Landscape Ecology

Landscape ecology is the study of pattern and structure across dynamic temporal and spatial scales. Patterns are created through biotic and abiotic processes and disturbances that occur within the environment. As changes throughout a landscape occur, the overall structure of ecological communities is affected, and thus revealing the need for landscape-level studies. According to Turner, "landscape structure must be identified and quantified…before interactions between landscape patterns and ecological processes can be understood (1989).” In order to understand landscapes, ecologists must be able to recognize the patterns that are created along both spatial and temporal scales. However, the ability to recognize and understand these patterns becomes more difficult to accomplish due to scale-related constraints.

There are several theories and techniques that can be employed in allowng us to understand landscapes. Since landscapes are very much dependent upon spatial and temporal patterns, the scale of such patterns must be considered (Turner, 1989). In addition, the relatively recent developments of many techniques such as landscape models and metrics give ecologists the ability to examine spatial and temporal patterns. The dynamic processes that take place at spatial and temporal scales are of importance to landscape ecology as the main causes of landscape patterns. The significance of addressing scale using these techniques are vital to many ecological studies, including landscape ecology, restoration ecology, and conservation ecology. The following will be a discussion of the advancements that have been made in landscape ecology through the applications of scale theory and one method, the geographic information system (GIS).

GIS and Landscape Ecology

In brief, a GIS is essentially a tool that may be used to store, manipulate, analyze, and display spatial data (Longley et. al, 2001). The possibilities that are presented through using a GIS can be illustrated by its many current uses, which ranges from simple mapping as found on www.mapquest.com to the digitizing of watershed management areas by state governments (www.state.nj.us/dep/gis/). Just as an electron microscope allows biologists to view certain organelles at the cellular level, a GIS allows landscape ecologists to view patterns at the landscape level. One simple requirement that must be considered when using GIS is that elements must be spatial. This unique property of GIS allows users to take into consideration scale and allow for scale theory to be applied to the study. This can be exemplified by the spatiality of all landscape studies. More importantly, the spatial component of GIS can then solve the difficulties ecologists face while attempting to simulate and test spatial hypotheses.

Data Modeling

Some models may be as simple as displaying land-use changes over time, while other analyses may show predictions, which support ecologists’ hypotheses. Several unique characteristics of most GIS packages enable certain types of studies to take place.

Overlays

For example, the ability for landscape ecologists to monitor pattern and process changes is demonstrated by Kienast of the Swiss Federal Institute of Forest Snow and Landscape Research in Switzerland (1993). The goal of the project was to conduct a temporal study of landscape history change using different methods, including a spatio-temporal model, landscape indices, and traverse analyses. The use of the latter two methods allowed Kienast to compare his GIS data model to other landscape methods. The ability of GIS to store data layers that can be imposed on top of each other allows for the same area of a map to be analyzed for changes over time. As changes are detected using the data processing tools within the GIS, new maps may be produced showing the locations of these changes. Although statistical analyses of fractal dimensions and traverse analysis allowed for certain calculations and measurements to be performed, and besides being time consuming, they did not reveal any information about the spatial changes that took place in relation to this landscape model. In this case, the GIS model proved to be the more reasonable tool for monitoring landscape changes over time.

Buffering

Cherrill and McClean used a buffering tool in their study of mapping habitats while detecting land cover change. Just as Keinast encountered issues with temporal scale while measuring land use change, the same problems were addressed here as well. To determine the correct width of buffering that would allow them to still recognize land cover changes at their extent of study, buffers from 5 to 25 meters were created to compare a 1992 field survey to a 1991 field survey. Differences in land cover changes revealed errors at each buffering interval. For example, many ecological processes such as old-field succession may only be detected after a few years and not in a single year (Barbour et. al., 1999). Cherril and McClean measured percentages of changes that were not possible and found that at the 25-meter buffer, 35% of land cover change could not have occurred, while at the unbuffered control, 41.2% of land cover changes could not occur. A decrease of almost 7% in error was produced using this buffering technique unique to GIS. The use of the buffers in larger sizes allows for detection of meaningful land cover changes at this scale, rather than include these impossible changes (Cherrill and McClean, 1995).

