Application of Association Rule Mining for Exploring the Relationship between Urban Land Surface Temperature and Biophysical/Social Parameters

This paper explores the relationship between remote sensing measurements of land surface temperature and biophysical/socioeconomic data by utilizing the association rule mining technique. The surfaces associated with urban uses typically radiate more heat as compared to its rural counterparts. There is a need to quantitatively analyze this contrast in temperature and the biophysical and social characteristics which influence it. Furthermore, in order to consider the urban heat island (UHI) effect, a parameterization is required to account for the urban surface characteristics impacts on the magnitude of land surface temperature (LST). The association rule mining model has demonstrated to bring in additional quantitative information concerning the relationships among urban parameters. The ASTER data from 2000 was used for the selection of appropriate variables to be used in the model. This information was then used for generating association rules between land-use land-cover (LULC) and LST information from 2000, 2001, and 2004. The results thus obtained quantitatively described the relationships between various urban parameters. It was found that there was little change in the percentage area of the LULC types from 2000 to 2004. This made the comparison of the results possible. In the case of the 2000 data, it was found that forest and impervious surfaces had strong association with temperature and scaled normalized difference vegetation index (SNDVI). Specific zones such as hospitals and universities had negative association with water. The comparison of data from 2000, 2001, and 2004 suggests that impervious surface and the zoning category of airport had a strong association. Nevertheless, the information extracted needs to be analyzed in greater detail in order to arrive at robust decision rules. Overall, the model so developed has demonstrated to be effective in predicting associations between urban LST and pertinent factors. This model could be useful for urban planners and environmental managers in quantifying rules that characterize a particular urban landscape. Introduction In the United States, the current urban growth rate, based on 1990 and 2000 census figures, is approximately 12.5 percent, Application of Association Rule Mining for Exploring the Relationship between Urban Land Sur face Temperature and Biophysical/Social Parameters Umamaheshwaran Rajasekar and Qihao Weng with 80 percent of population residing within urban areas. As cities continue to grow, urban sprawl creates unique problems related to land-use, transportation, agriculture, housing, pollution, and development for policy makers (Shepherd and Burian, 2003). Demonstrated links exist between land-use land-cover (LULC) and temperature (Hawkins et al., 2004, Weng et al., 2004, Kalnay and Cai, 2003). There has been a significant amount of research conducted using thermal infrared measurements from ground and remotely sensed data with respect to urban and suburban characteristics leading to the well-known urban heat island (UHI) effect (Voogt and Oke, 2003; Chudnovsky et al., 2004; Hung et al., 2006; Khaikine et al., 2006). Furthermore, changes in LULC and their impact on the UHI effect has been demonstrated by Golden (2004), Ghiaus et al. (2006) and Weng et al. (2004). Factors such as urbanization, the conversion of other types of LULC associated with growth of population, and the economy impact the overall land surface temperature (LST) change in a specific area. These characteristics when amalgamated, contribute to the increase in the spread of UHI effect. This relationship between LULC and LST change has been measured qualitatively and has been characterized as overall change at a micro level using ground sensors during the day and night time. Fast et al. (2005) elegantly detailed the quantifiable parameters by installing data loggers at specific locations within a landscape. In this study, they attempted to determine the interaction of the physical variables with the environmental variables. Stathopoulou et al. (2004) studied the spatial temperature distribution and the intensity of the UHI of selected cities in Greece. Over the last three decades, great progress has been made with the advent of the space program associated with Earth observations (Kafatos et al., 1998). Advancements in the field of UHI have been well reviewed by Arnfield (2003). Terra bytes of data are being collected from satellites, aerial sensors, telescopes, and other sensor platforms (Tan et al., 2002). Some of the data collected suggests that skin temperatures for large areas could be mapped and studied more effectively by using satellite remote sensing data in the infrared region as compared to ground based sensor (Stathopoulou et al., 2004). PHOTOGRAMMETRIC ENGINEER ING & REMOTE SENS ING Ap r i l 2009 385 Center for Urban and Environmental Change, Department of Geography, Indiana State University, 177 Science Building, Terre Haute, IN 47809 ([email protected]). Photogrammetric Engineering & Remote Sensing Vol. 75, No. 3, April 2009, pp. 385–396. 0099-1112/09/7503–0385/$3.00/0 © 2009 American Society for Photogrammetry and Remote Sensing 385-396_07-053.qxd 3/19/09 1:13 PM Page 385