Multiscale Assessment of Binary and Continuous Landcover Variables for MODIS Validation, Mapping, and Modeling Applications

Validation, mapping, and modeling efforts require accells. We recommend that aggregation of fine-grain measurements during validation of the Moderate Resolution curate methods to transform process rates and ecosystem attributes estimated from small field plots to the 250– Imaging Spectrometer (MODIS) products be based on continuous variables to reduce errors that originate from 1000-m-wide cells used by a new generation of land cover mapping sensors. We provide alternative scale uncertainties in binary maps. Elsevier Science Inc., 1999 transformations, each with attendant assumptions and limitations. The choice of method depends on spatial characteristics of the land cover variables in question and consequently may vary between biomes or with the intended application. We extend the fractal similarity diINTRODUCTION mension renormalization method, previously developed Accurate determination of global net carbon flux is perfor binary maps, to continuous variables. The method haps the most urgent issue facing the ecological commucan preserve both the mean and the multifractal propernity (Churkina and Running, 1998). Loading the atmoties of the image, thereby satisfying a major goal, namely, sphere with carbon in excess of the biosphere’s steady to provide accurate areal estimates without sacrificing instate cycling capacity has many well-known implications, formation about within-site variation. The scale transforincluding increases in global temperature (Schlesinger, mation enables the multifractal scaling exponents of land1991). The capacity of the biosphere for carbon involves scapes or individual spectral bands to be brought in and interactions among various processes such as water and out of register with each other, thereby opening another energy supply, nutrient cycling, production, storage, and dimension upon which to detect the scales at which varirespiration. Remote sensing combined with process modous land use or terrain processes operate. Alternatively, eling and field validation offers the only currently feasilandscapes can be selectively rescaled to highlight patble approach for assessing ecosystem capacities for carterns due to particular processes. We recommend geosbon over vast regions. However, many crucial tatistical procedures with which to assess spatial characassumptions are made in the course of relating field teristics both within a site and within individual image measurements to the grid cells used in remote sensing (Box et al., 1989). Accuracy assessment is virtually impossible without expressions for the discrepancies between *Department of Biology, University of New Mexico, Albuquerque estimates obtained from field plots and from remotely †USDA Forest Service, Pacific Northwest Research Station, Forestry Sciences Laboratory, Corvallis, Oregon sensed imagery. The problem is compounded when the Address correspondence to B. T. Milne, Department of Biology, grid cells are large (60–1000 m) as is the case for the University of New Mexico, Albuquerque, NM 87131. E-mail: bmilne@ next generation of sensors, such as the Moderate Resolusevilleta.unm.edu Received 15 May 1998; revised 30 March 1999. tion Imaging Spectrometer (MODIS; Ustin et al., 1991).