Ecological Land Classification Research Articles

Application of CART in ecological landscape mapping: Two case studies

Abstract The goal of this paper is to introduce a statistical concept to derive ecological classifications of terrestrial and marine environments. Such ecological regionalisations reflect spatial combinations of biotic and abiotic characteristics and therefore may serve for environmental planning and monitoring issues. Referring to two case studies the paper presents how to calculate and map ecological defined regions from geodata by use of decision tree models and GIS-techniques. The first study deals with marine environments, exemplified by benthic habitats in the North Sea. The second study is on an ecological land classification of Europe which was computed using surface data on potential natural vegetation, elevation, soil texture and several climatic elements. Both the marine and the terrestrial ecoregions maps were combined with exposure data provided by environmental monitoring activities. The ecological land classification of Europe was intersected with measurement data on the metal loads in mosses taken from the German part of UN ECE moss surveys. In this way, the temporal development of the metal bioaccumulation within ecoregions since 1990 was assessed. The benthic habitat map was used to regionalise the temporal trend of the temperature conditions near the sea floor for the months of July and January from 1995 to 2000 by analysing these measurements according to the spatial categories of the habitat map. In the future, both ecological regionalisations should be used as a spatial framework for the analysis of up-to-date meteorological and phenological data in order to disclose climate change induced impacts.

Statistical classification of terrestrial and marine ecosystems for environmental planning

Environmental planning is an instrument for the operationalisation of the precautionary principle in environmental law and, to this end, must rely on maps depicting the spatial patterns of ecological attributes of aquatic and terrestrial ecosystems and of environmental change effects, respectively. In this context, different mapping techniques are presented by example of three case studies covering terrestrial, coastal and marine environments. The first case study was selected to demonstrate how to compute an ecological land classification of Germany by means of CART. The resulting ecoregions were mapped by GIS. This CARTography enables to regionalise metal bioaccumulation data in terms of 21 ecological land categories and to prove the specifity of emission control measures as being part of environmental policies. The second investigation was chosen to apply for the first time in Germany the regionalisation approach to the research of climate change effects in terms of past, recent and potential future incidences of Anopheles sp. and malaria in Lower Saxony. To investigate whether malaria might be transmitted due to increasing air temperatures, data sets on past and future air temperatures were used to spatially model malaria risk areas. The third example demonstrates the transfer of the CARTography approach presented in the first case study from terrestrial to marine environments. We analysed the statistical relations between data on benthic communities and physical properties of their marine environments by means of CART and applied these rules to geodata which only describe physical characteristics of the benthic habitats. By this, those parts of the sea ground could be predicted where certain benthic communities might occur.

Arthropods as ecological indicators of sustainability in Canadian forests

The high functional and unmatched biological diversity represented by arthropods demand that these organisms be considered as ecological indicators of sustainable forest management. Successful use of arthropods in this capacity will require a systematic and rigorous process, including selection of potential indicators, definition of relationships between indicators and disturbance variables, optimization of the useful range of the indicator and application of the indicator(s) in monitoring. In Canada, the single greatest impediment to the use of arthropods as ecological indicators is the importance of accurate species-level identification and the difficulty achieving it. Consequently, most work has focused on a few relatively well-known groups (e.g., epigaeic carabid and staphylinid beetles and spiders, saproxylic beetles, butterflies and larger night flying moths).Many recent studies have provided baseline data about the range of natural variation and have begun to quantify arthropod responses to natural and anthropogenic disturbances in the context of preplanned experiments or through various retrospective approaches. Carabid beetles are the best-studied group and sufficient sets of data now exist to permit a meta-analysis of the robustness of carabids as indicators across multiple spatial scales and in terms of how well they represent broader ecological responses to disturbances. There is good potential to incorporate arthropod indicators into monitoring programs in Canada, but it is necessary to first complete a scientifically credible selection process for specific ecological indicators. Future research should focus on completing the process for taxa under current study as this develops the best presently understood opportunities for using arthropod indicators in assessing various aspects of environmental change. Researchers should also consider other means of monitoring arthropod biodiversity by the use of surrogate ecological parameters such as ecological land classification and habitat classification systems. Key words: arthropods, ecological indicators, monitoring, biodiversity, taxonomy, sustainability

