| Multi-layered and Statistically Based Ecosystem Mapping
The de facto standard for land resource planning in the 21st century
By John D. Beckingham, Michael Desilets, Darrin Nielsen, and Frank Johns As the century nears its close, the pace of industrial development is increasing along with public concern for the environmental integrity of not just the planet, but of small or regional areas as well. Simultaneously, population growth continues along with the trend towards higher densities of urbanization, particularly in developing countries. Concurrently, the accelerated exploration for oil and gas reserves and the exploitation of renewable resources (such as forests) requires that sustainable management strategies be developed. With the restructuring and capitalization of Russia and China and the eventual recovery of Asia, South America, and Africa, we can expect these converging trends to accelerate even further.
There are political, economic, and policy decisions that must be made and public consulta- tions held to discuss the issues surrounding conflicting demands on the Earth's landbase. But, these initiatives must be supported by relevant and defensible scientific data about the specific landbase under discussion-and the ecological function of the landbase in relation to surrounding areas.
In western Canada, sufficient concern over the past 10 years has caused new approaches to be developed that respond to the conflicting demands of industry and public concern about the environment. Geographic Dynamics Corp. (GDC) has created a truly innovative, predictive ecological mapping system, called SiteLogix, to identify functionally different ecosystems at very fine spatial scales of resolution. These ecosystem delineations provide a new framework for organizing and overlaying a large number of ecological variables.
The result of this effort is a predictive ecosystem map that allows efficient and ecologically responsible development of resources and clearer, more accurate communication between resource managers and the public. By the end of 1999, GDC will have successfully applied SiteLogix to over 2 million hectares (one hectare is an area 100x100m and is equal to 2.74 acres) of forest land for five international forestry companies operating in western Canada.
What makes SiteLogix unique is that rather than being a software package, it is a process that integrates GIS applications, statistical techniques, and ecological knowledge in order to create a seamless mapping product based upon statistically justifiable ecological relationships. Ecological Background
Ecological information is usually available in nested levels of detail. The crudest maps have scales of 1:100,000 to 1:1,000,000 and contain information about dominant landforms, major soil types, and geological formations. At the 1:15,000 to 1:20,000 scale are vegetation cover, elevation, and contour maps.
However, the map scale is not necessarily the determining factor in accuracy and usability of ecosystem mapping. The complexity of detail contained in underlying information layers is the key to distinguishing functionally different ecosystems. For example, a forest cover map may be at a scale of 1:15,000, but additional information extracted from a landform map (even though it is at a coarser map scale) may indicate ecosystem differentiation not apparent from the forest cover map. The use of multi-layered ecological information is fundamental to the predictive mapping of SiteLogix and can include such overlying ecological parameters as: landforms and topography; parent geological materials of the main soil types; soil, moisture, and nutrient characteristics; slope, aspect, and elevation; soil drainage characteristics; vegetation cover, including tree species distribution and productivity; climate factors; successional stages; as well as natural and human disturbances of the landscape.
These ecological parameters are used to map functional ecological units. The fundamental ecological unit is named differently in different jurisdictions, but in most mapping completed with the SiteLogix system, the appropriate ecological land unit has been called the ecosite. It can be as small as 1-hectare or as large as 50-ha, depending on the data available and management objectives. Generally, the ecosite is an area where the climate, soil moisture, and soil nutrient characteristics are similar. Ecosite phase is a subdivision of an ecosite and is defined on the basis of the dominant plant species in the canopy. Methodology
Given the vast quantities of ecological data that must be processed, GDC develops custom programming scripts to speed the analytical steps required to complete an ecosite or ecosite phase map in SiteLogix. More importantly, advanced statistical techniques are employed to identify ecological patterns and trends and assign probabilities to the ecological relationships.
The result is a detailed predictive ecosite or ecosite phase map from which many sub-maps can be generated for resource value estimates and management interpretations. Thus far, eight distinct ecological information packages have been identified for management purposes.
To handle the volume of data, SiteLogix employs a series of steps which build on the previous ones in order to produce the most accurate classification possible, within logistical constraints. What makes SiteLogix a particularly efficient process is the custom programming which, at critical steps in the process, connects five off-the-shelf applications: MicroImages TNTmips, TYDAC SPANS, Statistical Analysis System (SAS), ARC/INFO, and ArcView.
Custom scripts are created to automate: digitization of hardcopy maps; importation of digital terrain files; calculation of slope, aspect, and other topographic variables from digital terrain files; organization and classification of ecological plot and polygonal data; statistical identification of interrelationships among the ecological variables; merging of neighboring polygons that had common ecological parameters; and elimination of false precision polygons which had areas of less than 0.5-ha Collecting and Preparing the Data
The key variables are derived from four primary sources of spatial data: vegetation coverage, soils data, digital terrain model, and ecological land classification data. Additional data, which can be statistically modeled and queried, increases the applicability of the final model and include climate data, plant community inventories, habitat and wildlife inventories, disturbance history, management parameters and regulatory constraints, and riparian features and hydrological data.
