Mapping Forest Height for Mississippi Using Long-term Landsat Observations and GLAS Data
Ainong
Li, Department of Geography, University of Maryland, College Park, MD 20742, ainongli@umd.edu
(Presenting)
Chengquan
Huang, Department of Geography, University of Maryland, College Park, MD 20742, cqhuang@umd.edu
(Presenting)
Hua
Shi, ASRC Research and Technology Solutions, Contractor to USGS/EROS, Sioux Falls, SD 57198, hshi@usgs.gov
Guoqing
Sun, Department of Geography, University of Maryland, College Park, MD 20742, guoqing.sun-1@nasa.gov
Zhiliang
Zhu, U.S. Geological Survey, 12201 Sunrise Valley Drive, Reston, VA 20771, USA, zzhu@usgs.gov
Samuel
N.
Goward, Department of Geography, University of Maryland, College Park, MD 20742, sgoward@umd.edu
Jeffrey
G.
Masek, Biospheric Sciences Branch, NASA Goddard Space Flight Center, Greenbelt, MD 20771, jeffery.g.masek@nasa.gov
Forest height is a key structural attribute for estimating carbon stock and standing biomass. In general tree height can be determined reliably using lidar measurements. For most areas, however, such lidar measurements do not exist or are inadequate for developing spatially contiguous maps. While Landsat can provide spatially contiguous data for most land areas, it has been demonstrated in most studies that tree height is not highly correlated with spectral data in individual or limited Landsat acquisitions. To overcome these limitations we have developed an approach for mapping forest height using ICESat GLAS lidar data and Landsat time series stacks (LTSS). In this approach, LTSS are used to separate “old” forests that remained forested during the entire observing period of each LTSS from “young” forests that are regenerating from disturbances that occurred during that observing period, which are detected using a vegetation change tracker (VCT) algorithm. The GLAS lidar data are used to determine forest height at the footprints of the lidar data. These GLAS samples are then used to develop empirical relationships between GLAS height measurements and LTSS-VCT products. For the “young” forests whose age since disturbance can be calculated using the LTSS-VCT products, height is modeled using age since disturbance and post-disturbance spectral trajectories. The predicted values were found highly corrected with those derived from GLAS lidar data (R2 ~ 0.9, standard error ~ 2 m). For the “old” forests whose age can not be derived from the LTSS-VCT products, only Landsat data are used in the models. Still, good relationships (R2 ~ 0.73, standard error ~ 3.4 m) were found between predicted height values and those derived from GLAS data when all Landsat observations in the LTSS were used as model input. A wall-to-wall forest height map is created by combining the model predictions for both “young” and “old” forests. This approach has been used to produce a wall-to-wall forest height map for the state of Mississippi.
Presentation Type: Poster
Poster Session: Carbon Cycle Science
NASA TE Funded Awards Represented:
Huang, Chengquan
Integration of long term Landsat observations with DESDynI measurements for monitoring terrestrial carbon flux within and beyond the DESDynI mission