Principal Investigator (PI): Matthew Hansen, University of Maryland

Vegetation continuous fields (VCFs) depict the sub-pixel fraction of three land cover types: tree cover, bare ground and non-tree vegetation. The tree cover type is further partitioned into fractions of needleleaf, broadleaf, evergreen and deciduous tree cover. VCF depictions improve upon traditional land cover characterizations in which each pixel is assigned a single land cover class. Earth System Data Records (ESDRs) are a consistent time series of observations over decades, developed by NASA to answer critical Earth science questions.

It has been recognized by the science community that VCFs are an important and basic measurement of the land surface that should be included in the ESDR compilation. VCF products have been adopted by a significant and growing number of Earth science researchers, including use in global and regional climate models, global terrestrial carbon cycle models, hydrology, fire ecology, habitat studies and agricultural monitoring. Researchers frequently express the need for a longer, consistent time series suitable for identifying land cover change and for validating models.

The proposal team has successfully created global VCF products, first single products from Advanced Very High Resolution Radiometer (AVHRR) and Moderate Resolution Imaging Spectroradiometer (MODIS) data, and recently, annual products from MODIS Collection 5 at 250m resolution. The regression tree methods developed for these efforts are robust across temporal and spatial resolutions, and satellite instruments. In addition, we have developed reusable, flexible code, built upon open source libraries, which automates model building and data processing.

Building on the heritage of previous Making Earth System Data Records for Use in Research Environments (MEaSURES) projects, we will use the Land Long Term Data Records (LTDRs) of daily surface reflectance and normalized difference vegetation index (NDVI) to provide a coherent input data set. These new long-term data sets will provide the core products needed to create consistent multi-instrument VCF LTDRs from 1981 to the present.

We propose to create ESDRs of global vegetation continuous fields from AVHRR and MODIS LTDRs for 1981 through the most current data available. Annual data products of percent tree cover, percent non-tree vegetation cover and percent bare ground at 0.05 degree resolution will be produced for all time periods. We will also create five-year products that further partition tree cover into needleleaf, broadleaf, deciduous and evergreen classes. A supervised regression-tree method will be used with phenological and physiological metrics derived from the LTDRs and with training derived from Landsat supplemented by high-resolution data from Ikonos, RapidEye and QuickBird. A reference data layer of fractional water cover will be included in the product, and error estimates and quality flags will be provided for each data product pixel and for the product as a whole. Software and data developed for this project will be freely available to the scientific community.

Because of the extensive use of VCF representations by the carbon modeling community, we propose to evaluate the implications of using the VCF ESDRs in the CASA-GFED3 carbon cycle model. We will use the extended VCF time series to prescribe biomass allocation patterns in CASA-GFED3 and determine the sensitivity of modeled carbon fluxes and carbon flux uncertainties to the new VCF data.