US-IALE 2018 has ended
Back To Schedule
Monday, April 9 • 1:30pm - 1:45pm
SYMPOSIA-03: Measuring the Configuration Accuracy of Land Change Simulations Using Metrics Relevant to Landscape Ecology and Land Change Science

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

AUTHORS: Brian R. Pickard*, Environmental Defense Fund; Ross K. Meentemeyer, Center for Geospatial Analytics, North Carolina State University

ABSTRACT: For decades, land change scientists have applied various contingency table-derived measures to quantify the accuracy of simulated landscapes. These measures, primarily adapted from remote sensing, are chiefly concerned with determining if a given pixel is correctly classified in a specific location, and whether the overall mapped proportions are correct. Since the pixel is the fundamental unit of analysis, the accuracy of the configurations of mapped land cover patches is largely ignored. However, landscape ecology is founded on the premise that patch-level patterns influence ecological processes. The challenge remains for the land change profession to therefore reconcile the need to consider spatial patterning with the traditional accuracy methods adapted from remote sensing perspectives. We illustrate a methodology for capturing spatial configuration, termed configuration disagreement, designed to be complimentary to the traditional, widely accepted accuracy measures of quantity and allocation. Using the FUTURES model, we simulate multiple configurations of urban growth for ten counties surrounding the growing megaregion of Charlotte, North Carolina. By holding the proportion of new development consistent and only manipulating the spatial arrangements of simulated pixels, we are able to isolate how pattern influences allocation and configuration accuracy values. Our results demonstrate that by varying the spatial configurations, correctly classifying a pixel in a specific location may improve allocation accuracy but fail to capture the discrete mosaic of patches necessary to represent landscape elements across an entire region. The degree to which simulated and observed spatial configurations of land change match, irrespective of where the specific location of land change category occurs, is likely a more appropriate method for assessing land change model accuracy. This work illustrates the need to move from away from pixel to pixel accuracy classifications, and instead focus on developing accuracy metrics that are appropriate for spatial ecology and land change modeling research.

Monday April 9, 2018 1:30pm - 1:45pm CDT
Spire Parlor

Attendees (8)