Changes in grassland management that increase the photosynthetic uptake of CO2 and the subsequent decomposition and stabilization of plant residues in soil, may be a significant C sink option that can be applied to much of the grassland area of the earth at relatively low cost and with numerous environmental and socio-economic co-benefits (Cole et al. 1996, Bruce et al. 1998, Lal et al. 1998, Paustian et al. 1998, Rosenberg et al. 1999). However, the extent to which various grassland management practices that sequester atmospheric C are utilized is not well quantified. This proposal provides an experimental and theoretical framework to develop methods to assess grassland management and the implications on C dynamics and C cycling.
While land use and land management changes are widely recognized as key drivers of global C dynamics (Schimel 1995, Houghton et al. 1999), the role of grassland management has only recently received attention as a substantial potential C sink (Conant et al. 2000, Follett et al. 2000, Sampson et al. 2000). A recent literature review concluded that a variety of management practices including irrigation, fertilization, sowing improved grass and legume species, and improved grazing management all lead to C sequestration in grasslands (Conant et al. 2000). High rates of C sequestration (0.1-3.0 MgC ha-1 yr-1) coupled with large areas responsive to improved management prompted the IPCC special report on land use, land use change, and forestry (Sampson et al. 2000) to conclude that 70 Tg C could be sequestered annually in Annex I countries as a result of changes in grassland management, more than twice what is likely to be sequestered as a result of land use change (30 Tg C yr-1). Intensively managed pastureland is likely to sequester C following improved fertility and irrigation management and implementation of intensive grazing management, but these practices are generally uneconomical in extensively managed rangelands which account for 77% of all grasslands and nearly 20% of terrestrial area. However, extensively managed rangelands have the potential to sequester C in soils with improved grazing management at lower rates (0.05-0.15 Mg C ha-1 yr-1; Follett et al. 2000), but due to their large areal extent rangelands may be capable of sequestering as much as 45 Tg C yr-1 globally (Conant and Paustian 2000). Most current estimates of regional changes in terrestrial C stocks and fluxes are based on highly aggregated, non-spatial land use statistics. Examples include assessments based on forest inventories (Houghton et al. 1999), agroecosystem surveys (Paustian et al. 1997a, Smith et al. 1997) and expert assessment (Conant and Paustian 2000). Employing land survey statistics to address questions about regional land management implicitly assumes that land management change is unidirectional, and is distributed uniformly across time and space. Survey data collected on private land, containing information about specific farms, fields, or forests, is published only in aggregated form. Thus, changes in land management are not discernible directly, and net rather than gross land management change is typically evaluated. Furthermore, because the magnitude of change in C pool size is dependant upon its initial size, incorrect assumptions about the direction of land management change and initial land management will affect results. Aggregated survey data are, therefore, most useful as a first approximation of land management change at any particular location; better land management data are required for more thorough and accurate regional analyses.
Land use/cover information derived from satellite data are increasingly being used to generate model inputs to evaluate regional (Running 1990, Paruelo et al. 1997) and global (Running 1990, Field et al. 1995, Prince and Goward 1995, Williams et al. 1997) primary production, biospheric CO2 exchange (Potter and Klooster 1999), and regional rates of trace gas production (Davidson et al. 1998). A number of aspects inherent to remotely sensed data that make them advantageous for developing land use and land use change data sets are potentially useful for use in detection of grassland management. Remotely sensed data are spatially explicit, broad in extent, uniform for the entire area sampled (following radiometric and geometric preprocessing), repeatable over time, and capable of appraising the entire landscape (Roughgarden et al. 1991). Remotely sensed data, thus, are superior to survey data in a number of respects, some of which allow incorporation of more detailed information, at several points in time, into regional analyses of C dynamics (DeFries et al. 1999). Procedures quantifying land management in a spatially explicit manner, thereby enabling linkages to spatially varying soils, topography, and climate, will enable more accurate and insightful evaluations of the effects of changes in land management on ecosystem C dynamics.
