Making Landsat Time Series Consistent: Evaluating and Improving Landsat Analysis Ready Data
Abstract
The open and free policy of Landsat data has made time series analysis with moderate resolution images possible for the first time. Recently, U.S. Geological Survey has released a another higher version of Landsat data called Landsat Analysis Ready Data (ARD). This dataset is designed specifically for time series analysis such as environmental monitoring and change detection. In this study, we evaluated and improved the consistency of ARD in four aspects. First, we compared the consistency between ARD and the user-generated time series by re-projecting Collection 1 data to the same extent with ARD. Second, we evaluated whether the newly developed cloud and cloud shadow mask (Fmask version 4.0) could result in higher consistency of ARD compared to the Fmask used by USGS ARD (version 3.3). Third, we corrected the Bidirectional Reflectance Distribution Function (BRDF) effects of surface reflectance in ARD and evaluated whether the BRDF correction could improve the consistency of ARD. Finally, we correct the topographic effects by using the sun-canopy-sensor (SCS) topographic correction algorithm (Gu and Gillespie, 1998), the SCS+C algorithm (Soenen et al., 2005) and the rotation model (Tan et al., 2013), and assessed their influences on the consistency of ARD. Results showed that ARD was more consistent than the user-generated time series and would have a higher consistency when using Fmask 4.0 cloud and cloud shadow mask. BRDF correction could improve the consistency of ARD, while whether topographic correction could improve the consistency depends on the locations and the used algorithms.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFM.B31L2652S
- Keywords:
-
- 0434 Data sets;
- BIOGEOSCIENCESDE: 0480 Remote sensing;
- BIOGEOSCIENCESDE: 1632 Land cover change;
- GLOBAL CHANGEDE: 1640 Remote sensing;
- GLOBAL CHANGE