PE&RS April 2001

Highlight Article

By Kevin P. Price, Theresa J. Crooks, and Edward A. Martinko

Grasslands Across Time and Scale:  A Remote Sensing Perspective

The Kansas Applied Remote Sensing Program (KARS), home to the NASA Great Plains Regional Earth Science Applications Center (RESAC), has conducted a wide range of fundamental and applied remote sensing projects since the early 1970s.

In the October 2000 issue of PE&RS we highlighted our RESAC Program, and our applied remote sensing research and commercial applications development activities (Martinko et al., 2000). In this highlight article, we’ll focus on our fundamental research in grasslands.

In particular, we’ll highlight the remote sensing technologies we are employing across multiple spatial scales and some of the ways these technologies are enhancing our knowledge and understanding of grasslands.

Introduction
Grasslands are the largest of the Earth’s four major vegetation types and are among the world’s most agriculturally productive lands. The tallgrass prairies of the North American Great Plains are among the most biologically diverse in the world. However, since the mid-1800s tallgrass prairies have become highly fragmented by conversion to croplands and by widespread introduction of non-native grasses for cattle grazing.

Habitat fragmentation and the combined effects of various land management practices, like burning and grazing, continue to alter important grassland characteristics such as biodiversity, community structure, primary productivity, biogeochemical cycling, and soil stability.

From a remote sensing perspective, key research issues related to grasslands include assessing grassland condition and productivity, monitoring land use/land cover change and conservation practices, and modeling soil erosion. Of particular importance is the ability of remote sensing approaches to discriminate among subtle differences in grassland types and land management practices.

One reason that discriminating between grassland types is important is the role that different grasses play in biogeochemical cycling and storage of carbon in soils. Warm-season C4 grasses that are native to the tallgrass prairie, such as big and little bluestem, fix greater amounts of atmospheric CO2 than the introduced cool-season C3 grasses such as smooth brome. This has significant implications for understanding the relationship between land management, grassland productivity, and long-term sustainability.

Grassland community dynamics operate across numerous spatial and temporal scales. These dynamics are in turn controlled or altered by climate variation and human activities, which also operate across varying scales.

Remote sensing methods applied at different spatial and temporal scales provide critical and unique insights into the complex interrelationships between grasslands, climate, and human activity.

Close-Range Remote Sensing
KARS researchers have conducted close-range remote sensing with handheld spectroradiometers in conjunction with field measurement of plant biophysical factors and soil moisture to investigate grassland dynamics at the plot and field level.

One advantage of handheld spectroradiometers is that they facilitate collection of multi-date, or multitemporal, measurements at the plot and field scales. These data can then be integrated into multi-scale monitoring and modeling efforts.

Multitemporal measurements are particularly important because monitoring efforts that employ remote sensor data to assess plant biophysical characteristics and changes in land use/land cover (LULC) are often confounded by spectral patterns that change with climate from season-to-season and year-to-year. 

Changes in land management also alter plant species composition, productivity, and canopy characteristics, further complicating the relationships between climate, land management, and vegetation reflectance.

Integrating a multitemporal component into our hyperspectral spectroradiometer datasets has revealed important field-scale differences in canopy surface geometry and plant canopy architecture during the growing season and under different management regimes. Unique multi-seasonal hyperspectral patterns have been identified for several management regimes, including native, hayed, mowed, grazed, and burned grasslands (Fig 1).

Our close-range remote sensing research has been greatly enhanced by access to the Kansas Ecological Reserves (KER). KER is a unique biological field station, located near Lawrence in northeastern Kansas, encompassing 726 ha (1,794 acres) of tallgrass prairie and other diverse natural ecosystems, as well as manipulated habitats and experimental facilities (www.ker.ukans.edu). KER’s unique combination of protected natural ecosystems and experimental areas allows us to study a variety of grassland types in different stages of ecological succession under a variety of land management practices.

