Methods to characterize spatial-temporal patterns of land transitions using multi-temporal binary maps

Keywords: Spatial-temporal pattern, Land transition, Patch, Plum Island Ecosystems, Ponds
Session Type: Virtual Guided Poster Abstract
Day: Sunday
Session Start / End Time: 2/27/2022 11:20 AM (Eastern Time (US & Canada)) - 2/27/2022 12:40 PM (Eastern Time (US & Canada))
Room: Virtual 41

Abstract

Ecologists use spatial pattern metrics to characterize land cover because landscape pattern is a major theme in Ecology. However, the profession’s most popular metrics measure pattern at a single time point, e.g metrics in the FRAGSTATS software. I have derived generalizable methods to characterize spatial-temporal transitions of a land cover in a time series. My methods consist of two parts: 1) a pixel-based approach to quantify and categorize types of land transition across time, and 2) a patch-based approach that utilizes union patch, which is a new concept that I propose as the unit of analysis to characterize patterns of land transitions across time and space. I illustrate my methodology by measuring changes of ponds in the Plum Island Ecosystems site of the National Science Foundation’s Long Term Ecological Research network (PIE-LTER) across the years 1938, 1971, and 2013. Results reveal dramatic changes in ponds at PIE-LTER that raise questions about the reliability of the manually digitized dataset for the purpose of studying long-term change of ponds. Among the entire area of pond that is pond at any time, only 5% is the area of pond that is present at all three time points, while more than half of the ponds out of all union pond patches either appear or disappear completely during one of the time intervals. In conclusion, my methods provide powerful tools for ecologists to quantify transitions among categories in ways that popular metrics miss.