Browsing by Subject "CART"

Now showing 1 - 4 of 4

Open Access
Identification and prioritization of lands for restoration of Piedmont prairie in North Carolina
(2007-12-07T15:59:32Z) Taecker, Erica
In the central Piedmont of North Carolina, prairies and savannas were noted by European settlers to have covered a significant portion of the landscape. Piedmont prairie is valued for its extraordinary biodiversity; at least 277 plant species, some endemic, are associated with this unique area. Rich prairie ecosystems in the Piedmont were maintained by both naturally-occurring and human-ignited fires, which created open fields or patches of prairie within oak-pine-hickory or Piedmont longleaf pine forests. Anthropogenic changes to fire regimes and land use have fragmented the Piedmont prairie ecosystem, such that several of its plant species are now federally endangered. Effective conservation of this native ecosystem in our rapidly developing state depends on a solid understanding of its science. Just as importantly, it necessitates the ability for conservation agencies to act efficiently to protect and maintain areas of intact prairie, while quickly identifying and protecting other areas with restoration potential. This masters project compares the suitability of two multivariate modeling tools, CART (Classification and Regression Tree) and Maxent (Maximum entropy), for predicting the potential geographic distribution of the Piedmont prairie ecosystem in nine Piedmont counties of North Carolina. Natural Heritage “Element Occurrence” point location data of four prairie species were the basis for the models, which considered environmental variables such as elevation, topographic relative moisture index (TRMI), slope, relative aspect, soil clay content, and soil effective cation exchange capacity (ECEC) in the prediction of potential prairie extent. Further, a basic prioritization of the resulting prairie “habitat” patches mapped in GIS highlights areas adjacent to existing protected areas in which to focus conservation and restoration efforts. The results indicated that the habitat model of prairie created by Maxent reasonably predicts known prairie species occurrences without over-generalizing the possible distribution of prairie in the study area. Maxent also highlights that ECEC is the most important predictor variable of prairie distribution, followed by clay content. The CART technique resulted in similar accuracy and explanatory variables, but when mapped, “habitat” covered a large proportion of the study area, less useful for targeting regions for further study. The preliminary prioritization suggests that several zones around Charlotte, NC and in Davidson County warrant further investigation for prairie remnants. With sufficient additional information about current land use and cover, the prioritization can be further refined to reduce the effort needed to find suitable sites for the restoration and conservation of Piedmont prairie and its associated forest cover types.
Open Access
Land Cover Change and Ecosystem Services on the North Carolina Piedmont 1985 to 2005
(2008-04-25T20:38:43Z) Donohue, Michael John
Analyses of ecosystem processes are advanced through remote sensing and geostatistical modeling methods capable of capturing landscape pattern over broad spatial and temporal scales. Many ecological studies rely on land cover data classified from satellite imagery. In this, changes in land cover are often presumed to correlate with changes in ecosystem processes or services provided by ecosystems (e.g., watershed protection). Documenting changes in land cover requires that images be classified over time, often using historical images to document landscape change. But this is difficult to do for historical images because we cannot ground-truth old images, lacking actual land cover data from the past. I developed a land cover classification scheme using a classification and regression tree (CART) model generated from 2001 National Land Cover Dataset (NLCD) and Summer, Fall, and Winter triplets of Landsat 5 Thematic Mapper (TM) imagery. The model is robust to inter-annual variability in surface reflectance, and thus can be extended in time to classify land cover from images from any time, past or future. The model was used to predict land cover from 1985 to 2005, for a study region in the Piedmont of North Carolina. Temporal and spatial analyses focused on ecosystem services of carbon sequestration and biodiversity support as affected by forest fragmentation. This study offers a landscape-level identification of the relationships between spatial and temporal development patterns and the provision of ecosystem services. The project also represents the creation of a multi-annual land cover classification dataset of which few exist, thus providing a framework for further studies of landscape pattern and ecological processes.
Open Access
Restoring Brook Trout to North Carolina's National Forests
(2008-04-22T21:07:03Z) DiBacco, Sara
Stream systems in the Southern Appalachian Mountains represent the southern limit of brook trout (Salvelinus fontinalis) distribution in the eastern United States and are home to the region’s only native salmonid, the Southern Appalachian Brook Trout. Currently, the species occupies only 25% of its former range in the region, but in North Carolina, opportunities exist to restore brook trout to high quality National Forest watersheds. The purpose of this project is to provide the information necessary to design a watershed-specific restoration strategy for the Fires Creek Watershed, Nantahala National Forest and to develop a model that predicts natural migration barriers within a stream network. A geographic information system (GIS) is the primary analysis tool used to derive, interpret, and display relevant data. In the Fires Creek Watershed, migration barriers are identified and characterized to delineate potential brook trout reintroduction sites. The watershed is also assessed as a target for brook trout restoration according to five criteria. These are the historical presence of brook trout, the current distribution of trout in the basin, the genetic identity of potential donor populations, site accessibility, and current and future habitat suitability. Barrier data are also used to develop a classification and regression tree (CART) model to predict barrier locations. Results show that numerous opportunities exist to restore brook trout to the Fires Creek Watershed. The most effective restoration strategy combines the availability of protected habitat, as delineated by migration barriers, with information extracted from the comprehensive assessment. The model provides the framework for a tool that improves the efficiency of completing restoration projects, but results suggest that higher resolution data is necessary to increase prediction success. Overall, this work contributes to the development and implementation of brook trout restoration projects in North Carolina’s National Forests.
Open Access
Searching for Eastern Old Growth: Modeling Primary Forest in Western North Carolina Using Terrain Attributes and Multispectral Satellite Imagery
(2010-12-10) Hushaw, Jennifer
After centuries of timber harvest and conversion of forest to farmland and development, only small pockets of old growth forest remain in the eastern United States. These remnant portions of older forest have intrinsic value as a rare forest type and they play an important ecological function on the landscape. However, old growth forests in the eastern U.S. are less well-studied and documented than their counterparts in the Pacific Northwest. This study was undertaken to predict the geographic location, ecological and spectral characteristics of existing old growth, specifically in the southern Appalachian forests of western North Carolina. Stands of old growth previously field validated by the Southern Appalachian Forest Coalition were used as the response variable. Predictor variables included a range of landscape, topographic, and satellite indices derived from Landsat TM 7 satellite imagery and terrain analysis. Predictions were made using Classification and Regression Tree (CART) and Maximum Entropy (MaxEnt) modeling techniques. Model results were successful based on validation with existing field data. However, the MaxEnt model produced the most realistic estimate of potential old growth area given the inherent rarity of this forest type and suitability of the MaxEnt modeling technique for predicting the distribution of rare species. Results highlight over 54,000 hectares of potential old growth to be investigated by researchers on the ground. This analysis will contribute to the relatively limited body of knowledge about old growth in the eastern U.S. and is unique in terms of its broad geographic extent. Continued research on these remnant eastern old growth stands must be done to increase our understanding of this rare forest stage and to better inform related management decisions on both public and private land in the eastern U.S.