Attribute Editing

Another tool located within the GIS framework is the ability to add, delete, and edit attributes of data layers. Since there differences in interpretation of land covers during each field survey of 1992 and 1991, certain standards were needed to define each land cover category. After buffering, land cover fields were combined and aggregated at more a more coarse scale and more similar land cover types were considered as a single unit. The results of completing such a task led to a 27% decrease in area discrepancies (Cherrill and McClean, 1995). It is important to note that we then lose sight of finer scale patterns like distinguishing streams from rivers and other waterway networks. Nevertheless, the land cover scale at this level of study only considers type of land cover and not more specific.

Predictive Modeling

It is suggested that GIS may best be considered as a tool to be used in a study, rather than base the entire study on the GIS. This will prevent certain studies from becoming dependent upon the capabilities of the GIS. GIS has been widely used as a support system for simulation models and other investigations. In governmental applications, US Census data is used to map demographics and support policy decisions such as redistricting of congressional voting areas. GIS is just another tool, in addition to the hard numbers provided by the Census data that facilitates policy-making. This lesson also applies to landscape ecology. Ecologists are armed with a multitude of tools that enable them to support their studies. Advancements in methods such as fractal geometry, remote sensing, and GIS allow for models to assist ecologists in making predictions and performing hypothesis testing.

There are some applications of GIS that involves more than just displaying spatial data. This is especially noted when difficulty is encountered in conducting controlled experiments in real-world situations due to certain political, economic, social, and logistical constraints (Gustafson, 2001). Predictive modeling may be employed through the use of GIS for studying landscape-level processes and even decision making.

Modeling with Scale

Another study was conducted that relied heavily upon the capabilities of the GIS at hand. Scientists at ManTech Environmental Technology, Inc., in conjunction with the US EPA Environmental Research Laboratory, conducted an investigation into the use of GIS to monitor rice production and climate change and the data required for accurate portrayal. Information from monitoring stations throughout Asia and remotely sensed data was integrated into attribute layers of topography, hydrology, temperature, precipitation, and humidity, and several analytical techniques were conducted within the GIS that provided more data of rice production and climate of the study area. Through the use of a GIS tool, interpolation, the authors were able to create distribution maps of climate as a factor of rice production using the data from 1274 meteorological stations in Asia. One stipulation to this process, however, is that the level of detail would be greater if there were more data points. For example, fine-scale differences in temperature would be detected if there were more meteorological stations located in Asia. The authors do note that the purpose of the map is the most important factor when determining how many data points should be used or what scale should be resolved (Bachelet et. al., 1993). Subtle differences would be noticed and unnecessary because the scope of this study was conducted at the continental level. The reliability upon the data portrayed through the interpolation and other analyses is a limiting factor when using GIS.

Supplemental GIS

Vegetation is more easily studied using GIS due to the ease of mapping and detecting changes throughout the landscape. Wildlife can also have a major impact on landscapes. GIS can also be used as a support system for studying wildlife at the landscape scale. For example, the distribution of the Gila monster lizard in Utah was mapped using GIS (Longley, et. al., 2001). Comparisons of the range and protection area of this species were then applied to management decisions. One would assume that mapping distribution of wildlife would be a simple locate and plot procedure, however, applying further knowledge is needed to produce a model that would assist ecologists in hypothesis predictions.

This method has been tested with respect to bird distributions using GIS and Bayes Theory. The Bayesian rule-based approach allows for determining distribution of three bird species based on probability of encountering a species and known preferences of a species (Tucker et. al., 1997). The authors identified quantifiable habitat preferences and other distribution factors of three bird species and developed map layers in a GIS using Bayesian formulas, distribution factors, and satellite base-maps. Analyses were conducted and the authors concluded that this modeling could be used to investigate effects of land use policy at larger scales within the area of study.

