Analyzing the land cover of an urban environment using high-resolution orthophotos

https://doi.org/10.1016/S0169-2046(02)00165-2Get rights and content

Abstract

To estimate the impact of light-colored surfaces (roofs and pavements) and urban vegetation (trees, grass, shrubs) on meteorology and air quality of a city, it is essential to accurately characterize various urban surfaces. Of particular importance is the characterization of the area fraction of various surface-types as well as the vegetative fraction. In this paper, a method is discussed for developing data on surface-type distribution and city-fabric (land cover) makeup (percentage of various surface-types) using high-resolution orthophtos. We devised a semi-automatic Monte Carlo method to sample the data and visually identify the surface-type for each pixel. The color aerial photographs for Sacramento covered a total of about 65 km2, at 0.3-m resolution.

Five major land-use types were examined: (1) downtown and city center, (2) industrial, (3) offices, (4) commercial, and (5) residential. In downtown Sacramento, the top view (above-the-canopy) shows that vegetation covers 30% of the area, whereas roofs cover 23% and paved surfaces (roads, parking areas, and sidewalks) 41%. In the industrial areas, vegetation covers 8–14% of the area, whereas roofs cover 19–23%, and paved surfaces 29–44%. The surface-type percentages in the office area were 21% trees, 16% roofs, and 49% paved areas. In commercial areas, vegetation covers 5–20%, roofs 19–20%, paved surfaces 44–68%. Residential areas exhibit a wide range of percentages among their various surface-types. On average, vegetation covers about 36% of the area, roofs about 20%, and paved surfaces about 28%. Trees mostly shade streets, parking lots, grass, and sidewalks. In most non-residential areas, paved surfaces cover 50–70% of the under-the-canopy area. In residential areas, on average, paved surfaces cover about 35% of the area.

Land-use/land cover (LULC) data from the United States Geological Survey (USGS) was used to extrapolate these results from neighborhood scales to metropolitan Sacramento. Of an area of roughly 800 km2, defining most urban areas of the metropolitan Sacramento, about half is residential. The total roof area comprises about 150 km2 and the total paved surfaces (roads, parking areas, sidewalks) are about 310 km2. The total vegetated area covers about 230 km2. The remaining 110 km2 consist of barren land and miscellaneous surfaces.

Introduction

The Heat Island Reduction Initiative (HIRI) is a joint program sponsored by the US Environmental Protection Agency (EPA) and the Department of Energy (DOE) to encourage the use of strategies designed to reduce demand for cooling-energy use and prevent smog formation in US cities. As part of the initiative, the Urban Heat Island Pilot Project (UHIPP) was launched to quantify the potential impacts of Heat Island Reduction (HIR) strategies in terms of energy savings, economic benefits, and air-quality improvements. Sacramento, CA, Salt Lake City, UT, and Baton Rouge, LA were selected for this analysis.

One of the components of UHIPP research activities is to analyze the fabric of the pilot cities by accurately characterizing various surface components. Of particular importance is the characterization of the area fraction of various surface-types. These data are required to analyze the impact of HIR measures in reducing energy consumption and improving air quality. Thus, it is important to characterize the surface as accurately as possible, particularly in terms of surface-type distribution and vegetative fraction. An accurate characterization of the surfaces will allow a better estimate of potential increases in surface albedo1 (roofs, pavements) and urban vegetation, providing more accurate modeling of the impact of HIR measures on ambient cooling and urban air quality.

