Stratified random sampling generates the Total Operating Characteristic to compare flood index maps

Authors: Zhen Liu*, Clark University, Robert Pontius, Clark University, Graduate School of Geography
Topics: Geographic Information Science and Systems, Remote Sensing, Hazards and Vulnerability
Keywords: total operating characteristic (TOC), flood mapping, remote sensing, stratified sample, accuracy assessment
Session Type: Virtual Paper
Day: 4/9/2021
Start / End Time: 3:05 PM / 4:20 PM
Room: Virtual 14
Presentation File: No File Uploaded


The Total Operating Characteristic (TOC) is a procedure to compare an index variable to a binary variable. Previous methods required the data to derive from a census or a random sample. However, random sampling is not as efficient as stratified random sampling to collect data. Our article presents a new methodology that uses stratified random sampling to generate the TOC. An application to flood mapping illustrates how the TOC evaluates an index’s ability to diagnose the presence of flood at various thresholds. The TOC shows visually and quantitatively the index’s diagnostic abilities relative to baselines. Results show that the Modified Normalized Water Index has the greatest diagnostic ability across various thresholds, while the Normalized Difference Vegetation Index has diagnostic ability greater than the Normalized Water Index at the thresholds near where diagnosed quantity equals the estimated abundance. The TOC gives more information than the Relative Operating Characteristic and is more comprehensive than popular methods that measure diagnostic ability at exactly one threshold. We invite researchers to use our software that generates the TOC from a census, random sample or stratified random sample. The TOC Curve Generator is free at https://lazygis.github.io/projects/TOCCurveGenerator.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login