Further Considerations

Map Scale

Ecologists of the USDA Forest Service in Ohio used GIS as a supplement for their investigation of forest composition and productivity of Ohio forests. Their focus was on developing an integrated moisture index (IMI) using GIS to predict forest productivity and species composition. Using digital elevation models (DEM) at different map scales would then assist in determining at what spatial scales management should take place by displaying what patterns are detectable at each scale. Previous studies have discovered that distribution and growth of trees in Ohio may be correlated with local topography and soils, but difficult to map (Iverson et. al., 1997). IMI is a moisture index that predicts forest site productivity and composition based upon direct solar radiation, slope, curvature, and water holding capacity. Applying IMI to DEMs within a GIS allows the users to make these predictions about forests in Ohio. IMI was derived for Vinton Furnace Experimental Forest in Ohio at the following map scales: digitized 7.5 minute USGS topographic map, 7.5 minute DEM at 1:24,000 scale, USGS digital line graph at 1:100,000, and 1:250,000 scale USGS DEM. After several edits were conducted to address the change in map scale, it was found that at the 250K scale, ridges and valleys were misrepresented resulting in inaccurate IMI. At this coarse scale, predictability of forest composition is not feasible. The authors concluded that the other map scales were suitable for predictions based on correlation and statistical analyses (Iverson et. al., 1997). The beauty of this study is that no field assessments or current vegetation maps were needed. The use of remotely sensed data and GIS allows ecologists to accurately predict forest composition and productivity at the landscape scale using maps up to and including 1:100,000 scale.

Caution Using GIS

While the advantages of using GIS for landscape ecology studies cannot be ignored, there are definitely flaws as well. The quality of data output from the GIS is highly dependent on the quality of the data input. In Kienast's report, this can be mainly attributed to the digitizing of landscape elements. As in other landscape studies, detection of such elements depends highly on the definition of a boundary and what scales are considered. Urban et. al point out a fact that boundaries might be dependent upon topographic features at the landscape scale (1987).

Other sources of error within GIS that have potential impacts upon landscape patterns recognized, including inaccurate display of landscape elements from input maps, editing errors, and overlay algorithms (Kienast, 1993). Fortunately, there are many features within a GIS that can be used for correcting such problems. As mentioned before, the identification of patches and boundaries will have an effect on the landscape pattern or process seen. To control this effect, a GIS is able to create buffer zone areas around such polygons. This allows the user to concentrate their analyses more on the core of the data rather than discrete polygons.

Ecologists also recognize that disturbances within a landscape and the effects are important but difficult to predict (Turner, 1989). This has been demonstrated in one study where GIS was used to represent wildfires. Different data layers were used to represent points in time and locations of wildfires. This framework of spatio-temporal modeling based upon the GIS limited the scope of the wildfire study. The author of one article explains that support for modeling fires as a single entity is problematic as each wildfire is different in temporal and spatial scale. Difficulties in linking the different temporal and spatial components through queries are also encountered (Yuan, 1997).

While the inaccuracies of each map showed that data input accuracy is crucial to a GIS, there are ways to control errors. Cherril and McClean encountered difficulty in identifying land cover changes. They were able work at more coarse scales for more accurate measures, but they lost ability to detect fine scale changes.

Most of the problems encountered throughout the avian habitat study involved technicalities within the GIS. Inaccurate habitat variables were identified as one source of error. The high mobility of birds and coarseness of satellite data led to difficulty in determining habitat area and boundaries and although this problem is seen in other avian studies, the spatiality of GIS presents a major limitation. Another scale-related error source stemmed from defining the scale of the study itself. The authors detected habitats at 1 km, 2 km, and 10 km grain sizes, but should they produce results for the 1 km, 2 km, or 10 km scale? They concluded that in order to minimize the inaccuracies in probability, they should analyze and display the data at the coarsest resolution. In essence, they averaged out the details while maintaining the larger picture. Even though their original goal was to determine predictions to be made at local scales, this GIS-based model will be useful only at a larger scale.

