Natomas Basin, CA, Land Cover Change Analysis

by: Jesus Alvarez

Fall, 2009

Advanced Image Processing, Emporia State University



Table of Contents

Abstract
Natomas Basin
Change Analysis
Active Vegetation Change
Cluster Analysis
Land Cover Changes
Conclusions
References




Abstract

Land cover neighboring urban areas in northern California has been subject to land use changes in recent years. The Natomas basin located just east of Sacramento River between Sacramento and Sutter Counties (see Fig. 1) has changed dramatically between 1999 and 2009. Natural marshes have been diminished and creeks have been relocated and lined with concrete. Within incorporated areas, land owners have turned agricultural land use to more profitable urban land use. The changes affecting the Natomas Basin can be assessed using satellite remote sensing imagery. The size of the Natomas Basin (approx. 220 km2) is suitable for the datasets produced by Landsat 5 Thematic Mapper (TM). To conduct the change analysis, the visible, near-infrared (NI), and mid-infrared (MI) bands (see Table 1a) were processed for 1999 and 2009. Landsat 5 datasets with 10-year temporal resolution were used to conduct land cover change analysis. The land cover change maps were produced by ratioing bands to develop vegetation indices and unsupervised cluster classification. The final output from cluster analysis was then classified and change detection maps and graphs were generated using a pixel based land change classification.



Natomas Basin View Using Landsat 5 TM Natural-Color Composite

Figure 1. Natomas Basin Composite. July 10, 2009 Landsat 5, visible bands 1,2,3 (see Table 1a) were used to build the natural-color composite. The Natomas basin is deleniated by the Sacramento River to the west, Cross Canal to the north, the Natomas East Main Drainage Canal to the west, and the American River to the south. Three major freeways cross in the lower basin (Interstate 5, Highway 99, Interstate 80). The urban centers and new subdivisions branch out from the major freeways. Urban sprawl is expanding northwest from the lower basin.



Table 1a. Landsat 5 TM Spectral/Spatial Resolutions
Band Spectrum Spectral Resolution Spatial Resolution
1 Visible (Blue) (0.45 0.52 m) 30 m
2 Visible (Green) (0.52 0.60 m) 30 m
3 Visible (Red) (0.63 0.69 m) 30 m
4 Near-Infrared (0.76 0.90 m) 30 m
5 Near-Infrared (1.55 1.75 m) 30 m
6 Thermal (10.40 12.50 m) 120 m
7 Mid-Infrared (2.08 2.35 m) 30 m

Return to Table of Contents



Natomas Basin

The Natomas Basin located on the northeast confluence of the Sacramento and American Rivers. The basin covers about 220 Km2 of land area and is enclosed by levees (see Fig 1). The majority of the basin is unincorporated and is serviced by Sutter County on the north, Sacramento County on the south, and the City of Sacramento on the lower southeast. The elevations above sea level within the basin range between 0 meters in the lower south basin near the freeway intersections to about 14 meters on the upper east basin near the Natomas East Main Drainage Canal. Prior to the urban sprawl that changed the lower basin; the land cover was composed of agricultural land areas, wetland areas, and natural streams and lakes. The urban areas in the lower basin are inside the City of Sacramento have expanded and filled the entire incorporated areas with high-density residential development in the last 10 years. The remaining agricultural land areas and wetland areas are located on the unincorporated northern portion of the basin.

Prior to 1970, the land cover of the basin was dominated by grasslands, marshes, natural lakes and streams that drained runoff from the north and northeast highlands. Agricultural land uses occurred along the streams and lakes and the few residential areas were located along the American River. Between 1970 and late 1990's the south Natomas neighborhood was developed just south of Interstate 80 and farming land uses remain in the upper basin. Between 1999 and 2009, land cover changes that took place in the lower basin have impacted the habitat of the basin. Many of the species that lived in the Natomas Basin such as the gater snake and the Swainson's hawk, have been listed in the endangered species list. Land cover has changed from agricultural and open grass land to residential and commercial land use. Natural detention ponds have been filled and roads have been paved over. Natural streams have been relocated or filled to make way for subdivisions.

