Combining NASA and Commercial Satellite Data

Tracking Land Cover Change in California

The living cover types of our planet’s land masses are in constant change: forests are converted to farmland, cities expand into former woodlands and pastures, cleared parcels are replanted with green vegetation, and wild lands of all types are burned by fires. The use of satellite imagery has made the mapping of land cover change possible from local to global scales. This type of analysis has proven helpful to a variety of disciplines, including agronomy, urban planning, and forestry.

Chiristopher Potter, PhD
Research Scientist
NASA Ames Research Center
Moffett Field, Calif.

Vanessa Genovese &
Peggy Gross
Research Scientists
California State University Monterey Bay
Seaside, Calif.

Determining where, when, and why natural land cover is converted to human uses is a crucial scientific concern. Characteristics of the land cover can have important impacts on local climate, radiation balance, biogeochemistry, hyd-rology, and the diversity and abundance of terrestrial species. Consequently, understanding trends in land cover conversion at local scales is a requirement for making useful predictions about other regional and global changes. It is urgent that accurate, timely, and economical tools be made available to better document these conversions and to aid in the management of their impacts.

Land cover alterations can be assessed by comparing two or more images taken at different dates over the same location. The 1999 launch of NASA's Terra satellite platform with the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument on board has enabled the collection of daily images of the entire Earth land surface, with promising implications for land cover change research. MODIS satellite data can uniquely capture the temporal dynamics of vegetation distributions across the landscape. Operational MODIS algorithms have been used to generate the Enhanced Vegetation Index (EVI) at 250-meter spatial resolution from 2000 to the present. (1) EVI represents an optimized land cover index, whereby the vegetation reflectance properties in red and near-infrared spectral bands are designed to approximate canopy radiative transfer theory. EVI was developed to optimize the "greenness" signal (area-averaged canopy photosynthetic capacity), with improved sensitivity in high biomass cover areas. Global MODIS EVI data sets are available free of charge from NASA-supported data centers.

Regions experiencing rapid population growth and changing economic act-ivities are logical places to evaluate the MODIS time series for characterizing land cover changes. California’s population increased by 75 percent between 1970 and 2005, resulting in rapid and ext-ensive urbanization and loss of natural cover types. For example, of all land urbanized in 42 of the state’s 58 counties between 1984 and 1990, an estimated 13 percent occurred on irrigated prime farmland, whereas 48 percent occurred on wildlands or fallow marginal farmlands. (2)

The complete MODIS 250-meter resolution time series of EVI data (2001 to 2005) covering the majority of California were obtained to assess the usefulness of this satellite imagery for land cover change detection. Quality-control filter methods were first applied to the MODIS 250-meter EVI data set used in this analysis. The EVI usefulness index is a high-resolution quality indicator with value for a pixel determined from several conditions, including 1) aerosol quantity, 2) atmospheric correction conditions, 3) cloud cover, 4) shadow, and 5) sun-target-viewing geometry. About 80 percent of the EVI values in this California data set were found to have high-quality indicator scores.

Figure 1 - Month of maximum MODIS EVI observed during the calendar year 2005

MODIS EVI time series examined over individual January-December yearly periods show that natural vegetation cover in California has a characteristic seasonal profile of greenness levels that is influenced strongly by precipitation and temperature variations over an annual cycle. (3) A large portion of the state’s forested land cover reaches maximum greenness levels (EVImax) during the period of June through August when temperatures are highest, whereas rangelands and mixed grassland-woodland areas reach EVImax levels earlier in the calendar year, typically during the period of January through April when rainfall is highest (Figure 1). As regional examples, the month most commonly identified as EVImax was April in the Sacramento and San Joaquin Valleys and the Sierra Nevada Foothills, whereas the EVImax month was most commonly identified as June in the Sierra Nevada Mountain Range and the Coastal Redwood forest belt. (4)

Annual cycles of vegetation greenness can be characterized further by examining the changes in seasonal amplitude of EVI during a year. This attribute was computed as the EVI difference (EVIdiff) between the maximum (EVImax) and minimum (EVImin) greenness values measured over each of the single calendar years that comprise the entire 2001-2005 time series available to date. By identifying areas where both the annual EVIdiff and the EVImax values changed continuously over the five-year MODIS time series, locations of potential land cover change can be extracted from the statewide data set.

Figure 2 - Map of the locations (in red) where MODIS 250-meter time series showed continuous decreases in both EVIdiff and the EVImax over the period 2003-2005.