The proposed research will address the pressing need for broad-based assessment of grassland management. The major hypothesis underlying this research is that grassland grazing management is detectable through remote sensing of biophysical responses to management. This research will build upon recent work addressing the effects of land use change on C dynamics (Houghton et al. 1998, Houghton et al. 1999, Schimel et al. 2000) and research incorporating remotely sensed data into ecosystem models (Field et al. 1995, Prince and Goward 1995, Running et al. 1995, Williams et al. 1997, DeFries et al. 1999). Development of methods to assess grassland management using remote sensing will lead to accurate, extensive and timely maps of grassland management. Reliable surveys of grassland management that are spatially explicit can be used to drive ecosystem C models, which will advance capability to assess the impacts of land management decisions on the global C cycle.
1.2 HypothesesMy global hypothesis is that different types of grazing management influence NPP which is detectable using remote sensing of plant community biophysical characteristics. Different types of grazing management are likely to affect various biophysical parameters (Absorbed Photosynthetically Active Radiation (APAR), Leaf Area Index (LAI), and light-use-efficiency (e)) used to estimate NPP based on remotely sensed data; I hypothesize that these parameters will vary in a predictable manner in response to different types of grazing management.
Q1: Can pastures under intensive rotational grazing be distinguished from pastures that are managed less-intensively using remote sensing? H1: Pastures under intensive rotational grazing are (1) more productive (higher NPP) and (2) have a different seasonal distribution of aboveground biomass and can, thus, be identified using remote sensing. Rationale for hypothesisIntensive rotational grazing is widely believed to increase grassland forage production by ensuring more uniform forage removal and allowing a recovery period (Gammon 1978). A recent review found that rotational grazing in dry rangelands does not influence forage production, but in more humid regions, forage production increased by 20-30% (Holechek et al. 1999). There is a long history of using rotational grazing to increase production (Hudson 1929, Haynes and Neal 1943) and in some areas, such as New Zealand and Australia, rotational grazing is widely used (Gifford et al. 1992). However, the use of rotational grazing in the southeastern United States has not been well researched.
A key component of the benefit of intensive rotational grazing is more efficient use of forage that is produced. More frequent forage removal keeps plants from reaching slower growth phases associated with leaf maturity (Gifford and Marchall 1973). Therefore, while intensive rotational grazing increases annual forage production, standing aboveground biomass may actually be equal or greater under traditional, non-rotational grazing (Fig. 1; Chapman and Lemaire 1993). Seasonally integrated or one-time measurements of vegetation indices, such as the Normalized Difference Vegetation Index (NDVI) or soil adjusted vegetation indices (SAVI) are unlikely to be useful in identifying pastures under intensive grazing management. However, more frequent measurements throughout the growing season would enable quantification of changes in biomass over time and total biomass production.Removing aboveground biomass through grazing reduces LAI and APAR and should affect NDVI. Grazing decreases the gross photosynthetic capacity of plants, but prompts compensatory photosynthetic rates in remaining tissue exceeding that in ungrazed plants of the same age (Richards 1993). Reduction of photosynthetic capacity following defoliation can occur if damage is substantial or recurring, but defoliation through grazing generally slows or reverses declines in photosynthetic capacity associated with leaf senescence (Gifford and Marchall 1973, Richards 1993). Thus, the immediate effects of grazing are to decrease APAR and to increase e, both of which are important components for estimating NPP using remote sensing (Field et al. 1995, Prince and Goward 1995).
Experimental ApproachIntensive site-level measurements, including ground-based and remotely sensed measurements, and surveys across management and climatic gradients will be used in tandem to assess the effects of grazing management on remotely sensible variables. Remotely sensed data from various sources will be used in accordance with scale requirements of management units in different regions. All measurements are designed to (1) detect grazing management system and grazing intensity using remote sensing, (2) quantify NPP and management impacts thereon, and (3) generate data for use in assessing soil C and management impacts.
