1. NUCI results for Houston run on 30-plus years of Landsat imagery. Notice the legend on the left side of the image matching the year’s changes with pixel color in the NUCI layer.

 

Focusing on Four Years

As Figure 1 shows, Houston has changed a lot over the past 30 years. To visualize the type of historical urban development in the Houston area that PCM can detect, let’s narrow our look to four years—1996, 2001, 2006 and 2011—and observe the change at the pixel level using the NUCI dataset. I chose these years because there were many Landsat scenes collected in that timeframe by USGS for its land-cover mapping efforts. These years will have high temporal datasets for more accurate change detection.

1996

The flood of October 1994 had a substantial impact on Houston. When the remnants of Hurricane Rosa hit the area, the results were catastrophic. At least 17 people died. A ruptured pipeline spewed gas fires into the San Jacinto River, which then transported the flames downstream and set houses and cars on fire. Houston rebounded after the flood with considerable new development all around the metro region.

Figure 2 shows NUCI’s PCM results for 1996. These results include all changes observed in imagery collected that year. Note the especially concentrated areas of development, such as just east of Cypress, northeast of Sugar Land, and along the southwest coast of Houston Lake.

Figure 2. Houston NUCI change results for 1996. This is only change observed in imagery collected in 1996 specific to urban development.

 

Figure 3a. Image collected  December 30, 1995 of Cypress in northwest Houston.

Figure 3b. NUCI PCM results for the same area from October 11, 1996.

 

Figure 3c. Image collected December 30, 1996 showing urban development in an area identified as change by NUCI.

 

 

 

Figure 4a. Imagery collected January 14, 1995 of a bayside plot in the Clear Lake Shores area.

Figure 4b. NUCI results for the plot showing a continuous area of change in the area, detected March 10, 1996.

 

Figure 4c. Imagery collected on December 30, 2002 of the same area showing substantial development. This was the first high-resolution scene collected after 1996 we could find on Google Earth.

 

 

 

2001

On June 4, 2001, Tropical Storm Allison hit the coast of Texas. Allison was a long-lasting storm that dropped heavy amounts of precipitation along the coast, and especially on Houston. It peaked at 40 inches of rain, and the heavy flooding that resulted destroyed more than 2,700 homes, damaged 70,000 more, and killed 23 people in Texas alone. Allison continued to wreak havoc in Louisiana, the southeast states and up the mid-Atlantic, with severe damage reported as far away as Pennsylvania.

 

Allison caused $8.5 billion in damage even though it never reached hurricane strength. In the wake of the deluge and destruction of property, Houston continued building and rebuilding in all directions. There is the very obvious development of the George Bush Airport in North Houston, as well as some forest clearing north of the airport to make way for residential development.

Figure 5. Houston NUCI change results for 2001. This is only change observed in imagery collected in 2001 specific to urban development.

 

Figures 6a-6d show an area south of Houston called Pearland where NUCI identified change. Upon investigation, we see this is the start of construction of a new residential subdivision.

Figure 6a. Imagery collected January 18, 1995 for an area in Pearland, south of Houston.

Figure 6b. Area determined as change by NUCI on July 22, 2001.

Figure 6c. Image of the change area collected on December 30, 2002.

Figure 6d. Image of the change area collected on March 27, 2015 showing the residential area fully developed.

 

 

The largest apparent contiguous urban change during this year was the clearing of vegetation and paving at the George Bush Intercontinental Airport in the northern part of the city. Figures 7a-7c show the paving and expansion of the airport. Notice the areas between the runways are grass in 1995 and are paved in 2001. Natural vegetation was also removed during that same time to make way for additional airport infrastructure.

Figure 7a. George Bush Intercontinental Airport, January 22, 1995.

Figure 7b. Area determined as change by NUCI on September 24, 2001.

 

Figure 7c. The same area on April 26, 2002 showing cleared vegetation and new paved surfaces.

 

 

 

2006

Torrential downpours on June 19, 2006 caused major flooding throughout the Houston area. Up to 11 inches of rain fell and approximately 3,000 homes were flooded. However, as we see in Figure 8, there was still lots of urban development occurring in the same general area as in the previous two years.

Figure 8. Houston NUCI change results for 2006. This is only change observed in imagery collected in 2006 specific to urban development.

 

In Figures 9 and 10, we see typical examples of urban development for residential and industrial use observed in satellite imagery collected in 2006.

Figure 9a. Imagery collected in October 21, 2005 in the Bayview area southeast of Houston.

Figure 9b. NUCI results, change detected on November 1, 2006.

Figure 9c. Imagery collected on January 14, 2006 showing land cleared for development.

Figure 9d. Imagery collected on April 29, 2006 showing continued development.

Figure 10a. Imagery collected October 21, 2005 of Cummings, southwest of Houston.

Figure 10b. NUCI results, detected on December 16, 2006.

Figure 10c. Imagery collected April 29, 2006 showing land cleared for residential development.

Figure 10d. Imagery collected January 30, 2009 showing the final residential development.

 

 

2011

Texas had one of its driest stretches in recorded state history in 2011. Judging by Figure 11, there was a slowdown in urban development compared to the previous years we observed. At this scale, we can see there were some small concentrations of construction just south of Katy, south of Sugar Land, and south of the Woodlands.

Figure 11. Houston NUCI change results for the year 2011. This is only change observed in imagery collected in 2011 specific to urban development.

 

In Figures 12 and 13, we see some road expansions.

Figure 12a. Image from March 10, 2011 showing a small farm road near Richmond, southwest of Houston.

Figure 12b. NUCI observes road development October 29, 2011.

 

Figure 12c. Image from April 21, 2012 shows completed road construction.

 

 

 

Figure 13a. Image from March10, 2011 shows construction has not yet begun at Highway 58 and Croix Road.

Figure 13b. NUCI PCM results show road expansion was observed on August 27, 2011.

 

Figure 13c. Image from October 27, 2012 shows completed road.

 

 

 

In Figures 14a-14d, we see probably the largest concentration of new development observed by NUCI in 2011, south of Katy just west of Cinco Rancho.

Figure 14a. Imagery collected February 5, 2010 in the west Cinco Rancho area, near Katy and west of Houston.

Figure 14b. NUCI results for the same area detected October 29, 2011.

Figure 14c. Imagery collected January 27, 2011 showing a concentration of development.

Figure 14d. Imagery collected April 21, 2012 showing continued residential development.

 

 

What We Learned

We can surmise from these results that in some years there is more urban development than in other years. To find out why, I consulted other urban indexes. I found Figure 15, a graph of unemployment and crude oil prices, in a blog by Dr. Richard Green (a business professor at the University of Southern California). Interestingly, when oil prices are low, such as in 1996, 2001 and 2006, unemployment rates in Houston are lower as well. We also saw that NUCI reported higher amounts of urban development in those years. In 2011, however, crude oil prices and unemployment rates increased, and we noticed significantly less urban development. Could it be that Houston responds to floods with increased urban development and to drought with less development? Admittedly, four samples is not enough to go on, but it does provide enough evidence to warrant a full-scale study to describe the relationship between PCM-detected urban change, natural disasters and the economy of Houston or any other study areas.

Figure 15. Graph showing comparison of several urban economic health statistics for Houston. Published in Dr. Richard Green‘s blog: http://real-estate-and-urban.blogspot.com/2014/12/is-houston-really-vulnerable-to_30.html

Visit us at GEOINT this year for demonstrations and in-booth Smart Talks on our Persistent Change Monitoring Capability. For more about Radiant Solutions, visit www.radiantsolutions.com.