Classification is performed at two different scales, characterized by the size of their minimum mapping unit: 160 square meters for watersheds, and 40 square meters within 30-meter riparian zones. For source data, high-quality historical digital orthophotos are locally produced using existing aerial photography and a unique softcopy photogrammetric solution. Land use/land cover is interpreted from the orthophotos in a GIS environment using on-screen digitizing techniques. Data are interpreted from aerial photography dating back to 1937 (the year of the first statewide aerial survey of Wisconsin).
Upon completion of the GIS land use/land cover database, we will attempt to correlate the observed changes in land use/land cover over the prescribed time period with changes in lake chemistry for the lakes within the study watersheds. In addition, the database will be utilized to create a landscape-level model to capture the nature of the observed landscape changes over time. It is hoped that this model could be used to project future land use/land cover changes within the study regions under alternative planning scenarios.
The North Temperate Lakes LTER (NTL-LTER) site was established at the onset of the LTER program, and augmented in 1994 to expand the spatial scale of the site’s research as a whole. Originally focused on seven core lakes in North Central Wisconsin (Magnuson, Bowser and Kratz, 1984), research at the NTL-LTER site has since expanded to include landscape and regional analysis. Today, the program goals can be best described under the label of long-term regional limnology (Magnuson and Kratz 1999).
Today, the NTL-LTER site is comprised of two distinct research sites: the Northern Highland Lakes District, and the Madison Lakes Region. Despite their relative geographic proximity, the two sites exhibit highly diverse historical, ecological, and socio-economic legacies. A variety of cross-disciplinary research examinations are underway within NTL-LTER to examine the nature of, and relationships between, these differing characteristics. One such examination is the Land Use/Land Cover Change (LUCC) component of the NTL-LTER, which aims to examining the interactions between lakes, their landscapes, and humans over the 20th century in our two study sites.
Objectives
Landscapes, like other ecological constructs, are dynamic in structure, function, and spatial pattern. The spatial configuration of elements within a landscape can be attributed to a combination of environmental and human forces (Forman and Godron 1986). In our particular study sites, these forces can be said to consist of relationships between the landscape as a whole, the lakes themselves, and human interactions with both. The Land Use/Land Cover Change (LUCC) component of the NTL-LTER project is aimed at examining these lake-landscape-human interactions within our two study sites over the 20th century.
Specifically, the LUCC project strives to address the following questions:
Once completed, the GIS land use/land cover database will be used in tandem with an existing database of historical limnological data for the lakes within the study sites to perform various analyses. Utilizing specific landscape metrics, we intend to explore the nature of the relationship between the spatial/temporal changes in the landscape and the dynamic historical water chemistry data.
Ultimately, the land use/land cover data will be used to model the evolution of landscape change in northern and southern Wisconsin, and to project these models into the future. This will provide an understanding of the changes that we can expect in the study areas under various alternative future planning scenarios. Although the focus of the modeling will be on landscape change as it affects aquatic ecosystems, the value of the models will be largely heuristic – that is, they will help determine the importance of various drivers in determining changes in the landscapes in question, and suggest new questions and directions for further research.
Project Impacts
With over 15,000 lakes, and countless streams and wetlands, aquatic ecosystems are critical to the state of Wisconsin, historically, economically, and environmentally. In a state that bases a sizable fraction of its economy on the recreational values of its lakes, hydrologic resources play a critical role in the state’s well being (Riera 2000). It is of interest, then, on many levels, to quantify the effects of human activities on the region’s lakes, and vice-versa.
Additional project impacts may stem specifically from the unique methodology employed in this study. By developing and utilizing an in-house softcopy photogrammetric solution for producing historical orthophotos, we were able to create a low-cost and highly accurate GIS database of historical land use/land cover change for the research sites, including high-resolution data within riparian zones. This method minimizes many of the traditional shortcomings associated with the use of historical aerial photography as a data source for landscape-level research. The methods employed here could serve as a model for other entities interested in asking geospatial questions about a given landscape over a broad temporal scale.
Finally, the expanding availability for research applications using high-resolution remotely sensed data only serves to increase the value of just such a historical record of landscape change. With well-identified and frequent sampling intervals, the satellite platforms of tomorrow promise to support and enlighten the next generation of land use/land cover change examinations, both at our study sites, and around the world. However, just as current historical landscape examinations are limited by the scarcity of aerial photography prior to the 1930’s, future researchers will be limited by the fact that high-resolution orbital data were virtually non-existent in the 20th century. Improved facility for the creation of accurate, high-resolution historical land use/land cover datasets will serve as a complement to the landscape data of the future.
Methods
Study Area
The North Temperate Lakes (NTL) LTER site is comprised of two distinct research sites: the Madison Lakes Site, and the Northern Highland Lakes District (Figure 1). The two landscapes differ in their natural settings, their histories, and, presumably, their future development. In many ways, they represent the two extremes in the spectrum of Wisconsin landscapes, and therefore, we believe, provide and informative contrast of the range of interactions that occur between lakes and society.
