Urban dynamic change detection in Indianapolis using time series Landsat data from 1998 to 2017

Authors: Hang Li*, Department of Earth and Environmental Systems, Indiana State University, Qihao Weng, Center for Urban and Environmental Change, Department of Earth and Environmental Systems, Indiana State University
Topics: Land Use and Land Cover Change, Remote Sensing, Urban and Regional Planning
Keywords: Change detection, Landsat series images, CCDC, Indianapolis
Session Type: Paper
Day: 4/10/2020
Start / End Time: 9:35 AM / 10:50 AM
Room: Mineral Hall B, Hyatt Regency, Third Floor
Presentation File: No File Uploaded

Monitoring and classifying land use and land cover (LULC) with Landsat series images has made some progress since the Landsat images were free to access in 2008. Many studies attempted to detect LULC change with multiple temporal images including threshold rules, image differentiating, segmentation, trajectory and regression with dynamic boundary. Among them, Continuous Change Detection and Classification (CCDC) is a classical approach where the algorithm includes gradual change, seasonal change and abrupt change for reflectance. The judgment whether the abrupt change occurs is dynamic based on root mean square error (RMSE) of the regression. The regression can simulate how reflectance varies. On the basis of original CCDC models, our study developed three improvements: Firstly, the logistic function was used to describe gradual change. Secondly, more indices as features were input into the classifier. Lastly, spatio-temporal filtering was utilized in post classification. The first two improvements increased classification accuracy by 9.8% and 0.7% respectively. The third improvement improved the temporal consistency for a series of LULC maps from 1998 to 2017 and partly released the confusion between wetland and other similar LULCs. From the classified maps, it was found that the size of built-up area was increasing from 247.76km2 in 1998 to 293.92km2 in 2017 along the highway. The proposed approach not only increased the classification accuracy but also improved temporal consistency. The future study will apply the model to tropical, subtropical or coastal cities where weather conditions are complex and changeable in order to make the whole approach more robust.

Abstract Information

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

To access contact information login