Researchers involved in the analysis of urban climate have tried to estimate the composition of various urban surfaces. In Sacramento, CA, Myrup and Morgan (1972) examined the city data in progressively smaller integral segments of macro-scale (representative areas of Sacramento), meso-scale (individual communities), micro-scale (land-use ordinance zones), and basic-scale (city blocks). The data used included USGS photos, parks and recreation plans, city engineering roadways, and detailed aerial photos. The analysis covered 195 km2 of the urban area. Myrup and Morgan calculated the percentages of the land-use areas as follows: residential 35.5%, commercial 7.2%, industrial 13.5%, streets and freeways 17.0%, institutional 3.2%, and open space and recreational 23.6%. The analysis found the average residential area to be about 22% streets, 23% roofs, 22% other impervious surfaces, and 33% green areas. Overall, for the city, they found 14% streets, 22% roofs, 22% other impervious surfaces, 36% green areas, and 3% water surfaces. Their research defined the “other impervious surfaces” to include highway shoulder strips, airport landing runways, and parking lots. Streets included curbs and sidewalks.

McPherson (1998) has also analyzed the fabric of Sacramento using aerial photos. For the low-density residential areas constituting about 90% of the total residential area, he reports that buildings cover about 29% of the area, paved surfaces 27%, trees and shrubs 19%, and grass and soil 24%. For commercial and industrial areas, buildings cover 25% of the area, paved surfaces 50%, trees 6%, and grass and soil 19%.

Other researchers have used aerial orthophotos and satellite imagery to study various aspects of urban land-use. Small (2001a) has estimated the abundance of urban vegetation in several areas of New York City, using Landsat Thematic Mapper (TM) and 2-m resolution aerial photography. For some residential areas in Central Manhattan, he found a vegetation fraction of 20–40%. Small (2001b) has also conducted investigations using high-resolution satellite imagery for the analysis of urban reflectance. Luvall et al. (2001) have used high-resolution thermal infrared and visible data to measure characteristics of surfaces typical of the urban landscape for three US cities (Sacramento, Salt Lake City and Baton Rouge). The same data were also used to measure the vegetation fraction and the reflectivity of the urban areas. Nowak et al. (1996) reviewed several methods for determining urban cover from aerial photographs. They found that urban tree cover is highest in cities that developed in naturally forested areas (15–55%; mean 31%), followed by grassland cities (5–39%; mean 19%), and desert cities (0.4–26%; mean 10%). For Sacramento the total green space was estimated at 61% and the tree cover at 14%.

The objective of this study is to develop a high-quality data base of surface-type and city-fabric makeup (percentage of area covered by various surfaces) for various land-uses in each pilot city selected by the EPA for the UHIPP. An effort is made to develop a method that automates (objective analysis) most of this process to obtain accurate results in an efficient, reproducible manner.

Akbari et al. (2000) review and discuss the pros and cons of a variety of available data sources for analyzing the fabric of the UHIPP cities. These data sources include:

  • Advanced Thermal and Land Applications Sensor (ATLAS) used by NASA to collect high-resolution surface data in 15 channels.

  • High-resolution (0.5 m) black-and-white photography.

  • High-resolution (0.5 m) color infrared photography.

  • High-resolution (0.3 m) custom color digital orthophotos.

Of all the various data sources tested, the high-resolution custom color digital orthophotos offer the best platform for obtaining accurate estimate of urban fabric. To obtain these high-resolution photos a digital camera is flown aboard a low-altitude aircraft equipped with a global positioning system (GPS) and a computer for acquiring and storing data from both the camera and the GPS. The data collected by the GPS system along with topographical data are used in the process of orthorectification. Thus, errors created by the terrain and angle between the camera and location are minimized.

Using true color aerial photography at a 0.3-m pixel size, it is possible to identify clearly the materials and surfaces that make up the fabric of an area. Using a classification procedure similar to that used with ATLAS data, a semi-automatic procedure for classifying the surfaces of a city can be developed. In a color photograph, the red, green, and blue (RGB) band data can be used in a parallelepiped classification scheme in the same way the bands of ATLAS data were used. However, all three bands are in the visible spectrum and, thus, do not cover the entire solar and thermal radiative ranges. For this reason, limited information can be acquired from this data type.

