USING CONSERVATION GIS TO BUILD A PREDICTIVE MODEL FOR OAK SAVANNA ECOSYSTEMS IN NORTHWEST OHIO

 
 
 
 

USING CONSERVATION GIS TO BUILD A PREDICTIVE MODEL FOR OAK SAVANNA ECOSYSTEMS IN NORTHWEST OHIO

Show full item record


Title: USING CONSERVATION GIS TO BUILD A PREDICTIVE MODEL FOR OAK SAVANNA ECOSYSTEMS IN NORTHWEST OHIO
Author: Ricci, Marcus Enrico
Description: The Oak Openings Region in Northwest Ohio is one of the few remaining remnants of oak savanna and oak barrens, or “oak savanna complex.” It is a 33,670 ha complex of globally-significant ecosystems and has more listed species than any other similarly-sized region in the state. Agriculture, drainage and fire suppression have reduced its area by half, underscoring the need to locate and prioritize appropriate habitat for acquisition and conservation. Land managers often have difficulty in implementing regional conservation efforts due to a lack of detailed ecological knowledge or habitat quality data. I used ArcGIS 9.1 to build a predictive geographic model (PGM) to detect oak savanna complex remnants and restorable patches by determining significant ecological variables from known remnant patches. Software and data used was constrained to readily available sources and ecological variables investigated included soil type, elevation, slope, topographic position and aspect. This research used predictive modeling in a new way by using it to predict areas of high probability of a rare ecosystem, rather than its typical use for creating predictive habitat models for individual taxa, multiple taxa or vegetative communities. The resulting model succeeded in locating potential remnants and restorable patches at the landscape level, as well as creating a suitability index to rank the probability of accurately predicting oak savanna complex presence at the landscape level. Both simple statistics and regression analysis were used to determine the significant predictors of oak savanna complex presence: suitable soil types; mean elevation and topographic position. Single-variable predictive models reduced the county-wide search area as much as 93% with a predictive accuracy of 87–100%. However, combining these models into a multi-variable model reduced the search area as much as 99%. Regression analysis determined that the model explaining the highest amount of variance used only two ecological variables: suitable soils and mean elevation. Validation of this two-variable model on a randomly-generated data set proved it was 90% accurate in locating high-probability areas of oak savanna complex. This research produces a scientifically robust predictive ecosystem model that more simply and systematically locates and prioritizes conservation at a landscape scale.
Permanent Link: http://rave.ohiolink.edu/etdc/view?acc_num=bgsu1143490180
http://hdl.handle.net/2374.OX/16050
Date: 2006

Files in this item

Files Size Format View

There are no files associated with this item.

This item appears in the following Collection(s)

Show full item record