A Method of Aggregating Heterogeneous Subgrid Land-Cover Input Data for Multiscale Urban Parameterization

Abstract

method of representing grid-scale heterogeneous development density for urban climate models from probability density functions of subgrid-resolution observed data is proposed. Derived values are evaluated in relation to normalized Shannon entropy to provide guidance in assessing model input data. Urban fraction for dominant-class and mosaic urban contributions is estimated by combining analysis of 30-m-resolution National Land Cover Database 2006 data products for continuous impervious surface area and categorical land cover. The aim of the method is to reduce model error through improvement of urban parameterization and representation of observations employed as input data. The multiscale variation of parameter values is demonstrated for several methods of utilizing input. This approach provides multiscale and spatial guidance for determining where parameterization schemes may be misrepresenting heterogeneity of input data, along with motivation for employing mosaic techniques that are based upon assessment of input data. The proposed method has wider potential for geographic application and complements data products that focus on characterizing central business districts. It utilizes observations to obtain a parameterization of urban fraction that is dependent upon resolution and class-partition scheme, thus providing one means of influencing simulation prediction at various aggregated grid scales.