Over the entire period 2001 to 2005, locations that showed continuous decreases in both the EVIdiff and the EVImax made up more than 5170 km 2 over the MODIS image area (Figure 2). These locations were distributed across land cover types categorized mainly as forest (46 percent), followed by shrublands (29 percent) and herbaceous cover (7 percent). Forest and shrubland areas together remained the predominant land cover types over which most of the locations (60-70 percent) of continuous decreases in both the EVIdiff and the EVImax were detected, regardless of which three- or four-year sub-period of the 2001-2005 MODIS time series was extracted for analysis. In contrast, locations that showed continuous increases in both the EVIdiff and the EVImax made up only 366 km 2 over the MODIS image area analyzed for this study. Therefore, the remainder of this report will include only cases of continuous decreases in both the EVIdiff and the EVImax.

Visual analysis of high-resolution commercial images available, for example, in Google Earth offered more detailed examination of localized land use change records at locations where both EVIdiff and EVImax decreased over the MODIS time series. The 50 largest contiguous areas of concurrent EVIdiff and EVImax decrease (as mapped in Figure 2) were examined for visual evidence of disturbance in the land cover. In 15 of these 50 cases, the satellite imagery at spatial resolution of less than 10 meters revealed a recent wildfire burned area. In another 7 of these 50 cases, the satellite imagery revealed extensive cutting of the forested land cover. High-resolution imagery shows an example of what appears to be extensive forest harvest as a patchwork of land cover types near Sly Creek Reservoir in the Plumas National Forest in Northern California (Figure 3). Additional imagery readily identifies extensive forest cutting patterns similar in appearance to Figure 3 from areas of concurrent EVIdiff and EVImax decrease locations in Mendocino, Butte, and Amador counties.

Figure 3 - High-resolution satellite image showing cleared patches of land in the Plumas National Forest, California, near Sly Creek Reservoir. For scale, the white line at the top-center of the image (just below a large cleared patch of forest) is 250 meters in length. Red circle lines are 1-km buffer boundaries around the center locations of MODIS 250-meter areas where the most recent land cover change has been detected. Google Earth imagery © Google Inc. used with permission.

Land use change records in California counties with the highest rates of farmland conversion and urban development were also compared to EVIdiff and EVImax changes over the MODIS time series. The state’s Farmland Mapping and Monitoring Program (FMMP) produces maps and statistical data for use in analyzing impacts of land conversion on California’s natural resources. The FMMP covers 45.9 million acres, representing 91 percent of the state's private agricultural lands. Analysis of FMMP trends from 2002 to 2005 in the counties with the highest areas of agricultural land conversion, namely Kern, Santa Barbara, Tulare, and Ventura, revealed that the most common pattern of change in the annual EVImax (between 40 percent and 70 percent of land use change areas) was an increase followed by a decrease and then another year of EVImax increase.

This comparison suggests that the extensive agricultural land conversion taking place in California over the past five years has not resulted in a persistent loss of vegetation greenness in most cases. Instead, even areas converted to housing and business developments tend to increase greenness cover observed by MODIS EVI patterns, probably due to seasonal irrigation of replanted lots and parklands. The sustainability of this high demand for irrigation water in California’s increasingly developed agricultural counties is not a certainty, however.

The methods of image analysis that we describe in this study are developing as a leading fusion of NASA data with high-resolution commercial images to identify areas of recent land cover change over regional scales. Our approach has been to combine the frequent (daily) repeat time series of MODIS 250-meter image data with the detailed "snapshots" of the Google Earth imagery to show when, where, and what types of changes in land use have occurred.

This is among the first applications of its kind to take full advantage of the scientific importance of NASA satellite data in tandem with the compelling visual details of high-resolution color imagery from commercial sources. Using NASA MODIS time series to pinpoint extensive areas of land cover change enhances underlying images in tools such as Google Earth by giving unique, dynamic attributes, and in the process, makes interpretation much more exciting.


  1. Huete, A., K. Didan, T. Miura, E.P. Rodriguez, X. Gao, L.G. Ferreira. 2002. “Overview of the radiometric and biophysical performance of the MODIS vegetation indices,” Remote Sensing of Environment, 83, 195-213. Sensing of Environment, 83, 195-213.
  2. Charbonneau, R. and G. M. Kondolf. 1993. “Land use change in California, USA: Nonpoint source water quality impacts,” Environmental Management, 17, 453-460., 17, 453-460.
  3. Potter, C., P. Tan, V. Kumar, C. Kucharik, S. Klooster, V. Genovese, W. Cohen, S. Healey. 2005. “Recent history of large-scale ecosystem disturbances in North America derived from the AVHRR satellite record,” Ecosystems, 8(7), 808-827., 8(7), 808-827.
  4. U.S. Geological Survey (USGS ) and U.S. Environmental Protection Agency (USE PA). 1999. National Land Cover Data.
Sensors & Systems | Monitoring, Analyzing and Adapting to Global Change | Stay in tune with the transformation. Subscribe to the free weekly newsletter.