2001 NLCD Accuracy
The MRLC consortium expects that the new, improved mapping methods used for NLCD 2001 will result in improved accuracy of the land-cover data. Accuracy assessment is needed to document whether or not that expectation has been met. In addition, there is a proliferation of accuracy assessment objectives that arises from the development of new database elements (impervious surface, canopy density, land-cover change) and the anticipated widespread use of the data (link to NLCD 1992 applications table). Recognizing the proliferation in accuracy assessment objectives, MRLC has developed a design that will accommodate most if not all of the following objectives:
- Per-class thematic accuracy of land cover
- Accuracy of impervious surface and canopy density
- Accuracy of land-cover change, and
- Accuracy of land-cover composition.
The last objective was included to address a common application of the NLCD data (link to NLCD 1992 applications table) - summary estimates of the proportion of land-cover classes by watershed, county, or other summary unit. The basic geographic structure of the design is shown below. On average, several samples will be collected inside each of the solid black squares, which will allow us to address the multiple objectives described above. Users should note that underwriting for accuracy assessment is often insufficient (see Scepan (1999), Photogrammetric Engineering and Remote Sensing 65:1051-1060). Because of this reality, MRLC has prioritized the four objectives. Thematic accuracy assessment the NLCD 2001 land-cover data has been given the highest priority. Samples have been drawn for region 8, outlined in cyan in the figure below, and reference data collection is underway. Other objectives will be addressed when and if funding becomes available.