Logging-induced change (1930-2002) of a preindustrial landscape at the northern range limit of northern hardwoods, eastern Canada

Logging-induced changes from preindustrial (1930) to current conditions (2002) were studied in a landscape covering 13 550 ha in eastern Quebec. Age and types of forest cover were compared between 1930 and 2002 forest maps. In addition, we compared relative species abundance between living stems and coarse woody debris to study these changes at the stand scale. More than 90% of the 1930 preindustrial landscape was composed of forest stands older than 100 years. A balsam fir (Abies balsamea (L.) Mill.) – white spruce (Picea glauca (Moench) Voss) dominated conifer cover (77% of the landscape area) formed the landscape matrix across the lowlands and was intermingled with mixed stands of sugar maple (Acer saccharum Marsh.) and conifers on the highlands. As a result of recurrent logging, stands less than 70 years old accounted for 93% of the 2002 landscape. From 1930 to 2002, 37% of the landscape was converted from coniferous to mixed forest, and 19% evolved towards a deciduous cover. The total number of cover patches doubled to 193, whereas mean patch size decreased twofold to 65 ha. Sugar maple, red maple (Acer rubrum L.), striped maple (Acer pennsylvanicum L.), and white birch (Betula papyrifera Marsh.) probably experienced a greater increase in abundance, whereas balsam fir, white spruce, and eastern white-cedar (Thuja occidentalis L.) experienced a more pronounced decrease. Because it does not consider preindustrial landscape patterns, the system of ecological land classification currently in use in this area suggests that potential late-successional cover types should be more similar to present-day than to preindustrial conditions.

Synergistic Techniques for Better Understanding and Classifying the Environmental Structure of Landscapes

The desire to capture natural regions in the landscape has been a goal of geographic and environmental classification and ecological land classification (ELC) for decades. Since the increased adoption of data-centric, multivariate, computational methods, the search for natural regions has become the search for the best classification that optimally trades off classification complexity for class homogeneity. In this study, three techniques are investigated for their ability to find the best classification of the physical environments of the Mt. Lofty Ranges in South Australia: AutoClass-C (a Bayesian classifier), a Kohonen Self-Organising Map neural network, and a k-means classifier with homogeneity analysis. AutoClass-C is specifically designed to find the classification that optimally trades off classification complexity for class homogeneity. However, AutoClass analysis was not found to be assumption-free because it was very sensitive to the user-specified level of relative error of input data. The AutoClass results suggest that there may be no way of finding the best classification without making critical assumptions as to the level of class heterogeneity acceptable in the classification when using continuous environmental data. Therefore, rather than relying on adjusting abstract parameters to arrive at a classification of suitable complexity, it is better to quantify and visualize the data structure and the relationship between classification complexity and class homogeneity. Individually and when integrated, the Self-Organizing Map and k-means classification with homogeneity analysis techniques also used in this study facilitate this and provide information upon which the decision of the scale of classification can be made. It is argued that instead of searching for the elusive classification of natural regions in the landscape, it is much better to understand and visualize the environmental structure of the landscape and to use this knowledge to select the best ELC at the required scale of analysis.

Criteria to assess and select sites for long-term avian monitoring in an urbanizing landscape.