The lower incorporated areas of the basin have been subdivided and are now covered with paved surfaces and high-density residential. Small lakes and detention ponds have been build to compensate for habitat impacts, however, the species never adapted to their new environment and remain in the endangered species list. The few wetlands that remain are located on the upper northeast basin and based on the patterns observed in land cover change in the lower basin, it is only a matter of time before the wetlands are relocated or diminished.

Return to Table of Contents



Change Analysis

Satellite remote sensing datasets from Landsat 5 TM are suitable for detecting changes over a large region. The datasets used must have a suitable temporal resolution and similar characteristics as far as sun angle and atmospheric conditions. From the datasets, simple study land cover change can be done by studying image composites (see Fig. 2) or by conducting image processing using unsupervised/supervised classification techniques with visible, near-infrared, and middle-infrared bands. Vegetation is also an indicator for land cover or environmental changes. Vegetation can be assessed by developing vegetation indexes from band rationing (see Fig. 3-4).

For the Natomas Basin, land cover change analysis was performed using Landsat 5 TM datasets with a 10-year temporal resolution. The datasets were acquired for the summer of 1999 and 2009 and were collected around the same time of the month with only a few days off. The image datasets were collected under normal atmospheric conditions with less than 5% cloud cover and were corrected for haze effects. Differences observed in brightness values (BV's) are due to high water content in the 1999 dataset. The 1999 dataset shows signs of saturated land cover due to floods that occurred in previous years. Differences in BV's mainly on the near infrared bands are related to the high moisture content in the 1999 dataset from previous wet years in 1996/1997. In addition, other factors that need to be considered in the change analysis or image comparison has to do with the water content of the 2009 dataset. 2009 has been declared a year of drought, which may affect the BV's of the 2009 dataset mainly when comparing vegetation indexes (Fig. 3-4).

The datasets were corrected for haze effects and a mask was digitized and applied to the final image processes. The first image change detection was conducted by generating Normalized Differenced Vegetation Index (NDVI) for 1999 and for 2009. The index changes were performed by developing a standardized class image from NDVI differenced z scores (see Fig 3-4). The final analysis was conducted by developing clusters using unsupervised techniques to create land cover classes and assess areas (see Fig. 5-7). The land cover classes developed were compared using a pixel based analysis approach to develop a land cover change map and net changes graphs based on pixel change values (see Fig. 8-10). The land change map and graphs were processed with the Land Change Model (LCM) available in IDRISI Taiga.

Figure 2. Landsat 5, 1999 and 2009 false-color composites from bands 2,3,4.
19991999 False-Color Composite 20092009 False-Color Composite

Return to Table of Contents



Active Vegetation Change

Figure 3. Vegetation Indexes for 1999 and 2009. The indexes are use as a greenness indicator to assess active vegetation in the Natomas Basin. Active vegetation is shown with values above 0.0, some vegetation placed in the low values could be related to high moisture content or may be inactive. Negative values represent inactive vegetation, paved surfaces, residential land cover, bare soils, and water. The indexes are calculated using the following band equation NDVI = ((B4 - B3)/(B4 + B3)).
1999 Vegetation Index Index Legend 2009 Vegetation Index



Figure 4. NDVI Z-Scores. The values are based on the 1999 and 2009 NDVI values. The values are calculated from pixel values, mean, and standard deviation values. Z-Scores = ((observed value - mean value)/std deviation)). The values represented with negative standard deviations (sd) represent negative vegetation changes between 1999 and 2009. Positive sd values represent positive changes.
NDVI Z-Scores

Return to Table of Contents



Cluster Analysis

Figure 5. Cluster images were generated from bands 1,2,3,4,5 and 7. Clusters are built from BV's similarities and neighboring pixels. The number of clusters suitable for the Natomas Basin was set to 20. Before the number of clusters was set, several cluster images with different number of cluster groups were compared to the 1999 and 2009 natural and false-color composites to find a suitable grouping number based on visible land cover.
1999 Cluster Analysis 2009 Cluster Analysis



