Disaggregation of an Urban Population with M_IDW Interpolation and Building Information

Abstract

Human population distribution represents critical information used for urban planning, sustainable development, and human–environment interaction studies. A number of models for the mapping of population distribution have been developed at different scales (e.g., global, regional, and urban), but their respective performance may not be satisfactory for all applications, and methods that provide high accuracy at the scale of a built-up urban area are still rare for cities in developing world. This paper proposes a new method known as the modified inverse distance weighted model (M_IDW) for disaggregation of the population in a built-up urban area using only the geometrical information from ground-level houses as ancillary information. In particular, the method considers the population distribution in the target area as synthesized and mosaicked with a homogeneous density of houses and their associated land patches in addition to an inhomogeneous density for the outdoor areas. The authors chose the built-up area of Shanghai as the case study area, and geographic information system and remote sensing images processing software were used to derive the ancillary information. Subsequently, an interpolation routine was developed using the interface definition language (IDL) 8.0, and the resulting performance of the new method was compared against that of two other approaches. The results indicate that: (1) the proposed M_IDW is able to represent a detailed distribution of information from a population in a built-up urban area and yields good performance when applied to a single-type auxiliary data; (2) the traditional inverse distance weighted model interpolation (T_IDW) cannot be directly applied for population interpolation at the scale of a built-up urban area; (3) the inverse of the national population aggregation (INPA) method and the M_IDW interpolation can obtain detailed spatial information from a population distribution on a highly refined grid scale, and both methods produce a considerable and equivalent accuracy; however, the M_IDW has a greater ability to depict the details of the population distribution than the INPA. This novel method can substantially improve the disaggregation accuracy and spatial details that are important in this context.