A methodology was developed to prioritize the suitability of sites for long-term monitoring of avian populations, including vulnerable species, both to enhance assessment of changes in ecological resources and to facilitate land-use planning at the regional scale. This paper argues that a successful monitoring program begins with a site prioritization procedure that integrates scores based on spatial controls with ecological and socio-economic indicators, particularly those dependent on community involvement. The evaluation strategy in this study combines 1) spatial controls such as land ownership and accessibility, with 2) biological and habitat indicators such as vulnerable species and habitat connectivity, and 3) community and agency variables such as volunteer commitment and agency priorities. In total, a set of ten indicators was identified. This strategy was applied to predominantly agricultural landscapes, which are experiencing increasing human pressures, in three sub-watersheds of the Credit River, Southern Ontario. Specifically, bird populations were recorded during the breeding seasons of 2000-2002 in nine land units or habitat types including marsh, deciduous forest, and grasslands as mapped by Credit Valley Conservation (CVC) following Ecological Land Classification (ELC) guidelines. CVC selected sites for long-term monitoring in 2002 and the relationships between the scored (or ranked) sites and the selected long-term monitoring sites are discussed.

Ecosystem mapping in the Lower Foothills Subregion of Alberta: Application of fuzzy logic

Predictive ecosite mapping involves developing computer models that consistently identify and map ecosystems. This method of predicting ecosystem occurrence on the landscape uses basic inventory information and expert knowledge, and is an effective integrated planning tool for providing a record of the location and spatial distribution of ecosystems within a management area. Fuzzy logic technology can be used to computerize essential elements of ecosystem identification, and the outputs can be linked to a Geographic Information System for map production. A pilot study was undertaken on the application of this technology to the Alberta Ecological Land Classification database and the resulting ecosite map for a township located in central Alberta (Tp42R9W5). The range of attributes used in the program was constrained by the attributes recorded on mapped polygons. Three maps with suitable attributes were available for the township studied: a Digitized Elevation Model map, an Alberta Vegetation Inventory map, and a reconnaissance soil survey map. Attributes of all polygons from all three maps were compiled and seven attributes (humus form, Ah thickness, surface texture, aspect, organic thickness, slope angle, and Alberta Vegetation Inventory moisture regime) were chosen to produce a computerized program for ecosite identification. Four sets of data were used to calibrate the program, as well as a small-plot data set collected from the township studied. The computer program was used to analyze the polygon data corresponding to two sets of data collected in the field and resulted in 72% and 70% similarity between the choices of experts and of the computer program. The quality of the original polygon attributes contributed to errors in identification. In addition, the reconnaissance soil survey map gave only an estimate of four attributes (Ah horizon thickness, organic thickness, humus form, and surface texture). Key words: ecosystem classification, site classification, fuzzy logic, fuzzy sets, predictive ecosystem mapping, predictive site mapping

A comparison of pre-European settlement forest characterization methodologies

The characterization of "natural" or "presettlement" forest has become a relatively common practice in Canada as forest managers strive to put into practice concepts of sustainable forest management. Various methods have been developed to undertake such characterizations, leading to confusion about how to define "presettlement forest" and uncertainty over the approach that will best serve as a basis for management. We report on two methods of presettlement forest characterization: the "Witness Tree" and the "Potential Forests" approaches. We compare results from these approaches to the existing forest composition in the Fundy Model Forest, New Brunswick. Both approaches indicate a decline in the predominance of tolerant hardwood and eastern cedar since presettlement. However, the Potential Forests approach consistently suggests much higher presettlement frequencies of spruce (Picea spp.) and, in most cases, pine (Pinus spp.) than the Witness Tree method. Differences between frequencies of tree species estimated by the two methods probably result from biases associated with both methods and the different scales of reporting. If used critically, the combined use of both sets of presettlement forest information will allow managers to determine the historical frequency of individual tree species and forest communities. Such information will provide some guidance in maintaining the diversity of native species and community types. Key words: presettlement forest, ecological land classification, Witness Tree method, Potential Forests, forest management, biodiversity