These watersheds were selected to represent a cross section of 20th century South Central Wisconsin landscapes. The Upper Yahara Watershed represents a landscape that has been, and continues to be, largely agricultural. The Lake Wingra Watershed and Spring Harbor Storm Sewershed represent areas of transformation from agricultural to urban/suburban land uses. All three have exhibited different rates and types of landscape change in the past, and promise to do the same in the future.
The Northern Highland Lakes District is located in North Central Wisconsin. It is composed of seven core lakes located within a glacial landscape characterized by abundant forest cover, wetlands and lakes. As such, it is hoped that our findings will be easily transferable to adjacent lake districts in the Great Lakes Region. Today, the area is largely a recreational district, where the lakes and dense forest cover are the main attractions. The LUCC group is currently in the process of selecting specific watersheds for use in our evaluation.
Data sources
The interpretation of historical land use/land cover for the study areas was conducted using aerial photography. Unites States Department of Agriculture (USDA) black-and-white panchromatic photography was acquired for the years 1937 (the year of the first Wisconsin statewide aerial survey) and 1962. In addition, digital orthophotos were acquired for the study areas from Vilas and Dane Counties for the year 1995.
Current statewide digital hydrography data was acquired from the Wisconsin Department of Natural Resources (DNR) for the entire state, and subsequently improved to a sufficient level of resolution based on interpretation from the aerial photography. Historical hydrography layers were interpreted from the photography with the aid of the current DNR coverage. Digital hydrography layers were then used within a GIS environment to create the 30-meter riparian buffer zones for the three dates in question.
Thanks to the efforts of early limnologists Birge, Juday, and others, water chemistry data exists for over 500 lakes in the Northern Highland Lakes District between 1929 and 1941. Other regional surveys covered the area in the 1950’s, 1970’s, and 1980’s. The Madison Lakes area is centered on Lake Mendota, home to the University of Wisconsin, where the first limnology program in the United States was established in the 1880’s. Therefore, there is no shortage of historical limnological data for the Madison Lakes region. In fact, Lake Mendota itself has been called “the most studied lake in the world.” This wealth of historical limnological data at both of our study sites is invaluable to our project.
Digital Elevation Models (DEM’s), necessary for ortho-rectifying the historical aerial photography, were acquired from the USGS. The pixel size is 30-meters, a resolution fine enough to capture the subtle variations in the topography of the two sites.
Data generation
Although interpretation of aerial photography has historically been the most effective and accurate data source for obtaining temporal land use/land cover information (Lindgren 1985), one major drawback of this methodology is that problems with registration and distortion may be such that successive years cannot be readily superimposed or accurately analyzed (Turner 1991). This project’s specific needs for high-resolution data within riparian zones, and reliable assessment of change within these zones, served to further underscore this potential methodological shortcoming.
To remedy this problem, a softcopy photogrammetric software solution, developed by Dr. Frank Scarpace of the Environmental Remote Sensing Center (ERSC), University of Wisconsin-Madison, was employed. The program (OrthoMapper) uses the historical aerial photos (scanned at 700 dpi), a Digital Elevation Model, and an accurate map, orthophoto, or Digital Line Graph (DLG) of the area to be represented in the orthophoto. Together with information regarding the geometry of the original camera (either supplied or assumed), the software is able to create accurate digital orthophotos from historical aerial photography (Figure 2).
In the interior orientation phase, the user simply selects the four fiducial marks on the digital photograph (in cases where no fiducials exist on the photograph, as was the case with our 1937 photography, the four corners of the photographs may be used). This process establishes the location of the center of the photograph, and the position of the camera at the time of the photograph.
In the visual orientation phase, the user matches up identifiable ground control points from the historical photos with the same points from the map, orthophoto, or DLG. A minimum of four points is necessary, and a uniform spatial distribution of points is required to effectively perform the transformation. An RMS error value informs the user to the “goodness of fit” of the transformation performed, and points may then be edited, deleted or added accordingly.
A specific need of this methodology is the existence of identifiable ground control points that have remained constant over the time period in question. In our case, the centers of road intersections were found to be the most consistently reliable stationary points. In areas without a road network, natural features such as individual trees or boulders can be used as control.
Once the high-resolution digital orthophotos have been created and mosaicked to provide full coverage of the watershed at hand, the next step is to interpret the landscape. Land use/land cover is interpreted into a GIS database using ESRI ArcView (v. 3.1) software and on-screen digitizing techniques. Stereoscopic viewing of hardcopy photography is performed to aid in this interpretation.