Concurrent multitemporal measurements of spectral reflectance and key biophysical and physiologic variables at locations such as KER provide a practical means of studying the effects of changing land use and fluctuating annual weather patterns on plants and their associated spectral responses.

 

Figure 1. Multi-date spectral response curves for unplowed native prairie (N), and restored native grasslands under five management practices including: burned (B), hayed (H), mowed but not hayed (M), grazed (G), and untreated (U). The curves show both intra- and interannual spectral variability among the six types. The unplowed native prairie is most spectrally unique, especially in 1990 when it was first burned. The effects of haying and mowing can be seen in the late summer curves, in which high NIR reflects greater amounts of photosynthetically active leaf material (From Price, et al., 1993).

Field to Regional Scale: Airborne Remote Sensing
KARS’ Aerial Digital Image Acquisition System. Timely access to multitemporal imagery at multiple spatial scales is often critical to monitoring both large area LULC changes and small area changes in canopy condition across the growing season.

To provide such capability to our researchers, KARS has developed an aircraft-mounted multispectral digital imaging camera system that can be flown at various altitudes on a user-demand basis to support field and county level or subregional investigations (Fig 2).

Figure 2. The system consists of a DuncanTech MS3100 digital multispectral camera, a digital frame grabber, a 600 MHz Pentium III PC and monitor, and a self-contained power supply that provides over eight hours of system power. The assembly is custom mounted to a single-engine Cessna, either in the floor or in a specialized compartment in the door of the aircraft. The image acquisition and PC download process is controlled in-flight from the passenger seat by means of a mini-keyboard and touch pad mouse (Whistler et al., 2001).

The camera acquires high-resolution imagery in three co-registered bands in the blue (450-520 nm), red (630-690 nm), and near-infrared (760-900 nm) regions of the electromagnetic spectrum. All of the components have been integrated onto a single platform that can be custom-mounted into the floor of a specially modified single engine Cessna aircraft.

The most useful features of this imaging system are its low overall cost and its flexibility to meet user needs. Because it is mounted in a small airplane, imagery can be flown over specific regions at varying altitudes to produce whatever spatial and temporal resolutions are required to support the given research objectives. Planned research applications include:

• Identifying localized indications of stress in development of both natural and agricultural vegetation
• Identifying changes in grassland composition due to succession or anthropogenic stress
• Forest canopy monitoring
• Water quality assessment

The test flight also provided important supplemental information for interpreting HyMapTM hyperspectral imagery acquired over the same area three weeks previously.

Initial processing and classification of the data indicates that the camera will be useful for a variety of research objectives requiring on-demand multitemporal high-resolution data (Fig 3).
 

Figure 3. Image mosaic of multispectral aerial photographs taken on November 14, 2000. A) False-color composite images created using the NIR, red, and blue wavelengths. Actively growing vegetation (cool season grasses and evergreen trees) appears red, and dormant vegetation (warm season grasses and deciduous trees) appears as a blue/green color. B) Normalized Difference Vegetation Index (NDVI) image chain shown as a gray scale. Lighter shades represent vegetation with more active chlorophyll, and darker shades depict areas with very little chlorophyll. C) Classified image produced from an unsupervised classification on a four-band image chain (NIR, red, blue, NDVI) (From Whistler et al., 2001).

HyMapTM Hyperspectral Data. The evolution of a new generation of spaceborne hyperspectral sensors (e.g. MODIS, Hyperion) is expanding research horizons for investigating the utility of hyperspectral remotely sensed data in understanding grassland dynamics.

Fundamental questions about the utility of hyperspectral data are currently being explored using airborne data sources in order to understand the range of applications that might be possible as hyperspectral data from earth-orbiting satellites continues to become available.

For example, we’ve recently begun investigating the use of hyperspectral remote sensing data sets for quantifying key biophysical and biochemical plant characteristics at the field and county levels.