An advantage of custom aerial color photography is that flights can be scheduled as desired. Accordingly, the photos can be taken at solar noon, thus, minimizing the inaccuracies introduced by shadows. In addition, the high resolution allows for the calibration of photographs (RGB bands) with laboratory-measured reference panels that can be placed under the flight path in the field. Another practical advantage of these photos is that they cost only US$ 140–160 km−2 of coverage for most urban areas.

In this paper, we present a Monte Carlo approach for analysis of aerial color photography of Sacramento, CA. We apply the method to several representative areas in Sacramento to obtain urban surface characteristics data. Results for the analysis of representative areas are used to estimate the fabric of metropolitan Sacramento (to be used in meteorological and air-quality modeling). We conclude the paper by providing suggestions and recommendations for future work.

Section snippets

Method of analysis for custom color digital orthophotos

The color aerial photographs obtained for Sacramento covered a total of about 65 km2. At 0.3-m resolution, approximately 7×108 pixels of data were collected. It was impossible to review all these data visually in detail. Hence, an automated method to classify the data was deemed necessary.

Initially, we analyzed the three bands of RGB data for a selected set of data, searching for characteristic signatures for various surfaces. Unfortunately, since there were significant similarities between the

Results from Sacramento, CA

Three flights were performed on sunny, cloud-free and clear days around solar noon to minimize the impact of shadows (20 August, 7 September, and 4 November 1998). On all days, a specially-equipped aircraft took off from Sacramento Executive Airport and flew at approximately 1.5 km over areas selected by the Heat Island Group. The color aerial photographs of Sacramento covered a total of about 65 km2. Fig. 2 depicts a sample photograph of downtown Sacramento. All data were taken at 0.3-m

Extrapolation to metropolitan Sacramento

We used the land-use/land cover data from the United States Geological Survey (USGS) to extrapolate the limited data obtained from the analysis of aerial photos to the entire Sacramento area. LULC data classify surface characteristics, at 200-m resolution, into many different urban and non-urban categories. The LULC classification for urban areas includes: residential, commercial/service, industrial, transportation/communications, industrial/commercial, mixed urban or built-up land, and other

Discussion

This paper focused on the characterization of the fabric of a region in terms of surface-type makeup. The data obtained from the Sacramento flights suggest that it is possible to characterize the fabric of a region of interest accurately and cost-effectively. However, depending on the purpose of the application and the funds available for this task, a separate decision must be made for each UHIPP city or region as to the most appropriate combination of data, i.e. a combination of aerial

Conclusions

To estimate the impact of light-colored surfaces (roofs and pavements) and urban vegetation (trees, grass, shrubs) on the meteorology and air quality of a city, it is essential to accurately characterize various urban surfaces. Of particular importance is the characterization of the area fraction of various surface-types as well as the vegetative fraction. In this paper, a method is discussed for developing data on surface-type distribution and city-fabric makeup (percentage of various

Acknowledgements

This work was supported by the US Environmental Protection Agency and US Department of Energy under contract DE-AC0376SF00098. We acknowledge the support and guidance from Virginia Gorsevski and Jeanne Briskin of the EPA. Brian Pon of LBNL helped us with graphic arts.

References (8)

  • D.J. Nowak et al.

    Measuring and analyzing urban tree cover

    Landsc. Urban Plann.

    (1996)
  • Akbari, H., Rose, L.S., Taha, H., 2000. Characterizing the Fabric of the Urban Environment: A Case Study of Sacramento,...
  • ERDAS, 1997. Field Guide 1997. ADDRESS, p....
  • Luvall, J., Quattrochi, D., Rickmann, D., Estes, M., 2001. EPA/NASA Urban Heat Island Pilot Project. Global Hydrology...
There are more references available in the full text version of this article.

Cited by (191)

  • In-camera IMU angular data for orthophoto projection in underwater photogrammetry

    2023, ISPRS Open Journal of Photogrammetry and Remote Sensing
View all citing articles on Scopus
View full text