To allow for specific analysis of the effects of changes occurring specifically within the riparian zones of the study areas, the photo interpretation is performed at two different scales, characterized by the size of their minimum mapping unit: 160 square meters for watersheds, and 40 square meters within riparian zones. Riparian zones are defined as 30 meters on each side of a stream, or the shoreline of a lake, and are delineated by creating a GIS buffer layer around the digitized hydrography layers for each year.
For an interdisciplinary research project such as this, a knowledge of both land use and land cover can be important for understanding the landscape (Lillesand 1994). For this project, a set of mixed land use/land cover classes is employed for both regions with land-water interactions in mind. These classes include urban (high- and low-intensity), agricultural (row crops and forage), grassland, forest (evergreen, deciduous, aspen, tree farm, clear-cut and mixed), wetland (forested, shrub, and non-forested), cranberry bogs, barren land, farm complex, and open water.
To address changes in lake chemistry, we draw upon the long legacy of limnological research at both of our sites, which were first studied by Birge, Juday and collaborators at the beginning of the 20th century (Beckel 1987). Thanks to this research, and other studies, we have direct limnological data for hundreds of lakes in the study sites dating back to the 1930s.
In addition to a spatial-statistical analysis, it is hoped that the land use/land cover data from this project will be used to inform a landscape model for the region, which can be used to project landscape change into the future under various alternative planning scenarios. Although specific analysis and modeling methodologies have not yet been set, measures of temporal landscape change can be described in terms of changes in patch number, patch size, number and type of corridors, number and type of barriers, probability and spread of disturbance. Land use/land cover types can be assigned weights within a GIS environment based on factors like percent impervious surface, or amount of nutrients per unit area, etc. In terms of modeling, traditional stochastic and process-based landscape modeling methodologies, as well as neural network landscape models, are being examined for their efficacy to these particular landscapes.
Preliminary findings
To date, we have finished creating the historical GIS land use/land cover database for the Madison Lakes Region, and we are in the process of performing preliminary analyses the impacts of these landscape changes on the area lakes using historical limnological data. We are also examining several landscape-level modeling methodologies for their utility in modeling the long-term landscape change within the region.
We have found that, in general, over the sixty-year period, sub-watersheds located in the Madison Lakes Region have exhibited a great loss of agricultural land at the expense of suburban development (Figure 3). Riparian buffer zones have decreased in area in some cases, but forested area has increased within riparian zones. Transformations within this region have been occurring more dramatically at the watershed scale. That is, upland landscape transformation far outweighs changes occurring at the riparian zone scale. This kind of landscape transformation can have major impacts on nutrient loading in receiving waters (Perterjohn 1984).
Acknowledgements
I am grateful to Dr. Tom Lillesand, Dr. Monica Turner, Dr. Frank Scarpace, and Dr. Joan Luis Riera, for all of their help. I would also like to acknowledge the Environmental Remote Sensing Center (ERSC), Robinson Map Library, and the Institute for Environmental Studies (IES) at the University of Wisconsin-Madison, as well as the Unites States Geological Survey (USGS), and the Wisconsin Department of Natural Resources (DNR). This work was supported by a grant from the National Science Foundation (NSF), under the LTER program.
References
Beckel, A.L. 1987. Breaking New Waters. Wisconsin Academy of Sciences, Arts, and Letters. Madison, Wisconsin.
Christensen, D.L. 1996. Impacts of Lakeshore Development on Coarse Woody Debris in North Temperate Lakes. Ecological Applications, 6:1143-1149.
Forman, R.T.T., and Godron, M. 1986. Landscape Ecology. New York: John Wiley and Sons.
Lindgren, D.T. 1985. Land Use Planning and Remote Sensing. Dordrecht, Netherlands: M. Nijhoff Publishers.
Lillesand, T.M., and Kiefer, R.W. 1994. Remote sensing and Image Interpretation. 3rd Ed. New York: John Wiley and Sons.
Magnuson, J.J., C.J. Bowser, and T.K. Kratz. 1984. Long-term ecological research on north temperate lakes (LTER). Verh. Internat. Verein. Limnol., 22:533-535.
Magnuson, J.J. and T.K. Kratz. 1999. Lakes in the landscape: approaches to regional limnology. Verh. Internat. Verein. Limnol., 27:1-14.
Peterjohn, W.T. and Correll, D.L. 1984. Nutrient Dynamics in an Agricultural Watershed: Observations on the Role of A Riparian Forest. Ecology, 65:1466-1475.
Riera, J.L. , S.R. Carpenter, T.K. Kratz, T.M. Lillesand, J. Schnaiberg, M.G. Turner, P.R. Voss and M.W. Wegener. Nature, Society and history in Two Contrasting Landscapes in Wisconsin, USA: Interactions between Lakes and Humans during the Twentieth Century. Land Use Policy.
Turner, M.G. and Gardner, R.H., ed. 1991. Quantitative Methods in Landscape Ecology. New York. Springer-Verlag.