Of key importance is whether hyperspectral data, which has high spectral resolution but low temporal resolution, can provide equivalent or enhanced insights compared to results obtained with multitemporal data of lower spectral resolution but higher temporal coverage. In particular, we’re exploring whether hyperspectral imagery and endmember analysis approaches can be used to effectively identify subtle differences in grassland management practices and to discriminate mixtures of C3 and C4 grasses.

The HyMAP™ hyperspectral image shown on the front cover of this issue of PE&RS was flown in October 2000. The croplands in the lower third of the image had already been harvested, while the landscape north of the river is a mosaic of pastureland and deciduous forests. Using HyMAPTM data and associated ground measurements, we’re currently assessing the ability of hyperspectral data to:

• Discriminate cool-season C3 grasses from warm-season C4 grasses.
• Discriminate grasslands according to various land management practices, such as grazing, haying, or Conservation Reserve Program (CRP) management.
• Estimate biophysical properties of grasslands (e.g. LAI and canopy cover) for various grassland types and management practices.
• Quantify post-harvest crop residues and grassland standing litter components to support research into the relationships between surface litter, residue management, soil fertility, and carbon sequestration.

Multitemporal analyses of satellite remotely sensed data have yielded numerous insights into the relationships between land management and grassland dynamics.

Landsat Thematic Mapper. Our work with multitemporal Landsat Thematic Mapper (TM) imagery for general land cover classification has improved accuracy by as much as 30% over single date classification, and final accuracies often exceed 90% (e.g. Egbert et al., 1995).

Grassland discrimination and classification using multitemporal Landsat TM outperformed classification derived from similar analyses using ERS-2 Synthetic Aperture Radar. Integrating the two data sources in a multi-sensor approach does not seem to provide any additional advantages in regional discrimination of grassland types (e.g. Price et al., 2000).

A combination of spring, mid-summer, and late summer Landsat TM imagery has improved classification accuracy in land cover characterization and discriminating grasslands under various management regimes. For example, cool season grasses can be successfully discriminated from warm season grasses with very high accuracy and management practices with reasonable accuracy (Guo et al., 2000).

Discriminant analysis of multi-temporal data has identified Landsat TM band combinations that optimize classification of grassland types and management practices over a growing season (e.g. Price et al., 2001). For example, the NIR band produces optimal classification when compared to other bands. In contrast, biomass across the growing season is best predicted by visible bands (Guo et al., 2000).

These and other insights gained from multitemporal satellite data into discriminating grassland types and management practices provide improved classification, enhanced understanding of grassland productivity, and have begun to provide a basis for developing methods to model key grassland biophysical components.

AVHRR. To evaluate regional crop and grassland response to year-to-year climate variation, we’re using EROS Data Center AVHRR biweekly maximum Normalized Difference Vegetation Index (NDVI) composites (e.g. Wang et al., 2001).

Using a historical biweekly data set (1989-2001), we monitor grassland and agricultural phenological development state and compare current conditions to the previous week, year, and long-term average. Our findings are reported in map format as part of our weekly GreenReport®. Recently, we also developed the RangeReport®, which focuses on rangeland conditions in the U.S.

As part of our Great Plains RESAC technology transfer and remote sensing commercialization efforts, information derived from the GreenReport® and RangeReport® are combined with long-range weather forecasts to assist resource managers with their strategic planning and resource management efforts.

Our interest in grasslands extends to other parts of the world as well. For the past five years, we’ve been using AVHRR 4-km biweekly NDVI composites dating back to 1981 to study broad scale grassland dynamics in Inner Mongolia.

Multitemporal analysis has revealed that grassland phenology over large areas of the grassland steppe is changing (Fig. 4). Steppe vegetation is greening up later each year and the ecotone between desert and typical steppes appears to be shifting as desert steppe expands into the typical steppe.

