Jump to main content.
Virtual Field Reference Data Base
OVERVIEWThe availability of high quality reference data is essential to the performance of statistically rigorous accuracy assessments of land-cover (LC) maps derived from digital satellite remote sensor data. A quantitative validation using an independently derived LC map can only be realized when the reference data are acquired to correspond with the following criteria: (a) coincide with the temporally relevant period of interest; (b) spatial data resolution is commensurate with that of the data being evaluation; (c) an unambiguous match between classification systems exists or can be obtained; and (d) the variability of the reference data set has been characterized. To generate a reference data set with known variability we developed a AVirtual Field Reference Data Base (VFRDB)@ to provide a versatile source of field measurement data. The goal of the VFRDB was to provide an independent source of in situ measurement data that could be adapted to various classification systems to provide a robust source of reference data with known variability to assess the accuracy of current map products, and to provide a data archive for the future testing and evaluation of remote sensing based LC conversion methodologies within the Neuse River Basin (NRB) of North Carolina.
Four major elements contributed to the development of the NRB-VFRDB. These included the development of: (1) the sampling frame design; (2) field sampling protocols; (3) quality control (QC) and associated standard operation procedures (SOPs); and (4) the database design.
Sampling Frame Design
The sampling frame for the NRB-VFRDB was based on three separate design elements to provide adequate watershed and riparian buffer zone sampling points. The sampling design included the following three elements: (1) systematic unaligned samples based on the USGS Quarter Quadrangles to provide uniform distibution; (2) stratified random samples to capture rare resources; and (3) multiple stratified random samples for the characterization of riparian buffer zones.
Field Sampling Protocol
Field sampling was accomplished during May through September of 1998 and 1999. Field crews navigated to sampling points using Global Positioning Systems (GPS) operating in real-time (satellite broadcast) differential corrected mode. Sampling point locations were located to within a one-meter accuracy to provide for future revisit capability. Circular plots with a radius of 36.5 meters were measured and flagged to provide 0.42-hectare plots corresponding to the Minimum Mapping Unit (MMU), used for the development of LC maps. Subsequent plot delineation field measurements were made corresponding to location and time, physical parameter and biophysical measurements.
Location and Time. For each sampling point, data were recorded including the unique site identification number assigned, the corresponding USGS 1:24,000-scale quadrangle, calendar date and time of observation, crew members, the original site x,y coordinates and the actual coordinates of the sampling site. The differentiation between the original and actual coordinates was required because the protocol mandates the relocation of sampling sites under certain circumstances. For example, if greater than 20 percent of the area within the original plot coordinates contains a completely different land cover type (i.e., forest and agricultural), then two sampling sites are established within the two land-cover types.
Physical Measurements. Measurements corresponding to physical parameters included slope, aspect, elevation, landform, percent cover of predominate land cover type, water regime, soil moisture condition (descriptive), diameter breast height (DBH), and percent canopy cover were determined for each site, as appropriate. DBH was determined using direct measurement, tree heights were determined by triangulation, slope was measured using a clinometer, and percent canopy using both the vertical tube and spherical densiometer techniques.
Biophysical Measurements. The suite of biophysical site measurements included general site descriptors such as ecological system type, general vegetative cover type, successional state, and origin (i.e., natural or man-made). Detailed inventory of crop types, ground cover, shrubs, under-story tree and tree canopy constituents and relative abundances were determined for each site, as appropriate. Both genus and species identifications were made in most cases, except where species specific identifications were not possible.
Digital Photographic Documentation. A high-resolution (1040 x 840 pixels) natural color imagery series was acquired for each site. At a minimum, documentation included a panoramic site documentation set of six images. Also, as appropriate, a four image canopy image series and up/down stream two image series were collected. Additionally, representative vegetation key graphics were included in the database to aid users unfamiliar with NRB flora.
Field Data Sheets
Field data sheets were designed to capture pertinent location and time, physical, and biophysical attribute data to provide measurement and imagery (camera) data to document LC characteristics across the NRB. The basis for paper field sheet(s) and complementary computer sheet(s) designs were to develop an unambiguous record system for use by filed interns and to provide sufficient data to insure a high degree of data quality control. The ultimate objective was to provide a high quality source of data for subsequent interpretation by any potential end user.
Quality Assurance and Quality Control
A high priority was placed on Quality Control (QC) during the development of the NRB-VFRDB to provide Quality Assurance (QA) relative to data completeness, accuracy, and consistency between field crew members. QC elements included two weeks of combined classroom and field training, continuous evaluation of field methods and plant identifications throughout the field deployment, weekly independent confirmation of plant identifications, and independent review of field data forms and imagery. Independent species confirmations were made by the herbarium at North Carolina State University.
Database Design and FunctionThe database for the NRB-VFRDB was designed to provide comprehensive data storage, analysis, quality control, and display capabilities. Tight integration of the design elements listed above provides the following functionality.
Data Entry. The VFRDB was designed for both rapid and accurate data entry. The computer data forms use a sequence of tabbed pages so as to eliminate the need for scrolling through a single page to view all of the data fields.
Geographical Analysis. The NRB-VFRDB has the capability to export attribute data in a format readable by GIS software. The dynamic link between spatial and attribute information allows for extended geographic analysis of the data. Ancillary data themes (soil maps, for example) can be incorporated into the GIS to supplement data collected in the field. This gives near limitless analysis potential.
Data Display and Distribution. In addition to the display capabilities of the data forms and through the GIS software link, the NRB-VFRDB has been ported to a web/database server for distribution via the EPA Intranet.
ftp Access - Coming Soon!
Jump to main content.