Kansas Academy of Science

Modeling spatial dimensions of Bison preferences on the Konza Prairie landscape ecology: An overview

M. Duane Nellis1 and John M. Briggs2

Office of the Dean1 and Division of Biology2, Kansas State University, Manhattan, KS 66506 (mdngeog@ksu.edu)

This article is published in the Transactions of the Kansas
Academy of Science, vol. 100, no. 1/2, p. 3-9 (1997).

Table of Contents
Introduction Discussion
Study Area Summary & Conclusions
Related Research References


Within the Konza Prairie, Kansas (a Nature Conservancy Preserve), grazing, primarily by bison, directly affects primary production, nutrients, organic matter, species composition, and to a degree, drainage and depositional networks. Geographic information systems (GIS) and remote sensing have been observed twice per week from March through November. Through GIS modeling approaches, we determined bison preferred watersheds previously burned (late March/early April) in April through June, with soil (which influenced vegetation productivity) a more significant landscape component in bison grazing pattern from July through autumn.


The concept of landscape ecology emphasizes broad spatial scales and the ecological effects of the spatial scales and the ecological effects of the spatial patterning of ecosystems. Within the context of the Konza Prairie Research Natural Area, Kansas, most of the landscape ecology research focuses on development and dynamics of ecosystem spatial heterogeneous landscapes and, more importantly, how spatial heterogeneity influences and effects biotic and abiotic processes within the tallgrass prairie.

In order to facilitate quantitative measures of the tallgrass prairie landscape ecology, particularly as it relates to impact of bison (Bos bison), we have used remote sensing and geographic information systems (GIS) for capturing, manipulating, processing and analyzing spatial data. In our paper we provide an overview of selected examples of the use of remote sensing and GIS in an attempt to characterize bison grazing spatial patterns.

Return to Table of Contents


The Konza Prairie Research Natural Area is part of the Flint Hills physiographic region, an area of dissected uplands made up of chert and flint-bearing limestone layers interbedded with shale. The ridges are usually flat with shallow, rocky soils, whereas the larger and wider valleys have relatively deeper permeable soils. Local relief is as much as 130 meters in an area that receives approximately 830 millimeters of precipitation per year.

The Konza Prairie is an area of slightly less than 3487 hectares (8,600 acres) owned by the Nature Conservancy and leased to the Kansas State University Division of Biology. An important component of Konza Prairie research efforts relates to long term ecological research (LTER). The LTER program is funded by the National Science Foundation, and was initiated in 1981.

Although Andropogon gerardii is the dominant grass, there are over 60 species in the C4-dominated grassland. Galley forests (thin bands of forest along stream channels) dissect Konza with the dominant species usually Quercus macrocarpa.

Bison (Bos bison) were reintroduced to Konza in 1987. The stocking rate gradually increased to 5 ha AU-1 by 1992 through natural herd growth (Hartnett et al., 1996). Within the regional construct, the central research theme for Konza is that the structure, function and dynamics of the tallgrass prairie ecosystem are products of multiple limiting resources which vary in importance in space and time. This variability is caused by nonlinear interactions among three key factors: climate variability, fire regime, and grazing pressure, as they are expressed across the landscape. Through use of GIS and remote sensing we believe that the spatial variation in these three factors can be characterized across broad landscape units. A selected overview of the relationship between these three factors using GIS and remote sensing follows.

Return to Table of Contents


For the Konza Prairie Research Natural Area (KPRNA), satellite data have been acquired and analyzed at various spatial scales relative to landscape ecological dynamics since the mid-1980's. In addition, GIS landscape layers in both vector and raster format have been completed for soils, surficial geology, topography, watershed burning treatments, hydrologic networks, vegetation condition, and bison locations (Table 1). As illustrated by Nellis and Briggs (1989) and Briggs and Nellis (1991), the spatial scale of these datasets and temporal component influence the ability of researchers to characterize Konza landscape units and their dynamics.

Table 1. Key GIS-Parameters Influencing Bison Grazing Patterns on Konza Prairie
Factor Factor
Vegetation condition Soil quality
Hydrologic network Topography
Watershed burning treatments Surficial geology

Generally, satellite vegetation indices exhibit a strong linear relationship to above ground biomass at the time of peak biomass, but the scale and spectral resolution significantly impacts the strength of this relationship (Briggs and Nellis, 1989; Nellis and Briggs, 1992). At the same time burning of watersheds also affects the nature of the relationship obtained. The impact is further modified by larger ungulates (bison) removing variable quantities of aboveground biomass. Normalized Difference Vegetation Indices (NDVI) on grazed areas reflect this impact and exhibit a lower value relative to ungrazed areas, reflecting differences in standing crop at the time of data acquisition, rather than differences in biomass produced.

Since 1991, bison on KPRNA have been observed twice per week from March through November with one observation during the week in the morning and one in the afternoon to insure accurate and timely sampling. Large-scale aerial photographs overlaid with a 30 meter grid system (rectified using a global positioning system (GPS)) were used for recording bison location during each field observation session. Each 30 meter cell has a unique field within ARCINFO GIS software that corresponds to the grid system on the field sheets.