Post-fire vegetation of the Montane natural subregion of Jasper National Park

Ecological Land Classification information gathered in the mid to late 1970s was used to predict post-fire vegetation for the Montane natural subregion of Jasper National Park. Percentage similarity indices (SI) between vegetation types based on understory species composition and cover, calculated distances ( D) generated from altitude, nutrient and moisture, broad patterns of vegetation canopy based on moisture regimes, canonical correspondence analysis (CCA), and cluster analysis were used. Of the 33 Montane plant communities, grassland, low shrub and shrub communities were considered to regenerate after fire. Nineteen percent of the area was predicted as Potentilla fruticosa/ Arctostaphylos uva-ursi/ Galium boreale (L1), and 33% of the area as Pinus contorta var. latifolia/ Shepherdia canadensis/ Aster conspicuus (C6) or Pinus contorta var. latifolia/ Shepherdia canadensis/ Linnaea borealis (C19) after fire. Dominant post-fire understory species were Juniperus communis, Sherpherdia canadensis, Rosa acicularis, Elymus innovatus and Linnaea borealis. Some closed canopy forests were assumed to open after fire, while other communities reverted to closed canopy pine forests.

Application of GIS in Ecological Land Type (ELT) mapping—A case in Changbai Mountain area

This paper depicted the physiographic landscape features and natural vegetation situation of study area (the eastern Jilin Province), and expatiates the definition, basic characters and its development of Ecological Land Classification (ELC). Based on the combination of relief map, satellite photography for study area and vegetation inventory data of 480 sample sites, a 5-class and a 15-class ecological land type map was concluded according to 4 important factors including slope, aspect, vegetation and elevation. Ecological Classification System (ECS) is a method to identify, characterize, and map ecosystems. The Ecological Land Type (ELT) was examined and applied initially in eastern Jilin Province.

Canada's National Ecological Framework: An asset to reporting on the health of Canadian forests

The Canadian Forest Service, in cooperation with its partners, has a mandate to report on the health of Canada's forests and determine if, how, and why it is changing. A holistic perspective of forest health is taken whereby the ecosystem rather than a single element is considered. The use of the national ecological classification of Canada as a key reporting framework facilitates this task. Advantages for reporting purposes are several, including the use of ecological over jurisdictional boundaries to discuss ecosystems, wide national acceptance of the framework, and access to a wide array of other environmental databases that use the same framework. Compromises have to be made for forest health reporting as the ecological classification is not a forest ecosystem classification. However, advantages to using the framework for national reporting far outweigh these shortcomings. Key words: ecological land classification, forest health, national and international reporting

Evaluation of an ecological land type classification system, Natchez Trace State Forest, western Tennessee, USA

In an effort to aid future research and forest management decisions, we incorporated data on forest type, soils and topography into a geographic information systems (GIS)-based ecological land classification system for Natchez Trace State Forest (NTSF) in western Tennessee. The area is still recovering from the effects of cultivation, logging, forest clearing and grazing during the 1800s and early 1900s, but the data indicated that most stands now contain sawtimber-sized trees after more than 50 years of secondary succession. Our analyses showed that forest types were not randomly associated with soil characteristics or topography. For example, the most heavily gullied lands were strongly associated with pine forests while hardwoods were more associated with upland soils and certain types of sideslopes. Mature hardwood stands, which were deemed to be of the greatest conservation value, were classified into eight ecological land types based on soil type, aspect, orientation and topographic setting for further study. When these land types were evaluated with respect to data on forest composition and environmental conditions collected in the field, we found that the variables used to derive the eight classes were important in influencing floristic patterns. It was, however, clear that the classes themselves imperfectly captured patterns of overstory community composition for a number of reasons. Thus, while land types provide a potentially valuable management tool, their utility and successful implementation are dependent upon a recognition of their inherent limitations.