Figure 4. A vegetation phenology trend map (1981 and 1992), derived from time-series analysis of multi-date NDVI data, for annual onset-date of plant green-up over a portion of Central Asia. The Gobi Desert is shown in black and dark gray. As a consequence of warmer winter and fall temperatures, areas in red and yellow are experiencing later green-up while areas in green are greening up earlier each year (areas in gray have not changed). Inset photograph shows an area of typical steppe under grazing by sheep. (From Yu, et al., 2000).

Conclusion
There is much to learn from studying grasslands across a variety of spatial and temporal scales. We are continuing to explore the use of multi-sensor, multitemporal remote sensing analysis and classification methods. We also believe there will be a wealth of knowledge gained as we add the hyperspectral dimension to our study of grassland dynamics.

Insights derived from these investigations identify efficient methods for accurately mapping and monitoring land use/land cover change, and provide enhanced ability to model important components in the complex interrelationships between grassland dynamics, climate change, and human activities.

Ultimately, such work should lead to better management of grasslands to ensure their long-term productivity and sustainability.

Acknowledgement
We’d like to acknowledge the contributions of Stephen Egbert, Jerry Whistler, Xulin Guo, and Fangfang Yu to this article, and we thank the NASA Earth Science Applications Research Division, Kansas NASA EPSCoR Program, and the National Science Foundation for their continued support.

Selected References
Egbert, S.L., K.P. Price, M.D. Nellis, and R. Lee. 1995. Developing a land cover modeling protocol for the High Plains using multi-seasonal Thematic Mapper imagery. ACSM/ASPRS ’95 Annual Convention and Exposition, Charlotte, NC, February 27-March 2. pp. 836-845.

Guo, X., K.P. Price, and J.M. Stiles. 2000. Modeling biophysical factors for grasslands in Eastern Kansas using Landsat TM data. Transactions of the Kansas Academy of Science, 103(3-4): 122-138.

Martinko, E. A., K. Price, S. Egbert, M. Jakubauskas, J. Whistler, and T. J. Crooks, 2000. Building on three decades of remote sensing and decision support: The NASA Great Plains RESAC and the Kansas Applied Remote Sensing (KARS) Program. Photogrammetric Engineering and Remote Sensing, 66(10): 1158-1166.

Price, K.P. V.C. Varner, E.A. Martinko, D.C. Rundquist, and J.S. Peake. 1993. Influences of land management and weather on plant biophysical and hyperspectral response patterns of tallgrass prairies in Northeastern Kansas. Proceedings of PECORA12, Sioux Falls, SD, August 24-26, pp. 441-450.

Price, K.P., X. Guo, and J.M. Stiles. 2000. Comparison of Landsat TM and ERS-2 SAR data for discriminating among six grassland types in Eastern Kansas. Proceedings of the Second International Conference on Geospatial Information in Agriculture and Forestry, Lake Buena Vista, FL, January 10-12, I: 265-272.

Price, K.P., X. Guo, and J.M. Stiles. 2001. Optimal Landsat TM band combinations and vegetation indices for discrimination of six grassland types in Eastern Kansas. In press - International Journal of Remote Sensing.

Wang, J., K.P. Price, and P.M. Rich. 2001. Spatial patterns of NDVI in response to precipitation and temperature in the central Great Plains. (Accepted for publication) - International Journal of Remote Sensing.

Whistler, J. L., X. Guo, M. E. Houts, K. P. Price, E. A. Martinko, D. DePardo, 2001. Developing a multispectral data acquisition system for capturing hi-resolution airborne digital imagery. ASPRS 2001 Annual Conference, St. Louis, MO, April 2001

Yu, F., K. P. Price, R. Lee, J. Ellis, and P. Shi. 2000. Analysis of the relationships between climatic variation and seasonal grassland development in central Asia. ASPRS 2000 Annual Conference, Washington, D.C., May 2000.

For more information, visit www.kars.ukans.edu or contact:

Theresa Crooks
Project Coordinator, Great Plains RESAC
KARS Program
Phone: (785) 864-7369
E-mail: tjcrooks@ukans.edu

Top

Home