Conclusion

The importance of GIS to landscape ecologist cannot be ignored. Presented above only highlights a few uses that may be employed to landscape studies. It does not matter whether the subject of study pertains to land use patterns, agriculture, land cover, vegetation, or wildlife distribution. GIS should always be included as an option to landscape ecologists. It is also important for ecologists to recognize certain errors that may be involved when using GIS for their studies. It is known now that there is difficulty in building predictive models at finer resolutions (Costana et. al., 1994). Scaling up only blinds the conclusions to many details that may be important to the study. Map scale appears to be a continuing issue as well. GIS software will be able to adjust data layers according to appropriate coordinate systems but many times the coordinate system is unknown. Further edits and calculations must be performed which can often take too much time and the costs are too high. The most prevalent error that would be encountered is data source inaccuracy. There are not enough editing and buffering procedures that would prevent these errors from confounding an experiment. In Cherril and McClean's land cover mapping, larger buffers were needed to render fine scale patterns obsolete. Changing scales may assist in avoiding data source errors, however, they would still be present and doing so would only blind the user to them. It is important to note that while these errors will always exist due to differences in scale, GIS should never be totally excluded from being used. GIS is an extremely useful tool that enables landscape ecologists to take into account and address the important issue of scale.

Work Cited:

Bachelet, Dominique, Brown, Doug, and Herstrom, Andrew. "Rice Production and Climate Change: Design and Development of a GIS Database to Complement Simulation Models." Landscape Ecology 8 (1993): 77-91.

Blaiklock, John, Martin, Elaine B., Rushton, Stephen P., Sanderson, Roy A., and Tucker, Kenneth. "Modelling Bird Distributions--A Combined GIS and Bayesian Rule-Based Approach." Landscape Ecology 12 (1997): 77-93.

Cherrill, Andrew, and McClean, Colin. "An Investigation of Uncertainty in Field Habitat Mapping and the Implications for Detecting Land Cover Change." Landscape Ecology 10 (1995): 5-21.

Costanza, Robert, and Maxwell, Thomas. "Resolution and Predictability: An Approach to the Scaling Problem." Landscape Ecology 9 (1994): 47-57.

Dale, Martin E., Iverson, Louis R., Prasad, Anantha, and Scott, Charles T. "A GIS-derived Integrated Moisture Index to Predict Forest Composition and Productivity of Ohio Forests (U.S.A.)." Landscape Ecology 12 (1997): 331-348.

Goodchild, Michael F., Lngley, Paul A., Maguire, David J., David, and Rhind

W. Geographic Information Systems and Science. John Wiley & Sons,

Ltd.,Baffins Land, England, 2001.

Gustafson, E.J. 2001. "Simulating Changes in Landscape Pattern." Learning Landscape Ecology, a Practical Guide to Concepts and Techniques Ed. J.A. Bissonette. Springer, New York, 1997. 49-61.

Kienast, Felix. "Analysis of Historic Landscape Patterns with a Geographical Information System--A Methodological Outline." Landscape Ecology 8 (1993): 103-118.

Turner, Monica G. "Landscape Ecology: The Effect of Pattern on Process." Annu. Rev Ecol. Syst. 20 (1989): 171-197.

Turner, Monica G., Gardner, Robert H., O'Nell, Robert V.,. Landscape Ecology

in Theory and Practice. Springer-Verlag, New York, New York, 2001.

Urban, Dean, O'Neill, Robert V., and Shugart Jr., Herman. "Landscape Ecology: A Hierarchial Perspective Can Help Scientists Understand Spatial Patterns." Bioscience 37 (1987): 119-127.

Wiens, J.A. "Spatial Scaling in Ecology." Functional Ecology 3 (1989): 385-397.

Yuan, May. "Use of Knowledge Acquisition to Build Wildfire Representation in Geographical Information Systems." Int. J. Geographical Information Science. 8 (1997): 723-745.