Figure 6. Land cover maps developed from unsupervised classification. The cluster images shown on Figure 5, were classified by relating clusters to natural and false-color composite features. In addition, the 2009 land cover classes were field verified along Highway 99 and Interstate 5. The 2009 land cover map show an increase in urban areas in the lower Natomas Basin, which coincides with the City of Sacramento incorporated boundary. The areas on the northwest away from the urban center shown as populated/paved surface are open parking lots and new construction at the Sacramento International Airport. Some generalization of small features and/or linear features occurred due to spatial resolution of Landsat 5 datasets.
1999 Land Cover 2009 Land Cover


Figure 7. Land cover areas delineated for 1999 and 2009. Areas developed from land cover classes show a significant 40% decrease in vegetated areas and urban areas have double in the last decade. The total land areas total 227 km2 of which 7 km2 are river areas. The river areas were intentionally digitized on the mask to help classify water features using the river BV's.
Land Cover Changes

Return to Table of Contents



Land Cover Changes

Figure 8. Land cover gains and losses. Gains and losses are based on change of values between 1999 and 2009 land cover image pixels and exclude pixels with no change. Values are shown in Km2 some of the gains and/or losses are offset by pixel misclassification and generalization of cluster groups.
Land Cover Gains and Losses


Figure 9. Land cover net changes. The graph shows the total net changes in km2 for the 4 land cover classes. The net change is approximately the same as that obtained from subtraction of land covers in Figure 7. The graph shows the vegetated areas have declined more than 30 km2 and urban areas have increased more than 25 km2.
Net Change



Figure 10. Land cover change map. Detail land cover changes are shown on the legend. Changes are based on a pixel value change analysis. Pixels with no change have been assigned a no-data value and are not shown on the legend. The most noticeable changes are in the vegetated areas (orange) and populated areas (yellow).
Land Change Map

Return to Table of Contents



Conclusions

The urban areas in the Natomas Basin have more than doubled between 1999 and 2009. The incorporated areas were developed to capacity. Based on the observed trends, the land cover will continue to change as new areas are annex from unincorporated into incorporated areas. Croplands Vacant crop grasslands and croplands will continue to change to other more profitable types of land use. Habitat impacts will remain negative as more grasslands and marshes are relocated or eliminated. Based on a decade of land cover changes, the Natomas Basin could be transformed into an fully urbanized basin by the year 2040 at the current 1999-2009 growth rate.

Satellite remote sensing datasets provide the tools to monitor land cover change for various scales. Remotely sensed datasets are available in different spectral and spatial resolutions. The different resolutions allow for land cover change analysis for small or large scales. Landsat 5 TM datasets were suitable for analyzing land cover changes in the Natomas Basin 220 km2 area. Minor generalization of land cover occurred with small features such as small parks, lakes, linear features, and creeks during the cluster analysis classification. The overall cluster classification could have been improved by developing signatures based on field land cover verification for both 1999 and 2009, which would have developed more land cover classes. However, generalization of small features is inevitable since spatial resolution is limited to the 30 m pixel resolution of the Landsat 5 TM datasets. For a more accurate land cover change assessment a higher spectral and spatial resolution dataset needs to be considered.

Return to Table of Contents



References

  • Alvarez, J., 2009. Natomas Basin Area Field Trip, Land Cover Verification Along Freeway Corridors.

  • IDRISI Taiga. GIS/Image processing software, World Wide Web homepage <http://www.clarklabs.org/>

  • USGS, Earth Resources Observation and Science (EROS). Landsat 5 TM Imagery, World Wide Web homepage<http://eros.usgs.gov/#/Find_Data/Products_and_Data_Available/Satellite_Products> [retrieved on 20 Nov. 2009]