Seasonal (four to five times per year) and integrated annual ground-based measurements of NPP (above- and belowground), LAI, APAR, e, and reflectance will be conducted at all sites using standard exclosure-clipping and quantum measurement techniques. Spectrally-derived vegetation indices such as NDVI, which can be correlated with biophysical properties such as green biomass, LAI, and APAR (Tucker et al. 1986, Gamon et al. 1990), will be generated using these data collected at each site. Analyzing frequent multi-temporal satellite data using the relationships generated, will enable derivation of broad-scale information on community growth characteristics (Eidenshink and Haas 1992, Loveland et al. 1995, Eve and Peters 1999) and, thus, grassland management. Generating relationships between remotely sensible biophysical variables and management based on ground-based measurements will enable detection of grassland management using remote sensing.
The ability to observe grassland management using remote sensing is scale dependant. Management differences in rangelands in the drier western United States will be uniform for larger areas, and therefore management differences are more likely to be detected using coarse data. In the Southeast, however, fields are much smaller and coarse datasets limit ability to accurately distinguish different types of land use. Therefore, I plan to use a variety of sources of remotely sensed data depending on the scale of interest and the type of management assessment desired. The relatively fine resolution and very frequent sample interval (every one to two days) of the Moderate Resolution Imaging Spectroradiometer (MODIS) make it ideal for detecting different types of grassland management in pastures. Ground-based data will be used as training data to generate MODIS-derived estimates of pasture management using data acquired electronically through the MODIS Distributed Active Archive Center (DAAC). Likewise, linking ground-based measurements to spatially (Advanced Very High Resolution Radiometer) and temporally (LANDSAT-TM) coarser-resolution data in western rangelands will enable detection of range management and range condition using remotely sensed variables. Systematically corrected AVHRR and LANDSAT images will be georectified and evaluated for possible atmospheric attenuation; adjustments will be made where needed.
Broad- and fine-scale gradients have been used to investigate the influence of a number of variables on grassland ecosystem processes (e.g. Parton et al. 1987, Burke et al. 1989, Schimel et al. 1994, Epstein et al. 1996, Paruelo et al. 1997, Epstein et al. 1998, Paruelo et al. 1999). Grassland responses to grazing are dependant on intensity of grazing, but may also be affected by other ecosystem characteristics, such as competition and fertility, and may be species-specific (Oesterheld 1991, Alward and Joern 1993). Large variability in grassland productivity and variable responses to grazing across environments make assessments using fine-scale remotely sensed data difficult (Dyer et al. 1991b, Turner et al. 1992). Assessing grazing management across a variety of paired sites of differing grazing intensity using moderate resolution remotely sensed data should minimize climatic, topographic, edaphic, and seasonal variation and will minimize problems arising from inherent ecosystem variability.
A number of grassland research sites exist across the Great Plains on which grazing has been eliminated or limited (Fig. 2). While it is impractical to intensively study all of these sites, paired grazed and ungrazed sites can potentially be identified in many of these locations. Information about historical grazing practices within and around these sites will be collected from site scientists and local NRCS personnel. Site histories and current practices will be used to evaluate the influence of (1) current grazing (2) long-term exclosure, and (3) long-term heavy grazing on current productivity, peak standing biomass, interannual variability in standing biomass, and plant cover. Once suitable comparative sites are identified, MODIS data will be used to develop LAI and APAR estimates from NDVI data for sites with varying grazing intensity. Relationships between NDVI and plant growth parameters will be verified, when possible, with ground-based measurements of NPP, LAI, or APAR. Once methods have been established and assessed using MODIS data, three Landsat 7 images acquired throughout the 2001-2002 growing seasons will be utilized to estimate plant biomass and aboveground productivity for selected sites. Relationships between NDVI and NPP, LAI, and APAR developed using MODIS data are not transferable to Landsat TM data since bandwidths do not completely overlap. However, the techniques described above can be used to derive similar relationships. The ability to remotely sense grazing management using fine-resolution data has important implications for sensitive areas such as riparian zones and smaller land holdings, and may provide a new management tool to livestock producers similar to precision farming techniques used in crop production.