The majority of the available resource maps and data are in hardcopy; the analysis procedure in SiteLogix demands that all data be available in digital form. Thus, one of the first steps is determining the level of effort required to digitize the existing information. A secondary issue is converting between different file formats in which the ecological information is available. For example, ecological data is usually vector based while topological data is usually raster based. In one project, we used raster-to-vector conversion algorithms to produce a polygon coverage of a mapping area that indexed over 100 ecological and geomorphological variables.
We work with the client to determine the detail required and the accuracy desired for the predictive mapping. At the beginning of a project, this means that we must be able to clearly identify and justify any field surveys. To assist our ecologists in precisely identifying where to most effectively collect field data, we use statistical procedures to select field plots. In one project, 45 unique site types were sampled over an area of 385,000-ha by the strategic placement and description of 585 detailed ecological plots. Sampling intensity for each site type was based on the relative area of coverage, ecological heterogeneity encountered in the field, and the accessibility of the site.
An essential and time-consuming task after the ecologists return from the field is to ensure the validity of the location data. This requires that the ecological data be cleaned, GPS data is differentially corrected, GPS points and map plots are reconciled spatially, and plot locations are linked to digitized land resource data. Creating and Merging Ecological Layers
Digital terrain files are imported into the GIS where hydrology breaklines and a 50 meter grid are used to produce an elevation model, based on a 90% confidence that all elevation values are accurate to within 5m. The elevation model is then used to generate slope, slope curvature, aspect, and other topographic information.
In one project, in order to capture major slope trends, we grouped the slope model into 13 classes. Pixel noise (originating from the fine scale of the model) was removed by using a median matrix filter. Accumulated slope length and downslope curvature models were generated, filtered, and optimized for processing. The vector data were used to create an attribute table with nearly 130 ecological parameters.
Generally, the raster data are vectorized to create a data layer that holds all the topographic variables. These variables are important for classifying and ranking areas according to their potential for moisture retention, water runoff, and nutrient characteristics. Analytical Framework for Linking Spatial to Ecological Data
The data layers are now ready for integrating the GPS data, which provides a digital point reference file that is used to extract thematic information from the digital layers. Tabular databases are created for the ecological parameters, recorded on plot forms, and the data extracted from the GIS. The databases are subjected to statistical tests using Statistical Analysis System (SAS).
SAS is next employed to analyze the ecological plot and polygonal data in order to identify patterns and trends and define statistically significant ecological relationships between the variables. Some of the multivariate techniques used include ANOVA together with cluster, regression, discriminant, ordination, and correspondence analyses. Statistically justifiable relationships, patterns, and trends are identified by our professional ecologists. Based on the results of these analyses, the ecological model is modified to reflect the actual distribution of ecological units on the landscape. The advanced statistical techniques are iteratively applied to fine tune the mapping parameters to the point where further iterations do not significantly change the ecosite map. It is common for 10-30 iterations to be required to optimize the mapping. In this way, the best predictive model, with the available data, can be created. Refining the Mapping
The final steps in SiteLogix involve reducing the number of vector polygons by merging polygons that have shared ecological attributes. In one project, the number of polygons in the modeling area was close to 2 million. After merging, only 20,000 polygons remained.
The usual map scale is between 1:15,000 and 1:50,000. At these scales, polygons with areas of less than 0.5 ha are not appropriate. To eliminate these minimum size polygons, another merging algorithm is run, based (again) on shared ecological attributes of adjacent polygons. Mapping Products
The unique nature of SiteLogix is threefold: predictive accuracy of the resulting ecosystem map can be determined, geomorphological alometrics are intrinsic to the model, and advanced statistical analyses are used for pattern and trend recognition, as well as revealing ecological relationships. Using this framework, many different predictive databases and accompanying maps can be generated, depending on client needs: forest volume and productivity, bio-diversity estimates, wildlife habitat assessment, successional stages, soil characterizations, carbon sequestering, pipeline and road routing, and restoration of disturbed sites. The Future
Before the year 2001, satellite remote sensing data will have spatial resolutions of 1x1m. In addition, aerial remote sensing data with sub-meter spatial resolutions are currently available. The unprecedented level of detail offered by these technologies will make possible quantitative landscape studies that incorporate simultaneous measurements of environmental variables. The challenge facing researchers is to mine this enormous amount of data for the relationships that most accurately describe ecosystem functioning, given the fact that the data will contain inherently more variability and uncertainty. We have started adapting the statistical techniques developed for SiteLogix to high resolution images, which will allow us to take our modeling procedures to the next level of accuracy and applicability for addressing integrated resource management needs. About the Authors:
John D. Beckingham is CEO and senior ecologist, Michael Desilets is the senior GIS analyst, Darrin Nielsen is the senior ecological analyst, and Frank Johns is the communications manager at Geographic Dynamics Corp. Back |