Ecological land classification in the Southern Loam Hills of south Alabama

A landscape scale classification of ecosystems was undertaken on the Conecuh National Forest and Solon Dixon Forestry Education Center in south Alabama. Nine landtypes (LTs) were identified in the study area. Each LT had a unique assemblage of plant species and environmental variables. In the Pine Hills, three LTs were identified with diagnostic species including Quercus incana, Ipomoea sp., Q. stellata, Rhus copallina, Vaccinium myrsinites, and Pityopsis graminifolia. The diagnostic environmental variables were landform index, slope, B horizon depth, percent B horizon nitrogen (N), percent A horizon fine sand, and percent A horizon silt. In the Dougherty Plain, two LTs were identified with diagnostic species including Diospyros virginiana, Elephantopus tomentosus, Vaccinium corymbosum, Lechea minor, Aristida stricta, and Q. virginiana. The diagnostic environmental variables were slope and percent B horizon fine sand. In the Wet Pine Flatwoods, four LTs were identified. The diagnostic species included Diospyros virginiana, Smilax glauca, Vaccinium stamineum, Hibiscus aculeatus, Pinus elliottii, Drosera brevifolia, Clethra alnifolia, and Aristida stricta. The diagnostic environmental variables were slope, landform index, depth to mottling, percent B horizon clay, and drainage class. The vegetation of the LTs is similar to longleaf pine ecosystems throughout the southeastern Coastal Plain.

Preliminary ecological land classification of the Chauga Ridges Region of South Carolina

An ecological method of integrated multifactor land classification was applied in the Chauga Ridges, a subregion of the Blue Ridge Mountain Province, in northwest South Carolina. Phase I sampling involved measuring 61 reference plots in undisturbed late-successional upland hardwood stands. Four distinct site units (xeric, intermediate, submesic, and mesic) were identified using ordination, cluster analysis, discriminant function analysis, and canonical correspondence analysis. Using cross-validation of a discriminant function, a combination of five environmental features (landform index, terrain shape index, root mat thickness, slope gradient, and slope position) identified site units with an accuracy rate of 82%. A predictive GIS model which used spatial variables only (landform index, terrain shape index, topographic-related moisture index, distance-to-bottom, and slope gradient) was also developed and identified site units with the same success rate of 82%.

A Quantitative Approach to Developing Regional Ecosystem Classifications

Ecological land classification systems have recently been developed at continental, regional, state, and landscape scales. In most cases, the map units of these systems result from subjectively drawn boundaries, often derived by consensus and with unclear choice and weighting of input data. Such classifications are of variable accuracy and are not reliably repeatable. We combined geographic information systems (GIS) with multivariate statistical analyses to integrate climatic, physiographic, and edaphic databases and produce a classification of regional landscape ecosystems on a 29 340‐km2 quadrangle of northwestern Wisconsin. Climatic regions were identified from a high‐resolution climatic database consisting of 30‐yr mean monthly temperature and precipitation values interpolated over a 1‐km2 grid across the study area. Principal component analysis (PCA) coupled with an isodata clustering algorithm was used to identify regions of similar seasonal climatic trends. Maps of Pleistocene geology and major soil morphosequences were sued to identify the major physiographic and soil regions within the landscape. Climatic and physiographic coverages were integrated to identify regional landscape ecosystems, which potentially differ in characteristic forest composition, successional dynamics, potential productivity, and other ecosystem‐level processes. Validation analysis indicated strong correspondence between forest cover classes from an independently derived Landsat Thematic Mapper classification and ecological region. The development of more standardized data sets and analytical methods for ecoregional classification provides a basis for sound interpretations of forest management at multiple spatial scales.

Environment and biogeography of the western boreal forest

The western boreal forest of North America (Manitoba through Alaska) has a typical boreal climate, but the largely sedimentary Interior Plains and the northern Cordillera (part of which was ice-free in the Pleistocene) are physiographically and geologically very different from the Canadian Shield that underlies most of the eastern boreal forest. The mountainous nature of much of the region, with pronounced topography, local climates, aspect differences, sharp drainage, and elevational clines, gives some of its boreal landscapes a different character from those of the Shield. The region's forests are structurally consistent with boreal forests everywhere, but floristic differences are apparent, with the loss of eastern species and the addition of Cordilleran and Beringian elements. The southern margins of the western boreal forest also have a distinctive character, as they merge with either the Cordilleran subalpine forests or with aspen parkland and grassland—not with Great Lakes-St. Lawrence conifer-hardwood forests.Several different biogeographic or ecological land classifications have been applied to the western boreal forest. Although there are a variety of classifications, several common themes are apparent. Broad-scale zonation reflects 1) latitudinal change (boreal forest to subarctic woodland to forest-tundra ecotone, from south to north), 2) physiography and biogeography (Plains/Shield to northern Cordillera to Alaska-Beringia, from east to west), and 3) the physiographic vs. vegetation bias of the classification. Regardless of higher-level, broad-scale differences, the various classifications tend to be consistent at the site level, and therefore all have some value and utility for foresters.

