|
Time Series Fusion of Optical and Radar Imagery for Improved Monitoring of Activity Data, and Uncertainty Analysis of Emission Factors for Estimation of Forest Carbon Flux
Josef
Kellndorfer, Woods Hole Research Center, josefk@whrc.org
Curtis
Woodcock, Boston University, curtis@bu.edu
Richard
A.
Houghton, The Woods Hole Research Center, rhoughton@whrc.org
Pontus
Olofsson, Boston University, olofsson@bu.edu
Oliver
Cartus, Woods Hole Research Center, ocartus@whrc.org
(Presenter)
Neeti
Neeti, Woods Hole Research Center, neeti@whrc.org
Chris
Holden, Boston University, ceholden@bu.edu
A core need for the development of REDD+ Monitoring, Reporting, and Verification (MRV) systems is the annual assessment of land area change from forest lost or gained, and forest degradation. The goal of this project is 1) to develop an approach for accurate assessment of land area change (i.e., activity data) making optimal use of spaceborne optical and radar time series (Landsat, ALOS PALSAR), and 2) to assess the uncertainty of carbon emission estimates derived from Landsat/PALSAR-derived Activity Data and various sources for Emission Factors for test sites in Mexico, Colombia and Peru.
The work carried out in year 1 of the project was focused on:
1) Identification of test sites in Mexico, Peru, Colombia.
2) Preprocessing of Landsat and PALSAR data for the selected test sites.
3) Analysis of Landsat time series based on the change detection algorithm that is under active development at Boston University (Zhu & Woodcock, 2014) to understand the multitemporal signal over different types of land cover/use change in the tropics, and to evaluate where the fusion of optical and radar data can help to obtain timely and accurate activity data (e.g., in areas with persistent cloud cover).
4) An algorithm was implemented that identifies persistent backscatter changes across a time series of PALSAR observations. Initial tests of the algorithm demonstrated 1) the potential of PALSAR data to complement Landsat time series for detecting change, and 2) PALSAR time series are often not sufficiently dense to effectively distinguish between changes in backscatter that are a consequence of actual land cover/use changes and those related to small-scale moisture dynamics, agriculture, etc.
A crosscheck of change signals in the Landsat and PALSAR time series improves the accuracy/timeliness of change detection.
Associated Project(s):
|