Remote Sensing for Forest Dynamics and Its Implications for Tree Outside Forest over Maryland, U.S.A.
Quan Shen, University of Maryland, qshen@umd.edu (Presenter)
George Hurtt, University of Maryland, gchurtt@umd.edu
Lei Ma, University of Maryland, lma6@umd.edu
Rachel L Lamb, Maryland Department of Environment (DEP), rachel.lamb@maryland.gov
Ralph Dubayah, University of Maryland, dubayah@umd.edu
Jarlath O’Neil-Dunne, University of Vermont, jarlath.oneil-dunne@uvm.edu
Matthew Hansen, University of Maryland, mhansen@umd.edu
Chengquan Huang, University of Maryland, cqhuang@umd.edu
Elliott Campbell, Maryland Department of Natural Resources, elliott.campbell@maryland.gov
Reem Aida Hannun, University of Pittsburgh, reem.a.hannun@nasa.gov
Nancy L Harris, World Resources Institute, nharris@wri.org
Haley Leslie-Bole, World Resources Institute, haley.leslie-bole@wri.org
Susan Minnemeyer, Chesapeake Conservancy, sminnemeyer@chesapeakeconservancy.org
Forest conservation and restoration are important goals for ecosystem services and climate mitigation. In the state of Maryland, USA, the Forest Conservation Act and Greenhouse Gas Emissions Reduction Act set legal limits on forest extent, and track carbon emissions from forest dynamics. In this study, we compared four remote sensing datasets for forest extent, tree cover, and disturbance. Three of the datasets were multi-temporal land cover datasets based on 30-m Landsat imagery (GFW, NAFD and NLCD). One was a one-time tree cover map derived from a combination of 1-m resolution airborne lidar data and NAIP optical imagery (CMS). We found that the 1-m lidar-derived dataset captured ~20% more forest area and tree cover area in 2010 than the other three Landsat-based datasets, implying the spatial resolution of data source (1-m vs 30-m) primarily determines the ability to fully capture tree cover.
The lidar-derived dataset detected >2,000 km2 tree cover over areas which were not classified as forest in a Landsat-based dataset (trees outside forests, TOF). The three Landsat-based datasets were comparable regarding forest extent and tree cover area but remarkably differed in forest disturbance rates between 2000 to 2010 (i.e., 0.28 ~ 0.8%/yr). We found that challenges exist in comprehensive monitoring of forest dynamics with these four datasets due to limitations including large areas of unclassified TOF and its unknown temporal dynamics. Future efforts are needed to develop a high spatial and temporal resolution remote sensing products for more comprehensive tree cover detection, and to integrate these datasets with models to determine the consequences for carbon.
Poster: Poster_Shen_2-43_113_35.pdf
Associated Project(s):
Poster Location ID: 2-43
Presentation Type: Poster
Session: Poster Session 2
Session Date: Wed (May 10) 5:15-7:15 PM
CCE Program: TE