The collected field data were then entered into ARCINFO via a custom macro (SML) developed by the researchers. Once the bison data were entered into the GIS, the data were then easily summarized for various time periods and analyzed relative to other GIS spatial data attributes (e.g. soil quality or vegetation condition based upon NDVI measurements). Although the complete range of findings from our research are not complete, the following section provides a selected sample review of our analysis results to date.

Return to Table of Contents


Through GIS and remote sensing analysis we have been able to characterize bison grazing preferences with regard to: burned versus unburned watersheds, soil quality underlying the preferred grazing areas, the relationship to biomass (as measured to NDVI), the relation to species richness, and concentration on specific watersheds (as a measure of localization).

From earlier research based on data collected 1989-1991 (Nellis et al., 1992) using the GIS modeling approach, we were able to determine that bison preferred burned watersheds during April, May, and June, but showed less interest in these watersheds later in the growing season. This is probably explained in relation to the dynamics of watershed green-up. Burned watersheds (normal burning is late March/early April) have a more rapid green-up process in contrast to unburned watersheds. Therefore bison move to the tender, green grass as watershed rapidly responds to the field burning process. In contrast, soil, which influenced the potential productivity of vegetation, was a more significant component in bison grazing pattern from July through autumn based on GIS overlaying techniques. Generally, deeper, more productive soils become targets of bison grazing later in the summer. As grasses and forbs on burned and unburned watersheds develop toward senescence, soil factors, which created variability in the overall growth and vigor of the vegetation, tended to be more important in bison preferences. In addition, according to our GIS overlay procedures, access to water had no regular impact on bison grazing behavior.

In comparison with the pre-1992 data, we analyzed a similar series of GIS data sets for April through August of 1993. Although results were consistent with early analysis, bison appeared to stay on burned watersheds for a longer period of time (through August), and to follow watersheds that exhibited the highest NDVI values. Such NDVI values relative to bison grazing pattern are also consistent with findings of Harnett, et al. (1995), who found that bison grazing tended to increase plant species richness, evenness, and diversity on sites grazed by bison over a 4-year study period. In addition, increases in plant diversity components associated with bison grazing were generally greater in watersheds annually burned (as part of a control fire experimental treatment) than in watersheds burned on a 4-year rotation (Hartnett et al. 1996). Part of the longer bison lag on burned watersheds may be related to record rainfall amounts during the summer of 1993.

In order to further verify the relationship between bison grazing and tallgrass prairie heterogeneity we also used the fractal dimension. A test of the fractal dimension of patches using satellite data from the GIS database for multiple years by the researchers also supported these findings. The fractal dimension measurement suggested that bison grazing increased spatial heterogeneity in the tallgrass prairie landscape.

In combination with analysis of bison impact on tallgrass prairie landscape spatial heterogeneity is an assessment of degree of concentration of bison. Using a coefficient of localization that describes the relative concentration of bison within grazed watersheds, it appears through GIS analysis, that there is a general increase in bison concentration during the five month period April through August 1996--see bison trends. Values of localization vary from 0 (even distribution) to 1 (concentrated in one watershed). This particular component of our analysis further confirms bison movements to watersheds of higher NDVI, especially during the late summer grazing period.

Return to Table of Contents


This research demonstrates through selected examples the potential applications of GIS and remote sensing for more fully understanding the relationships between bison grazing and landscape characteristics. Although watersheds burned annually tended to be places of bison preference early in the season, areas of higher soil quality tended to be preference areas in late season. These areas of higher quality soils, and related net higher forage production, also resulted in more localization of bison onto specific watersheds late in the grass growing season.

As we look to the future it will be important to refine our approach. To understand landscape dynamics associated with bison grazing will require more accurate spatial information associated with GIS approaches. Further, to adequately scale up beyond Konza Prairie to the Flint Hills to measure the impact of grazing on this ecosystem, it is imperative that researchers understand the impact of scale in their measurements. At sensor spatial resolutions of greater than 100 meters, drought and fire affect multiple contiguous pixels while grazing occurs on a smaller scale and occurs at sub-pixel scales. While proposed finer spatial resolution sensors could offer a potential solution to largescale grazing analysis, the associated costs seem prohibitive in the near future for use on the bison project. Therefore field observation data and its manual input into the GIS, as well as linkages to Landsat TM datasets (30 meter spatial resolution) will continue to offer important information on the grazing-landscape interface. Overall our research approach demonstrates some examples of how bison seasonal movements and general location in relation to watersheds and sub-watered units relate the watershed burning treatments, soil variability, and grassland species diversity.

Return to Table of Contents


We wish to thank graduate students Ken Marshall, Sean Hutchinson, and Kazi Rahman for their field assistance. We also wish to thank Nature Conservancy and the National Science Foundation through grant #BSR-8514327 for Long Term Ecological Research on Konza Prairie.


Return to Table of Contents

Return to Geospatial Analysis Symposium

Comments to mdngeog@ksu.edu