Introduction-global to local: Ecological Land Classification

Ecological Land Classification (ELC) is a scientific endeavour which attempts to organize, stratify and evaluate ecosystems (and complexes of ecosystems) for the purposes of land resource management. Since ecosystems themselves are not easily defined in practical terms, ELC is likewise not a trivial concept. Nonetheless, ELC is a prerequisite for ecosystem management and the conservation of biological diversity simply because ecosystems must be described, characterized and spatially-located before they can be managed. Regarding the current status and future direction of ELC, mainly in relation to forest management: 1) approaches to ELC construction and utilization have shifted considerably over the past 2 decades; 2) there appears to be a current consensus regarding basic approaches to ELC; 3) spatial scale is a critical variable that must be addressed by ELCs; 4) ELCs must strive to more directly address management objectives; 5) natural ecosystem functions need to be better integrated within ELC frameworks; and, 6) the need for quality, georeferenced ELC-related data will continue to grow.

Developing an Ecological Land Classification for the Fundy Model Forest, southeastern New Brunswick, Canada

The methodology for developing and mapping a hierarchical Ecological Land Classification (ELC) is presented. The classification provided a systematic methodology that explained the distribution and composition of southern New Brunswick's forested landscape. The nested structure of the ELC identified and provided a hierarchical linkage between ecosystems from the size of forest stands to climate regions. This framework made the collection and analysis of data efficient and gave confidence that tree species distributions, which were central to understanding the influence of abiotic factors on the forest systems, were controlled by the factors examined at each level of the hierarchy. This ELC methodology, developed for the Fundy Model forest, was successful in describing and mapping the Climate, Geomorphologic, and Regolith controlled forest ecosystems. Preliminary classification indicates that spatial referencing of the Site Level is achievable.

Spatial scale-dependent policy planning for land management in southern Europe

This study outlines and original tool for rural policy planning in southern Europe. This new tool is a process-based, scale-dependent, rural policy-making approach, which is designed to address increasing land degradation problems in southern Europe. Seven important processes are identified (land abandonment, devegetation, intensification in agriculture, global climate change, accelerated soil erosion, increasing water demands, urbanisation) and plotted on a space-time diagram, which clearly shows the spatial and temporal scales for which these processes are significant for landscape change in southern Europe. Conclusions are derived concerning, in particular, sustainable (optimal) rural policy-making for southern Europe's problematic land management. An optimal spatial-temporal scale for land management in southern Europe may range spatially from the "farm" (0.5 km(2)) to "sub-provincial" level (450 km(2)) and temporally from 7 to 30 years. The study delineates methods and results derivable from such a new policy-planning approach and suggests the usefulness of combining this approach with ecological land classification at the landscape level.

Primary databases for forest ecosystem management-examples from Ontario and possibilities for Canada: NatGRID

This paper identifies some scientific impediments to ecosystem management and describes bio-physical databases required to help systematically and empirically address the ecological sustainability challenge. Examples are drawn from ongoing work in Ontario. This work has implications for efforts in ecological land classification, landscape ecology, more efficient locating of research and monitoring plots, wildlife management and ultimately trade-off analyses. We conclude with the recommendation that the key primary databases, as currently evolving for Ontario, could and should be developed nationally, thereby creating a "NatGRID database", i.e., Nationally Georeferenced Resource Information for Decision-making. NatGRID could be used to help address, in a more quantitative manner, fundamental questions regarding ecological sustainability and trade-offs in forest management.